RICERCA ITALIANA     

Financial, credit and labour markets in business cycle models for policy evaluation. Theory and empirics.

Abstract

Quantitative DSGE models rooted in microeconomic foundations have become the workhorse tool for the analysis of monetary policy in the current macroeconomic literature (Gali and Gertler (2007)).

At the theoretical level, the baseline DSGE framework, however, suffers of a number of shortcomings:
i) It is incapable of generating fluctuations in unemployment (all variations in the labor input take place only at the intensive margin, i.e. in hours per worker).
ii) It usually assumes perfect capital markets, and therefore cannot provide any assessment of the likely role of financial factors in shaping households’ consumption and saving decisions. As a consequence, the role of financial intermediation is very little studied in this specification.
iii) It is unable to match some important features of asset price dynamics, in particular stock price dynamics.
iv) It studies monetary and fiscal policies assuming either the existence of a benevolent social planner or of an exogenously given fiscal policy, and it therefore neglects both the possibility of having separate monetary and fiscal decisions and the one of endogenizing the conduct of fiscal policy.
v) It neglects the existence of uncertainty concerning model specification and parameters and the possibility of learning by agents.
vi) It is linear or linearized around the steady state.

At the empirical level the effort on model evaluation has been limited to the consideration of very simple alternative such as VAR models, while more sophisticated alternative specification such as FAVAR (Factor Augmented VAR models) or auxiliary forecasting models have been very rarely considered as alterntive benchmarks for evaluation.

The objective of our research programme is to extend the baseline DSGE framework to explore the importance of the omitted theoretical dimensions and to provide a thorough empirical evaluation. <<<

Principal Investigator

Carlo Ambrogio Favero Università Commerciale "Luigi Bocconi" MILANO

Research Objectives

The objective of our research programme is to extend the baseline DSGE framework to explore the importance of the omitted theoretical dimensions and to provide a thorough empirical evaluation for a better use of DSGE models for (monetary and fiscal) policy evaluation.
The final objective of the project is to deliver inputs and research results useful to the construction of models of the business cycle for policy simulation capable of handling the main puzzles and the unresolved issues for the standard DSGE models. <<<

First Results

The explicit inclusion in DSGE model of labour and credit markets frictions, the intractions between monetary and fiscal policy, non-linearities and learning should enhance the potential of this type of models for policy simulation analysis.
Furthermore, better model evaluation practice should allow us to assess more precisely the degree of confidence to be put in the results of simulation of different policies. <<<

Timescale

24 months

National and international background

DSGE models and labor market imperfections

In recent years, monetary business cycle models with monopolistic competition and staggered price setting have been widely used to study the implications of alternative specifications of monetary policy (see Woodford, 2003, and Gali, 2007, for textbook treatments). One shortcoming of these models, however, is that they do not include a very detailed description of the labor market, and are therefore not suited to discuss the relationship between monetary policy and unemployment, a relationship that lies at the heart of practical monetary policy discussions. In the labor market literature, on the other hand, search and matching models with equilibrium unemployment have been fairly successful in explaining aggregate labor market fluctuations (see, for instance, Mortensen and Pissarides, 1994, Pissarides, 2000, or Shimer, 2005). Such labor market specifications have recently been extended to monetary business cycle models by Trigari (2004, 2006) and Walsh (2005) and thus present a natural alternative to the standard monetary framework.

Using traditional business cycle models with staggered price and wage setting, but without search and matching frictions on the labor market, Christiano, Eichenbaum, and Evans (2005) and Smets and Wouters (2003, 2006) have demonstrated that nominal wage rigidities are a crucial ingredient when explaining business cycle fluctuations in the U.S. and in Europe. Similarly, Levin, Onatski, Williams, and Williams (2005) have shown that wage rigidities account for the main welfare cost of business cycle fluctuations. Although these results are obtained in models without equilibrium unemployment, they suggest that the specification of the labor market has important consequences for monetary policy.


DSGE models and credit market imperfections

The baseline DSGE model employed for the quantitative analysis of monetary policy analysis typically assumes perfect financial markets. This amounts to assuming that (in equilibrium) the structure of financial markets is immaterial for the consumption and investment decisions taken by households.
In many instances, however, it seems implausible that financial factors play such a minimal role in the economic decision-making, especially as regards the role of housing wealth on economic activity. The role of the latter has recently attracted considerable attention among academic researchers, policy-makers and press commentators. This attention is partly explained by the sizeable rises in property prices and household indebtedness in several industrialized countries over recent years and the need to understand both the determinants of such rises and their potential implications for monetary policy and financial stability. Beyond these policy considerations, there is growing interest in the effects of changes in property prices on consumption decisions, given the predominance of housing in total household wealth

The existing literature on the role of financial market imperfections in DSGE models is scant. It originates from the seminal work of Bernanke and Gertler (1989), who emphasize the role of collateral requirements in affecting aggregate fluctuations. In their work, collateral constraints are motivated by the presence of private information and limited liability. More recently, Kiyotaki and Moore (1997) build a general equilibrium model in which two categories of agents (borrowers and savers) trade private debt. In KM, collateral requirements are motivated by the presence of limited enforcement. Both BG and KM, despite some differences, share the central implication that the wealth of the borrower influences private spending. Campbell and Hercowitz (2005) analyze the implications for macroeconomic volatility of the relaxation of collateral requirements in the U.S. (dated around 1980) in a general equilibrium environment. However, their real business cycle framework is not suitable for a study of monetary policy, and it abstracts from any role of asset prices. Iacoviello (2005) and Aoki, Proudman and Vlieghe (2004) are recent exceptions that extend the work of KM to build a bridge with the recent New Keynesian monetary policy framework. Crucially, their work does not feature any analysis on the role of the institutional characterisctics of mortgage markets, nor any exploration of the empirical fit of the employed model.

Learning, asset prices and monetary policy design

The standard NK model seems ill-equipped to account for the stock price volatility observed in the data and, in general, for how stock prices may be related to real macroeconomic instability.
Bernanke et al. (1999) and Bernanke and Gertler (1999) provided this missing channel by enriching the standard NK-DSGE model with an informational friction in the credit market: in the face of aggregate shocks, a “financial accelerator” amplifies the macroeconomic volatility generated in a frictionless environment. As also Girlchrist and Leahy (2002) and Cecchetti et al. (2002) argue, a central bank should try to offset these fluctuations (and bring the economy back to the frictionless equilibrium) only if stock price swings were driven by non-fundamentals (e.g. fads, bubbles) and the bank knew their underlying stochastic process. As pointed out by Cogley (1999), a central bank can hardly cannot possess this information; therefore the optimal monetary policy is likely to correspond to a regime of flexible inflation targeting (Bernanke and Gertler (1999)).

However, optimality is not the only metric to judge policies. Some authors have recently pointed out that the existence of a stock price channel might sensibly alter the properties of standard monetary policy rules as to the determinacy of equilibrium under Rational Expectations (RE). Bullard and Schaling (2002) and Carlstrom and Fuerst (2007) have shown that a systematic response to stock price fluctuations in a NK model might induce non-fundamental fluctuations driven by self-fulfilling revisions of expectations (sunspot equilibria). Airaudo et al. (2007) confirm these results extending the equilibrium determinacy analysis to an economy with real demand-side wealth effects from stock price changes. Furthermore, following the line of Evans and Honkapoja (2001), they depart from RE and assess whether the fundamental RE equilibrium can be eventually attained if agents form expectations via recursive least squares (E-stability of REE).

The interaction between monetary and fiscal policy

Conventional wisdom holds that fiscal policy may influence price stability; notably, excessive deficits and increasing public debt may be harmful for price stability (Aiyagari and Gertler, 1985; Sargent and Wallace, 1981).
Recent New Keynesian dynamic stochastic general equilibrium (DSGE) models study monetary and fiscal policies assuming either the existence of a benevolent social planner or fiscal policy as exogenously given (Benigno and Woodford 2003; Woodford, 2003). To the best of our knowledge this strand of literature has neglected the possibility of endogenizing the conduct of fiscal authorities and to separate fiscal from monetary policy decisions. Governments, however, typically have short planning horizon and parties alternate in office as a result of elections. The implications for fiscal policy of alternating parties and on the strategic use of public debt as a mean to redistribute the burden of fiscal adjustment to future governments have been extensively studied in the literature (see e.g. Alesina and Tabellini 1990; Beetsma, and Bovenberg, 1997a, 1997b; Tabellini; 1986).


Non-linearities in DSGE models
There are many important sources of non-linearity that have been neglected by traditional DSGE models.
One issue that has recently received renewed attention is the analysis of the effects of a disinflationary policy. Namely, the study of monetary policies aimed at lowering the level of trend inflation or the inflation target. Interestingly,while many empirical works find evidence that disinflation policies are costly in terms of foregone output, there is a consensus in the literature that in the basic New Keynesian model a credible disinflation can be achieved without any output costs. This is due to the fact that at the core of the model lies the Calvo technology of price staggering that implies price level stickiness, but not inflation stickiness. Hence inflation can adjust immediately to a step jump in the rate of growth of money (Ball, 1994 and Mankiw, 1999). Then, researchers started enlarging the model with features that enhance inflation persistence, as a necessary condition for delivering output costs of a disinflation (Blanchard and Galí, 2006), such as the assumption of backward-looking indexation (Christiano et al., 2005).
Mankiw and Reis (2002) argue for a different price setting mechanism. Ball (1995) and more recently, Erceg and Levin (2003), Nicolae and Nolan (2006) introduce lack of credibility. Other papers in the literature instead show that non-linearities are extremely important in assessing the effects of a disinflationary policy (Ascari, 2004, Yun, 2005, Ascari and Merkl, 2007). Non-linear simulations show that the interplay between long-run effects and short-run dynamics is crucial in the adjustment path after a disinflation. It may then be very misleading to use a log-linearization of the model around a steady state to judge the effects of a policy that implies a movement to another steady state.

Several considerations confirm the relevance of nonlinear DSGE model approximations. First, they are well suited to model dynamics in presence of large deviations from the steady state.
Secod, nonlinear DSGE models provide sharper estimates than their linearised counterparts. Fernandez-Villaverde et al.(2006) highlight a more general empirical advantage of the estimation of nonlinear models. Amisano and Tristani (2007a) show that second order approximations might help to reduce the extent of under-identification typical of DSGE models (Canova and Sala, 2005).
Third, in a nonlinear DSGE model filtering is done by simulation (Doucet et al., 2001 and Arulampalam et al., 2002). Amisano and Tristani (2007b) investigate simulation-based filtering in a DSGE framework. Many issues remain to be explored to find computationally efficient ways to sequentially estimate nonlinear DSGE models to assess their structural stability (Fernandez and Rubio , 2007b) or to assess their forecasting performance in real time. Fourth, non-linearities are important to solve DGSE models in which innovations are not IID Gaussians. Methods based on perturbation theory assume that the innovations are IID. Many studies, however, have argued for the presence of heteroskedasticity in macroeconomic data (Cogley and Sargent, 2005, Primiceri, 2005 and Fernandez and Rubio, 2007a).
One difficulty with regime-switching within DSGE models is whether these models can be solved using perturbation methods. Davig et al.(2004) argue that linearisations produce inaccurate results in regime-switching models. Models have been solved either exactly or using global solution methods. Exact solutions are available for very simple models, and global approximation methods are too slow for estimation. Amisano and Tristani (2007c) show that perturbation methods can be applied to the regime switching case, and illustrate how to construct a second-order approximation in this case.

Model Evaluation

Econometric model evaluation in the Bayesian camp has recently witnessed some progress with special reference to DSGE models (see An and Schorfeide(2007), Canova and Sala(2007) and Consolo et al.(2007)). In virtually all cases the DSGE model evalution is conducted by using a VAR as the benchmark. Some early effort has been put in considering more sophisticated empirical alternatives to VARs, such as Factor Augmented VARs or auxiliary forecasting models. FAVARs are an important addition to macroeconometrics, allowing to easily handle the estimation of large-scale dynamic econometric models in the convenient VAR framework. Recent contributions due to, among others, Bernanke and Boivin (2003), Bernanke et al. (2005), Favero et al (2005), and Stock and Watson (2005a), share the intuition that the information contained in very large data sets of economic variables can be summarized by a small number of factors, that can then be employed to augment VAR systems, in order to allow for a better description of the economic dynamics. In addition, auxiliary forecasting models can be introduced to provide the policy models with better initial conditions about the main macroeconomic variables (namely GDP and inflation) by using the relevant short term information timely available, i.e., well before the macromodel-relevant figures are officially released. <<<