A central bank reaction function tells us how the central bank sets its monetary policy instrument (for example a nominal interest rate) as a function of various indicators of the state of the economy (for example inflation and output). Suppose an econometrician is estimating a central bank reaction function. Under what circumstances would the econometrician find a structural break in the estimated reaction function? And what would the performance of the economy look like during those structural breaks? Would it be better or worse than normal?
Here are two examples:
Example 1. Suppose a central bank is initially targeting 10% inflation. Then it decides that 10% is too high, and decides to target 5% inflation instead. If the central bank is right about 5% inflation being better than 10% inflation, we would expect the economy eventually to do better when it changes its target. But most economists would say that the immediate impact of the change in target could well be worse performance, because it takes time for the economy to adjust to the new inflation target, especially if that new target is not well-communicated or is not fully credible. An econometrician who estimated the central bank's reaction function would notice a structural break in that reaction function, and that structural break being associated with worse economic performance.
Example 2: Suppose a central bank is targeting 2% inflation. Then a big shock hits, that causes a permanent fall in the (unobserved) natural rate of interest. That shock may itself cause worse economic performance. Plus, if the central bank is not immediately aware of that shock, or its magnitude, and fails to adjust its reaction function quickly enough, that will also cause worse economic performance. An econometrician who estimated the central bank's reaction function would notice a structural break in that reaction function, and that structural break being associated with worse economic performance.
Those two examples show that it is hardly surprising that structural breaks in an estimated central bank's reaction function should be associated with worse economic outcomes. The NPP econometric results do not surprise me. But those results do not support their policy conclusions.
Ask yourself this question: do those two examples support the argument that the central bank's reaction function should be fixed by law?
Example 1: If the central bank's reaction function had been fixed by law, it would have been unable to change the inflation target from 10% to 5%. Unless the law had changed.
Example 2: If the central bank's reaction function had been fixed by law, it would have been unable to change the reaction function in response to the shock. Unless the law had changed.
If it were impossible to change the law, would the results have been better or worse? In the first example it would have been impossible to move to a better inflation target. In the second example it would have been impossible for the central bank to respond to the shock to the natural rate, and inflation would have fallen a long way below target.
And if it were as quick and easy for lawmakers to change the law as for the central bank to change its reaction function, and if the lawmakers views on monetary policy were the same as the central bank's, would the results have been any different?
But in that case, why bother with the central bank at all? Why not just have the lawmakers set interest rates every month?
There is always a tension between rules and discretion, and not just in monetary policy.
We want rules to solve the Kydland and Prescott time-consistency problem, which means creating the expectations of our future actions we want to create, even if it might not be in our future interest to carry out those actions as promised. That's why promises exist, why laws exist, and why societies exist, and why we are not all living in the Hobbesian State of Nature.
But at the same time, our views on what are the best laws may change over time. Things change, and they change our views, and they sometimes change in ways we could not foresee and so could not have written into the laws as contingencies. There are unknown unknowns, and even the known unknowns would be very difficult in practice to completely specify and write into the laws. Could we state a law for monetary policy that would be contingent on all possible future results for all possible future theoretical and applied research in monetary policy? Of course not.
Most advanced countries have squared this rules vs discretion circle for monetary policy. In Canada, for example, the 2% inflation target has a "quasi-constitutional" status. It is not something we would change on a whim, because people have made their plans for the future based on that expectation. It gets re-examined every 5 years, jointly by the government and by the Bank of Canada. It is a rule. The Bank of Canada uses its discretion to set the interest rate instrument, but must do so to bring the inflation rate back to the 2% target. The Bank of Canada is accountable for keeping inflation at or near the 2% target. We can all hold the Bank of Canada accountable for following that rule because we can all observe the inflation rate. Most people don't have a clue about what interest rate would be best for hitting the 2% inflation target; they are too busy doing their own jobs to do the Bank of Canada's job as well. That's why we delegate it to the Bank of Canada.
Canada has a rule for the target but discretion for the instrument. And the empirical evidence shows that instrument discretion hits the target rule better than any simple instrument rule like the Taylor Rule. Most advanced countries have something very similar. It's a bit like judges. We tell judges what rules to enforce, but leave it to their discretion how best to enforce those laws in each particular case. That's how it normally works in societies with the rule of law and a reasonable amount of trust and professional ethics. The president/prime minister/monarch/parliament/congress/boss does not decide everything himself. Because it doesn't work very well. He delegates, to other people with greater or lessor autonomy of action. Things normally seem to work better that way. That's probably why market economies usually work better than centrally planned ones.
I can't help thinking that there is something peculiarly American to this question. 1776 and all that. Why don't they get the distinction between target rules and instrument rules?