Symmetric or Asymmetric Monetary Policy Rules. in Different Countries?

Transcription

1 Symmetric or Asymmetric Monetary Policy Rules in Different Countries? Bengt Assarsson 1 Department of Economics, Uppsala University Paper to be presented at 10 th Annual SNEE European Integration Conference, May 21, 2008 his version May 6, 2008 mail@bassarsson.com Abstract his paper discusses and estimates monetary policy rules for a number of countries. he paper departs 2 2 from the conventional symmetric policy objective function L ( π π ) v( Y Y ) = +, from which reduced form policy rules can be derived. Since the socially desirable output Y > Y there is a case for asymmetric policy rules, since a positive output gap is to prefer to a negative gap. I estimate policy rules for 14 inflation and 5 non-inflation targeting countries and test for symmetry. he results show that in the tests of symmetry done in this paper only one country, South Korea, has a significantly asymmetric policy with a smaller weight put on positive output gaps. One country, the UK, has a significantly asymmetric policy with smaller weight put on negative gaps. For the other countries a symmetric policy cannot be rejected. 1 he views expressed here is solely the author s and may only by accident coincide with the views of the Board of Governors (Direktionen) of Sveriges riksbank.

2 1. Introduction heoretical monetary policy analysis views central banks as optimizing agents. his view is also used in practice by many central banks or is at least part of the rhetoric or models used by the central banks. In the conventional theory the central bank maximizes (or minimizes) a quadratic objective function (loss function) conditional on private sector behavior. his results in an optimal monetary policy rule, describing the actions of the central bank. Such simple rules (aylor rules) can also be postulated or estimated empirically and evaluated in general equilibrium models in which the optimal rule in itself may be difficult to derive. In the most common aylor rule for a central bank using the interest rate as its instrument the arguments used are an output and an inflation gap. For an inflation targeting central bank the inflation gap is the deviation from a set inflation target. Most often a smoother is used (lagged interest rate) that dampens the reactions in the interest rate. Such a simple rule often gives an accurate description of the historical development of the interest rate. Sometimes it also comes close to what some modern dynamic stochastic general equilibrium model considers optimal. A recognized problem with estimated aylor rules is that the central bank behavior is not identified. he reason is that the estimated parameters are functions of the structural parameters of central bank preferences as well as of the parameters of private sector behavior, i.e. in Phillips curves and aggregate demand curves. Hence, a low weight on the output gap might not reflect low priority on part of the central bank but rather illustrate the behavior of consumers and price setters. Macroeconomic models assume monopolistic firms and frictional labor markets. his means that equilibrium volumes are undesirably low. While a negative output gap lowers welfare a positive gap is socially desirable. Hence, one could suspect that some central banks would raise interest rates less in response to a positive output gap than they would lower the interest rate in response to a negative output gap. In other words, the aylor might be asymmetric. Since the conventional theory generally does not consider this asymmetry this would be a case in which practice (for some central banks) might be ahead of theory. In this paper I test this symmetry hypothesis. I assume that it is this (optimal) behavior of the central bank that is the sole reason for the asymmetry, i.e. there is no asymmetry in private sector behavior. hen I can identify potential differences between central bank behavior in different countries as well as between different types of policies; inflation targeting vs. non-inflation targeting countries. he next section describes the monetary policy problem and why it is reasonable to depart from an objective function with output and inflation gaps. Section 3 gives a more technical discussion about the objective function and explains the case for an asymmetric objective function/nonlinear policy rule. Section 4 presents data and econometric methods while section 5 gives the empirical results. Finally, there is a concluding section. 2. he Monetary Policy Problem Government policies, whether tax, fiscal, social, monetary, aim at affecting resource allocation or distribution one way or the other. Money facilitates transactions in the economy but comes at a cost. he opportunity cost of money is the nominal interest rate foregone. Hence, the higher the interest rate, the higher the cost of holding money. Since the cost of producing money is negligible it seems as if a zero interest rate would optimize money holdings the so called Friedman rule. In view of the

3 Fisher equation this sets long run inflation to minus the real interest rate, deterrent for many. But in a frictionless world this seems as the ultimate and simple monetary policy. However, macroeconomic theory acknowledges several frictions, among which rigidities in nominal prices and wages are regarded as most important. When the economy is hit by an aggregate shock some nominal prices and wages respond slowly while yet other prices may respond quickly. Apparently, relative prices change and affect the resource allocation, which becomes then suboptimal. his suboptimal allocation is the result of the price rigidities. he resource allocation would be left unaffected (and possibly optimal) in the case of perfect price and wage flexibility. 2 So, the monetary policy problem is twofold: determine the optimal rate of interest (rate of inflation) reduce the resource allocation problems associated with nominal price and wage rigidity using an appropriate monetary policy instrument, such as the short term interest rate now used in many countries (interest rate policy). One can also view this as deciding upon a normal rate of interest consistent with the optimal rate of inflation i.e. zero in the Friedman case above and changing this interest rate up and down so as to mitigate the allocation problems associated with price and wage rigidities. his description of the problem seems far from the rhetoric exercised by the monetary authorities in many countries, even though it is standard in the literature, e.g. in Woodford s book. here are differences among central banks in their effort to communicate their policies. Most central banks have explicit inflation targets and do not stress their role in stabilizing the real economy. Rather, as in Woodford s book, it is (implicitly) assumed that a central bank with inflation target fulfils the role above. he concept of flexible inflation targeting as put forward by e.g. (Svensson and Woodford 2004; Svensson 2006; Woodford 2007) has also been applied by e.g. the central banks in Norway and Sweden. By this is meant a policy in which the central bank acknowledges the tradeoff between the inflation target and some real target and put some weight on the real target as well. echnically, this is formulated as a social welfare, or loss, function to be optimized by the central bank. here is a widespread consensus about this setup but less so about the details, for instance what are the optimal targets, how to measure these, and so forth. 3. Defining the Social Loss Function he legislation underlying monetary policy in many countries is based on a lexicographic preference ordering 3, while economic theory most often postulates a quadratic loss function. his loss function can be used to derive an optimal policy rule for the central bank, taking account of optimizing private 2 his description is similar to Woodford, M. (2003). Interest and Prices. Foundations of a heory of Monetary Policy. Princeton and Oxford, Princeton University Press. 3 his is clear from, for instance, the Swedish legislation: he target for monetary policy shall be to maintain a fixed value of money. his target is outspoken in the law. As an authority under the parliament the Riksbank shall also, without setting the target of price level stability aside, support the targets for economic policy in general with the purpose of achieving balanced growth and high employment level. See Finansdepartementet (1997). Riksbankens st llning. Stockholm, Regeringskansliet, Finansdepartementet.. he Riksbank has interpreted maintain a fixed value of money as 2 percent inflation as measured by the Consumer Price Index.

4 sector behavior. Alternatively and in practice, aylor rules can be estimated and actual central bank behavior be evaluated in terms of a loss function. Central bank legislation typically favors price level stability, though the rhetoric as well as actual behavior of many central banks seems to take account of the real sector as well. A typical loss function would be ( π π ) ( ) 2 2 L = + v Y Y, (1) where π is the inflation target, Y is potential GDP and v is the relative weight put on output gaps. A central bank using (1), unless v=0, violates lexicographic preferences, since (1) implies invoking a tradeoff between inflation and output deviations. Hence, there seems to be a wedge between legislation and actual behavior in many countries which to some extent seems unhealthy and problematic, particularly as it comes to evaluating the policy. Sometimes it is argued that the rate of interest should be included in the loss function. he reason is that the higher the nominal interest rate, compared to some risk-free rate, the higher the social cost of money. his cost would be at the lowest when the interest rate is zero, according to the Bailey- Friedman rule. he discussion of optimal monetary policy has its origins in (Bailey 1992; Friedman 2006). hey argued that when a government has access to lump-sum taxation to finance its expenditure, the optimal monetary policy should adopt a rule (later called the Friedman rule) that generates a zero interest rate, corresponding to a zero inflation tax. he logic is straightforward. Because there is a wedge between the private marginal cost of holding money (which is the nominal interest rate) and the social marginal cost of producing money (which is approximately zero), a positive nominal interest rate generates social losses. hus, to achieve optimum, the monetary authorities should set the nominal interest rate to zero in order to eliminate the private opportunity costs. (Chari, Christiano et al. 1996) and (Correia and eles 1996) provide a general conclusion whereby the validity of the Friedman rule crucially depends on the property of consumers' preferences and the role of money for producing transaction services. In the case lump-sum taxation is unavailable the optimal nominal rate of interest might be some positive number, since then the inflation tax can reduce the excess burden imposed by other taxation. Hence, an alternative loss function should include the deviation of the nominal interest rate from some risk-free rate of interest, such as in ( π π ) ( ) ( 0) L = + v Y Y + i, (1 ) where the risk-free interest is assumed to be zero as in the Bailey-Friedman case. It is straightforward to derive a policy rule using (1) and substituting a Phillips curve for π and an aggregate demand curve for Y. he linear curves and the quadratic form (1) yields a linear policy rule, as illustrated in Diagram 1, where the left out demand curve is in the interest rate/output ( i, Y ) space. he point A illustrates the equilibrium in which the targets ( π, Y ) in (1) are fulfilled. he diagram also illustrates Y as the maximum long run attainable output level given the existing institutions such as monopolized firms and labor unions and is Y rather than Y as the level of output that maximizes welfare. Clearly, it Y that is commonly referred to in the literature though Y is the more relevant concept in (1). If the monetary authorities were to aim at Y, however, then the inflation target would not be credible and hence an inflation bias would occur. his is illustrated in the diagram as the distance AD, where D is the point at which the authorities aims at Y and actual and expected inflation is above target, i.e. monetary policy lacks credibility.

5 Instead of announcing an incredible target the central bank can use different weights on the real variable depending on the output gap being positive or negative. In other words, the central bank would raise the interest rate by less if the output gap is positive compared to lower the interest rate if the output gap is negative. he reason of course is that welfare is raised temporarily if lowered in the case Y like Y < Y Y but > Y. A simple generalization of the loss function could then look something ( π π ) 1 ( ) 2 ( ) L = + v Y Y + v Y Y, (2) Y > Y > Y Y < Y Y π PK PK D bias C A B PPR PPR Y Y Y Diagram 1. PK is the Phillips curve and PPR the monetary policy rule. with different weights depending on if the actual output level raises or lowers welfare compared to the attainable level Y. he policy rule derived from (2) should be credible and give no inflation bias. 1 1 i = in + η α ( π π ) α ( Y Y ) + + αv 1 for Y > Y > Y (3a) 1 1 i = in + η α ( π π ) α ( Y Y ) + + αv 2 for Y < Y Y (3b) and v1 > v. his could be simplified to the reduced form 2 n ( π π ) ( ) i = i + a + b Y Y for 1 1 Y > Y > Y (4a)

6 n ( π π ) ( ) i = i + a + b Y Y for 2 2 Y < Y Y (4b) in which the hypothesis of asymmetric response could be tested. he form (4) lacks dynamics usually introduced into aylor rules. A very simple form is ( ) ( ) i = k + λi + a π π + b Y Y + ε for Y > Y > Y (5a) t t 1 1 t 1 t t 1t ( ) ( ) t t 1 2 t 2 t t 2t t t t i = k + λi + a π π + b Y Y + ε for Y < Y Y (5b) t t t k for which we could derive the normal rate of interest as in = = π + r, where r is the real or 1 λ natural rate of interest. ε1 and ε are i.i.d. stochastic errors in the policy process. We could merge 2 (5a)-(5b) into ( ) ( ) ( ) ( ) i = k + λi + a π π + b Y Y + a π π + b Y Y + ε (6) t t 1 1 t Yt > Yt > Y t t t Yt > Y t t t t t > Yt Yt < Yt Yt Yt < Yt Yt where the RHS variables are treated as dummies, equal to zero when the condition at hand does not hold. he symmetry restriction then is a1 = a2 & b1 = b, under which (6) collapses to a conventional 2 aylor rule. Even if we cannot identify the structural parameters in (6), the only reason for asymmetry is to be found in the acknowledgement of the particular nonlinear form of the loss function. herefore, the above test is valid and we can disregard that the parameters are functions of the slope of the aggregate demand and Phillips curves. he dynamics in (6) is very simple and in particular it assumes that the central bank reacts instantaneously to deviations from targets. 4 Due to the lags in the transmission mechanism central banks are forward-looking and hence would rather act on forecasts of future expected deviations. he exact dynamics is of course unknown and therefore I try different dynamic patterns in the estimations below. 4. Data and Econometric Methods he model (6) is used to test the symmetry hypothesis for 19 countries, divided into 14 inflation targeting and 5 non-inflation targeting countries. he countries used in the study are presented in ables 1 and 2. able 1 show the first year of the inflation targeting regime and whether the target is set by the government, the central bank or by both. he exact target and the inflation measure are also shown and whether the central bank is partly or fully instrument independent. As can be seen in the table the starting years vary a lot among the countries, from 1990 in New Zealand to 2001 in Hungary. I have done the estimations for the period in which he 3-month interest rate is used as proxy for the policy interest rate. Inflation targeting countries use different measures of inflation, for instance CPI in Sweden and the harmonized consumer price (HCP) index in the EMU countries. As is common elsewhere GDP is used as the output variable, though it is less obvious which real variable should be used. GDP was used by aylor in his original paper and is the most common 4 o the extent that present inflation reflect expectations about future inflation (6) might be reasonable.

7 measure in most of his followers. In addition, there is the problem of defining the attainable or potential equilibrium Y and the optimal equilibrium Y, both of which are unobservable. For simplicity, I shall assume that historical shocks are sufficiently small, such that Yt < Y materializes in the data. In the empirical literature various measures of Y have been used. Most common is probably various mechanical filters, such as the Hodrick-Prescott filter. Recently, some researchers advocate a more model consistent measure, the so called flex price output gap, in which Y is the output level that would emerge in absence of nominally rigid prices. Whether we use a flex price output gap or a HP filter might affect the results. In particular, the HP filter gap does not recognize the difference between a gap generated by a demand or by a supply shock. he gap generated by a supply shock implies a policy conflict while the demand shock does not. his can be illustrated in diagram 1 where the policy rule implies a downward shift in the interest rate in case of a positive productivity shock (Y shifts to the right) and an upward shift in the interest rate in the case of a demand shock (both π and Y above target). he use of HP filter would indicate a rise in the gap in both cases though it would be more accurate with a decline in the former case. Hence, the choice of output gap might be critical. I still use the shortcut approach and use the HP-filter, focusing on the policy asymmetry problem. 5 Another issue is whether to use ex post or real-time data. Real-time data is what you would expect the central banks to have used, but is more difficult to collect, in particular for the large number of countries considered in this study. It is also possible that the difference is small, as suggested by (Osterholm 2005). Sticking for a moment to the simple dynamic form in (6) one could assume a fixed λ but a timevarying k, such that the real interest would change by int = = π + rt 1 kt λ his implies that (6) could not be estimated by OLS. I use a Kalman filter in order to estimate the variation over time in k. he variation is assumed to follow [yet to be done]. Even though the loss function depends only on variables dated at t, the policy rule depends on the dynamics of the Phillips and the demand (IS) curves. A New Keynesian Phillips curve would typically be π = κ π + κ π + κ (7) t e [ mc p] t+ 1t t t ] t where the real marginal cost [ mc p is often approximated by the labour share or some measure of π + t e the output gap (such as the HP filter). he problematic variable is, the future expected rate of t 1 e inflation, which however by assuming rational expectations can be defined by πt+ 1 = π t t + et + 1 where e + is a Gaussian disturbance. he dynamic form of the policy rule is scrutinized in the literature, see t 1 e.g. (Ambler, Dib et al. 2004; Amin 2005; Eusepi 2005; Marzo 2006; Piergallini 2006). In particular, forward-looking elements are introduced into the policy rule A so called flex price gap might be advocated by most researchers, but it depends on the particular model from which it is generated. his would most likely be a dynamic stochastic general equilibrium (DSGE) model but there are large differences within this class of models.

8 he simplest possible demand curve would be an IS type curve where demand depends on the interest rate, possibly with some lag. Estimation of (6) with OLS is not possible since both inflation and output are endogenous variables but we can use variables in the Phillips curve (7) and the lagged interest rate as instruments. Since it is assumed that the central banks smoothes interest rates and the lagged interest rate is included in the estimation it is likely that the model is subject to temporal aggregation. herefore, I include a moving average term and estimate the model with two stage least squares. 5. Empirical Results he development of GDP growth, inflation and interest rates for the periods and are shown in ables 3 and 4. he most successful country would have low variability and high GDP growth rates and low inflation variability with inflation close to target. Success countries are the Czech Republic and South Korea who introduced inflation targeting in 1997 and 1999, respectively. hese countries increased GDP growth rates while simultaneously reduced its variability. hey also reduced inflation towards the targeted levels and reduced inflation variability. he Czech Republic also had a low normal interest rate, 3.6 percent, while South Korea showed a higher rate, 5.0 percent. However, the inflation targeting countries usually experienced a higher inflation during earlier periods while the non-inflation targeting countries had lower inflation rates in the earlier period. Japan and Switzerland are good examples. If we pool the data and estimate aylor rules (based on (6)) for inflation and non-inflation targeting countries, respectively, they are remarkably similar and explain about 98 percent of the variation in interest rates. While the coefficient on the inflation gap is similar, the inflation targeting countries put relatively lower weight on positive output gaps. It s the other way around for the non-inflation targeting countries. able 5 shows the estimation results for the inflation targeting countries for the periods these countries have exercised inflation targeting. he second column shows the value of the smoother λ indicating an aggressive policy for countries with low values. Countries with a long experience with inflation targeting have large values, such as Australia, Canada, Sweden and the UK, though New Zealand has 0.6. he EMU also shows an aggressive or flexible policy with 0.5. he non-inflation targeting countries also smooth their interest rates showing lambdas around 0.9, as shown in able 6. he estimated aylor rules also deliver estimates of the normal rate of interest, as shown by the third k columns of tables 5 and 6 and estimated as. he normal rate of interest clearly differs among the 1 λ countries, mainly due to the period for which it is estimated. he normal rate of interest is generally higher in the earlier periods. As can be seen in column (5) in able 5, for the inflation targeting countries there are only one country which has an asymmetric policy rule with lower weight put on positive output gaps, South Korea. here are another 9 countries with lower but insignificant weight on positive output gaps; Australia, Canada, EMU, Hungary, Mexico, New Zealand, Norway, South Africa and Sweden. UK has a larger weight put on positive gaps. Looking at the non-inflation targeting countries Switzerland and the US have an insignificantly smaller weight put on positive gaps, as shown in able 6.

9 I also test for weights on both negative and positive weights being zero. he p-values for the tests are shown in column (6). his hypothesis is rejected for 7 inflation targeting countries; Canada, EMU, Mexico, New Zealand, Poland, South Korea, and Sweden. It is more interesting to compare the countries for a balanced sample. We show that in ables 7 and 8 which show the results for the common time period Again, only South Korea shows up with significantly asymmetric weights, smaller for positive than for negative output gaps. However, South Korea has a relatively high normal interest rate. hree countries have weights on the output gaps significantly different from zero but insignificantly asymmetric weights with smaller weights on the positive gaps; EMU, Mexico and Sweden. 6. Conclusions he fact that a positive output gap increases welfare while a negative output gap decreases welfare might make central banks follow an asymmetric policy. aylor rules describe the behaviour of the interest rate policy of central banks. he tests of symmetry done in this paper show that only one country out of the 19 countries considered, South Korea, has a significantly asymmetric policy with a smaller weight put on positive output gaps. One country, the UK, has a significantly asymmetric policy with smaller weight put on negative gaps. For the other countries a symmetric policy cannot be rejected.

10 able 1. Inflation targeting countries. (1) (2) (3) (4) (5) (6) Country Since arget set by Instrument independent arget % Inflation measure Australia 1993 G & CB Partly 2-3 CPI Brazil 1999 G Yes 4.5 ± 2 CPI Canada 1991 G & CB Partly 2 ± 1 CPI Czech Rep G & CB Yes 2-4 (3) CPI EMU 1999 CB Yes 2 HCP Hungary 2001 G & CB Yes 4 ± 1 CPI Mexico 1995 CB Yes 3 ± 1 CPI New Zealand 1990 G & CB Partly 1-3 CPI Norway 2001 CB Partly 2.5 CPI Poland 1998 CB Yes 2.5 ± 1 CPI Rep. of Korea 1998 G & CB Partly 3 ± 1 Core South Africa 2000 G & CB Yes 3-6 Core Sweden 1993 CB Yes 2 ± 1 CPI UK 1992 G Partly 2 HCP Note. he 15 EMU countries are Austria, Belgium, Cyprus, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Malta, the Netherlands, Portugal, Slovenia and Spain which entered the union in 1999 with the exception of Greece (2001), Slovenia (2007) and Cyprus and Malta (2008). However, I use EMU-12 and exclude the three most recent members.

11 able 2. Non-inflation targeting countries. (1) Country (2) Instrument independent Denmark India Japan Switzerland USA Yes Partly Partly Yes Yes

12 able 3. GDP growth, inflation and interest rates in inflation targeting countries. Average figures for the periods and Standard deviation in parenthesis (1) Country (2) GDP growth (3) Inflation (4) Interest rate (5) GDP growth (6) Inflation (7) Interest rate Australia 3.70 (1.15) 1.99 (1.02) (1.03) 2.22 (1.10) 5.41 Brazil 2.93 (2.67) (641) (1.97) 7.19 (3.39) Canada 3.31 (1.42) 1.59 (0.54) (1.39) 1.74 (0.55) 3.59 Czech Rep (2.90) 5.34 (5.94) (2.02) 1.88 (1.46) 3.60 EMU 2.24 (0.96) 2.02 (0.60) (1.09) 1.95 (0.52) 3.12 Hungary 3.82 (1.41) (7.96) (1.18) 5.74 (2.76) 9.64 Mexico 3.06 (3.63) (11.5) (2.27) 6.66 (4.10) New Zealand 3.42 (1.50) 1.78 (0.83) (1.34) 1.68 (0.90) 6.22 Norway 2.98 (1.63) 2.22 (1.03) (1.30) 2.05 (1.08) 4.82 Poland 4.90 (2.31) (11.07) (1.98) 3.55 (3.09) 9.30 South Africa 3.53 (1.35) 6.81 (2.11) (1.07) 6.04 (2.20) 9.38 South Korea 5.16 (4.16) 3.75 (1.80) (2.58) 2.70 (1.04) 5.01 Sweden 3.11 (1.24) 1.58 (0.80) (1.26) 1.44 (0.56) 3.05 UK 2.95 (0.72) 2.22 (0.68) (0.67) 1.92 (0.51) 4.83 able 4. GDP growth, inflation and interest rates in non-inflation targeting countries. Average figures for the periods and Standard deviation in parenthesis (1) Country (2) GDP growth (3) Inflation (4) Interest rate (5) GDP growth (6) Inflation (7) Interest rate Denmark 2.47 (1.30) 1.91 (0.63) (1.51) 1.92 (0.64) 3.32 India 6.73 (2.34) 6.51 (1.58) (2.02) 4.48 (3.37) Japan 1.40 (1.33) (0.47) (1.60) (0.75) 0.15 Switzerland 1.56 (1.40) 0.71 (0.51) (1.33) 0.79 (0.59) 1.54 USA 3.18 (1.25) 2.07 (0.58) (1.22) 2.23 (0.61) 3.67

13 able 5. Estimation of simple aylor rule for inflation targeting countries. Estimation by 2SLS. Variables in Phillips and IS curves are used as instruments. (1) (2) (3) (4) (5) (6) Country Smoother λ Normal rate of interest (std. error) Standard error of rule p-value for test of symmetry(-,+) p-value for test of inflation nutter Australia (2.5,2.00) (2.644) (2.382,-0.019) Brazil ) (5.5,7.19) (3.424) (-0.014,-0.158) Canada (2,1.87) (0.993) (0.193,0.158) Czech Rep (3,3.24) (1.487) (0.030,0.192) EMU (2,1.95) (0.411) (0.415,0.217) Hungary (4.5,4.54) (2.132) (0.021,-0.021) Mexico (3.5,4.69) (1.124) (0.069,0.046) New Zealand 1990 (2,1.95) (1.225) (0.261,0.250) Norway (2,1.78) (1.742) (0.100,-0.029) Poland (2.5,4.30) (1.538) (0.023,0.605) South Africa 2000 (4.5,5.85) (0.829) (0.216,-0.337) South Korea (3,3.22) (0.682) (0.368,-0.163) Sweden (2,1.58) (1.065) (0.074,0.021) UK (2,1.96) (2.499) (-0.301,0.613)

14 able 6. Estimation of simple aylor rule for non-inflation targeting countries. Estimation by 2SLS. Potential variables in NKPC and AD curves are used as instruments. (1) (2) (3) (4) (5) (6) Country Smoother λ Normal rate of interest (std. error) Standard error of rule p-value for test of symmetry(-,+) p-value for test of inflation nutter Denmark (-,1.92) (1.248) (0.082,0.201) India (-,6.52) (7.550) (0.030,0.124) Japan (-,-0.47) (0.990) (-0.556,0.062) Switzerland (-,0.72) (0.861) (0.207,0.053) USA (-,2.08) (1.844) (0.795,-0.158)

15 able 7. Estimation of simple aylor rule for inflation targeting countries. Estimation by 2SLS. Variables in Phillips and IS curves are used as instruments. Common period (1) (2) (3) (4) (5) (6) Country Smoother λ Normal rate of interest (std. error) Standard error of rule p-value for test of symmetry(-,+) p-value for test of inflation nutter Australia (2.5,2.00) (5.560) (-0.114,0.012) Brazil ) (5.5,7.19) (3.424) (-0.014,-0.158) Canada (2,1.87) (0.606) (0.046,0.126) Czech Rep (3,3.24) (0.683) (0.030,0.074) EMU (2,1.95) (0.411) (0.415,0.217) Mexico (3.5,4.69) (1.124) (0.069,0.046) New Zealand 1990 (2,1.95) (2.110) (0.034,0.049) Norway (2,1.78) (2.672) (0.075,-0.026) Poland (2.5,4.30) (1.255) (0.055,0.362) South Africa 2000 (4.5,5.85) (0.829) (0.216,-0.337) South Korea (3,3.22) (0.682) (0.368,-0.163) Sweden (2,1.58) (1.147) (0.063,0.018) UK (2,1.96) (0.443) (-0.039,0.598)

16 able 8. Estimation of simple aylor rule for non-inflation targeting countries. Estimation by 2SLS. Potential variables in NKPC and AD curves are used as instruments. Common period (1) (2) (3) (4) (5) (6) Country Smoother λ Normal rate of interest (std. error) Standard error of rule p-value for test of symmetry(-,+) p-value for test of inflation nutter Denmark (-,1.92) (0.926) (0.034,0.235) India (-,6.52) (5.646) (-0.020,0.133) Japan (-,-0.47) (0.224) (-0.509,1.627) Switzerland (-,0.72) (0.393) (0.269,0.378) USA (-,2.08) (1.011) (0.451,0.323)

17 References Ambler, S., A. Dib, et al. (2004). Optimal aylor Rules in an Estimated Model of a Small Open Economy. Amin, W. (2005). aylor Rules and Interest Rate Dynamics in Emerging Market Economies, Kansas State University. Bailey, M. J. (1992). he Welfare Cost of Inflationary Finance. Studies in positive and normative economics, Economists of the wentieth Century series.aldershot, U.K.:Elgar; distributed in the U.S. by Ashgate, Brookfield, Vt: Chari, V. V., L. J. Christiano, et al. (1996). "Optimality of the Friedman Rule in Economies with Distorting axes." Journal of Monetary Economics 37(2): Correia, I. and P. eles (1996). "Is the Friedman Rule Optimal When Money Is an Intermediate Good?" Journal of Monetary Economics 38(2): Eusepi, S. (2005). Comparing forecast-based and backward-looking aylor rules: a 'global' analysis. Finansdepartementet (1997). Riksbankens st llning. Stockholm, Regeringskansliet, Finansdepartementet. Friedman, M. (2006). he Optimum Quantity of Money, With a new introduction by Michael D. Bordo. New Brunswick, N.J. and London: ransaction, Aldineransaction. Marzo, M. (2006). "Optimal Monetary Policy with Price and Wage Rigidities." Economic Notes, February 2006, v.35, iss.1, pp Osterholm, P. (2005). "he aylor Rule and Real-ime Data--A Critical Appraisal." Applied Economics Letters 12(11): Piergallini, A. (2006). "Fiscal Deficits, aylor Rules, and Price Dynamics." Atlantic Economic Journal 34(4): Svensson, L. (2006). he Instrument-Rate Projection under Inflation argeting: he Norwegian Example, Princeton University, Department of Economics, Center for Economic Policy Studies., Working Papers: 75. Svensson, L. E. O. and M. Woodford (2004). Implementing Optimal Policy hrough Inflation- Forecast argeting, C.E.P.R. Discussion Papers, CEPR Discussion Papers: Woodford, M. (2003). Interest and Prices. Foundations of a heory of Monetary Policy. Princeton and Oxford, Princeton University Press. Woodford, M. (2007). "he Case for Forecast argeting as a Monetary Policy Strategy." Journal of Economic Perspectives 21(4): 3-24.