1. When I talk about debt relations that are arranged through our banking/monetary system, I don't refer to any bounding, enforceable debt contracts between two pre-defined entities. Through the banking system, one debtor always faces millions of potential creditors, and one creditor always faces millions of potential debtors. A net debtor (i.e. someone with a net debit balance) doesn't need to find a net creditor in order to pay his debt; he only needs to find someone with the ability to credit the debtor's account (by debiting his own account). Likewise, a net creditor doesn't need to find a net debtor to get what is due to him personally; everyone accepts the credit balance (in bank ledger) which he has to offer in exchange for goods and services. I don't talk about a debt which would be always defined and enforced by law (as in a contract between two parties), but a debt that is slightly closer to reciprocity, "give and take", debt incurred and paid on the market. All this works, when it works, through transactions between free individuals, and the transactions are based on mutual consent. But the sums (prices) owed are clearly expressed in the abstract unit of account and law can be used to enforce these debts, so I'm not describing some abstract "debt" (eg, to one's country, forefathers or humanity in general). Schumpeter used the expression social bookkeeping system when he talked about something very similar to what I'm describing.
2. At macroeconomic/aggregate level, debt is paid only when a net debtor sells something to a net creditor. If he, instead, sells to someone who wasn't a net creditor before the transaction, then aggregate debt obviously stays the same (increase in buyer's debt matches decrease in seller's debt). Likewise, new debt is created when a buyer becomes a net debtor, or increases his net debit balance. If he bought from someone who was a net debtor himself, then aggregate debt obviously doesn't increase (as the seller's debt is reduced). But if the seller is not just paying his debt (i.e. he either is a net creditor or at least ends up with a net credit balance after the transaction), aggregate debt increases as a consequence of the transaction. Note: It is this macroeconomic viewpoint which is perhaps most important to me, because the pressing question (to me, the economic question of our generation) I've been trying to find an answer to is basically this: "What the hell is going on with the increasing private and public indebtness worldwide?"
3. I do not say that a commercial bank, or the central bank, or the government owes something. This is a point that seems to escape people, although I thought I had already clearly stated it. What I say --when I adopt the Bookkeeping View -- is that the "agents" with a net debit balance in a bank ledger owe something. I also say that if the government has a net debit balance (like it always seems to have, in form of cumulative deficits and current spending), then it is mostly the duty of taxpayers (there are other fees than taxes, though) to credit that balance (i.e. to pay the debt); in other words, government's debit balance is an indirect obligation of private agents. (This seems to clash with some people's view of the government; a view which suggests that the government is something separate from the people. I, instead, view the government as an entity formed by citizens, (trans)acting and signing contracts on behalf of citizens.)
4. Some have told me that a debit balance alone doesn't imply any real debt (obligation). They seem to back this statement by saying that the central bank/government doesn't owe anything (and I agree; see point 3 above). But by describing the Bookkeeping View, I try to substantiate my claim that a debit balance is always ultimately connected to a debt (perhaps not in a way explicit enough to some, but isn't it those previously hidden, underlying "rules" a scientist should try to discover?). I apply the Bookkeeping View to the real economy and study how these debit (and credit) balances behave. And by doing this I derive empirical evidence which suggest that all these debit balances actually behave as if they were part of a bookkeeping system, the purpose of which is to track real debt relations. It seems quite obvious to me that had there been a clear, explicit accounting link -- even defined by law, as some seem to demand -- between government spending (credit) and tax obligations (debit), then there would be absolutely nothing for a scientist to try to discover here.
5. Finally, Naryana Kocherlakota in Money Is Memory (1996) -- my understanding is that he builds his ideas in the article mainly on J. Ostroy's The Informational Efficiency of Monetary Exchange, 1973 -- has suggested something that seems like a clear step towards the Bookkeeping View. He writes (somewhat "wonkish", as Krugman likes to warn; I underline and comment in brackets):
Within the class of environments under consideration, I show that adding money may expand the set of incentive-feasible allocations. However, I also show that the set of incentive-feasible allocations can be expanded by adding collective memory [a "social bookkeeping system"] to these environments. Here, memory is defined to be a historical record which reports to any agent at any date the full histories of all agents with whom he has had direct or indirect contact in the past. The main result of the paper is that in all of these environments, the set of incentive-feasible allocations generated by adding memory contains the set of incentive-feasible allocations generated by adding money. In this sense, in all of these environments, money is merely a primitive form of memory.
There is a simple reasoning behind the main proposition. John and Mary meet. John has apples and wants bananas. Mary wants apples but doesn't have bananas. In monetary economies, this problem is solved by Mary's giving John money in exchange for apples. John then uses the money to buy bananas from Paul; if John doesn't give the apples to Mary, John doesn't get the money and can't buy the bananas from Paul.
But of course the money itself is intrinsically useless. In terms of the reallocation of intrinsically valuable resources, we can think about the situation as being one in which John is considering making Mary a gift [see my point 1 above: this is reciprocity] of apples. If he makes the gift, Paul will give him bananas in the future; if he doesn't make the gift, Paul won't give him the bananas [point 1: "give and take"]. The money that John receives from Mary is merely a way of letting Paul know that John has fulfilled his societal obligations and given Mary her apples [Here the Bookkeeping View disagrees strongly: John possessing "money" obviously doesn't prove that he has fulfilled his "societal obligations"; he could be in debt. In other words, the possession of "money" doesn't reveal anything about the holder's trading history].
Thus, if we account for the fact that money itself is useless, monetary allocations are merely large interlocking networks of gifts [sounds like a "pure credit economy", or a "complicated and perfected system of barter"]. The point of this paper is to show that these same reallocations of resources are feasible if agents knew the past history of all actions: Paul could react to different histories of gifts on John's part in the same way that he reacts to John's having different amounts of money. It follows that any function performed by money can be provided by an ability to access the pasts of one's trading partners, their trading partners, and so on [our banking system makes, or at least should make, it unnecessary for individuals to access each others' trading histories; but the system itself tracks our historical, cumulative debits and credits, incurred and earned, respectively, by trading with others].
I have mentioned this before: I think that A. Mitchell-Innes's view on "money" was very much in line with mine, but he obviously didn't put it in explicit enough terms (see Innes here and here). The future will show if I manage to do better.