The Financial Accelerator and the Real Economy: A Small Macroeconometric Model for Norway with Financial Frictions

Hammersland, Roger; Træe, Cathrine Bolstad Research Report The Financial Accelerator and the Real Economy: A Small Macroeconometric Model for Norway with Financial Frictions Staff Memo, No. 02/2012 Provided in Cooperation with: Norges Bank, Oslo Suggested Citation: Hammersland, Roger; Træe, Cathrine Bolstad (2012) : The Financial Accelerator and the Real Economy: A Small Macroeconometric Model for Norway with Financial Frictions, Staff Memo, No. 02/2012, ISBN 978-82-7553-648-6, Norges Bank, Oslo, https://hdl.handle.net/11250/2507201 This Version is available at: https://hdl.handle.net/10419/210240 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc-nd/4.0/deed.no No. 2 | 2012 The financial accelerator and the real economy: A small macroeconometric model for Norway with financial frictions Roger Hammersland and Cathrine Bolstad Træe, Statistics Norway, Institute for New Economic Thinking at the Oxford Martin School and Norges Bank Financial Stability Staff Memo Staff Memos present reports and documentation written by staff members and affiliates of Norges Bank, the central bank of Norway. Views and conclusions expressed in Staff Memos should not be taken to represent the views of Norges Bank. © 2012 Norges Bank The text may be quoted or referred to, provided that due acknowledgement is given to source. References to the staff memo should read: “Roger Hammersland and Cathrine Bolstad Træe (2012): "The fiancial accelerator and the real economy: A small macroeconometric model for Norway with financial frictions", Norges Bank Staff Memo No. 2“ Staff Memo inneholder utredninger og dokumentasjon skrevet av Norges Banks ansatte og andre forfattere tilknyttet Norges Bank. Synspunkter og konklusjoner i arbeidene er ikke nødvendigvis representative for Norges Banks. © 2012 Norges Bank Det kan siteres fra eller henvises til dette arbeid, gitt at forfatter og Norges Bank oppgis som kilde. ISSN 1504-2596 (online only) ISBN 978-82-7553-648-6 (online only) The financial accelerator and the real economy A small macroeconometric model for Norway with financial frictions by Roger Hammersland and Cathrine Bolstad Træe Statistics Norway, Institute for New Economic Thinking at the Oxford Martin School1 and Norges Bank Financial Stability2,3 Abstract This paper studies the salient features of a core macroeconometric model that allows for selfreinforcing co-movements between credit, asset prices and real economic activity. In contrast to the economic literature that cultivates highly stylized model representations aimed at illustrating the workings and the implications of such features, the model of this paper integrates no less than two mutually reinforcing financial accelerator mechanisms in a fullfledged core macroeconomic model framework. Noteworthy, the impulse responses of such a model turns out to be very much in line with the ones one would have expected using a typical SVAR/DSGE model, though the amplitude of shocks is in most cases stronger than the ones pertaining to these kinds of models. This is due to the workings of the financial accelerators that contribute to the magnification of the effects of shocks to the economy. Furthermore, a forecast comparison undertaken between our model and an alternative macroeconometric model without a financial block, suggests that financial feedback mechanisms may improve the forecasting properties of theory-informed macroeconometric models. Keywords: The Financial Accelerator, Structural Vector Error Correction Modelling, Impulse response analysis, Forecasting JEL classifications: E1, E32, E44 1 This paper was partly written when Roger Hammersland was a visiting scholar at the Institute for New Economic Thinking at the Oxford Martin School. 2 E-mail addresses: Roger Hammersland, r[email protected] and Cathrine Bolstad Træe, [email protected] 3 We are indebted to many colleagues and seminar participants. In particular we would like to thank David F. Hendry, Myron Kwast and Steinar Holden for valuable comments on former drafts of the paper. We would also like to thank participants at Eurostat’s 6th Colloquium on Modern tools for Business Cycle Analysis and participants at internal staff seminars at Statistics Norway and Norges Bank for many insightful comments and suggestions. A special thanks goes to Haakon Solheim for commenting on and proofreading several drafts of the paper, and to Farooq Akram for his comments and contribution to the forecast evaluation in Section 4. Finally we would like to address a word of thanks to the Institute for New Economic Thinking at the Oxford Martin School, University of Oxford, for providing us with support for this research project. 2 Introduction There is much to indicate that financial frictions could have an important bearing on the transmission mechanism of shocks. This paper presents a small macroeconometric model, called Small Macro Model (SMM), which is designed to incorporate self-reinforcing comovements between credit, asset prices and real economic activity. As a case in point, the effects of a shock to the monetary policy rule governing the key policy rate could be reinforced through several channels in the presence of frictions. Such a contingency can be illustrated by spelling out the transmission mechanism of a monetary policy shock in the presence of self-reinforcing feedback loops between credit, asset prices and real economic activity. An interest rate hike would through affecting the propensity to save on the part of households, lowering real investments and reducing net trade – the latter as a consequence of an appreciating real exchange rate − lead to a drop in activity that could potentially be reinforced by a procyclical correction to asset prices. Such kind of a selfreinforcing feedback mechanism is given support by standard theory. For instance in the case of Tobin’s Q (Tobin (1969)), lower asset prices lead to a drop in the ratio of the market value of capital to its replacement cost and thus reduced investment. The permanent income hypothesis (Friedman (1957)) can likewise argue for a similar mechanism based on a negative wealth effect on consumption. However, in the presence of financial frictions this is only part of the story. Lower asset prices that affect net worth of firms and household wealth would also have a negative effect on the value of collateral. In the presence of asymmetric information that raises the cost of external finance relative to the cost of internal finance, this would affect the borrowing capacity of wealth constrained entrepreneurs and households and thus reduce investments. Through the working of a credit-asset price spiral where lower asset prices spur lower credit and lower credit in turn leads to a reduction in investment − and thus further reductions in asset prices due to their procyclicality − this amounts to a mechanism that in the end will lead to a self-reinforced procyclical drop in domestic absorption and output, asset prices and credit. Such a mechanism is often referred to as a financial accelerator in the literature (See e.g. Kiyotaki and Moore (1997) and Bernanke, Gertler and Gilchrist (1999)). Figure I.1 presents a simplified flow diagram of the financial accelerator mechanism referred to in the text.4 4 In Section 2, we present a more comprehensive flow diagram that spells out the whole transmission mechanism of a monetary policy shock in relation to the small macroeconomic model developed in this paper. 3 Figure I.1The Financial Accelerator In contrast to highly stylized model representations aimed at illustrating the workings of a financial accelerator, the SMM model integrates such mechanisms in a full-fledged macroeconomic structural model. The model is used by the financial stability wing of Norges Bank, for the purpose of forecasting, constructing risk scenarios and to illustrate the relative importance of different transmission channels (see also Andersen et al. (2008)). The model presented herein is based on an augmented and revised version of the model documented in Bårdsen and Nymoen (2009), and implies a model for the real economy that includes a financial block. The role of the financial block is to take account of the co-movements and procyclicality of credit, asset prices and real economic activity that typically characterises a financial accelerator. The model differs from optimizing representative agent models in several respects, the main reason for this being a wider and less stringent theoretical framework, and the fact that data is given a more central role in the shaping of the longand short-run structure of the model.5 This notwithstanding, the impulse response pattern of this model turns out to be in line with the ones from a typical SVAR/DSGE model, though the amplitude of shocks is in most cases stronger than the ones pertaining to the latter kinds of models. This is due to the working of the financial accelerators that contribute to magnify the effects of shocks to the economy. Furthermore, a forecast comparison undertaken between our 5To be more explicit this means that data in this framework has played the role of distinguishing between admissible structures lying in a hypothetical extended possibility set. This is a possibility set that in addition to span an exhaustive catalogue of theory-admissible subject matter structures also is intended to cover relationships with a less solid theoretical foundation, like relationships regarded to be admissible only because they make sense. For a more comprehensive account of such an approach, the reader is referred to Section 1.2 and the discussion therein. Interest rate Asset prices Credit GDP 4 model and an alternative macroeconometric model without a financial block, suggests that financial feedback mechanisms may improve the forecasting properties of theory informed macroeconometric models. In the following, we start with a presentation of the model and its methodological foundation in section 1. That is, Subsection 1.1 starts out with a brief account of the principles behind the construction of our data-based model. In Subsection 1.2, this is followed by a more extended account of the methodology used in design and estimation. Particular emphasis is in this respect given to a discussion of a pragmatic and non-dogmatic approach to model design. Subsection 1.3 ends the section with a more comprehensive account of the model’s main features, including a full account of all the model’s behavioural equations. In Section 2 we present the model’s transmission mechanism to a monetary policy shock. Special emphasis has in this respect been placed on describing the role of the financial accelerators. In Section 3, we proceed to a description of the model’s longand short-run responses to a wide range of different shocks.6 In this section, particular importance has been attached to describing the dynamic transmission mechanism of shocks. Section 4 addresses the model’s forecast properties, comparing the model’s forecasts to forecasts of simple autoregressive and vector autoregressive models and an alternative econometric model designed and estimated on Norwegian data. Finally, Section 5 offers some concluding remarks. 6A structural shock is often taken to mean a shock with a clear structural interpretation, in the sense of referring to shocks to structural model representations derived from an explicit utility maximizing rational representative agent (RA) framework. However, in this case, a shock is given a wider interpretation, and refers to shocks to theory-driven structural representations in general, be that structures based on more old fashioned type of macro informed models, so-called emergent models or structural representations based on an explicit representative agent utility maximizing framework. A consequence of this is that the concept of “a structural shock” loses its un-ambiguity as several types of shocks can rightly be claimed to have a structural interpretation, though the way they are defined or interpreted as structural will differ across models. In spite of this, Section 3 reveals a great degree of conformity between our impulse responses and those following from a typical SVAR or DSGE framework. 5 1 The model and its construction 1.1 The construction of the model: Design and estimation The SMM model is an estimated equilibrium-correction model with in general backwardlooking rather than forward-looking rational expectations and a credit channel for monetary policy (see Bårdsen and Nymoen (2009)). At Norges Bank the model is mainly used for constructing risk scenarios related to low-probability events. A model with backward-looking expectations and on estimated reduced-form has proved to be useful for this purpose so far (see e.g. Bårdsen et al. (2006)). Economic policy enters the model through public expenditures as an exogenous variable in a reduced form GDP equation, as well as through an estimated Taylor-type rule for money market interest rates. The model uses quarterly data from 1978 to date. However, some equations are estimated over a shorter time period due to lack of data or difficulties in finding stable relationships over periods with shifts in policy regimes. The model’s variables can be decomposed into fully model endogenous variables, and weakly and strongly exogenous variables. The weakly exogenous variables are not only a function of variables characterized as strongly exogenous, but depend also on lagged variables classified as model endogenous. The strongly exogenous variables consist of non-modelled variables and policy variables, and include domestic tax rates, world market prices, real foreign demand and government expenditures. As far as the specification of the possibility set is concerned, the approach is closely related to − and compatible with − the design strategy proposed by the general to specific strategy of Hendry (1993), though it clearly is more restrictive than indicated by a completely atheoretical modelling strategy. Thus, as a backdrop for model design we have sought to start out with the most general specification given support by what we a priori perceive to be a sensible possibility set7, and then to simplify such a point of departure down to a parsimonious representation. Ideally, this process of reduction should have taken place within the framework of a simultaneous structural system setup. However, a general lack of degrees of freedom due to short time series makes such a strategy unfeasible and restricts us to follow a mixture of strategies. One of these involves splitting the model up into blocks perceived to be sufficiently autonomous to be treated separately from the rest of the system without 7 Proper account taken to subject matter theory. 6 invalidating the outcome of a modelling exercise. Another strategy implies to resort to individual equation model design procedures, proper account taken to the fact that some of the explanatory variables might be characterized as endogenous. In designing and estimating the model of this paper, a variety of strategies has thus been utilized. In estimating the wage, price and productivity block of the model we have used full information maximum likelihood in estimating the final structural specification, while the final structure itself is the outcome of a general to specific reduction process on the block’s individual equations separately. A potential bias in design due to simultaneity – and as indicated by proper tests of exogeneity – has in this context been taken into account by utilising appropriate instruments. Moreover, an automatic general-to-specific modelling algorithm called Autometrics (Doornik (2009)) has been used extensively as a device for controlling for a potential path dependence in the chosen simplification scheme (crosschecking). As far as the simultaneous system consisting of asset prices and corporate credit is concerned, this block has been designed and estimated jointly with real activity, utilising a fully simultaneous procedure of Simultaneous Structural Model Design.8 In this procedure, the whole structure9 of the subsystem has been designed and estimated jointly by full information maximum likelihood procedures, based on an exactly identified point of departure utilizing structural dummies. Noteworthy, and as distinct from the other equations of the model, this sequence of reductions has entirely been undertaken manually due to the lack of an automatic general-to-specific modelling algorithm for structural systems. Other equations of the model, like the equations for the nominal exchange rate, household credit, interest rates and importand house prices, have on the other hand all been designed by utilising ordinary least squares in an ordinary general to specific sequence of simplifications, proper account taken to alleviating the threat of a simultaneity bias in design by proper testing and utilising instruments if deemed necessary. As was the case for the single equation general-to-specific scheme followed to arrive at the final wage-, productivityand pricesystem, potential path dependence in the chosen simplification scheme has here been controlled for by using Autometrics. 8 See Hammersland and Jacobsen (2008) for a more detailed account of such a procedure. 9 That is all the equations of the structural model. 13 in real domestic credit to firms is contemporaneously affected by asset price growth ()pa . As asset prices in turn are affected contemporaneously by credit growth, Equation (14), this gives rise to a dynamic interaction between credit and asset prices that turns out to create a transmission mechanism by which the effects of real shocks could persist and amplify. This feature is fully in accordance with Kiyotaki and Moore (1997), where a financial accelerator mechanism is reinforced by a creditasset price spiral. As regards the long-run structure of our model, there is a link between household debt and output. However, according to Equation (1) there is no such link between enterprise debt and activity. Hence, while innovations to asset prices and firm credit do cause short run movements in production, and while real activity spurs credit of firms, such innovations do not precede real economy movements in the long run. Otherwise, according to Equation (13), higher oil prices ()po affect credit negatively in the short run, only mitigated partially by its positive effect on asset prices. Such an effect of higher oil prices on credit is interpreted to represent a cost effect. In the long-run, however, the effect of higher oil prices on credit comes exclusively via its effect on asset prices and is strongly positive. In fact a one percent rise in oil prices is estimated to increase credit in the long run by approximately 0.26 percent, the same effect that an oil price hike is estimated to have on asset prices in the long run. The equations of default16 by households and firms in (15) and (16), respectively, are based on Berge and Boye (2007). Households’ default rate () hh dc , i.e., default as a share of total household bank debt, depends on households’ real income ()inc p , unemployment ()u , the real interest rate ()RL and real house prices ()ph p . As regards firms’ default, there is no homogeneity between default and debt in the short run, only in the long run. Firms’ default, measured in real terms () e dp depends on the level of debt () e cr p , the real interest rate ()RL , domestic demand, proxied by the unemployment rate ()u , the real exchange rate * ()v p p as a measure of competitiveness and the real oil price ()po usd p . The latter variable captures that the level of activity and investments in the oil sector affect other industries. 16 Our data on defaults include both defaults and loans with a very high probability of default as reported by the banks (problem loans). These are all loans where banks have made write offs. The actual recorded losses by the banks are then denoted as a fraction alpha of these problem loans. 14 In addition to the equations commented on above comes a technical relation for the determination of the consumer price index adjusted for energy and taxes () c p and a set of identities defining various transformations of the model variables. 2 The Transmission Mechanism As commented on above the SMM model includes financial accelerators for both firms and households, see Figure 2.1; where procyclical fluctuations in houseand asset-prices affect borrowing capacity of, respectively, households and non-financial enterprises and hence real activity through an increase in both real and housing investments. As far as both accelerators are concerned, this feedback mechanism is reinforced by an assetprice credit spiral where higher asset prices chases more credit and vice versa. As will be shown in Section 3 on impulse responses, these feedback effects are significant in both the shortand long-run. Figure 2.1 - The transmission mechanisms of the SMM model Through the mechanisms outlined in Figure 2.1, the SMM model is able to represent procyclical co-movements between asset prices, credit growth and the real economy. House prices and credit to both households and firms directly affect GDP growth. Corporate and Wages(w) and prices(p) Productivity z GDP y Unemployment u Import prices pi Exchange rate v Interest rates R,RL Real economyvariables Financial sectorvariables House prices PH Household credit crh Credit nonfinancial enterprises cre Market capitalisation for enterprises pa 15 household credit affect GDP in the short run, possibly reflecting frictions in the credit market, while the long-run effect of household credit points towards some form of persistent rationing of the household sector. The house price effect can be interpreted as a wealth effect. As GDP growth spurs house prices and credit, in both the short and long run, a financial accelerator emerges that contributes to the amplification of shocks through a credit asset price spiral enhanced feedback mechanism between output, credit and house prices. To get a sounder grasp on the transmission mechanism of the model, and to facilitate the identification of the models’ chain of causation, we will take a closer look at the transmission mechanism of the monetary policy shock alluded to in the introduction and trace the entire dynamic response of a monetary policy shock in the model. A negative monetary policy shock in terms of a positive shock to the rule governing the policy rate, will lead to a decline in activity through several channels. First, given that a positive shock to the policy rule will lead to a jump in the money market interest rates (longand short-term), bank lending rates will to a varying degree follow suit. In the model, this will lead to a downward credit, house and asset price spiral. Combined with an enhanced propensity to save on part of households, lower real investments and reduced net trade – the latter as a result of a stronger real exchange rate − this will initiate a feedback mechanism that in the end leads to a self-reinforcing procyclical drop in domestic absorption and output, asset prices and credit. As output declines (relative to a baseline scenario) unemployment will also increase. In the model, this will dampen the pressure in the labour market and lead to a restrain in wage and consumer price inflation. Combined with a negative output gap this will result in a reversal of the central bank’s monetary policy stance and thus to lower interest rates. Lower interest rates on the other hand will contribute positively to household credit and house prices. Together with lower domestic inflation and a weakening of the exchange rate this will lead to a significant slowing down of the feedback mechanism initiated by the monetary policy shock in the first place. Eventually, this course of progress will in the model partially reverse the decline in employment. However, before this happens, wage and price inflation have already reached their turning point as a consequence of productivity gains related to the high level of unemployment. Lower unemployment on the other hand will eventually contribute to the amplification of this process of higher wage and price inflation and we enter a new period of policy tightening on part of the central bank. This tightening will so initiate a new round of cyclical oscillations to take place and so it continues until the oscillations in the long run gradually die out. 16 3 Impulse Responses In this section, we illustrate the model’s shortand long-run properties by adding a series of shocks. The shocks are considered one at a time, entered as shocks to a baseline scenario of the model. Noteworthy is the fact that the impulse response patterns overall are very much in line with the ones one would have expected using a representative agent (RA) modelling framework, though the amplitude of the responses in most cases is stronger and the responses are more volatile than the ones in for instance the Dynamic Stochastic General Equilibrium Models of the Euro area (SW) and Norway (NEMO) (see respectively Smets and Wouters (2003) and Brubakk et al. (2006)). The stronger amplitude is mainly due to the working of the financial accelerators that contribute to magnify the effects of shocks to the economy. The more volatile pattern comes as a consequence of utilizing unadjusted data, a richer dynamic structural model specification and a policy rule with an interest rate persistence that differs somewhat from the ones present in SW and NEMO. A shock to interest rates Figure 3.1 shows the responses to a shock to the equation governing the money market interest rate, calibrated such that the money market interest rate increases by 1 percentage point in 2010q4, and letting the full system play out freely after the shock. The impulse responses are based upon the monetary policy reaction function of a Taylor-type rule, see Equation (8) in Appendix 2.17 The interest rate increase is channelled to the real economy through an increase in the banks’ lending rates, as well as through a currency appreciation, both having a contractionary effect on activity and employment, amplified by the financial accelerators. As a consequence, consumer price inflation and wage inflation are reduced. Credit demand and house price growth falls. In the quarters following the shock, the interest rate gradually reduces to its previous level, and at the same time the exchange rate depreciates. GDP growth strengthens, with inflation and credit growth also picking up again. The effect diminishes in the course of the 20 quarters covered by the graph, indicating that the system is stable (For a more comprehensive account of the transmission mechanism, see the last part of the previous section). 17 See appendix 4 for an alternative impulse response analysis based upon an augmented Taylor rule with interest rate smoothing. 17 In contrast to the impulse responses of a monetary policy shock in a typical DSGE model, the real quantitative consequences of the shock are amplified while real prices including the real wage are less affected by the interest hike. For instance, in SW a 1 percentage point increase in the sight deposit rate is estimated to reduce output by approximately 0.4 per cent after about 4 quarters. The impulse responses of a corresponding monetary policy shock using NEMO imply a less pronounced fall in GDP of 0.25 per cent and the outputresponse is somewhat quicker than in SW. In the case of the SMM model, the same type of shock is simulated to reduce output by almost 0.6 percent already after a couple of quarters. However, a prompt policy response contributes to quickly reverse the drop in output such that output is back on trend already after 6-8 quarters. Unemployment on the other hand shows a more protracted course of progress as the effect does not reach its maximum of almost 0.07 percentage points − corresponding to an increase of about 2 ½ per cent − before after 4-5 quarters. However, as regards real wages these are in SW simulated to be reduced by approximately 0.25 per cent after about 8 periods, while the SMM model predicts a more modest drop of about 0.15 per cent in a slightly shorter time span (6-7 quarters). Noteworthy, real wages in the SMM model initially rise in the wake of a monetary policy shock. This is due to nominal prices being more flexible than nominal wages in the short run. Overall, though, the pattern is largely the same as in both SW and NEMO, with hump shaped responses to output, prices and wages. 18 -0.4 0.0 0.4 0.8 1.2 08 10 12 14 16 18 20 3 months effective nominal money market rate -.10 -.08 -.06 -.04 -.02 .00 08 10 12 14 16 18 20 Consumer Price Index -.10 -.08 -.06 -.04 -.02 .00 .02 08 10 12 14 16 18 20 Inflation (CPI) -.06 -.04 -.02 .00 .02 08 10 12 14 16 18 20 Core inflation (CPIJAE) -.25 -.20 -.15 -.10 -.05 .00 .05 08 10 12 14 16 18 20 Wage Income per hour -.20 -.15 -.10 -.05 .00 .05 .10 08 10 12 14 16 18 20 Real Wage Income per hour -.6 -.5 -.4 -.3 -.2 -.1 .0 08 10 12 14 16 18 20 GDP Mainland Norway -.6 -.4 -.2 .0 .2 .4 08 10 12 14 16 18 20 GDP Mainland Norway, growth rate -.04 -.02 .00 .02 .04 .06 .08 08 10 12 14 16 18 20 Unemployment rate -4 -3 -2 -1 0 1 08 10 12 14 16 18 20 Real Exchange rate -2.0 -1.5 -1.0 -0.5 0.0 0.5 08 10 12 14 16 18 20 Real house prices -2 -1 0 1 2 08 10 12 14 16 18 20 Real house prices, growth rate -.8 -.6 -.4 -.2 .0 08 10 12 14 16 18 20 Real domestic credit to households -.8 -.6 -.4 -.2 .0 08 10 12 14 16 18 20 Real credit to non-financial enterprises -.02 .00 .02 .04 .06 08 10 12 14 16 18 20 Banks problem loan share, total Deviation Figure 3.1 – A rise in the money market interest rate1 1 Quarterly data, the numbers on the time axis refer to observations from 2008Q1 onwards. Most impulse responses for the variables in levels are displayed as deviations from the baseline in percent. Interest rates, the rate of unemployment and the problem loan share are displayed as deviations in percentage points from the baseline scenario. Moreover, growth rate responses are all displayed as deviations in percentage points from the baseline scenario. 19 0.0 0.2 0.4 0.6 0.8 1.0 1.2 08 10 12 14 16 18 20 3 months effective nominal money market rate 0.00 0.25 0.50 0.75 1.00 1.25 1.50 08 10 12 14 16 18 20 Consumer Price Index -0.4 0.0 0.4 0.8 1.2 1.6 08 10 12 14 16 18 20 Inflation (CPI) 0.0 0.2 0.4 0.6 0.8 1.0 08 10 12 14 16 18 20 Core inflation (CPIJAE) -0.4 0.0 0.4 0.8 1.2 1.6 08 10 12 14 16 18 20 Wage Income per hour -1.2 -0.8 -0.4 0.0 0.4 08 10 12 14 16 18 20 Real Wage Income per hour -.6 -.5 -.4 -.3 -.2 -.1 .0 08 10 12 14 16 18 20 GDP Mainland Norway -.6 -.4 -.2 .0 .2 08 10 12 14 16 18 20 GDP Mainland Norway, growth rate -.04 .00 .04 .08 .12 08 10 12 14 16 18 20 Unemployment rate -5 -4 -3 -2 -1 0 1 08 10 12 14 16 18 20 Real Exchange rate -6 -4 -2 0 2 08 10 12 14 16 18 20 Real house prices -6 -4 -2 0 2 4 08 10 12 14 16 18 20 Real house prices, growth rate -2.5 -2.0 -1.5 -1.0 -0.5 0.0 08 10 12 14 16 18 20 Real domestic credit to households -2.5 -2.0 -1.5 -1.0 -0.5 0.0 08 10 12 14 16 18 20 Real credit to non-financial enterprises -.02 .00 .02 .04 .06 .08 .10 .12 08 10 12 14 16 18 20 Banks problem loan share, total Deviation A price shock Figure 3.2 below, shows the response to a permanent shock to the equation governing consumer prices, calibrated such that the CPI index increases by 1 percent in 2010q4, and letting the full system play out freely after the shock. The impulse responses are based upon using a Taylor rule, Equation (8) in Appendix 2. Again the similarities to the responses in SW are striking. Figure 3.2 – A shock to the price level1 1 Quarterly data, the numbers on the time axis refer to observations from 2008Q1 onwards. Most impulse responses for the variables in levels are displayed as deviations from the baseline in percent. Interest rates, the rate of unemployment and the problem loan share are displayed as deviations in percentage points from the baseline scenario. Moreover, growth rate responses are all displayed as deviations in percentage points from the baseline scenario. 20 A price-shock in the SMM model will lead to an instantaneous appreciation of the real exchange rate, and via the Taylor rule, to higher real money market interest rates. Higher real interest rates and a stronger real exchange rate will on the other hand contribute to reducing the level of real activity in the economy, as a result of both lower consumption, investments and a drop in net exports. As a result, the rate of unemployment will start to increase. Together with reduced real income, these effects combined will spur a financial accelerator where lower credit (both among households and firms), reduced activity and increased unemployment contribute to mutually reinforcing each other. As far as both sectors are concerned this financial accelerator is boosted by a creditasset price spiral, where falling asset prices and credit mutually contribute to reinforce each other. Gradually, the level of unemployment will have increased so much that the pressure on wages and prices starts to abate. Together with a lingering nominal depreciation and an expansionary monetary policy, there will be a gradual pick-up in activity and employment. As with the decline, this process will be characterized by the credit asset price spiral enhanced financial accelerator where asset prices, credit and activity contribute to mutually reinforce each other according to the mechanism described above, only that this time the process will be put in reverse. When activity and inflation have recovered sufficiently, time has come for a new bout of interest rate increases and real exchange appreciations. In other words, we have started on a new cycle exactly like the one we have just described, the only difference being that the amplitude this time is smaller. In accordance with Figure 3.2, this process of subsequent cycles will continue until the cycles become so small that they eventually die out (In the figure this does not seem to happen before after the end of the simulation period). In contrast to the impulse responses in SW, the real volume effects of a shock to the price mark-up are all amplified due to the working of the financial accelerators. For instance, a oneperiod shock that is calibrated to lead to an instantaneous rise in consumer prices of one per cent is in SW estimated to reduce output by approximately 0.12 per cent after 4-5 periods. A similar exercise using the SMM model on the other hand would, according to Figure 3.2, lead to a reduction in the order of magnitude of 0.5 per cent after 5-6 quarters. As was the case with a monetary policy shock the unemployment response of the SMM model is protracted and does not reach its maximum increase of about 0.1 percentage points − corresponding to a increase in unemployment of about 3.5 percent − before after about 8 quarters time. As regards wages, we see that the pass-through of the price shock is rather slow and protracted in the SMM model leading to an instantaneous drop in real wages of almost the same order of magnitude as the shock to inflation itself. Compared to an instantaneous drop of 21 approximately 0.2 percent in SW this illustrates the comparable high degree of nominal wage stickiness in the SMM model. However, Figure 3.2 shows that the real wage gradually rises towards its equilibrium level – given by its level in the baseline scenario − in the long run. Noteworthy, this is a characteristic that SMM shares with the impulse responses in SW. Another feature that SMM seems to share with the impulse responses of a price shock in SW is the relatively protracted and slow adjustment of real output as neither in SW nor in the SMM, output seems to have reached its longrun equilibrium level within the simulation period. While SW though clearly demonstrates that output converges to its baseline level in the long run this is more unclear in the case of the SMM model. In fact, an extension of the simulation period shows that output in the case of the SMM model remains below its baseline level for a considerable time. Figure 3.2 and Equation (1) in the appendix suggest that this comes mainly as a consequence of a persistent drop in household and firm credit. Otherwise, we do again see that the impulse response pattern is largely the same as in SW, with hump shaped responses to output, prices and wages. A wage shock Figure 3.3 shows the response to a permanent shock to the equation governing the wage rate, calibrated such that wages increase by one percent in 2010q4, and letting the full system play out freely after the shock. The impulse responses are based upon using the Taylor rule, Equation (8), in Appendix 2. A wage-shock will in the model feed into higher inflation. Higher inflation in turn will lead to an instantaneous appreciation of the real exchange rate and via the Taylor rule, higher real interest rates. Higher real interest rates and a stronger real exchange rate will on the other hand contribute to reducing the level of real activity in the economy, as a result of both lower consumption, investment and a drop in net exports. As a result, the rate of unemployment will start to increase. Together with reduced income these effects combined will ignite the financial accelerators of the model where lower credit (both among households as among firms), reduced activity and increased unemployment contribute to mutually reinforce each other. As for both sectors this financial accelerator is boosted by a credit asset price spiral, where falling asset prices and falling credit mutually contribute to a reinforce each other. Gradually, the level of unemployment will have increased so much that the pressure on wages and prices starts to abate. Together with a lingering nominal depreciation and a monetary policy put in reverse, this will in turn lead to a gradual pick-up in activity and employment. 22 .00 .02 .04 .06 .08 .10 .12 08 10 12 14 16 18 20 3 months effective nominal money market rate .00 .04 .08 .12 .16 08 10 12 14 16 18 20 Consumer Price Index -.04 .00 .04 .08 .12 .16 08 10 12 14 16 18 20 Inflation (CPI) .00 .02 .04 .06 .08 .10 08 10 12 14 16 18 20 Core inflation (CPIJAE) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 08 10 12 14 16 18 20 Wage Income per hour 0.0 0.2 0.4 0.6 0.8 1.0 1.2 08 10 12 14 16 18 20 Real Wage Income per hour -.04 -.02 .00 .02 .04 .06 08 10 12 14 16 18 20 GDP Mainland Norway -.08 -.06 -.04 -.02 .00 .02 .04 .06 08 10 12 14 16 18 20 GDP Mainland Norway, growth rate -.005 .000 .005 .010 .015 .020 .025 08 10 12 14 16 18 20 Unemployment rate -.5 -.4 -.3 -.2 -.1 .0 .1 08 10 12 14 16 18 20 Real Exchange rate -.3 -.2 -.1 .0 .1 08 10 12 14 16 18 20 Real house prices -.3 -.2 -.1 .0 .1 .2 .3 08 10 12 14 16 18 20 Real house prices, growth rate -.2 -.1 .0 .1 .2 08 10 12 14 16 18 20 Real domestic credit to households -.16 -.12 -.08 -.04 .00 .04 08 10 12 14 16 18 20 Real credit to non-financial enterprises -.004 .000 .004 .008 .012 .016 08 10 12 14 16 18 20 Banks problem loan share, total Deviation Figure 3.3 – A shock to wages1 As with the decline, this process will be characterized by the financial accelerators ,where asset prices, credit and activity contribute to mutually reinforce each other according to the mechanism described in earlier sections, only that this time the process will be put in reverse. 1 Quarterly data, the numbers on the time axis refer to observations from 2008Q1 onwards. Most impulse responses for the variables in levels are displayed as deviations from the baseline in percent. Interest rates, the rate of unemployment and the problem loan share are displayed as deviations in percentage points from the baseline scenario. Moreover, growth rate responses are all displayed as deviations in percentage points from the baseline scenario. 29 4.4 Results Detailed results of the forecast comparisons are presented in respectively, Table 1 for the forecasts starting each quarter, and Table 2 for the forecasts starting every 4th quarter. The figures in the respective cells show variable-specific RMSE-values for SMM compared to corresponding values for an indicated model. Numerical values larger than 1 indicate higher RMSE-values and thus poorer accurateness for SMM. Missing values due to non-comparable variables or missing variables in the VAR or EMod model are indicated by a "-". The forecasts are worked out for 4, 8 and 12 quarters over the period 2001Q1 to 2009Q4 and the start period has been advanced, respectively, one and 4 quarters.19 One can draw the following conclusions based on the forecast comparisons of these tables: The accuracy of SMM is better than AR and VAR for forecasting wage inflation and GDP growth on all horizons. However, for forecasting core inflation, the rate of unemployment, the lending rate and the real exchange rate, the VAR model is the best on all horizons. For forecasting core inflation, GDP growth, the rate of unemployment, short-term interest rates and the nominal exchange rate, the SMM model is better than the corresponding forecasts of the EMod model on all but the 4 quarter horizon, where the core inflation forecast of the EMod model does a slightly better job than the SMM model. However, for wage inflation the EMod model makes it clearly better than the SMM model. The relative advantage of the SMM model seems to increase with the forecast horizon. 19 Noteworthy, only part of these forecasts can be characterized as true “out-of-sample” forecasts, as all the models have used data up to and including 2007Q4 in their design. The forecasts made for the period after 2007Q4 though could be classified as close to true “out-of-sample” forecasts. 30 Table 1: SMM’s forecast properties when forecasts are made each quarter. Relative RMSE. Table 2: SMM’s forecast properties when forecasts are made every 4 quarter. Relative RMSE. 31 5 Conclusions In this paper, we have studied a small macroeconomic model that allows for self-reinforcing co-movements between credit, asset prices and real economic activity, often denominated a financial accelerator in the literature. The model considered in this paper has tried to integrate a financial accelerator mechanism in a full-fledged macroeconomic model framework. New in this context, is the fact that the model presented in this paper contains no less than two interdependent financial accelerator mechanisms; i) one with a firm side origin where asset prices affect borrowing capacity and hence real activity through an increase in real investments, and ii) another based on a similar procyclical feedback mechanism between household credit, house prices and housing investment. Noteworthy, the impulse response patterns overall are very much in line with the ones based on a typical SVAR/DSGE model, though the amplitude of shocks are in most cases stronger than in the latter models. This is mainly due to the working of the financial accelerators that contribute to magnify the effects of shocks to the economy. Taken at face value this suggests that the absence of an explicit financial accelerator can lead to underestimation of effects of shocks in a macroeconomic model. As regards the forecast properties of the model, the model clearly outperforms simple univariate autoregressive time series models. It also outperforms an alternative econometric model designed on Norwegian data without a financial block. As far as the last finding is concerned this suggests that the incorporation of financial accelerators can improve forecasting properties of a macroeconomic model. For some variables, though, a multivariate data driven VAR seems to be preferable to our model when it comes to forecasting. However, the relative advantage of the SMM model seems to increase with the forecast horizon. 32 References Akram, Q.F. (2004): Oil prices and exchange rates: Norwegian evidence, Econometrics Journal, Vol. 7, pp. 476-504. Andersen, A., T.O. Berge, E. Bernhardsen, K.G. Lindquist and B.H.Vatne (2008): A suite-of models approach to stress-testing financial stability, Staff Memo 2/2008, Norges Bank. Berge, T. O. and K. Boye (2007): An analysis of banks’ problem loans, Economic Bulletin 2/2007, Norges Bank. Bernanke, B., M. Gertler and S. Gilchrist (1999), The financial accelerator in a quantitative business cycle framework, in J. Taylor and M. Woodford, eds, ‘Handbook of Macroeconomic’, Vol.1, Elsevier Science B.V., pp. 1341-1393. Bernhardsen, T. and G. Bårdsen (2004): “Sammenhengen mellom styringsrenten og makroøkonomiske variable: Noen enkle ligninger for Norge”, Staff Memo 2/2004, Norges Bank. Brubakk, L, T.A. Husebø, J. Maih, K. Olsen and M. Østnor (2006): Finding NEMO: Documentation of the Norwegian economy model, Staff Memo 6/2006, Norges Bank. Bårdsen, G., Ø. Eitrheim, E.S. Jansen and R. Nymoen (2005): The econometrics of macroeconomic modelling. Oxford University Press. Bårdsen, G., Lindquist, K.G. and D.P. Tsomocos (2006): Evaluation of macroeconomic models for financial stability analysis, Working Paper No. 1, Norges Bank. Bårdsen, G. and R. Nymoen (2009): Macroeconometric modelling for policy, Ch 17, Palgrave Handbook of Econometrics Vol 2, Palgrave Mac-Millan. Colander, David C. (1993): The macro foundations of micro, Eastern Economic Journal, vol. 19, no. 4, 1993, pp 447-457. Doornik, Jurgen A. (2009) Autometrics. In: The Methodology and Practice of Econometrics: A Festschrift in Honour of David F. Hendry. Oxford University Press. ISBN 978-0-19923719-7. Friedman, M. (1957): A theory of the consumption function. Princeton, NJ: Princeton University Press. Goodfriend, M. (1991): “Interest rates and the conduct of monetary policy”, CarnegieRochester Series on Public Policy, pp. 7-30. Goodhart, Charles (1999): “Central Bankers and uncertainty”, Bank of England quarterly bulletin, 39(1). pp. 102-121. ISSN 0005-5166. Hammersland, R. and D. H. Jacobsen (2008): The Financial Accelerator: Evidence using a procedure of Structural Model Design, Discussion Paper No. 569, Statistics Norway. 33 Hendry, D. F. (1993) Econometrics: Alchemy or Science? Essays in Econometric Methodology, Blackwell Publishers, Oxford. Jacobsen, D.H and B.E. Naug (2004): What influences the growth of household debt? Economic Bulletin 3, Norges Bank, pp. 191-201. Jacobsen, D.H and B.E. Naug (2005): What drives house prices? Economic Bulletin 1, Norges Bank, pp. 29-41. Jacobsen, D.H., K. Solberg-Johansen and K. Haugland (2007): Housing investments and house prices, Economic Bulletin 1, Norges Bank, pp. 33-46. Kiyotaki, N. and J. Moore (1997): Credit Cycles, Journal of Political Economy 105, pp. 211248. Kolsrud, D. and R. Nymoen (1998): Unemployment and the open economy wage-price spiral, Journal of Economic Studies 25 (6), pp. 450-467. Kuttner K.N.(2004): “The role of policy rules in inflation targeting”, Federal Reserve Bank of St. Louis, vol. 86, no. 4, pp. 89-111. Layard, R., S. Nickell and R. Jackmann (1994): The Unemployment Crisis. Oxford University Press. Smets, F. and R. Wouters (2003): An estimated Dynamic Stochastic General Equilibrium Model of the Euro Area, Journal of the European Economic Association, no 5, pp. 11231175. Tobin, J. (1969): A general equilibrium approach to monetary theory, Journal of Money, Credit and Banking Vol 1, pp. 15-29. 34 Appendix 1 – Variable descriptions and data sources crte Credit to non-financial enterprises, mainland Norway. Source: Statistics Norway pt Consumer Price Index (CPI). Source: Statistics Norway ptc Consumer Price Index Adjusted for Taxes and Energy Prices (CPI-ATE). Sources: Statistics Norway and Norges Bank vt Nominal Exchange Rate, import-weighted 44 countries (I-44). Sources: Statistics Norway and Norges Bank pt* Consumer Price Index Trading Partners (25 countries). Sources: Statistics Norway and Norges Bank pit Imports, deflator He Household expectations (Norsk Trendindikator). Source: TNS Gallup crth Domestic credit to households (C2). Source: Statistics Norway πt CPI inflation. πtc Core inflation (CPI-ATE) πt* CPI inflation trading partners gt Public consumption. Source: Statistics Norway zt Productivity. GDP divided by hours worked. Source: Statistics Norway pet CPI Electricity Component. Source: Statistics Norway µRLM Long-run lending margins at banks (calibrated) pht House prices. Thousand NOK per square meter. Sources: NEF, NFF, Finn.no, Econ Pöyry inct Wage income households. Source: Statistics Norway hst Value of households housing stock. Source: Statistics Norway jt Gross investment housing. Source: Statistics Norway pjt Deflator housing investment. Source: Statistics Norway pat Price shares Oslo Stock Exchange (OSEAX). Source: Ecowin and Statistics Norway dth Banks’ problem loans, households. Sources: Statistics Norway and Norges Bank dte Banks’ problem loans, non-financial enterprises. Sources: Statistics Norway and Norges Bank 4 4 100 t t P P 4 4 100 t t P P4 4 100 t t P P 35 usdt Nominal spot exchange rate NOK/USD. Source: Norges Bank RLt Average spot interest rate on bank loans (total). Sources: Statistics Norway and Norges Bank Rt 3 months effective nominal money market rate. (NIBOR). Source: Norges Bank Rt* 3 months effective nominal money market rate, euro area. (EURIBOR). Source: Norges Bank pot Oil prices Brent Blend USD per barrell. Source: Norges Bank ut Registered unemployment rate. Number of unemployed people registered at NAV. Source: Statistics Norway pit* Producer Price Index, Norway’s 25 largest Trading Partners. Source: Norges Bank wt Wage Income per hour Mainland Norway. Source: Statistics Norway yt Gross domestic product Mainland Norway. Measured in million NOK at fixed market value prices. Source: Statistics Norway Dummy for inflation targeting. Equals one starting from 2001Q2 The mean of the steady-state relationship for equation i 36 2 22 1 4 4 11 (0.01) (0.08) (0.33) 0.03[ 11.1 ( ) ] 11 0.4 1.5( ( ) 22 Estimated by OLS, T=119. Estimation period 1979Q3-2009Q1 u t t t t t j t j jj u u w p u y mean y Appendix 2 – The main equations of SMM Variables in small letters denote the natural logarithm of the variable. j denotes the j-period difference operator, and foreign variables are denoted with starred superscripts. Standard errors are reported in brackets. The estimation method, which is either ordinary least squares (OLS), or full information maximum likelihood (FIML) is indicated below each equation, along with the sample size ,T. Intercept terms, dummies and seasonal dummies are omitted due to space considerations. Identities are not reported. Real economy block Aggregate demand 2 2 1 4 1 1 1 1 1 3 (0.05) (0.1) (0.06) 0.2[( 0.8 0.1( ) 0.1( ) 0.01( ) ] 0.6 0.7 0.4 0.1 ( ) 0.1 ( ) 0.2 ( ) h t t t t t t eh t t t t t t y y g v p p cr p RL y g g ph p cr p cr p (0.1) (0.04) (0.04) (0.08) Estimated by OLS, T=64. Estimation period 1991Q1-2006Q4 (1) Exchange rate 1 1 1 1 (0.04) (0.01) (0.01) (0.02) 0.1[( ) 0.03(( ) ( ) ) 0.1( ) ] ( 0.04 0.05 0.1 ) Estimated by OLS, T=63. Esti t t t t t v t t t v v p p R R po usd p R R po mation period 1994Q2-2009Q4 (2) Import prices 11 (0.1) (0.08) (0.2) 0.4[( ) 0.6( ) ] 0.4 0.8 Estimated by OLS, T=70. Estimation period 1990Q1-2007Q2 t t t tt pi pi pi v p p v v pi (3) Unemployment (4) 37 Wages, prices and productivity Wages 2 1 2 1 11 (0.05) (0.09) 0.4[ 0.1 ] 0.8 ( ) t t t t t w t t t w w p z u w z z (5) Consumer prices 3 3 2 1 2 1 2 (0.01) (0.06) (0.04) (0.04) (0.009) 0.05[ 0.7( ) 0.3 ] + 0.3 0.1 0.1 ( ) 0.1 t t t t t p t t t t t p p w z pi p y w z pe (6) Productivity 2 1 3 3 3 2 (0.08) (0.1) (0.07) 0.5 0.15[ 0.52( ) 0.0004 0.0025 ] 0.12 Estimated by FIML, T=117. Estimation period 1978Q4-2007Q4 t t t t t t z tt z z z w p u T wp (7) Financial sector block Money market interest-rate21 (0.2) (0.3) 1.3( 2.5) 0.7( 3) 5.5 Estimated by OLS, T=43. Estimation period 1999Q1-2009Q1 c t t t Ru (8) Banks’ lending rate 1 1 1 0.8 0.2 0.35[ ] Calibrated t t t t t RLM RL R R RL R (9) 20 In the forecasting version of this equation, 8 b) the terms for foreign money market interest rates and interest rate differentials are included. 38 Household debt 1 4 4 2 2 2 2 3 (0.01) (0.0 ( ) 0.06[( ) 0.9( ) 0.03 0.4( ) ] 0.006 0.2 ( ) 0.1( ( ) ( ) ) hh t t t t t t t t cr p cr p ph p RL inc p RL inc p ph p ph p 01) (0.05) (0.02) Estimated by OLS, T=73. Estimation period 1991Q1-2009Q1 (10) House prices 1 1 1 1 1 (0.03) (0.2) (0.006) (0.005) (0.02) 0.1[ 0.07 0.4 1.1( ) 0.2 ] +0.1 0.02 0.01 0.07 Esti h t t t t t t e t t t t ph ph RL u inc hs cr inc RL RL H mated by OLS, T=75. Estimation period 1990Q2-2009Q1 (11) Housing investments 1 10 4 1 4 4 (0.08) (0.006) (0.005) 0.2[( ) ( ) ( ) ( ) ] 0.01 ( ) 0.02( ) Estimated by OLS, T=73. Estimation period 1991Q1-20 t t t t t t tt j j hs ph p inc p pj p RL RL 09Q1 (12) Non-financial enterprise debt and asset prices Non-financial enterprise debt 11 1 (0.08) (0.01) (0.17) (0.035) (0.02) ( ) 0.38( 0.007 ) 0.04[( ) ] 0.58 0.075 0.047 0.15 ( ) ee tt e t t t t cr p y Trend cr p pa y pa po cr p (0.09) (13) Asset prices 11 11 (0.04) (0.09) (0.33) 0.2( 0.25 0.01 ) 0.46 ( ) 0.44 ( ) 0.16 0.75 t t t ee t t t t pa pa po Trend pa cr p cr p po y 22 (0.26) (0.06) (0.21) Estimated by FIML , T=82. Estimation period 1986Q2-2006Q3 (14) 21 With GDP mainland Norway as an endogenous variable in the estimation, see Hammersland and Jacobsen (2008) for details. 45 * 1 2 1 0.25( 1.5( 2.5) 0.7( 3) 5.5) 0.3 0.1 c t t t t t t R R u R R Appendix 4 – A shock to interest rates, alternative monetary policy rule The monetary policy rule is, as one would expect, of great importance for the amplitude and dynamics of the model’s impulse responses to shocks. Often it is assumed a more sluggish adjustment of the monetary policy rate, by including a lagged term for the interest rate. The Taylor rule is simple compared to the actual information set central banks consider. Furthermore, data is often available with a lag, and can be subject to revisions. It can also be argued that by changing policy rates gradually, interest rate smoothing can ease the communication of policy to financial markets (see for example Goodfriend(1991)). Moreover, for example Goodhart (1999) argues that policymakers would move interest rates more slowly in order to avoid the need for frequent policy reversals. In the following, we look at the impulse responses based upon a Taylor rule augmented with a partial adjustment process for the key policy rate22: We also include foreign interest rates in this rule. This term can be thought of as a proxy for exchange rate effects. When foreign interest rates are included, we can also take contagion effects from increased money market risk premiums abroad into account. Figure A4.1 shows the impulse responses of a permanent shock to the equation governing the money market interest rate, calibrated such that the money market interest rate increases by 1 percentage point in 2010q4, and letting the full system play out freely after the shock. As was the case when using the Taylor rule, the interest rate increase is again channelled to the real economy through an increase in the bank lending rate, as well as through a currency appreciation, both channels leading to a contraction in real output and employment. As a consequence, consumer price inflation and wage inflation are also this time reduced and credit demand and house price growth fall. In the quarters following the shock, the interest rate gradually falls back to its previous level, and at the same time the exchange rate depreciates. GDP growth strengthens, with inflation and credit growth also picking up again. As was the case using the policy rule of Equation (8), the effect diminishes in the course of the 20 22 We have calibrated the smoothing coefficient and the remainder of the long-term relationship in brackets. The short-run effects are estimated over the time period 1999Q1 to 2007Q4. The chosen smoothing coefficient in levels is broadly in line with the results of Bernhardsen and Bårdsen (2004) for Norway. The coefficient is also in line with what eg. Kuttner (2004) finds for Sweden, US, UK and New Zealand (0.75-0.95). 46 quarters covered by the graph, indicating that the system is stable (For a more comprehensive account of the transmission mechanism, see the previous section). However, in contrast to the impulse responses of a monetary policy shock in a typical DSGE model, real effects of the shock are this time both strongly amplified. As we have commented on earlier, a 1 percentage point increase in the sight deposit rate is in SW estimated to reduce output by approximately 0.4 per cent after approximately 4 quarters. Furthermore, a similar experiment using NEMO, leads to a less pronounced fall in real activity of 0.25 per cent over a slightly shorter time span. Noteworthy, and which should be evident by looking at Figure A4.1, a similar experiment using a version of the SMM model were the Taylor rule has been replaced by an augmented policy rule with interest smoothing, leads to a considerably stronger drop in activity of almost 1 per cent. As was the case with a Taylor rule, the policy response contributes to reverse the drop in output, though output this time is not back on trend before after about 4 year’s time. Unemployment also shows a relatively protracted increase as the effect does not reach its maximum of about 0.3 percentage points − corresponding to an increase of close to 10 per cent − before after about 6-8 quarters. Also, as regards real wages these are in SW simulated to be reduced by approximately 0.25 after about 8 periods, while the SMM model predicts a larger fall of about 0.5 per cent over a somewhat longer time span (about 10 quarters). Interestingly, as was the case using the Taylor rule, real wages in the SMM model initially rises in the wake of a monetary policy shock. This is due to nominal wages being less flexible than prices in the short run. Overall, though, the pattern is largely the same as in both SW and NEMO, with hump shaped responses to output, prices and wages. 47 Figure A4.1 – A rise in the money market interest rate, augmented Taylor rule1 -0.4 0.0 0.4 0.8 1.2 08 10 12 14 16 18 20 3 months effective nominal money market rate -.5 -.4 -.3 -.2 -.1 .0 08 10 12 14 16 18 20 Consumer Price Index -.3 -.2 -.1 .0 .1 .2 08 10 12 14 16 18 20 Inflation (CPI) -.20 -.15 -.10 -.05 .00 .05 .10 08 10 12 14 16 18 20 Core inflation (CPIJAE) -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 08 10 12 14 16 18 20 Wage Income per hour -.6 -.4 -.2 .0 .2 .4 08 10 12 14 16 18 20 Real Wage Income per hour -1.2 -0.8 -0.4 0.0 0.4 08 10 12 14 16 18 20 GDP Mainland Norway -1.2 -0.8 -0.4 0.0 0.4 0.8 08 10 12 14 16 18 20 GDP Mainland Norway, growth rate -.2 -.1 .0 .1 .2 .3 08 10 12 14 16 18 20 Unemployment rate -4 -3 -2 -1 0 1 2 08 10 12 14 16 18 20 Real Exchange rate -4 -3 -2 -1 0 1 08 10 12 14 16 18 20 Real house prices -4 -3 -2 -1 0 1 2 08 10 12 14 16 18 20 Real house prices, growth rate -1.6 -1.2 -0.8 -0.4 0.0 08 10 12 14 16 18 20 Real domestic credit to households -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 08 10 12 14 16 18 20 Real credit to non-financial enterprises -.1 .0 .1 .2 .3 08 10 12 14 16 18 20 Banks problem loan share, total Deviation 1 Quarterly data, the numbers on the time axis refer to observations from 2008Q1 onwards. Most impulse responses for the variables in levels are displayed as deviations from the baseline in percent. Interest rates, the rate of unemployment and the problem loan share are displayed as deviations in percentage points from the baseline scenario. Moreover, growth rate responses are all displayed as deviations in percentage points from the baseline scenario.