Leverage and Deepening Business Cycle Skewness

We document that the United States and other G7 economies have been characterized by an increasingly negative business-cycle asymmetry over the last three decades. This finding can be explained by the concurrent increase in the financial leverage of households and firms. To support this view, we devise and estimate a dynamic general equilibrium model with collateralized borrowing and occasionally binding credit constraints. Improved access to credit increases the likelihood that financial constraints become nonbinding in the face of expansionary shocks, allowing agents to freely substitute inter-temporally. Contractionary shocks, however, are further amplified by drops in collateral values, since constraints remain binding. As a result, booms become progressively smoother and more prolonged than busts. Finally, in line with recent empirical evidence, financially driven expansions lead to deeper contractions, as compared with equally sized nonfinancial expansions. (JEL D14, E23, E32, E44)

We document that the U.S. economy has been characterized by an increasingly negative business cycle asymmetry over the last three decades. This finding can be explained by the concurrent increase in the financial leverage of households and firms. To support this view, we devise and estimate a dynamic general equilibrium model with collateralized borrowing and occasionally binding credit constraints. Higher leverage increases the likelihood that constraints become slack in the face of expansionary shocks, while contractionary shocks are further amplified due to binding constraints. As a result, booms become progressively smoother and more prolonged than busts. We are therefore able to reconcile a more negatively skewed business cycle with the Great Moderation in cyclical volatility. Finally, in line with recent empirical evidence, financially-driven expansions lead to deeper contractions, as compared with equally-sized non-financial expansions.

Previous articleNext article FreeCommentStephanie Schmitt-GrohéStephanie Schmitt-GrohéColumbia University and NBER Search for more articles by this author Columbia University and NBERPDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreBeaudry and Lucke’s paper “Letting Different Views about Business Cycles Compete” is a contribution to the empirical literature on the estimation of the sources of business cycles. It uses various five-variable vector error correction models (VECMs) to estimate the relative importance of anticipated total factor productivity (TFP) shocks, unanticipated TFP shocks, investment-specific technology shocks, preference shocks, and monetary policy shocks. An innovation relative to the large related literature on structural vector autoregression (SVAR)–based estimation of the sources of fluctuations is the focus on anticipated TFP shocks and on imposing cointegration relationships. Further, Beaudry and Lucke use time series on TFP, the relative price of investment, stock prices, federal funds rates, and a measure of aggregate activity in their estimation. This set of observables is slightly different than that used in the related literature. The main finding of their paper is that anticipated TFP shocks explain the majority of fluctuations in aggregate activity and stock prices at business cycle frequencies in the United States.Many authors have studied the question of what the sources of short-run fluctuations are. Yet this fundamental question in macroeconomics remains largely unresolved. Cochrane (1994), in a piece written for the Carnegie-Rochester Conference Series on Public Policy, starts his article on this topic as follows: “What shocks are responsible for economic fluctuations? Despite at least two hundred years in which economists have observed fluctuations in economic activity, we still are not sure” (295). Fifteen years later in business cycle research this statement is still a valid description of the state of the literature.Cochrane interpreted the findings of his (1994) study as suggesting that contemporaneous shocks to technology, money, credit, and oil cannot account for the majority of observed aggregate fluctuations.1 More recent SVAR-based papers using long-run restrictions such as Galí and Rabanal (2004) find, like Cochrane, a small role for permanent technology shocks in accounting for business cycle variations in hours and output. In table 2 of their paper Galí and Rabanal report that the share of variance due to technology shocks lies between 7% and 37% for output and between 5% and 36% for hours. Most important, under their favored interpretation, the technology shock accounts for less than 10% of the variance of output and hours. They therefore conclude that “nevertheless, it is safe to state that the bulk of the evidence reviewed in the present paper provides little support for the initial claims of the RBC literature on the central role of technological change as a source of business cycles” (Galí and Rabanal 2004, 228).On the other hand, there are papers presenting evidence that suggests that technology shocks are the major source of fluctuations, and the Beaudry and Lucke paper fits into this group. For example, the empirical paper of Fisher (2006), using SVAR methods, comes to the conclusion that neutral and investment-specific “technology shocks account for 73 percent of hours’ and 44 percent of output’s business cycle variation before 1982, and 38 percent and 80 percent afterward. The shocks also account for more than 40 percent of hours’ and 58 percent of output’s forecast errors over a three- to eight-year horizon in both samples. The majority of these effects are driven by the investment shocks” (413). Using Bayesian methods to estimate a dynamic stochastic general equilibrium model, Smets and Wouters (2007) find that at least 30% of the forecasting error variance of output is attributable to a combination of neutral and investment-specific technology shocks, with the majority of this share explained by neutral technology shocks. Justiniano, Primiceri, and Tambalotti (2008), like Smets and Wouters, using Bayesian estimation of a dynamic stochastic general equilibrium model, find an even larger share of fluctuations driven by technology shocks. Contrary to Smets and Wouters, however, Justiniano et al. find that most of the output variance is accounted for by the investment-specific technology shock rather than the neutral technology shock. Justiniano et al. attribute their finding of a larger role for the investment-specific shock to data differences, such as differences in the treatment of inventories and consumer durables. These differences in the definition of the data can increase the estimated share of the variance of output due to investment-specific technology shocks at business cycle frequencies from 18% to 53% for output and from 21% to 61% for hours.2The paper of Beaudry and Lucke is most closely related to Beaudry and Portier’s paper (2006). In that paper, Beaudry and Portier introduce a novel identification scheme to estimate (in the context of a VECM framework) anticipated TFP shocks. Most of the analysis is carried out for bivariate systems of TFP and stock prices. Under one identification scheme, the news TFP shock is that shock that does not affect TFP contemporaneously and under the other scheme the news TFP shocks is the one that has a long-run effect on TFP. Beaudry and Portier show that the correlation between the news TFP shock series identified by these two alternative schemes is very high and that impulse responses to them of measures of economic activity are quite similar. Therefore, Beaudry and Portier conclude that the common component of these two shocks represents an anticipated TFP shock. Most important for the relation to the paper of Beaudry and Lucke is the fact that Beaudry and Portier show that the so identified news TFP shock explains more than 50% of the forecast error variance of consumption, hours, investment, and output (measured as the sum of investment and consumption).1At the same time, Cochrane showed that VARs estimated using artificial data from a real business cycle (RBC) model driven by contemporaneous and news shocks to technology produce responses to consumption shocks that resemble the corresponding responses implied by VARs estimated on actual U.S. data. And thus his paper is often cited as one of the first to revive the idea of Pigou or news-driven business cycles.2One caveat to the results of Justiniano et al. is that their estimates imply a volatility for the relative price of investment, which they exclude from the set of observables, that is significantly larger than the observed standard deviation of this variable.I. Interpretation of Structural DisturbancesThe current paper by Beaudry and Lucke extends the work of Beaudry and Portier by moving away from bivariate SVAR systems to larger ones. Within a larger SVAR/VECM system the identification assumption of Beaudry and Portier must be modified. Specifically, Beaudry and Lucke estimate a VECM model of the form where the vector yt contains period t observations for TFP, the relative price of investment, stock prices, hours, and the federal funds rate, β denotes the cointegration vector, $$\Gamma ( L) $$ denotes a lag-polynomial, and $$\varepsilon _{t}=[ \varepsilon ^{1}_{t};\varepsilon ^{2}_{t};\varepsilon ^{3}_{t};\varepsilon ^{4}_{t};\varepsilon ^{5}_{t}] $$ denotes the vector of structural shocks that are the focus of interest. To identify the VECM, in particular the matrix B and the vector $$\varepsilon _{t}$$, Beaudry and Lucke impose the following identification restrictions. Identification restriction A1 says that only $$\varepsilon ^{1}_{t}$$ may have a contemporaneous effect on TFP. Therefore, $$\varepsilon ^{1}_{t}$$ is labeled the TFP shock. Implicitly it is therefore assumed that TFP is measured without error and that TFP is exogenous. Identification restriction A3 says that $$\varepsilon ^{5}_{t}$$ does not affect economic activity contemporaneously and is therefore interpreted as a monetary policy shock. Identification restriction A2 imposes that $$\varepsilon ^{4}_{t}$$ and $$\varepsilon ^{5}_{t}$$ have no long-run effect on TFP. Under identification scheme 1, denoted ID1, $$\varepsilon ^{3}_{t}$$, $$\varepsilon ^{4}_{t}$$, and $$\varepsilon ^{5}_{t}$$ are assumed to have no contemporaneous effect on the price of investment. Identification assumptions A1 and B1 then imply that $$\varepsilon ^{2}_{t}$$ must be the contemporaneous innovation to the relative price of investment. While in principle this identification scheme allows for $$\varepsilon ^{4}_{t}$$ to represent an anticipated temporary TFP shock or an anticipated temporary or permanent shock to the relative price of investment, the estimation results show that $$\varepsilon ^{4}_{t}$$ has very little effect on either TFP or the relative price of investment and thus it is unlikely that it represents a technology shock. Beaudry and Lucke therefore interpret it as a preference shock.Under identification scheme 2, ID2, assumption B1 is replaced by imposing that $$\varepsilon ^{2}_{t}$$ has no long-run effect on TFP. This would still allow for the possibility that $$\varepsilon ^{2}_{t}$$ is an anticipated temporary TFP shock. But the estimation assigns almost no role to $$\varepsilon ^{2}_{t}$$ in accounting for the variance of TFP. Identification scheme 2 leaves open the possibility that $$\varepsilon ^{2}_{t}$$, $$\varepsilon ^{3}_{t}$$, or $$\varepsilon ^{4}_{t}$$ affect the price of investment contemporaneously and thus could be called investment-specific shocks. Beaudry and Lucke, however, interpret $$\varepsilon ^{3}_{t}$$ as an anticipated TFP shock. The reason is that the identification assumption imposes that $$\varepsilon ^{3}_{t}$$ does not affect TFP on impact—thus it could not be an unanticipated TFP shock—and that the estimation yields that (at horizons not shown in figs. 3 or 5, namely 60 quarters) $$\varepsilon ^{3}_{t}$$ explains about three-fourths of the forecasting error variance of TFP under ID2. However, for horizons of 32 quarters or less (the time horizon shown in fig. 5) $$\varepsilon ^{3}_{t}$$ explains less than 20% of TFP and thus the interpretation as a TFP shock is less immediate. I want to entertain whether one could with equal plausibility interpret $$\varepsilon ^{3}_{t}$$ as an investment-specific shock. As shown in my figure 1, under identification scheme 2, $$\varepsilon ^{3}_{t}$$, explains between 40% and 60% of the forecasting error variance of the price of investment for forecasting horizons between 8 and 32 quarters. And $$\varepsilon ^{3}_{t}$$ explains less of the forecast error variance of TFP than that of the price of investment at any of these forecasting horizons. This might lead one to interpret $$\varepsilon ^{3}_{t}$$ as an investment-specific technology shock rather than, as maintained by Beaudry and Lucke, a TFP shock. The figure further shows that $$\varepsilon ^{3}_{t}$$ explains 60% of the forecasting error variance of hours and stock prices for forecasting horizons greater than 4 quarters. And thus one might be led to conclude that an investment-specific technology shock is the most important source of fluctuations in stock prices and real activity. This interpretation of $$\varepsilon ^{3}_{t}$$ would therefore be less at odds with the findings of Fisher (2006) on the importance of investment-specific technology shocks.Fig. 1. Share of forecast error variance due to ε3,t, VECM with ID 2View Large ImageDownload PowerPointII. Are Anticipated Shocks Identified in the Vector Error Correction Model?To be convinced by the interpretation given to the paper’s findings regarding the importance of anticipated shocks one needs to be convinced that the empirical strategy employed indeed is able to identify such shocks. Because it is not immediately obvious that this is the case, in what follows I will discuss some concerns one may have regarding the ability of SVAR/VECM methods to identify anticipated shocks.Beaudry and Lucke address the question of identification by presenting a theoretical model of the business cycle and checking whether the empirical identification strategy they employ, that is, a VECM analysis, would uncover the true structural shocks from data generated by this theoretical model. In particular, figure 4 (of the Web appendix of Beaudry and Lucke) shows the population forecast error variance decomposition (FEVD) of hours in the theoretical model with respect to the four structural shocks of the theoretical model: the unanticipated innovation to the growth rate of TFP, $$\varepsilon _{A,t}$$; the 8-quarter anticipated innovation to TFP, $$\varepsilon _{NA,t}$$; the unanticipated innovation to the growth rate of the price of investment, $$\varepsilon _{Z,t}$$; and the unanticipated innovation to the preference shock, $$\varepsilon _{\psi ,t}$$. (This is a real model and hence the fifth structural shock, which had the interpretation of a monetary policy shock, is dropped.) Then figures 4–6 of the Beaudry and Lucke Web appendix show the FEVD one would obtain were one to feed data generated by the calibrated real business cycle (RBC) model through the VECM machinery and impose the various identification schemes labeled ID1–ID3. In figure 2, I repeat this exercise for the case of identification scheme ID1. One difference between my figure 2 and Beaudry and Lucke’s figures is that I show the population FEVD implied by the calibrated theoretical model and the FEVD stemming from applying identification scheme ID1 to artificial model generated data in the same graph and for all four variables, that is, TFP, the price of investment, stock prices, and hours, whereas Beaudry and Lucke show this only for hours and in two different graphs.3 The purpose of this exercise is to check whether the VECM identified innovation $$\varepsilon ^{3}_{t}$$ does indeed explain the same share of variance in all four observables as the anticipated TFP shock, $$\varepsilon _{NA,t}$$, that it is meant to identify. A convincing case that the ID1 scheme, or any other of the identification schemes considered, is able to recover the true structural shocks is incomplete unless it does so for all four variables considered. After all, the fact that $$\varepsilon ^{3}_{t}$$ is interpreted as an anticipated TFP shock by Beaudry and Lucke is based on the finding that at very long forecasting horizons (60 quarters), longer than those shown in the graphs, it explains a large fraction (60%) of the forecasting error variance of TFP. (In the artificial economy, given the calibration of Beaudry and Lucke, the anticipated TFP shock explains 99% of the FEV of TFP for horizons greater than 8 quarters.) It follows that one needs to show that the ID1 scheme also picks up a similar share of the variance of these other variables as is true in population. The horizontal axis of each panel of figure 2 shows the forecasting horizon, which takes values between 1 and 32 quarters; the vertical axis measures the share of variance explained by the particular shock considered. The solid line corresponds to the FEVD implied by the structural vector error correction model (SVECM), and the dotted line corresponds to the population FEVD implied by the log-linearized approximation to the calibrated model. If the identification strategy were perfect, the solid line and the dotted line should be identical to each other. The figure shows that the SVECM delivers FEVDs that are very close to the population ones and hence suggests that the SVECM, with identification scheme ID1, is able to identify the contribution of all four structural shocks quite closely—as argued by Beaudry and Lucke. In particular, in the theoretical model most of the variance of hours of work, the measure of economic activity used by Beaudry and Lucke, at short horizons is due to the preference shock, $$\varepsilon _{\psi ,t}$$, and $$\varepsilon ^{4}_{t}$$ of the VECM reproduces this fact. Further, at longer forecasting horizons the most important source of fluctuations in hours are 8-quarter anticipated TFP shocks and the SVECM-identified innovation, $$\varepsilon ^{3}_{t}$$, is consistent with this feature of the theoretical model.Fig. 2. Forecast error variance decompositions in the baseline model: theoretical versus VECM estimates. Solid lines show the share of the forecasting error variance for horizons 1–32 quarters due to $$\varepsilon ^{i}_{t}$$, for i = 1, 2, 3, 4, which are the error terms identified with scheme ID1 by estimating a VECM on artificial time series simulated from the calibrated theoretical model. Dotted lines show the population forecasting error variance shares due to the true structural shocks εA,t, εN A,t, εZ,t, and εψ,t, respectively, and were computed from the log-linear approximation to the baseline model.View Large ImageDownload PowerPointI next consider a small variation in the model to see how well the SVECM methodology identifies the structural shocks in a slightly more complicated but empirically equally realistic environment. The only change I introduce is that the relative price of investment now is also subject to anticipated disturbances. And to keep it similar to the structure assumed by Beaudry and Lucke for anticipated TFP shocks, I will assume that the innovations to the investment price growth rate are also anticipated 8 quarters. Formally, this yields a process for the relative price of investment of the form where $$\varepsilon _{NZ,t-8}$$ denotes the 8-quarter anticipated innovation to the growth rate of investment. The innovation $$\varepsilon _{NZ,t-8}$$ enters the information set of private agents in period $$t-8$$ and thus will lead to changes in the endogenous variables included as observables, namely, the of hours and the growth rate of the stock in period but will only in an observed change in the price of investment 8 agents about I the structural of the model as only the standard of the shocks as follows: and _{\psi Under this calibration of the relative TFP is in equal due to and anticipated shocks, and the same is true for the relative price of investment. As we have in the stock prices to TFP shocks the assumed and hence stock prices will almost in equal be explained by and anticipated TFP I this calibration so that hours are in the long almost in equal driven by all shocks. It out that under this calibration in the short preference shocks are the most important source of As I artificial time series of the first and subject each of the data to the SVECM with the ID1 identification scheme and the that there are structural shocks and the VECM methodology only can identify it is less what the identification restrictions will Identification assumption A1 of Beaudry and Lucke imposes that $$\varepsilon ^{1}_{t}$$ is the only shock TFP suggesting that it identifies $$\varepsilon identification assumption 2, $$\varepsilon ^{4}_{t}$$ cannot have a long-run effect on TFP, thus only the possibility that it is $$\varepsilon $$\varepsilon or $$\varepsilon _{\psi ,t}$$, or a combination identification assumption B1 that $$\varepsilon ^{3}_{t}$$ $$\varepsilon ^{4}_{t}$$ have a contemporaneous effect on the price of investment. It follows that $$\varepsilon ^{2}_{t}$$ is to identify $$\varepsilon and $$\varepsilon ^{4}_{t}$$ cannot have a long-run effect on TFP, only $$\varepsilon ^{3}_{t}$$ has a of $$\varepsilon this leaves $$\varepsilon ^{4}_{t}$$ to identify either $$\varepsilon or $$\varepsilon _{\psi or some combination 3 the FEVD As in figure 2, each panel with a solid line the of the FEVD from applying the VECM methodology to the artificial data and with a dotted line the population FEVD implied by the theoretical model. The figure shows that in this economy, it is no longer the case that the structural identified by of the VECM methodology identify the structural shocks of the RBC model The VECM methodology delivers large in the FEVD of TFP, the price of investment, and in particular to news shocks. in this example, it that the share of variations in TFP explained by anticipated TFP shocks is estimated by the VECM methodology to be than the population most important, the figure shows that the of the contribution of news TFP shocks and news investment price shocks to economic activity, identified using the VECM is very different from the true or population The VECM methodology to that the share of anticipated investment-specific shocks in the FEV of the relative price of investment is The VECM assigns equal importance to the anticipated TFP shock and the anticipated investment-specific shock in the FEV of the relative price of investment. This case provides an of a in which the VECM methodology to identify the importance of sources of business Forecast error variance decompositions in model with anticipated investment-specific theoretical versus VECM estimates. Solid lines show the share of the forecasting error variance for horizons 1–32 quarters due to $$\varepsilon ^{i}_{t}$$, for i = 1, 2, 3, 4, which are the error terms identified with scheme ID1 by estimating a VECM on artificial time series simulated from the calibrated theoretical model. Dotted lines show the population forecasting error variance shares due to the true structural shocks εA,t, εN A,t, εZ,t, and εψ,t, respectively, and were computed from the log-linear approximation to the theoretical model with anticipated investment-specific Large ImageDownload FEVD from the VECM shown in my fig. 2 is the of FEVDs on simulated data with observations The simulated time series are and I the first I the calibration of Beaudry and Lucke by = β = = = = = = = = = and _{\psi I measure the stock price as the of the the of the be denoted by output by by and the of by stock prices can be as Identification and than VECM and true FEVD one could check for identification by whether the theoretical model with anticipated shocks to a in the the baseline model without anticipated investment-specific shocks shown in figure 2. that even figure 2 contains differences between the true population variance decompositions and those implied by the SVECM This could be due to or due to the fact that the particular theoretical model to have a of the assumed in the VECM analysis and given in and of the of the Beaudry and Lucke In particular, yt the vector of observables, that is, the of TFP, the of the relative price of investment, the of hours, and the of the stock the VECM analysis is the assumption that the vector yt has a A log-linear approximation to the of the theoretical model takes the form where is a 4 matrix a of the variables, denoted to the vector of state variables, denoted which in of and variables and has The state vector over time to \varepsilon where is an matrix and an 4 The 4 1 vector contains the four structural shocks. In the $$\varepsilon _{t}=[ \varepsilon _{\psi A over a denotes from the would be to the vector of observables of the growth rate of TFP, the growth rate of the price of investment, the growth rate of the stock and the of the of hours, that is, first a strategy to whether there a for in which the errors are indeed One can this question by applying the methods for example, et al. But the to this question should be for in the theoretical model there is a between the of the stock the price of investment, and TFP. Therefore, the differences of these variables, that is, should not have a This is the reason all Beaudry and Lucke a VECM rather than a model in consider the following vector of of four observables, where denotes the in which is given by Then is and is equal to where is the of In this case, we were able to show that has a then we would conclude that the of the observables also have a and thus we would have shown that indeed estimates from a VECM model should be able to recover the true structural shocks, et al. a model with this structure is that is, has a of the form L) only all the of the matrix ( are less than one in I a check of this for the model under and find that the is In particular, I find that more than of this matrix are greater than thus that the model to have a But in the of it is to interpret the of the VECM as the true shocks the model that the fact that we have four observables and four structural shocks and further that the matrix is reason for the of could be the large of state variables that an 8-quarter anticipated innovation is considered. If this were the case, this would support the that VECM methods are not well to identify news shocks. I this by the anticipated innovation to TFP by Then the model is driven by shocks only and we have $$\varepsilon _{t}=[ \varepsilon _{\psi To have any of the model a in we to thus consider only I hours from the vector of observables and set For this I the and I first check whether $$ is and find that it I then as before following et al. ( and I find that all are less than one in It follows that the model without news shock has a in and therefore the VECM methodology should be able to identify the true structural shocks. I stock prices from the vector of observables and set I can show that the theoretical model is that is, it has a in These results that at least in the it is the of news shocks that led to the of the I these findings as further evidence that VECM methods may not be well to the identification of news shocks even in where they a valid identification of unanticipated could also the preference shock, $$\varepsilon _{\psi ,t}$$, and $$\varepsilon _{t}=[ \varepsilon one can show that the theoretical model to have a for the case that the observables are $$ as well as for the case that for the of Anticipated the to the identification of news shocks by of SVAR/VECM methods that I have some recent authors have alternative empirical to estimate the importance of anticipated shocks as a source of business cycles. and (2008), for example, that methods a to the estimation of the importance of anticipated shocks. methods the that the empirical literature on the importance of news shocks has for it does not the dynamic stochastic general equilibrium model to have a in the is, it can be even methods allow to estimate what of anticipated shock is important we the VECM could not TFP and anticipated investment-specific and they allow to estimate how quarters in the main of business are In the VECM all we have is the between an innovation that the fundamental contemporaneously unanticipated and innovations that are and that will affect the fundamental in the anticipated But the VECM methodology is by about the In and we a structural Bayesian estimation of the contribution of anticipated shocks to business in the United in the context of an RBC model, which is slightly more than that by Beaudry and Lucke. assume four real investment in consumption, and in and allow business to be driven by permanent and neutral productivity shocks, permanent investment-specific shocks, and shocks. of these is by four of structural unanticipated innovations and innovations anticipated 1, 2, and 3 quarters in find that anticipated shocks account for more than of aggregate 1 estimation uses U.S. data on hours, investment, consumption, and the relative price of investment for the period which is very similar to the period in Beaudry and Lucke. 1 shows that to of the population variance of hours is due to anticipated shocks. further show that the forecasting error variance of hours explained by anticipated shocks with the forecasting horizon from 20% at a forecasting horizon of 2 quarters to 60% at a forecasting horizon of 32 which is similar to the in Beaudry and Lucke. 2 the decomposition of forecasting error at horizon 32 quarters by Beaudry and Lucke and those by and are differences between the two the most important one that Beaudry and Lucke an VECM whereas and using Bayesian methods, a dynamic stochastic general equilibrium model. use U.S. data on the price of investment, consumption, investment, and hours. Beaudry and Lucke use in data on TFP, stock prices, and Further, Beaudry and Lucke allow one measure of aggregate activity to the estimated system at the In and information on investment, consumption, and hours is used at the same In and use data on 2 shows that these differences the estimated shares of forecast error explained by anticipated shocks are rather similar the two that anticipated technology shocks explain the majority of short-run fluctuations in U.S. time 1 Share of by Anticipated Shocks and (2008), table 2 Share of of Error to Anticipated and and decompositions for the labeled Beaudry and Lucke are based on VECM estimation and should the information in fig. of Beaudry and Lucke. decompositions for the labeled and are from table of and These authors report FEVD for growth rates, with the of hours, which is in and FEVD at the of the of the estimated structural and and in Carnegie-Rochester Conference Series on Public in and of in of and of in and Shocks and the Business U.S. NBER in Primiceri, and Shocks and Business in and in Business Columbia in and and in U.S. Business A Bayesian in Previous articleNext article by NBER by the of on this are from by the of the following articles this of of and for

The article presents a dynamic stochastic general equilibrium model (DSGE-model) for the Russian economy. The model describes the behavior of the following macroeconomic agents: households, real sector, banking sector, Central Bank, as well as the interactions between them and the world. Household modeling uses the external habit formation approach to account for the inertia of preferences. To model the real sector, we abandoned the most common approach which assumes that the decision on investments is made by the households as the owners of production factors. Instead, we took the firm-specific capital approach which assumes that the decision on investment is made by the firms themselves. The study also considers that in Russia, fixed assets are mostly invested from the firms' own funds. To account for the investment inertia in the fixed asset in a real sector model, the expenditures are transferred to the commissioning of new facilities, the Calvo model is applied to describe the price setting under the monopolistic competition. A banking sector which defines the loan and debt interest rates to the key Central Bank interest rate is chosen to be a link between the households and firms in the model. The Taylor equation is used to describe the monetary policy of the Bank of Russia under the inflation targeting, while an inertia factor is included into the equation with the uncovered interest parity for the budget rule which regulates the purchases (or sales) of the currency by the National Welfare Fund. The final linearized model is a system of 23 difference equations with rational expectations. Based on the proposed model, calculations were made and key macroeconomic indicators were forecasted for 2020–2021 on a quarterly basis for the Russian economy. The calculations account for the relevant recessionary factors: oil price fall, oil production cut in OPEC+ deals, quarantine measures aimed to prevent the spread of the corona virus infection, anti-recessionary measures of the RF Government. The findings show that the economic downturn in 2020 can be from 5 to 7% under COVID-19 pandemic. Growth in 2021 is estimated to be within 3–5%. The developed model can be used for scenario projecting for the Russian economy, upgrading the monetary policy of the Bank of Russia, and for developing applied quarterly projection models (QPM). The model could be further modified by including more elements: decomposing the household sector into the Ricardian and non-Ricardian ones, identifying the resources industries and industries in the real sector which manufacture the invested goods, including the key taxes and budget expenses into the model. One more promising area is to analyze the equilibrium of the interest rates when large firms could accumulate their own financial resources. This prerequisite decreases the demand for the bank loans from the real sector and, thus, leads to lower, including the negative, interest rates. The proposed approach enhances the quality of a DSGE model as a predictive tool for making the political and managerial decisions.

Using the tools of regional dynamic models for analyzing the economy of the constituent entities of the Russian Federation, in particular, for studying regional business cycles is currently an urgent task. It is determined by the need to develop system conceptions about the factors, conditions and prerequisites for the development of regions, about the features and trends of the dynamics of their sector and territorial structure. The purpose of the article is to develop a dynamic stochastic multi-sector model for analyzing the effects of regional economic policy. The scientific novelty of the research concerns the development and implementation of dynamic models with microeconomic justification to formalize the processes of regional development, the sustainability of regional policy and spatial development. Similar class of models, that forms the theoretical foundation of contemporary macro-economics, is currently used for the analysis of national economy mostly. Models of such class that describe the processes in the regional economy are practically absent. The original tools for the construction of a regional dynamic stochastic general equilibrium model, suggested by the authors, describe the structure of a real sector of the economy of Sverdlovsk region. Parameterization of the model was made on the empirical data basis about the economy of Sverdlovsk region for 2003–2016. The behaviour of the following economic operators has been considered in the model: households; firms operating in the real sector of economy, the regional and federal government, and the Central Bank. Fiscal multipliers for three sectors of the economy – tradable goods sector, non-tradable goods sector and resource sector have been calculated with impulse response functions. The analysis of fiscal multipliers has revealed that the shock of the effective tax rate on individual income and the sock of regional costs have the most significant effect on the output in the above considered sectors of economy among all the rest fiscal shocks. The use of the tools in the form of a historical decomposition of regional variables demonstrates the results of the impact of supply and demand shocks in a time perspective on the output in the three sectors of the regional economy. The results of temporal decomposition of the variations of the endogenous variables mentioned above suggest that the cyclic processes in the regional economy of Sverdlovsk region during the study period are largely due to factors of supply rather than demand. The research results may be used both for the analysis of the regional economic policy priorities and for the development of measures aimed at the decrease of possible crisis phenomena in the regional economy. The trend to the construction of multi-sector models of regional economy in the framework of general equilibrium approach with the macroeconomics justification and rational expectations of economic operators described in the article stresses the importance and prospects of further studies. In particular, to reflect the specifics of the regions, it is necessary to take into account the institutional factors of each region in the model. This issue is an interesting theme for further research in the field of modeling of regional social and economic systems. Keywords region, regional economic policy, dynamic stochastic model, tradable and non-tradable goods sector, resource sector, fiscal multipliers, demand shocks, supply shocks, impulse response functions, historical decomposition of variations of endogenous variables. Acknowledgements The article has been written according to the Plan of Research and Development of the Institute of Economics, the Ural Branch of the Russian Academy of Sciences for 2019–2021.

We develop a dynamic general equilibrium model of imperfect competition where a sunk cost of creating a new product regulates the type of entry that dominates in the economy: new products or more competition in existing industries. Considering the process of product innovation is irreversible, introduces hysteresis in the business cycle. Expansionary shocks may lead the economy to a new ‘prosperity plateau,’ but contractionary shocks only affect the market power of mature industries.

Both, the popular opinion and the academic literature share the belief that the ongoing globalization with deepening trade channels is at least partly responsible for the appearance of a common business cycle across countries. Accordingly, one can observe a large and still growing literature that is concerned with trade as a medium of international business cycle transmission in general (cf, e.g., Kose et al. (2003), AER, Imbs (2003), IMF WP, Ambler et al. (2002), EER and Backus et al. (1992), JPE). One strand of this literature is completely empirical. Kose et al. (2003), for example, investigate annual data of a sample consisting of 75 industrial and developing countries over the last four decades. They demonstrate that the strength of the trade linkage with the G7 countries increases the correlation of domestic macroeconomic variables with the respective world variables. Typically, the empirical literature on trade as a means of international shock transmission does not reveal the particular characteristics of the trade linkage that can be seen as the causes for the certain sign of the transmission channel. In this context, the word sign refers to the assessment of the transmission channel. If, e.g., a home country H suffers an inefficiency shock and its trading partner, a foreign country F, gains - however gains are measured -, the sign of the transmission channel is termed positive. If, on the other hand, country F looses, the sign of the transmission channel is termed negative. A second strand of the literature draws conclusions from dynamic general equilibrium trade models. These models mostly intend to solve the puzzle of the so-called quantity anomaly described by Backus et al. (1992). The anomaly refers to the size of the correlation of macroeconomic variables between countries: Standard one-good aggregated dynamic general equilibrium models predict a cross-country correlation of output that is smaller than the cross-country correlation in the technology shocks, the latter in turn being smaller than the cross-country correlation in consumption. Empirical investigation, however, results in a reversed ranking of cross-country correlations. Ambler et al. (2002) succeed in removing this discrepancy between theoretical models and reality. By constructing a multi-sectoral dynamic general equilibrium model, they exploit the features that may result from a shock-induced change in the production structure. Anyhow, Ambler et al. (2002) only demonstrate that a certain correlation of macroeconomic variables between countries may exist. All the above mentioned contributions, however, are completely silent on the exact process of transmission and its welfare consequences. This is the starting point of the present study. By analyzing a dynamic multi-sectoral general equilibrium model numerically, it intends to reveal the actual character and causes of a positive or a negative transmission channel between two countries that are linked via goods trade. The base case model is characterized by Heckscher-Ohlin assumptions. This base case model is then extended by incorporating the households' labor supply decisions into the model and assuming imperfect substitutability between home and foreign produced goods. Both observed countries are characterized by a 4-sector production structure, in which the output of two sectors is used as an intermediate input in the other two sectors that each produce one final good. The model parameters are calibrated according to input-output tables from the G7 countries and China as an example of an emerging country. This results in countries as well as technologies for producing both intermediate and final goods that are quite different with respect to relative factor endowments and factor intensities. Therefore, the model may reflect business cycle transmission in the context of North-South trade. The business cycle in the model is driven by negative technology shocks. These technology shocks may be national shocks, if they occur in one country only. The shocks are termed global, if they occur in both countries. The model is solved with help of the software GAMS.

This article discusses the evolution of dynamic macroeconomic models from calibrated Real Business Cycle models to estimated dynamic stochastic general equilibrium models. The purpose is to suggest the usefulness of these models as a tool for policy analysis, with a particular emphasis on aspects of monetary policy. (JEL classification: D58, E50) This article gives an overview of the literature that has led to the emergence of dynamic stochastic general equilibrium (DSGE) models. This approach to macroeconomic modelling has gained widespread support among researchers and has recently started to be taken seriously by policy-making institutions as a modelling framework which is useful for policy analysis and the conceptual support of decision making. Modern macroeconomics is the result of an intense, and at times passionate, scientific debate that has taken place over the last decades. In the early 1980s, a new approach to the business cycle analysis was introduced by Kydland and Prescott (1982). The main tenet of their approach was that a small model of a frictionless and perfectly competitive market economy, inhabited by utility-maximising rational agents which operate subject to budget constraints and technological restrictions, could replicate a number of stylised business cycle facts when hit by random productivity shocks. This so-called real business cycle (RBC) approach to macroeconomic modelling was early on criticised on various aspects. Nevertheless, as it is now widely acknowledged, the RBC agenda has made a lasting methodological contribution. Most of today’s DSGE models

Neftci (1984) proposed a nonparametric test procedure for determining whether business cycles are asymmetric in the sense that contractions are steeper than expansions. He found evidence of this type of asymmetry in postwar quarterly U.S. unemployment variables.' This comment identifies a probable error in Neftci's empirical work that reverses the significance of his results for the unemployment rate, and it suggests that this test may have low power and be sensitive to measurement error.

Labor market indicators such as unemployment and labor force participation show a significant amount of heterogeneity across demographic groups, which is often not incorporated in monetary policy analysis. This dissertation is composed of three essays that explore the effect of labor market heterogeneity on the design and conduct of monetary policy. The first chapter, Effect of Monetary Policy Shocks on Labor Market Outcomes, studies this question empirically by looking at dynamics of macroeconomic outcomes to a monetary policy shock. I construct a measure of monetary policy shock using narrative methods that represent the unanticipatory changes in policy. Impulse response of unemployment rates for high and low-skill workers show low-skill workers bear a greater burden of contractionary monetary policy shock. Their unemployment rates increase by almost four times that of the high-skill group. Even though we see differences in dynamic response of unemployment rates, the empirical analysis shows some puzzling results where effects of contractionary shock are expansionary in nature. Moreover, these results are plagued by the “recursiveness assumption” that the shock does not affect current output and prices, which is at odds with theoretical models in the New Keynesian literature. In the second chapter, Skill Heterogeneity in an Estimated DSGE Model, I use a structural model to better identify these shocks and study dynamic responses of outcomes to economic shocks. I build a dynamic stochastic general equilibrium model, which captures skill heterogeneity in the U.S. labor market. I use Bayesian estimation techniques with data on unemployment and wages to obtain distribution of key parameters of the model. Low-skilled workers have a higher elasticity of labor supply and labor demand, contributing to the flatness of the wage Phillips curve estimated using aggregate data. A contractionary monetary policy shock has immediate effects on output and prices, lowering both output and inflation. Moreover, it increases unemployment rates for both high and low-skill groups, the magnitude being larger for the latter group. The presence of labor market heterogeneity will have new implications for the design of monetary policy, that I study in the third chapter, Optimal Monetary Policy with Skill Heterogeneity. I design an optimal policy for the central bank where policymakers respond to the different inflation-unemployment trade-off between high and low-skill workers. The monetary authority must strike a balance between stabilization of inflation, GDP and outcomes of high and low-skill workers separately. This optimal policy can be implemented by a simple interest rate rule with unemployment rates for high and low-skill workers and this policy is welfare improving.

This paper examines international linkages amongst G7 economies, in terms of co-movements in output growth and fluctuations, in the frequency domain. The paper has identified patterns in international business cycle co-movements among the G7, offering a general outlook of international business cycle co-movements. Moreover, the paper details the lower frequency, higher frequency and middle range characteristics of international co-movements in output growth and fluctuations. The main findings of the study are that co-movements among G7 economies are considerably stronger at lower frequencies, and G7 economies have become more synchronized considerably in episode 2 since the new millennium. The results and findings show support for real business cycle theory being extended to an international arena, with long effect real shocks impacting economies across borders. The three euro economies become more integrated since the new millennium, against a background that all G7 economies have also become more synchronized.

Previous articleNext article FreeCommentPaul BeaudryPaul BeaudryUniversity of British Columbia and NBER Search for more articles by this author PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreI. IntroductionInspired by ideas presented in Pigou’s 1927 book on Industrial Fluctuations, Pigou cycles refer to economic fluctuations that are driven by changes in firms’ belief about the future profitability of current investment decisions. A recent literature has emerged exploring whether Pigou’s ideas may offer a reasonable explanation to business cycle episodes (revival inspired in part by the episode of the tech boom-bust of the 1990s). This literature has many challenges: theoretical, conceptual, and empirical. For example, what is the source of the change in beliefs, what are the transmission mechanisms, and are such forces empirically relevant?One of the immediate and less obvious challenges of this literature is to identify environments in which such changes in beliefs can actually cause business cycles, that is, positive comovement between investment, consumption, and employment. Although such a possibility sounds very intuitive, it is nontrivial to build fully specified (and reasonable) dynamic stochastic general equilibrium models in which changes in fundamentals that affect the future profitability of current investment actually generate business cycle phenomena. In a recent paper, Den Haan and Kaltenbrunner (2009) have shown that news about future productivity growth can generate business cycle phenomena in an environment in which jobs are subject to matching frictions. However, their results are somewhat fragile since they depended on, among others things, assuming a high degree of intertemporal elasticity of substitution in consumption. In the current paper, Den Haan and Lozej examine whether extending the analysis of the earlier paper to an international setting allows Pigou cycles to emerge for more reasonable parameter values. The main result of the paper is to show that an international setting helps a search and matching framework generate Pigou cycles for reasonable values of the intertemporal elasticity of substitution in consumption. However, Den Haan and Lozej also find that for Pigou cycles to arise in an international context, it is important for trade flows to affect relative prices of imported/exported goods.In these comments, I will begin by reviewing why it is difficult to produce Pigou cycles in a simple general equilibrium environment. I will then discuss why matching frictions helps generate Pigou cycles and how the introduction of open-economy elements affects the result. Finally, I will provide a general assessment of the success and remaining challenges of this line of research.II. Expectation-Driven Fluctuation in a Two-Period EnvironmentConsider a two-period version of a standard one-sector macro model in which output is given byThe representative agent has per-period preferences given byThe driving force is taken to be news regarding θ2, with θ1 and K1 given. The question is, how does this economy respond to news about θ2 when the news is known to individuals at time 1? The equilibrium conditions for this two-period economy areWhen agents get positive news about θ2, this increases demand for current consumption, increases the demand for investment, lowers labor supply, and leaves labor demand unchanged. Hence to equilibrate, market prices will need to adjust. In the new equilibrium, it is easy to verify that R will be higher and consumption and wages in period 2 will be higher. But instead of creating a boom in period 1, the news causes a recession with lower employment, output, and investment. The one quantity that increases is consumption. This result can be partially reversed if preferences are instead given bywith σ < 1.In this case, if σ is sufficiently small, then news can give rise to an expansion in output, but now consumption will reduce. The problem with this case is that it again does not look like a business cycle. This simple example illustrates the difficulty in producing Pigou cycles in a standard equilibrium environment.Before I discuss the effects of introducting search frictions, it is useful to review both why simply placing the above setup in an open-economy environment will not solve the problem and why introducing nominal frictions is not a clear solution either. Under the assumption of a small open economy, good news about future productivity will now lead to an increase in both consumption and investment, but it will create an even greater domestic recession since employment and production will decrease more significantly in response to the consumption effects on labor supply. Introducing nominal frictions can help to generate Pigou cycles, but this has drawbacks also. In particular, results will depend on the nature of monetary policy. If monetary policy is set optimally, then the economy generally behaves as a flexible price economy, and therefore the problems associated with producing Pigou cycles illustrated above remain.III. Adding Matching FrictionsHow does adding matching frictions help? Matching frictions help on two dimensions: they affect labor demand and they affect the wage determination process. First, consider the effect on labor demand. Matching frictions act like an adjustment cost to labor. Hence, if the arrival of news causes a boom in employment next period, then this creates a need to hire more workers today since workers cannot be immediately hired at zero cost tomorrow. Given that more employment today means more output, this allows both consumption and investment to increase in response to the news. Second, the matching friction breaks the close link between the wage and marginal value of leisure. For example, the real wage can be thought as being quasi-fixed in such a case as long as it is in the bargaining set. Den Haan and Lozej do not go as far as assuming a fixed wage, but they do exploit this weaker link, thereby causing wages to respond little to wealth effects of labor supply. Accordingly, with matching frictions, news of a future boom leads to increased employment today at quasi-fixed wages.So what is the problem with this case? If σ is sufficient low, we get business cycle properties. However, if σ is set at an empirically reasonable value, we get a fall in investment as consumption absorbs all the extra output and interest rates remain high. In this case we are getting consumption and employment to move together, which is generally a difficult comovement to get; however, the model is not causing business cycle–type fluctuations since investment is not increasing.Now we can see how adding an international dimension helps. The matching friction is getting employment and output to increase, whereas a pinned-down interest rate allows both consumption and investment to increase. This all looks goods for generating Pigou cycles. However, as recognized by Den Haan and Lozej, a new problem arises in such a setting: the trade balance becomes extremely volatile. Hence, Den Haan and Lozej address this problem by adding trading frictions to better match the observed volatility of the trade balance. They consider two cases: interest rates that respond to the trade balance and import/export prices that react to import/export volumes. In the first case, they find that the economy mimics very closely the closed-economy case. In the second case, they find that the endogenous response of the price of investment exports and consumption imports greatly helps the model generate Pigou cycles. For example, the increase in the cost of imported consumption in response to news reduces the response of aggregate consumption and thereby favors even more production. The less obvious force is with respect to the price of investment goods. If the price of imported investment goods did not change, the country experiencing a positive news shock would not want to invest during the anticipation phase, knowing it can wait for the realization of the news. Hence, it would export investment goods instead of using them domestically to build up the capital stock. But when this price falls as the country tries to export (because of the friction), this creates incentives to invest domestically, and this favors the emergence of Pigou cycles in the sense of having consumption, investment, and output respond positively to news.IV. Remaining ChallengesThis leads me to the question: Does this paper provide a credible mechanism for Pigou cycles, or at least is it on the right track? To answer this question one needs to keep in mind that Pigou cycles are quite difficult to generate in a reasonable equilibrium setting. Most current proposals involve questionable departures from the baseline macro model. So in comparison to the literature, this paper nicely highlights some attractive features of how matching frictions help generate Pigou cycles. These frictions generate an expansion through a shift in labor demand—due to an adjustment cost–type mechanism—along a rather unchanging wage (in the bargaining set). That is intuitive. Such a mechanism seems at least as plausible for explaining Pigou cycles as those found in the literature, and moreover, it is shown to be quantitatively quite strong.The mechanism is nevertheless still somewhat weak on other dimensions. In particular, the theoretical impulse responses reported in the paper show that most of the movement in output comes about when actual productivity is increased, not when the news arises. Why am I worried about this? For this we need to look at evidence on news shocks. Although the evidence related to “news-driven” business cycles is controversial, my work with Franck Portier (2006) and more recent work with Bernd Lucke (2010) give some indication regarding what a model of Pigou cycles needs to explain. In this work, we have been trying to identify fluctuations induced by news shocks using structured vector autoregression methodology. Our findings, using different identification schemes, suggest that news of future growth in total factor productivity (TFP) is preceded by a period of approximately 2 years in which the economy is expanding without any increase in TFP. Most of the expansion period arises well before the increase in TFP. When TFP finally starts increasing, this is not associated with a further boom. Throughout this process, consumption, investment, and hours worked are high. These patterns are quite hard to match, and I believe that the current model does not fit them very well. In this sense, I think further modeling work is needed to explain business cycles as a response to news. Nonetheless, I view the model of Den Haan and Lozej as potentially being the right path.ReferencesBeaudry, Paul, and Bernd Lucke. 2010. “Letting Different Views about Business Cycles Compete.” NBER Macroeconomics Annual 2009:413–56.First citation in articleGoogle ScholarBeaudry, Paul, and Franck Portier. 2006. “News, Stock Prices and Economic Fluctuations.” American Economic Review 96, no. 4:1293–1307.First citation in articleGoogle ScholarDen Haan, W. J., and G. Kaltenbrunner. 2009. “Anticipated Growth and Business Cycles in Matching Models.” Journal of Monetary Economics 56:309–27.First citation in articleGoogle ScholarPigou, A. C. 1927. Industrial Fluctuations. London: Macmillan.First citation in articleGoogle Scholar Previous articleNext article DetailsFiguresReferencesCited by Volume 7, Number 12011 Article DOIhttps://doi.org/10.1086/658318 Views: 86 © 2011 by the National Bureau of Economic Research. All rights reserved.PDF download Crossref reports no articles citing this article.

We present an estimated dynamic stochastic general equilibrium model of stock market bubbles and business cycles using Bayesian methods. Bubbles emerge through a positive feedback loop mechanism supported by self-fulfilling beliefs. We identify a sentiment shock that drives the movements of bubbles and is transmitted to the real economy through endogenous credit constraints. This shock explains most of the stock market fluctuations and sizable fractions of the variations in real quantities. It generates the comovement between stock prices and the real economy, and is the dominant force behind the internet bubbles and the Great Recession.

Gold returns: Do business cycle asymmetries matter? Evidence from an international country sample

In this paper we examine the relationships between business cycles in the G7 countries. We focus on whether recessionary periods in one country are independent of the timing of recessions in other countries in the G7, using three different methods for dating recessions. We find that the evidence is mixed on whether phases of the business cycle in North America and in European countries are independent, or whether there is a common phase structure in the business cycle across all the G7 economies. NBER dates suggest that business cycles are synchronised, while other methods for generating business cycle chronologies are more consistent with regional, rather than international cycles. We also find mixed evidence on whether the UK is synchronised with European countries, while Japan quite clearly has the cycle that is most independent of other G7 countries.