I want to try to answer Simon Wren-Lewis' good question. Let me first restate his question in my own words. (Always a good way to begin, to see if I've understood him properly.)
Take a very simply model of an economy, in which a government issues a nominal stock of (high-powered) money M, and a nominal stock of bonds B. Money pays interest Rm (usually Rm=0 for currency, but I want to keep this fairly general), and bonds pay interest Rb.
We can write the model as:
1. M/P = L(Rm,Rb,stuff) real money supplied = real money demanded
2. B/P = H(Rm,Rb,stuff) real bonds supplied = real bonds demanded
3. dM/dt + dB/dt = P.deficit the nominal deficit is financed by issuing a flow of money and/or a flow of bonds
The first thing you should notice is the symmetry. There is nothing in the above three equations that tells us that P is determined by M rather than by B. Only if H( ) were a degenerate function (as it would be under Ricardian Equivalence, where 2 collapses to Rb = some function independent of B/P) would we say that P is determined by M not B.
The second thing you should notice (and this is where Simon's question comes in) is that if you have a central bank setting Rm and Rb you can ignore equation 1 and determine P from equation 2 alone. Even if the central bank does adjust Rb by adjusting M according to equation 1. Equation 1 tells you what M is, but you only need Rm, Rb, B, and equation 2 to solve for P. This is now the Woodfordian Neo-Wicksellian orthodoxy.
So why do some of us insist on the importance of M in determining P?
Here is my inadequate and unclear answer:
I don't like equation 1. It is heresy to say this, but there is something seriously wrong with equation 1, precisely because it treats the demand and supply of the medium of exchange exactly like the demand and supply of any other asset, like bonds, or houses, or antique furniture. The symmetry between equations 1 and 2 is a false symmetry.
Take an extreme case. Suppose we considered a government that did not issue money and could only issue bonds. A local or provincial or Eurozone national government, for example. So we can delete M, Rm, and equation 1 from the model. How many degrees of freedom would such a government have? It would have one degree of freedom. It could choose the deficit, or it could choose Rb, but it cannot choose both. If it chooses the deficit, it must pay a rate of interest on its bonds that is determined by the market (equation 2) if it wishes to persuade people to buy the requisite number of bonds. Or if it chooses the rate of interest it pays on bonds, the quantity of bonds it sells will be determined by the market, which in turn determines the deficit it can finance.
A government that can only finance a deficit by selling bonds must persuade people to buy its bonds by offering a sufficiently high rate of interest on those bonds to make them willing to hold the additional bonds it needs to sell.
Now let's take the opposite extreme case. Suppose we consider a government that never issues bonds and that finances its deficit only by issuing money. So we can delete B, Rb, and equation 2 from the model. How many degrees of freedom would such a government have? I say it would have two degrees of freedom. It can choose the deficit, and it can choose Rm. Take a simple example: suppose that M is currency, and it is administratively impractical to pay interest on currency (and you can even suppose expected inflation is always zero, if you are worried about the distinction between real and nominal interest rates). If you believed that equation 1 is symmetric with equation2, if in other words you believed that money was just another asset, like bonds, you would be forced to conclude that a government that could only issue money could never run a deficit, unless people wanted to hold more money. You would be forced to conclude that deficits are determined by the demand for money. And Zimbabwe would tell you you were hopelessly wrong.
A local government, which can only issue bonds to finance its deficit, would not be able to run deficits if it were unable to offer a competitive rate of interest on its bonds sufficiently high to persuade people to want to hold its bonds. Zimbabwe couldn't get anyone to buy its bonds, but could run deficits financed by printing money without any thought of offering a competitive rate of interest on its money. It paid zero percent interest nominal on its currency, and in real terms an interest rate that was very very negative. Yet Zimbabwe could still run a deficit. No local government with a binding constraint on the rate of interest it paid on its bonds could have done the same thing.
The symmetry between money and bonds in equations 1,2,and 3 is a false symmetry. Money is different from other assets. Every individual will accept more money in exchange for goods and services even if no individual wants to hold more money; that's because money is the medium of exchange, and by definition people accept a medium of exchange because they plan to pass it on to someone else, who in turn plans to pass it on, like the proverbial hot potato. Any increase in the stock of money is supply-determined; any increase in the stock of all other assets is determined by whichever is less: supply or demand.
A government that issues only money can issue more money any time it wants, even without changing Rm, and so P will be forced to adjust to eventually ensure that equation 1 holds.
A government that issues only bonds can only issue more bonds if it changes Rb to ensure that equation 2 holds and the demand for bonds increases in proportion to the increased supply, so there is nothing that would force P to adjust.
I think that's the gist of it. Not as clear as I want it to be, because I'm still thinking this through myself. (Laidler and Yeager have been the biggest influences on my thinking here.)