The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Authorized for public release by the FOMC Secretariat on 02/09/2018 Federal Reserve Bank of Chicago Subject: Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers From: Jeffrey R. Campbell Date: June 8, 2011 Jonas D.M. Fisher Alejandro Justiniano Overview In this memo, we describe the Chicago Fed’s estimated dynamic stochastic general equilibrium model. This framework yields a history of identified structural shocks, which we apply to illuminate recent macroeconomic developments. To aid the understanding of these results, we follow them with summaries of the model’s structure, the data and methodology employed for estimation, and the estimated model’s dynamic properties. In several respects the Chicago Fed DSGE model resembles many other New Keynesian frameworks. There is a single representative household that owns all firms and provides the economy’s labor. Production uses capital, differentiated labor inputs, and differentiated intermediate goods. The prices of all differentiated inputs are “sticky”, so standard forward-looking Phillips curves connect wage and price inflation with the marginal rate of substitution between consumption and leisure and marginal cost, respectively. Other frictions include investment adjustment costs and habit-based preferences. Monetary policy follows a Taylor rule with shocks. The model’s decomposition of the price of investment goods with respect to consumption into two components distinguishes it from other similar frameworks. We empirically identify the technological component with the investment-consumption relative deflator from the NIPA, and we tie the financial component to the High Yield-AAA corporate bond spread. Another distinguishing feature of the Chicago model is the use of multiple price indices. For this, alternative available indices of inflation are decomposed into a Federal Reserve Bank of Chicago/ June 8, 2011 / Page 1 of 22 1 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Table 1. Model Forecasts Q4 over Q4 Real GDP Federal Funds Rate PCE Core Consumption Investment Marginal Cost 2010 2.71 0.19 0.80 1.63 10.40 0.93 2011 3.22 0.11 1.46 2.52 7.63 ‐1.46 2012 4.40 1.24 0.68 2.68 11.25 ‐0.81 2013 4.24 2.73 0.34 2.56 11.27 0.15 2014 4.17 3.44 0.54 2.68 10.34 0.56 single model-based measure of consumption goods and services inflation and idiosyncratic (series specific) disturbances that allow for persistent deviations from this common component. Estimation uses a factor model with the common factor derived from the DSGE framework. Forecasting Methodology Constructing forecasts based on this model requires us to assign values to its many parameters. We do so using Bayesian methods to update an uninformative prior with data from 1987:Q1 through 2008:Q4. All of our forecasts condition on the parameters equaling their values at the resulting posterior’s mode. These parameter values together with the data yield a posterior distribution of the economy’s state in the final sample quarter. For the calculation of this initial state’s distribution, we add a sequence of forward guidance shocks that signal the future path of the Federal Funds rate. These shocks begin arriving in 2009:Q1 and continue to the present. We construct them so that model-based expectations of the policy rate equal actual market-based expectations for the first five quarters of each quarter’s forecast horizon. The forecasts begin with 2011:Q3 and extend through 2014:Q4. Our plug for 2011:Q2 GDP growth was set at 2.4 percent. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 2 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 2 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Figure 1. Quarterly Model Figure 1: Forecasts Forecasts starting 2011q3 Consumption Real GDP 5 4.5 3 4 3.5 2.5 3 2 2.5 2 1.5 1.5 2011 2012 2013 2014 2011 Federal Funds Rate 5 2012 2013 2014 Investment 10 4 5 3 2 0 1 −5 0 2011 2012 2013 2014 2011 PCE Core 2012 2013 2014 Marginal Cost 3 0 2.5 2 −1 1.5 −2 1 −3 0.5 −4 0 2011 2012 2013 2014 2011 2012 2013 Federal Reserve Bank of Chicago/ June 8, 2011 / Page 3 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 3 of 22 2014 Authorized for public release by the FOMC Secretariat on 02/09/2018 Table 1 presents data from 2010 and forecasts for the following four years. The first three rows correspond to three key macroeconomic observables, Real GDP growth (Q4-over-Q4), the Federal Funds Rate (Q4 average), and growth of the Core PCE deflator (Q4-over-Q4). The following rows report forecasts of Q4-over-Q4 growth for three model-defined aggregates of importance: Consumption of nondurable goods and non housing services, Investment in durable goods, residential housing, and business equipment and structures, and the marginal cost of production. Figure 1 complements this with quarter-by-quarter data and forecasts of these series. The plots’ dashed grey lines indicate the series’ long-run values. The economy’s long-run GDP growth rate – which we identify with potential growth – equals 2.6 percent. Our forecast for 2011 only exceeds this by 62 basis points, so an unfortunate series of negative shocks could easily send the economy below potential. To place our forecasts from 2012 onward into context, Figure 2 plots the level of GDP for the three recoveries in our sample. All have been normalized so that their values equal zero at the NBER trough, and we extend the current recovery with our forecast, given by the dashed line. It is well-known that the ongoing recovery has been tepid when one considers the depth of the most recent recession. The plot indicates that this comparison will become less favorable over time. Recall that we hard-wire the current values of monetary-policy news shocks to match current market expectations. Currently, these date the tightening of monetary policy at 2012:Q2. Thereafter, the forecast rate begins to rise as the conventional Taylor rule dynamics take over. Although the forecasted path for core PCE is nearly deflationary from 2012 through 2014, the Taylor rule sees expected output growth as strong enough to merit the removal of the extraordinary accommodation in place since 2008. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 4 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 4 of 22 5 of 22 -0.1 -10 -0.05 -5 Federal Reserve Bank of Chicago/ June 8, 2011 / Page 5 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 0 0.05 0.1 0.15 0.2 0.25 0.3 Log GDP (centered at trough) Figure 2. Comparison of Recent Recoveries 0 5 10 Quarters After Trough of Recession 15 1991 Q1 2001 Q4 2009 Q2 Forecast 20 Authorized for public release by the FOMC Secretariat on 02/09/2018 Authorized for public release by the FOMC Secretariat on 02/09/2018 Shock Decompositions Our analysis identifies the structural shocks responsible for past fluctuations. To summarize this information, we follow a suggestion of Charlie Evans: Fix an object to be forecast, such as Q4-over-Q4 real GDP growth. Then, pick a date in the past and forecast the object conditional on the information as of that date. This is not a real-time forecast, because it uses revised data and model parameters estimated with the complete sample. The model can be used to decompose the forecast error into structural shocks. (A detailed explanation of the forecast error decomposition procedure begins below on page 16) We repeatedly advance the forecast date, decompose the forecast error, and finally plot the results. In total, the model features nine structural shocks and four idiosyncratic disturbances without structural interpretations. For parsimony’s sake, we group the shocks according to a simple taxonomy. Demand These are the structural non-policy shocks that move output and inflation in the same direction. The model features two of them. One changes the households rate of time discount, and the other alters the rate at which consumption goods can be transformed into capital goods. We call these the Time Discount and Marginal Efficiency of Investment shocks. Supply Five shocks move real GDP and inflation in opposite directions on impact. These supply shocks directly change – Neutral Technology, – Investment-Specific/Capital-Embodied Technology, – Markups of Intermediate Goods Producers, – Markups of Labor Unions, and – Households’ Disutility from Labor Policy The model’s monetary policy follows a Taylor rule with interest-rate smoothing and an i.i.d. policy shock. Additionally, we have incorporated Federal Reserve Bank of Chicago/ June 8, 2011 / Page 6 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 6 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Table 2. The Model’s Decomposition of Business-Cycle Variance Real GDP Growth Federal Funds Rate Core PCE Inflation Demand 0.72 0.83 0.10 Supply 0.21 0.07 0.81 Monetary Policy 0.03 0.07 0.01 Residual 0.05 0.02 0.08 Note: For each variable, the table lists the fraction of variance at frequencies between 6 and 32 quarters attributable to shocks in the listed categories. forward guidance shocks since 2008:Q4. These are revealed to the model’s agents one to five quarters before they effect the federal funds rate, and they allow our forecasting exercise to match model-based expectations with information from futures markets. Residual We group other shocks that are usually of small importance into a residual category. These include a shock to the sum of government spending and net exports. Table 2 reports the fraction of business-cycle variance attributable to shocks in each category for three key variables, the growth rate of Real GDP, the Federal Funds Rate, and Core PCE Inflation. Three facts stand out here. First, demand shocks dominate business cycles. Supply shocks account for only 21 percent of the GDP growth rate’s total business-cycle variance, and the non-systematic part of monetary policy shocks makes only a minor contribution. The accounting for the Federal Funds Rate’s variance is similar. Perhaps this is unsurprising, because we classify the shock that directly moves households’ rate of time preference as “demand.” Nevertheless, supply shocks’ unimportance for the Federal Funds Rate contrasts sharply with their dominance of inflation fluctuations. The shock to intermediate goods’ firms optimal markup accounts for about half of supply shocks 81 percent contribution, and the Hicks-neutral technology shock accounts for another quarter of this. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 7 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 7 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 The Model’s Specification and Estimation Our empirical work uses eleven variables, measured from 1987:Q1 through the present: • Growth of nominal per capita GDP, • Growth of nominal per capita consumption, which sums Personal Consumption Expenditures on Nondurable Goods and Services; • Growth of nominal per capita investment, which sums Business Fixed Investment, Residential Investment, Personal Consumption Expenditures on Durable Goods, and Inventory Investment; • Per capita hours worked in Nonfarm Business, • Growth of nominal compensation per hour worked in Nonfarm Business, • Growth of the implicit deflator for GDP, • Growth of the implicit deflator for consumption, as defined above, • Growth of the implicit deflator for investment, as defined above, • Growth of the implicit deflator for core PCE, • The interest rate on Federal Funds, and • High Yield-AAA Corporate Bond Spread. We do not directly use data on either government spending or net exports. Their sum serves as a residual in the national income accounting identity. To construct series measured per capita, we used the civilian non-institutional population 16 years and older. To eliminate level shifts associated with the decennial census, we project that series onto a fourth-order polynomial in time. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 8 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 8 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Our model confronts these data within the arena of a standard linear state-space model. Given a vector of parameter values, θ, log-linearized equilibrium conditions yield a first-order autoregression for the vector of model state variables, ζt . ζt = F (θ)ζt−1 + εt εt ∼ N (0, Σ(θ)) Here, εt is a vector-valued innovation built from the model innovations described above. Many of its elements identically equal zero. Table 3 lists the elements of ζt . Habit puts lagged nondurable consumption into the list, and investment adjustment costs place lagged investment there. Rules for indexing prices and wages that cannot adjust freely require the state to include lags of inflation and technology growth. The list includes the lagged policy rate because it appears in the Taylor rule. Gather the date t values of the eleven observable variables into the vector yt . The model analogues to its elements can be calculated as linear functions of ζt and ζt−1 . We suppose that the data equal these model series plus a vector of “errors” vt . yt = G(θ)ζt + H(θ)ζt−1 + vt vt = Λ(ϕ)vt−1 + et et ∼ N (0, D(ϕ)) Here, the vector ϕ parameterizes the stochastic process for vt . In our application, the only non-zero elements of vt correspond to the observation equations for the two consumption-based measures of inflation, the GDP deflator, and the High Yield-AAA spread. The idiosyncratic disturbances in inflation fit the high-frequency fluctuations in prices and thereby allow the price markup shocks to fluctuate more persistently. These errors evolve independently of each other. In this sense, we follow Boivin and Giannoni (2006) by making the model errors “idiosyncratic”. The other notable feature of the observation equations concerns Federal Reserve Bank of Chicago/ June 8, 2011 / Page 9 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 9 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Table 3. Model State Variables Symbol Ct−1 It−1 p πt−1 Description Lagged Consumption Lagged Investment Lagged Price Inflation Kt At at at−1 Stock of Installed Capital Hicks-Neutral Technology Growth rate of At Lagged Growth Rate of At Zt zt zt−1 Investment-Specific Technology Growth rate of Zt Lagged Growth Rate of Zt φt bt λw,t Labor-Supply Shock Discount Rate Shock Employment Aggregator’s Elasticity of Substitution Intermediate Good Aggregator’s Elasticity of Substitution Marginal Efficiency of Investment Shock Government Spending Share Shock Lagged Nominal Interest Rate Monetary Policy Shock λp,t µt gt Rt−1 εR,t Disappears without Habit-based Preferences Investment Adjustment Costs Indexing “stuck” prices to lagged inflation Autoregressive growth of At Indexing “stuck” wages to lagged labor productivity growth Autoregressive growth of Zt Indexing “stuck” wages to lagged labor productivity growth Time-varying Wage Markups Time-varying Price Markups Interest-rate Smoothing Federal Reserve Bank of Chicago/ June 8, 2011 / Page 10 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 10 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 the GDP deflator. We model its growth as a share-weighted average of the model’s consumption and investment deflators. We denote the sample of all data observed with Y and the parameters governing data generation with Θ = (θ, ϕ). The prior density for Θ is Π(Θ), which we specify to be similar to that employed by Justiniano, Primiceri, and Tambalotti (2011). Given Θ and a prior distribution for ζ0 , we can use the model solution and the observation equations to calculate the conditional density of Y , F (Y |Θ). To form the prior density of ζ0 , we apply the Kalman filter. The actual estimation begins with 1987:Q1. Bayes rule then yields the posterior density up to a factor of proportionality. P (Θ|Y ) ∝ F (Y |Θ)Π(Θ) We calculate our forecasts with the model’s parameter values set to this posterior distribution’s mode. Three Key Equations This section summarizes the inferred parameters by reporting the estimates of three key log-linearized equations: the Taylor Rule, the Price Phillips Curve, and the Wage Phillips Curve. Taylor Rule 3 5 X 1X 1.77 s Rt = 0.84Rt−1 + (1 − 0.84) 1.63 πt−3 + (yt − yt−4 ) + mp + ξt−s , t 4 j=0 4 s=1 Besides the lagged interest rate, the variables appearing on the right-hand side are the four-quarter average of consumption inflation, the most recent four-quarter output growth rate, the current monetary policy shock (mp t ), and s the five previous quarters’ signals of the current monetary policy stance, ξt−s for s = 1, . . . , 4. (These signals play a prominent role in forecasting, but we do not yet use them during estimation.) Note that Federal Reserve Bank of Chicago/ June 8, 2011 / Page 11 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 11 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 • Holding the economy’s growth rate fixed, the long-run response of Rt to a permanent one-percent increase in inflation is 1.63 percent. Thus, the model satisfies the Taylor principle. • Since the four-quarter growth rate of output replaces the usual output gap in the rule, it is difficult to compare the estimated coefficient of 0.44 with the typical calibrated output response of 0.5. Price Phillips Curve p p + 0.019st + pt + 0.20πt−1 πtp = 0.80Et πt+1 Here, st represents intermediate goods producers’ common marginal cost. • The associated Calvo probablity of an individual firm not updating its price in a given quarter equals 0.86, which is well in line with other calibrations. • Producers unable to update their price with all current information are allowed to index their prices to a convex combination of last quarter’s p in the inflation rate with the steady-state inflation rate. This places πt−1 Phillips curve. The estimated weight on steady-state inflation is 0.76. Wage Phillips Curve The Wage Phillips curve can be written as p p w + jt−1 = βEt πt+1 + πt+1 + jt+1 − ιw (πtp + jt ) +κw xt +w πtw +πtp +jt −ιw πt−1 t , where πtw and πtp correspond to inflation in real wages and consumption prices respectively, jt = zt + α 1−α µt is the economy’s technologically determined stochastic trend growth rate, with α equal to capital’s share in the production function, zt the growth rate of neutral technology, and µt the growth rate of Federal Reserve Bank of Chicago/ June 8, 2011 / Page 12 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 12 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 p investment-specific technical change. The term πt−1 + zt−1 + jt arises from indexation of wages to a weighted average of last quarter’s productivity-adjusted good’s price inflation and its steady state value. The estimated weight on the steady state equals 0.64. The log-linearized expression for the ratio of the marginal disutility of labor, expressed in consumption units, to the real wage is xt = bt + ψt + νlt − λt − wt , where bt and ψt are disturbances to the discount factor and the disutility of working, respectively, lt hours,λt the marginal utility of consumption and wt the real wage. Finally, w t is a white noise wage markup shock. Note that without indexation of wages to trend productivity, this equation says that nominal wage inflation (adjusted by trend growth) depends positively on future nominal wage inflation (also appropriately trend-adjusted), and increases in the disutility of labor-real wage gap. The estimated equation is given by p p w +jt+1 −0.32 (πtp + jt )]+0.0055xt +w + jt−1 = 0.9988×Et [πt+1 +πt+1 πtw +πtp +jt −0.32 πt−1 t , The estimated Calvo probability of a wage remaining unadjusted in a given quarter underlying the estimate of κw = 0.0055 equals 0.73. The Model’s Shocks Our discussion of recent macroeconomic developments above featured the following shocks prominently: The discount rate shock, the “financial” shock to the marginal efficiency of investment, and both anticipated and unanticipated monetary policy shocks. In this section, we provide greater detail on the model’s responses to these four shocks. Figure 3 plots responses to a discount rate shock that increases impatience and tilts desired consumption profiles towards the present. The variables examined Federal Reserve Bank of Chicago/ June 8, 2011 / Page 13 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 13 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 are real GDP, the Federal Funds Rate, inflation, consumption, investment, and hours worked. The responses are scaled so that the change in GDP after 16 quarters equals one percent. In a neoclassical economy, this shock would be contractionary on impact. Upon becoming more impatient, the representative household would increase consumption and decrease hours worked. To the extent that the production technology is concave, interest rates and real wages would rise; and regardless of the production technology both real GDP and investment would drop. Increasing impatience instead expands activity in this New Keynesian economy. As in the neoclassical case, consumption rises on impact. Habit causes the consumption growth to persist for two more quarters. Adjustment costs penalize the sharp contraction and recovery of investment from the neoclassical model, so instead investment remains unchanged on impact. Market clearing requires either a rise of the interest rate (to choke off the desired consumption expansion) or an expansion of GDP. By construction, the Taylor rule prevents the interest rate from rising unless the shock is inflationary or expansionary. Therefore, GDP must rise. This in turn requires hours worked to increase. Two model features overcome the neoclassical desire for more leisure. First, some of the labor variants’ wages are sticky. For those, the household is obligated to supply whatever hours firms demand. Second, the additional labor demand raises the wages of labor variants with wage-setting opportunities. This rise in wages pushes marginal cost up and lies behind the short-run increase in inflation. After inflation has persisted for a few quarters, monetary policy tightens and real rates rise. Since the discount rate shock moves output and prices in the same direction, a Keynesian analysis would label it a shift in “demand.” In the neoclassical sense, it is also a demand shock, albeit a reduction in the demand for future goods. The matching neoclassical supply shock in our model is to the marginal efficiency of investment. A positive shock to it increases the supply of future goods. Figure 4 plots the responses to such a shock. Barro and King (1984) and Greenwood, Federal Reserve Bank of Chicago/ June 8, 2011 / Page 14 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 14 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Figure 3. Responses to a Discount Rate Shock GDP level 8 6 6 4 4 2 2 0 0 5 10 15 Federal Funds Rate 8 20 0 0 Core PCE inflation 3 8 2 6 1.5 4 1 2 0.5 0 0 0 5 10 15 20 15 20 0 5 10 15 20 Hours level Investment level 2 −2 10 Consumption level 10 2.5 5 8 0 6 −2 4 −4 2 −6 −8 0 5 10 15 20 0 0 5 10 15 20 Note: The impulse was scaled to yield a one percent increase in GDP after 16 quarters. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 15 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 15 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Hercowitz, and Huffman (1988) consider the analogous responses from a neoclassical model. Increasing the marginal efficiency of investment raises investment, hours worked, GDP, and the real interest rate but decreases consumption. Two aspects of our model stop consumption from falling on the same shock’s impact. First, habit-based preferences penalize an immediate decrease in consumption. Second, monetary policy responds to the shock only slowly, so real interest rates actually fall slightly on impact. The investment adjustment costs give the responses of GDP, hours, and investment their hump shape. Although this shock changes the economy’s technology for intertemporal substitution – and therefore deserves the neoclassical label “supply” – it makes prices and output move in the same direction. For this reason, it falls into our Keynesian taxonomy’s “demand” category. As noted above, monetary policy shocks partially offset negative demand shocks in 2010. Figures 6 and 5 present the impulse response functions for two of these, the unanticipated “contemporaneous” shock and the shock revealed to all agents five quarters in advance. The responses to the unanticipated shock are standard, but those following an anticipated shock require more explanation. At the announcement date, the expected value of the policy rate five quarters hence increases by 300 basis points. Because both Phillips curves are forward looking, this expected contraction causes both prices and quantities to fall. This anticipated weakness then feeds through the Taylor rule to create a gradual easing of policy. When the anticipated tightening arrives, it mostly offsets the prior endogenous easing. The policy rate rises only 85 basis points above its pre-shock value. These responses are clearly worthy of further study. Shock Decomposition Methodology We credit Charles Evans with the original ideas behind this decomposition. For the shock decomposition, we set the model’s parameters to their values at the ˆ Using all available data we use the Kalman posterior distribution’s mode, θ. smoother to extract sequences of estimated states {ζˆt }Tt=1 and a innovations Federal Reserve Bank of Chicago/ June 8, 2011 / Page 16 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 16 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Figure 4. Responses to a Marginal Efficiency of Investment Shock GDP level 2 Federal Funds Rate 2 1.5 1.5 1 1 0.5 0.5 0 0 0 5 10 15 20 −0.5 0 5 Core PCE inflation 0.5 10 15 20 Consumption level 1 0.4 0.5 0.3 0 0.2 0.1 0 5 10 15 20 0 5 10 15 20 Hours level Investment level 15 −0.5 2 1.5 10 1 5 0.5 0 −5 0 0 5 10 15 20 −0.5 0 5 10 15 20 Note: The impulse was scaled to yield a one percent increase in GDP after 16 quarters. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 17 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 17 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Figure 5. Responses to an Unanticipated Monetary Policy Shock GDP level 2.5 Federal Funds Rate 2 2 0 1.5 −2 1 −4 0.5 0 0 5 10 15 20 Core PCE inflation 1.5 −6 0 5 10 15 20 Consumption level 1.5 1 1 0.5 0.5 0 0 0 5 10 15 20 −0.5 5 Investment level 8 10 15 20 Hours level 2.5 6 2 4 1.5 2 1 0 −2 0 0.5 0 5 10 15 0 20 0 5 10 15 20 Note: The impulse was scaled to yield a one percent decrease in GDP 16 quarters after that date. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 18 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 18 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Figure 6. Responses to an Anticipated Monetary Policy Shock GDP level 0 −0.5 0 −1 −1 −1.5 −2 −2 0 5 10 15 20 −3 0 10 15 20 Consumption level 0.5 0.5 0 −1 −0.5 0 5 10 15 20 −1 −0.5 −2 −1 −4 −1.5 5 10 15 5 10 15 20 0 0 0 0 Hours level Investment level 2 −6 5 Core PCE inflation 0 1.5 Federal Funds Rate 1 20 −2 0 5 10 15 20 Note: The monetary policy shock is revealed to all agents five quarters before its realization, and the impulse was scaled to yield a one percent decrease in GDP 16 quarters after that date. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 19 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 19 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 {ˆ εt }Tt=1 . By construction, these satisfy the estimated transition equation for the state. ˆ ζˆt−1 + εˆt , ζˆt = F (θ) To keep this discussion simple, we henceforth suppose that the “error” shocks in vt equal zero. Incorporating them into the analysis changes the actual calculations only little. For concreteness, suppose that the forecasted object of interest is Q4-over-Q4 GDP growth for 2010. We position ourselves in 2009:Q4 and calculate 2009:Q4 ˆ ζˆ2009:Q4 ζˆ2010:Q1 ≡ F (θ) 2009:Q4 ˆ ζˆ2009:Q4 ζˆ2010:Q2 ≡ F (θ) 2010:Q1 ˆ ζˆ2009:Q4 = F 2 (θ) .. . 2009:Q4 ˆ ζˆ2009:Q4 ζˆ2010:Q4 ≡ F (θ) 2010:Q3 These are the “expectations” of the model’s states in each quarter of 2010 conditional on the state at the end of 2009 equalling its estimated value. With these “state forecasts” in hand, we can construct corresponding forecast errors by comparing them with their “realized values” from the Kalman smoother. For the period t state forecasted in 2009:Q4, we denote these with ηˆt2009:Q4 = ζˆt − ζˆt2009:Q4 . These forecast errors are related to the structural shocks by ηˆt2009:Q4 = t−2009:Q4 X ˆ ε2009:Q4+j . F j−1 (θ)ˆ j=1 ˆ 2009:Q4 for The shock decomposition is based on four alternative forecasts, ζ(ι) t t = 2010:Q1, . . . , 2010:Q4 and ι ∈ {D, S, M, R}. Here, ι indexes one of the four groups of structural shocks. For these, let εˆ(ι)t denote a version of εˆt with all Federal Reserve Bank of Chicago/ June 8, 2011 / Page 20 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 20 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 shocks except those in group ι set to zero. With these, we construct ˆ 2009:Q4 ≡ F (θ) ˆ ζˆ2009:Q4 + εˆ(ι)2010:Q1 , ζ(ι) 2010:Q1 .. . 2009:Q4 ˆ ζˆ2009:Q4 + εˆ(ι)2010:Q4 , ζˆ2010:Q4 ≡ F (θ) 2010:Q3 and ˆ 2009:Q4 . ηˆ(ι)2009:Q4 ≡ ζˆt − ζ(ι) t t By construction, ηˆt2009:Q4 = X ηˆ(ι)2009:Q4 . t ι∈{D,S,M,R} That is, each forecast error can be written as the sum of contributions from each of the shock groups. Using the observation equations, we transform these into components of the forecast error for observable variables. With this completed, we can then move the forecast date forward to 2010:Q1. The decomposition for that date proceeds similarly, except that we treat growth in 2010:Q1 as data. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 21 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 21 of 22 Authorized for public release by the FOMC Secretariat on 02/09/2018 Bibliography Barro, R. J. and R. G. King (1984). Time-separable preferences and intertemporal-substitution models of business cycles. The Quarterly Journal of Economics 99(4), pp. 817–839. Boivin, J. and M. Giannoni (2006). DSGE models in a data-rich environment. Working Paper 12772, National Bureau of Economic Research. Greenwood, J., Z. Hercowitz, and G. W. Huffman (1988). Investment, capacity utilization, and the real business cycle. The American Economic Review 78(3), pp. 402–417. Justiniano, A., G. E. Primiceri, and A. Tambalotti (2011). Investment shocks and the relative price of investment. Review of Economic Dynamics 14(1), 101–121. Federal Reserve Bank of Chicago/ June 8, 2011 / Page 22 of 22 Summary of Chicago Fed DSGE Model for Federal Reserve System Researchers 22 of 22