I think it is important to distinguish between two arguments why the Zero Lower Bound (ZLB) for nominal interest rates matters. I will label these the first and second, and economists and statisticians will soon  see why these labels have some significance. The first argument is debatable, but the second is I believe very difficult to argue against.
The first argument why the ZLB matters is that unconventional monetary policy either does not work, or hits limits on what it can do, and these constraints bite. In short, monetary policy at the ZLB cannot fully achieve monetary policy goals. The second argument about why the ZLB matters is that the impact of monetary policy becomes more uncertain. Under the second argument, it is possible for some particular dose of unconventional monetary policy to duplicate what interest rate policy might otherwise achieve, but there is more uncertainty about what that appropriate dose is. The impact of any unconventional monetary policy action is therefore more unpredictable than the impact of conventional policy.
Much of the discussion of unconventional monetary policy involves the first argument. An important feature of Quantitative Easing (QE) is that its size is potentially unbounded – the central bank can create as many reserves as it likes. So if the impact of each unit of QE on the economy is constant, even if this constant is small, if we knew what that constant was we just scale up the programme so it has the desired effect. However, it seems to be much more likely that the policy involves diminishing returns, but to be honest I have no idea whether that means there are limits to what the policy can currently do. There are also some who worry that unconventional monetary policy has dangerous side effects on financial stability if it becomes too large. As I said, the first argument is debatable.
What seems clear to me is that we know much less about the impact of QE, or other kinds of unconventional monetary policy, than we do about conventional monetary policy. This almost follows by definition: we have well established models for conventional policy, and much more data to check these models against. What data we have also suggests the impact of unconventional monetary policy is more uncertain. (This is the conclusion drawn by John Williams, who has done a good deal of work on their impact, and I have never heard anyone argue against this view.) That is why I think it is very difficult to deny that the impact of monetary policy at the ZLB is much more uncertain compared to monetary policy outside the ZLB. 
Why do I make this distinction? Good policy tries not only to achieve the best outcome for the economy, but it also tries to reduce the uncertainty associated with that outcome. Indeed, we might well be prepared to sacrifice some of the former for some of the latter: uncertainty is in general undesirable. A standard way to judge the merits of a particular rule for macro policy, for example, is to ask whether it reduces the variance of output or inflation when the economy is hit by a standard set of shocks.
Let me return to an old friend. Suppose you are a doctor, and you have two medicines to treat a disease. One is reliable, but the other requires trial and error to get the correct dose, and occasionally has nasty side effects. In these circumstances, you would rather not run out of the reliable medicine. Indeed, you would want to go out of your way to avoid running out of the reliable medicine.
Seen from this perspective, it becomes almost undeniable that fiscal austerity at or near the ZLB is a dangerous policy. By making us more reliant on unconventional monetary policy it increases macroeconomic uncertainty. It makes it more likely that we will have to resort to the unreliable medicine. I think too much of the argument over whether monetary policy is all you need focuses on the first reason why the ZLB may be important, and ignores the second. 
 I was tempted to write ‘in a moment’ rather than ‘soon’, but decided a bad pun might detract from the main text.
 The Williams paper also elaborates on a well known idea that uncertainty about the impact of policy instruments should make policy makers cautious in using those instruments. I think this is an additional consideration, which would reinforce the point I’m making here. I also agree that good policy should take into account the characteristics of uncertainties arising from the economy, and have used this to argue - for example – that policymakers should over rather than under estimate the size of the output gap at the ZLB.
 This is not an argument about whether fiscal stimulus is more or less reliable than monetary stimulus, interesting though that question is. All I require is that an austerity policy reduces short run aggregate demand, which it clearly does. It also does not say that there are no circumstances in which we should undertake austerity at or near the ZLB – in theory the benefits of austerity could be so great that they outweigh the costs of putting us in a dangerous place. In the absence of a significant and rising default premium on debt, I do not think those benefits exist, but that is a separate argument.
This blog was first published on Mainly Macro