I posted a comment on another blog a little while ago and, as it covers my (current) position on the question of whether bank deposits are money, I thought I would repost it here (after tidying up the spelling).

This one largely follows an old post here – but does expand in some ways on it.

I do not want to labour this point and I would be the first to say that my opinion is not in full agreement with the mainstream of economics. It does, however, come out of a long history of working in bank treasury functions in several parts of the world.
The question of whether or not banks create money revolves around whether or not a bank deposit counts as “money”, properly so called. Personally, I doubt it is, as I cannot conduct payment transactions with my bank account – I need to be able to withdraw the funds first. This is a contractual question between me and the bank. Most of the time the bank will honour the demand to pay the funds, so under most circumstances the funds in my bank account are a close cash substitute and can therefore be regarded as money, and I agree with Mark above on that. In extremis, though, the bank may not deliver the funds – so bank accounts are clearly less liquid than cash. That situation is only likely to occur when the bank itself cannot deliver on the demand for cash – i.e. there are more demands for withdrawals than the individual bank has access to cash to meet those demands.
Really, then, the total pool of liquidity represented by any bank is not the total amount on deposit with the bank, but the (much smaller) amount they have in ready cash or other sources of funds to meet calls for withdrawals.
You are right that a loan disbursement (or deposit withdrawal) from one bank will often, even usually, result in a deposit with either that bank or another – but this is generally not immediate, so the withdrawal or disbursement is separated by time from any re-deposit, nor certain. It may be that the funds withdrawn are not redeposited in any bank for some time – they could be kept under the bed, used multiple times for transactions or even go offshore.
The point here is that a particular bank cannot be sure where any disbursed loan funds may end up and can only count on a small proportion returning. This means that they cannot simply inflate their balance sheet by making loans and counting on the funds to come back as deposits. Additionally, if they did this their liquidity ratios (one of the key measures of any bank’s soundness) would steadily deteriorate, restricting their ability to make further loans. Their capital ratios would also steadily deteriorate, reducing their ability to raise funds.
To me, then, the question of whether you measure the total money supply to include or exclude deposits in the banking system is really irrelevant. That may well be a standard (and useful) measure for economic policy questions. However, to me, the total amount of funds actually available for transaction purposesin an economy does not include bank deposits (whether chequing or otherwise) but does include total liquid funds (cash, government bonds etc.) held by the banks – a much smaller figure. It is not increased by the actions of banks in depositing and lending – in fact it is reduced by the amount of the solvency ratios.
I appreciate that this is not purely mainstream, but I know how these things operate in every bank I have ever worked in.
You may (OTOH) argue that, in the event of a likely bank failure to deliver due to insolvency, the government is likely to step in – the (either implicit or explicit) “printing press” contingency. That may well be so – but if you are going to count the likelihood of the government stepping in and printing as part of the money supply then that should be explicitly acknowledged, not implicitly included by counting all bank deposits. That then becomes further government created money, not bank

created money.