Looking at Fickett's Example

In my slow non-economist way, I've been pondering the following explanation of borrowing's impact on the economy (from the previously linked GDP growth is compatible with shrinking credit):

Imagine a person who has an income of $5000/month. If this person has no debt and no savings, he can spend $5000/month. Suppose, now, that he wants to spend $6000/month. He has to borrow $1000/month, so, in order to maintain a fixed spending level, his debt has to steadily increase. In this example, a change in the level of spending requires not a one-time change in the level of credit, but a change in the growth rate of credit. So, again, in this very simplified case,

* A level of spending corresponds to a growth rate of credit
* A change in the level of spending requires a change in the growth rate of credit

This gets more interesting when we look at a case where credit is shrinking. Imagine a person who is earning $5000/month, using $2000/month to pay down debt, and spending $3000/month. Next suppose that this person decides to only pay down $1000/month while spending $4000/month. Both before and after the change, credit for this person was shrinking. But it was shrinking less quickly after the change than before. Here again, an increase in the level of spending required a change in the growth of credit, in this case from more negative to less negative.

* An increase in the level of spending can come about either through an increase in the growth of credit or through a decrease in the shrinkage of credit

I believe that Fickett is guilty here of a fallacy of composition. While our hypothetical person only borrows and spends each extra $1000/month once, the money is then continually respent by others in the economy. It becomes a semi-permanent part of the money supply, and circulates. So borrowing $1000 and spending it causes $1000 worth of growth in the economy (factored by the velocity of the money).

Of course, this makes the question of why the "impulse" correlates with GDP growth even more interesting. As Biggs et al are careful to point out in their paper, correlation is not causation.