In trotting around the country giving versions of DeLong and Summers (2012), “Fiscal Policy in a Depressed Economy”, I have found that a point that seemed completely obvious to us is not obvious at all to many.
Here is the point: an optimizing central bank that cares only about inflation and unemployment because it does not find itself at the zero nominal lower bound and does not fear engaging in nonstandard monetary policy will engage in full fiscal offset: it will take care to make sure that if fiscal policy becomes more stimulative then it will make monetary policy less stimulative by the same amount.
Here is the argument:
Suppose that we have a system in which future unemployment ut+1 is a function of variables V (which include past, present, and expected future unemployment and inflation rates), of the stimulative impact of monetary policy m, and of a shock:
(1) ut+1=ƒ(Vt) + mt + εut+1 ;
in which inflation is a function of variables V (which include past, present, and expected future unemployment and inflation rates) and of a shock:
(2) πt+1 = j(Vt) + επt+1 ;
and in which there is an objective function h depending on present and future values of unemployment and inflation (but not directly on monetary or fiscal policy variables):
(3) h(πt, πt+1, …; ut, ut+1, …)
At each point in time the monetary authority will choose an optimal m't in order to maximize the expected value of that objective (3) subject to (1) and (2), and the unemployment rate will consequently be:
(4) ut+1=ƒ(Vt) + m't + εut+1
Now consider what happens if we replace (1) with (1a), with the difference between (1) and (1a) being that fiscal policy g also affects unemployment:
(1a) ut+1=ƒ(Vt) + mt + gt + εut+1
Then as long as m't is the optimal choice of m for the system (1), (2), and (3):
(5) m"*t = m't − gt
is the optimal choice of m’ for the system (1a), (2), and (3).
Why? In system (1a), (2), and (3), define:
(6) nt = mt + gt
Then (1a) becomes:
(7) ut+1=ƒ(Vt) + nt + εut+1
By hypothesis, the objective h is maximized for the system of (7), (2), and (3) by choosing:
(8) n't =m't
And so the objective is maximized for the system of (1’), (2), and (3) by choosing:
(9) m"t =m't - gt
You take your optimal policy if fiscal policy were neutral--if gt=0--and you then subtract the stimulative value of the actual gt from it.
The important point here is that m and g cannot enter into the objective function h directly, but only indirectly through their effects on inflation and unemployment.
For this reason this argument breaks down at the zero nominal lower bound. At the zero lower bound the central bank does not care only about inflation and unemployment. It cares as well about the magnitude of the non-standard monetary policy measures it must take in order to achieve its net monetary policy impetus value m.