A permanent increase in the money supply (or one that is expected to be permanent) will have a different, and bigger, effect today than a temporary increase in the money supply (or one that is expected to be temporary). To say the same thing a different way, an increase in the expected future money supply has an effect today. I am not saying anything novel or unorthodox; just reminding us of something we already knew. Yet this distinction between temporary vs. permanent increases in the money supply seems not to occupy our attention much in discussions of quantitative easing.
Take one example: Paul Krugman's response to my post (where I argued the uncontroversial point that Neo-Wicksellian models could not help us tell whether quantitative easing would work). Paul links to a 1998 paper he wrote. The model in that 1998 paper is what you get when you take an otherwise Neo-Wicksellian model and introduce money via a cash-in-advance constraint. The CIA constraint is non-binding today (the natural rate of interest is negative, and so the nominal rate of interest hits the zero bound) but will be binding in the future (when the natural rate of interest rises, and the nominal interest rate will rise above zero). Paul uses that model to argue that quantitative easing will have no effect.
I have argued that Paul Krugman's 1998 model, with its assumption of perfect credit markets (nobody is borrowing-constrained), and its assumption of only one rate of interest (ignoring the spectrum of yields, some well above zero, on assets of differing liquidity) stacks the deck against quantitative easing.
But even if we ignore all that, it is not correct that quantitative easing will not work, even in Paul's model. Temporary quantitative easing will not work; but permanent quantitative easing will work. A permanent 10% increase in the money supply will raise the future price level by 10% (the CIA constraint will be binding in future), which raises the expected rate of inflation today, and lowers the real rate of interest today, and impacts the consumption-Euler equation today, resulting in more consumption and higher output and employment.
And I know I'm not saying anything that Paul Krugman doesn't already know.
Regardless of the model (at least, regardless of any vaguely reasonable model I can think of), quantitative easing will be more effective if it is expected to be at least partly permanent, rather than temporary. The trick is, how do we ensure that the policy is expected to be at least partly permanent?
If I read the economic historians like Scott Sumner correctly, one of the policies that lead to recovery from the Great Depression was countries' raising the price of gold. And when they raised the price of gold, people expected it would be, at least in part, permanent. It was permanent quantitative easing.
Do people today, in the UK for example, expect the Bank of England's current policy of quantitative easing to be permanent? My guess is they don't, unfortunately.
We want to have it both ways. We want quantitative easing to work, and yet we don't want to damage central banks' balance sheets, and we don't want people to fear that the policy will cause excessive inflation. We can't have it both ways. Or at least, there is a trade-off.
Take a helicopter increase in the money supply, which is permanent, and done holding present and future taxes and government expenditure constant. This damages the central bank's balance sheet. Of course it damages the central bank's balance sheet; it is giving money away! The central bank's liabilities expand, and its assets do not. And the government (by holding the paths of taxes and expenditure constant) guarantees that it will not indemnify the central bank against any losses. The losses to the consolidated government/central bank balance sheet will have to be paid for by the inflation tax. Contrast helicopter money to the UK's quantitative easing, where the Treasury has indemnified the Bank of England against any losses it might suffer.
Helicopter money is a powerful method of increasing aggregate demand. It can be thought of as a one-time lump-sum transfer payment (or tax cut) paid for by printing base money, where the government guarantees that it will not pay for the transfer by raising future taxes (so Ricardian Equivalence is moot). The current transfer will be paid by the future inflation tax, in long-run equilibrium, but that inflation tax only kicks in if the helicopter money does indeed shift the aggregate demand curve to the right and cause (eventually) a higher price level.
If quantitative easing is to have no permanent effect on the time path of the future price level, then it must be purely temporary. Any increase in the money supply today will have eventually to be fully reversed some time in the future. But if it is perceived as purely temporary, it won't work as well, and we will need a much bigger amount of quantitative easing today, to have the same effect, and so we will need a much bigger reversal in the future, and will therefore need to worry much more about the central bank's balance sheet, so that the assets it will need to buy back the money cannot be allowed to fall in value.
The more we can scare people that quantitative easing will damage the central bank's balance sheet and raise the future price level, the bigger the quantitative easing multiplier, and the less quantitative easing we will need. (Estimates of those multipliers that ignore that temporary/permanent distinction may be wildly innaccurate.) We want to scare people (up to a point). That's the whole point of the policy: to scare people out of money and safe short nominal bonds, so they buy something else instead.
It's a weird sort of game. A bit like one of those Christmas cracker games where you have to tilt the convex surface enough to get the little silver ball rolling, but then quickly tilt it back the other way before the ball rolls to far. Except the ball knows that's the game you are playing, and responds to the expected future tilt, as well as the current tilt. The central bank would rather point an inflationary gun at the money markets and say "Well, punk, are you feeling lucky?"
I can think of two ways to scare the money market by the right amount.
The first is for the central bank to buy some really bad assets. Or rather, assets that will be bad if the central bank fails, and good if it succeeds in preventing a deflationary recession. And the Treasury will guarantee that it will not bail out the bank.
The second is for the central bank to announce a price level target (or price level path target), and raise that target (relative to current expectations) in conjunction with quantitative easing.
The second policy will ensure that quantitative easing will be seen as (at least partly) permanent. The first policy ensures that the perceived degree of permanence is inversely correlated with the policy's perceived success.
There are almost certainly other ways, some possibly better, to make sure quantitative easing is seen as at least partly permanent.