Nice title, huh? I'll try to explain.
Many people have argued that a "bank loan" is not really a loan, because the bank is not lending any existing money; it creates the money it "lends" on the "borrower's" account and the money is destroyed when the loan is repaid. Still, from the bank customer's perspective, it looks like a loan: he receives money on his account ("from the bank"), spends the money and has to get the (same sum of) money back on his account to be able to return it ("to the bank") when the loan becomes due.
Not so in the case of an overdraft.
By agreeing with his bank on an overdraft facility (a "credit line") connected to his checking account, the customer can spend money he doesn't have. The customer incurs debt by using the overdraft – in the textbook case – to purchase goods and repays that debt by selling goods.
I cannot stop people imagining that there was money first borrowed and later returned in the case of an overdraft, too. One can describe an overdraft in terms of what I call a "traditional bank loan".
What I want to argue here is that this works the other way, too; that one can describe a traditional bank loan in terms of an overdraft. Furthermore, I argue that we can consider an overdraft – I'm talking about the accounting treatment – as the "base case" and view the accounting treatment of a traditional bank loan as an unnecessary complication. (If this sounds unnecessarily provocative, take it as an argument against any kind of primacy, other than chronological/historical, given to the traditional bank loan. An overdraft doesn't involve "netting" something more than a traditional bank loan involves "grossing" something.)
Imagine that the accounting treatment of all kinds of debt had always been like it is in a case of an overdraft. (Perhaps it's just a historical accident that it hasn't been?) The credit contract – which should be viewed separately from the accounting treatment – could state that the overdraft limit is to be reduced monthly by a certain sum and that the limit will reach zero in, say, 10 years (for a 10-year "loan"). The contract would also state when interest is paid (through a debit entry on the checking account, as usual).
To help you imagine that kind of world, I have created a document where I map the corresponding figures in the two alternative accounting treatments. The color of the circle tells which figures correspond. (Please note that where I write "Unused overdraft" it should actually say "Unused ovedraft + initial checking account balance". In that way it would also cover the most common scenarios where the initial balance is not zero but positive.)
Had the accounting treatment always been that of an overdraft, not of a traditional loan, the words and concepts we use might differ from what they are today.
For instance, we would repay debt when we sell goods (item #3 in my document), not when the overdraft/credit limit is reduced (item #4). The latter is how we see repayment when we think of a traditional loan. If the accounting treatments are identical in their outcome, then why this discrepancy?
As far as I know, this discrepancy is one of the main reasons why Nick Rowe thinks an overdraft is "negative money" while a traditional bank loan isn't. Another reason perhaps has to do with the maturity of the credit limit, but that is something which is specified in the credit contract, regardless of the accounting treatment. There are bank loans without any clear, pre-agreed maturity, just like there are overdrafts with a pre-agreed maturity. And even if the maturity is specified in the contract, there is nearly always a lot of flexibility involved – after all, it is quite often in the bank's interest that a customer repays his debt more slowly.
If we had never seen a "traditional bank loan", would we talk about the bank lending us money, or would we only talk about the bank extending credit to us? Would we still think that there is money on our account, and that we need that money to pay for things? Or would we just talk about "the number" on our account?
The world of overdrafts looks quite different from the world of traditional bank loans – at least to me and, so I've understood, Nick Rowe. If you don't see the difference, it might be because you try, quite successfully, to interpret overdrafts in terms of traditional bank loans. (Or then you're JKH, who is probably comfortable in both worlds.)
 I'm talking about a pre-authorized overdraft, which is in practice more common in Europe than in the US. One can also "overdraw" one's checking account without any prior agreement with the bank, but in that case there will usually be penalty charges.
 Alternatively, we could say that the unused overdraft is money he has, although it doesn't reside on his account – and never will.
Keynes (“A Treatise on Money”, Bk 1, Ch 3, 8ii(i.) “Deposits and Overdrafts”) was of the opinion that unused overdrafts should, logically, be included in monetary aggregates:
... it is the total of the cash-deposits and the unused overdraft facilities outstanding which together make up the total of Cash Facilities. Properly speaking, unused overdraft facilities – since they represent a liability of the bank – ought, in the same way as acceptances, to appear on both sides of the account. But at present this is not so, with the result that there exists in unused overdraft facilities a form of Bank-Money of growing importance, of which we have no statistical record whatever, whether as regards the absolute aggregate amount of it or as regards the fluctuations in this amount from time to time.
Thus the Cash Facilities, which are truly cash for the purposes of the Theory of the Value of Money, by no means correspond to the Bank Deposits which are published.
I see that we have two options if we want to be logical: either unused overdrafts are money, too, or then "cash-deposits" are not money, either. I opt for the latter, not least because there is only a fine line between a pre-agreed unused overdraft and an overdraft – or a "traditional bank loan" for that matter – that the bank would be willing to provide if only the customer wanted it. Trying to find out the size of Keynes' "Cash Facilities" is much like trying to find out the total amount of purchasing power in the economy – a hopeless task, if you ask me.
 He could be buying and selling financial assets, but it's best to leave that alternative out of scope for now. The interpretation of that case is somewhat more complicated within the broader framework/theory I'm promoting in my posts. One day I'll get back to that.