776 LEO. Svensson/ European Economic Review 46(2002)771-780 A specific targeting rule is a commitment to set the instrument rate so as to achieve specific target criterion for the target variables. The Bank of England and Sweden be expressed as"set interest-rates so the inflation forecast about two years amead o Riksbank have formulated a simple specific targeting rule to guide policy, which ca both simple and operational, it is not necessarily optimal. An optimal specific targeting rule expresses the equality of the marginal rates of transformation and substitution between the target variables in an operational way As an example(from Svensson, 2001b, where the details are explained), consider a variant of the popular New Keynesian model, where infation and the output gap are predetermined one period(a small concession to realism relative to the standard variant when both inflation and the output gap are treated as forward-looking variables, jump variables) and, in particular, judgment"matters. The aggregate-supply/Phillips 丌+1=7+2|+ax+1+x=21+1, where T1+2]r denotes expectations in period t of inflation in period t+ 2,etc,ox is a positive constant, a, is a row vector, and zu+I is a column vector (of the same dimension), the"", to be explained below. Thus, inflation in period t +I is determined by expectations in period t of inflation in period t +2 and of the output gap in period t l and by the deviation in period t+1. The aggregate-demand/ Is x+1=x1+2|-B(2+1-x+2)+2z+1, where Bx is a positive constant, it+It is the expectation in period t of the nominal interest rate in period t I and Bz is a row vector. Thus, the output gap in period t+ l is determined by expectations in period t of the output gap in period t+2 of the output gap in period t +2 and of the real interest rate in period t+1 and by the deviation in period t +1 The deviation nts the difference between the true model and this simplified New Keynesian model and includes all other determinants of inflation and the output gap. For simplicity it is treated as an exogenous variable. The central banks"judg nent,2=z1+r, o, is the central bank's best forecast of the deviation. This is a way to represent the importance and inevitability of judgment in monetary policy. Con- ditional on the central banks judgment, the banks forecasting model in period t is then 兀1+:t=丌t+t+1,t+axx+t,t+x271+r,t x+rt=x1+t+1-B(i+rt-+r+1)+B2z1+xt, for forecast horizons t> 1(where T+tt refers to the central banks t-period-ahead forecast of inflation in period t, etc. The optimal specific targeting rule for the loss function(2.1)and the model(2.6) and(2.7)can then be found by finding the marginal rate of transformation(MRT) and substitution(MRS) between (the forecasts of) the target variables(inflation and the776 L.E.O. Svensson / European Economic Review 46 (2002) 771 – 780 A speci@c targeting rule is a commitment to set the instrument rate so as to achieve a speci@c target criterion for the target variables. The Bank of England and Sweden’s Riksbank have formulated a simple speci@c targeting rule to guide policy, which can be expressed as “set interest-rates so the in ation forecast about two years ahead is on target” (Goodhart, 2001; Heikensten, 1999). Although this speci@c targeting rule is both simple and operational, it is not necessarily optimal. An optimal speci@c targeting rule expresses the equality of the marginal rates of transformation and substitution between the target variables in an operational way. As an example (from Svensson, 2001b, where the details are explained), consider a variant of the popular New Keynesian model, where in ation and the output gap are predetermined one period (a small concession to realism relative to the standard variant when both in ation and the output gap are treated as forward-looking variables, jump variables) and, in particular, “judgment” matters. The aggregate-supply=Phillips curve is t+1 = t+2|t + xxt+1|t + zzt+1; (2.4) where t+2|t denotes expectations in period t of in ation in period t + 2, etc., x is a positive constant, z is a row vector, and zt+1 is a column vector (of the same dimension), the “deviation”, to be explained below. Thus, in ation in period t + 1 is determined by expectations in period t of in ation in period t + 2 and of the output gap in period t + 1 and by the deviation in period t + 1. The aggregate-demand=IS curve is xt+1 = xt+2|t − x(it+1|t − t+2|t) + zzt+1; (2.5) where x is a positive constant, it+1|t is the expectation in period t of the nominal interest rate in period t + 1 and z is a row vector. Thus, the output gap in period t + 1 is determined by expectations in period t of the output gap in period t + 2 of the output gap in period t + 2 and of the real interest rate in period t + 1 and by the deviation in period t + 1. The deviation represents the diNerence between the true model and this simpli@ed New Keynesian model and includes all other determinants of in ation and the output gap. For simplicity it is treated as an exogenous variable. The central bank’s “judg￾ment”, zt ≡ {zt+;t}∞ =0, is the central bank’s best forecast of the deviation. This is a way to represent the importance and inevitability of judgment in monetary policy. Con￾ditional on the central bank’s judgment, the bank’s forecasting model in period t is then given by t+;t = t++1;t + xxt+;t + zzt+;t; (2.6) xt+;t = xt++1;t − x(it+;t − t++1;t) + zzt+;t; (2.7) for forecast horizons  ¿ 1 (where t+;t refers to the central bank’s -period-ahead forecast of in ation in period t, etc.). The optimal speci@c targeting rule for the loss function (2.1) and the model (2.6) and (2.7) can then be found by @nding the marginal rate of transformation (MRT) and substitution (MRS) between (the forecasts of ) the target variables (in ation and the