CORDIS - Resultados de investigaciones de la UE

Integrated Macro-Financial Modelling for Robust Policy Design

Final Report Summary - MACFINROBODS (Integrated Macro-Financial Modelling for Robust Policy Design)

Executive Summary:
In the wake of the global financial crisis and the ensuing Great Recession, economists at policy-making institutions had little choice but to augment macroeconomic models with ad-hoc assumptions and mechanisms in order to provide analysis and advice for policy makers. This approach is unsatisfactory as those exogenous ingredients do not contain any explanatory power of the mechanisms that have led to such a crisis, and induced such economic and social costs. The MACFINROBODS consortium aims to move policy-focused macroeconomic modeling beyond this approach, towards the endogenous modeling of the dynamics resulting from financial risks and the related decision making of all the actors in the economy.

The MACFINROBODS project brings together four broad lines of research to systematically develop new behavioral and institutional building blocks, integrate them into policy-focused macroeconomic models and use the resulting models for policy evaluation. In terms of building blocks, one line of research moves beyond the assumption of representative and homo-oeconomicus-type of agents, and incorporates micro-behavioral realism into decision making. A second line of research advances the modeling of financial institutions, their fragility and the dynamics of systemic risk. The third line of research integrates these new building blocks (including a selection of those developed by researchers outside the consortium) in a new generation of policy-focused macroeconomic models. In parallel, in the fourth line of research, new policy evaluation tools are developed, with a focus on robust tools that aim to contain financial contagion and boom-bust cycles, maintain fiscal sustainability and coordinate monetary, fiscal and regulatory policies in normal as well as in crisis regimes.

The consortium comprises researchers with a strong track record in advancing the frontier on behavioral and institutional modeling, highly influential macroeconomic modelers as well as seasoned veterans of model-based monetary, fiscal, and regulatory policy evaluation and design. The members of the consortium have strong academic backgrounds as well as substantive practical experience at policy-making institutions.

Project Context and Objectives:
In the wake of the global financial crisis in 2007-08 and the ensuing Great Recession, economists have been widely criticized for failing to predict the occurrence and the severity of those events, or at least for being unable to come up with early warning signals of such a global financial disruption. Moreover, their macroeconomic models appeared of limited help to explain the mechanisms that led to such a massive economic contraction and account for these mechanisms in guiding policy design.

Those criticisms have come both from the academic side, through influential researchers such as the Nobel prize winners P. Krugman and J. Stiglitz, and from the practitioner side, in the words of leading policy makers, especially at central banks, such as the President of the ECB at that time, J.-C. Trichet. The critics have questioned the usefulness of macroeconomic research and academic training over the last three decades, and blamed researchers’ focus on dynamic general equilibrium modeling for misdirecting the attention of policymakers, especially at central banks.

Economists at policy making institutions had little choice but to augment macroeconomic models with ad-hoc assumptions and exogenous adjustment processes in order to provide analysis and advice for policy design to fight the financial crisis and the deep recession. Of course, this state of affairs is not satisfactory, as ad-hoc and exogenous elements lack the necessary explanatory power about fundamental economic mechanisms that can produce such crises and induce such economic and social costs.

Against this background, researchers have been debating whether to re-invent macroeconomic and financial modeling from scratch, or whether to improve existing frameworks that have already been used to inform policy makers. The MACFINROBODS consortium has pursued a systematic comparative approach to model design, in a way that is fully open to the existing competing paradigms, with the eventual aim of being able to deliver robust policy prescriptions. By robust, we mean policy recommendations that can be obtained whether one uses one model or another different, and even competing, one. This way, the policy makers could be more confident in their policy design, as the “true” model of the macroeconomy will probably remain unknown. The MACFINROBODS project has addressed this uncertainty by searching for policy settings and rules that perform well across a range of diverse representations of the macroeconomy. This approach acknowledges and exploits diversity.

In his opening address to the European Central Bank’s annual conference on 18 November 2010, former ECB President J.-C. Trichet expressed the need for such an approach very clearly : “We need macroeconomic and financial models to discipline and structure our judgmental analysis. How should such models evolve ? The key lesson I would draw from our experience is the danger of relying on a single tool, methodology or paradigm. Policymakers need to have input from various theoretical perspectives and from a range of empirical approaches. Open debate and a diversity of views must be cultivated ... We do not need to throw out our DSGE and asset-pricing models: rather we need to develop complementary tools to improve the robustness of our overall framework”.

The main objective of the MACFINROBODS project has therefore been to offer policy frameworks that account for the endogenous dynamics generated by the decision making under financial risk and economic uncertainty of the major actors in the economy : banks, households, firms and public institutions. The project has been designed to proceed by different building blocks, by systematically developing new behavioral and institutional modeling devices and integrating them into policy-oriented macroeconomic and financial models. Most importantly, the consortium aims to help improve model-based policy analysis at a wide range of policy institutions, namely the European Commission, the European Central Bank, the National central banks within the euro area and other EU and related countries, and large international organizations such as the OECD and the IMF.

The improvements upon existing macroeconomic models particularly concerns two lines of research. The first one aims to offer a more detailed and realistic representation of the financial sector, including the interdependence between banks, their risk-taking decision making, measurements of systemic risk and the possibility of sovereign crises. The second line of research aims to move beyond the assumption of representative, perfectly rational agents, and incorporate micro-behavioral realism into decision making. This research focuses on expectation formation, as they play a key role in shaping economic dynamics and determining the effectiveness of policies. This line of research also emphasizes agents’ heterogeneity and learning in contexts of high levels of uncertainty, in order to capture shifts between normal, boom and crisis times.

Once augmented by these two lines of research, the resulting macro-financial models account for features that are essential to understand the current economic situation, especially in the Euro area, develop new policy tools and provide policy recommendations. The project has been designed to address a wide scope of policy issues, including monetary policy design with large asset price fluctuations and financial imbalances, fiscal policy design once interest rates are constrained by the zero-lower bound and public debt is potentially large, macro-prudential and regulatory policies to avoid financial contagion between strongly interconnected and interdependent banks, and innovative policy evaluation tools.

Project Results:
The MACFINROBODS project has been particularly successful, both in terms of the quantity of work performed, the quality and the variety of the outputs. Beyond the management of the project and the dissemination activities, the 12 international partners have produced 75 deliverables, corresponding to 348 persons/month (PMs), which corresponds to less than 5 PMs per deliverable. Those deliverables encompass mainly articles, software components and policy briefs. Most of the deliverables are academic articles that present theoretical, empirical and experimental work that has been published, or is intended to be published in leading international peer-reviewed journals. The related software packages and computer codes allow other researchers to reproduce the results and apply the utilized methodologies to various related data, research questions and models. Finally, the policy briefs are based on the conclusions achieved during the work for the project and the corresponding policy recommendations obtained from the models developed and the data analyzed.

The consortium has worked along the two lines of research described above – namely developing more realistic representations of the financial sector and the micro behavior of economic agents, especially regarding the formation of their expectations within DSGE models – in order to offer more realistic models that can both provide an explanation of the combined mechanisms that have led to the financial crisis and the Great Recession, and provide policy guidance.

The consortium has been successful in developing a number of extensions of these DSGE models along those lines. The researchers have performed extensive empirical fit of these models, especially based on European data in order to compare the situation of the Euro area with the one prevailing in the US for instance. This work has led to extensive and deep policy analysis and exercises with those richer models, and has come up with a wide range of policy recommendations, insisting on their robustness to the modeling assumptions. All the models developed within the consortium are collected in the so-called Macroeconomic Model Database provided by the consortium. Such a database serves as a repertoire to assess macroeconomic and regulatory policies against a wide range of potential economic models.

It is worth highlighting that solving and simulating those models have posed quite a number of technical challenges, as the researchers had to operate at the frontier of the existing research in the field. Indeed, the non-linearities that this type of models involves, as well as their degree of complexity cannot be dealt with using standard available numerical tools. To this aim, the consortium has involved researchers from the highest level of expertise who have developed a number of new software toolboxes. These new tools especially allow one to obtain third-order approximations or adapt global sensitivity analysis-based procedures. They are of critical importance to capture the effect of time-varying risk shocks on the economy, to validate these models against empirical data and to compare the forecasting and fit performances of different models. All those innovative tools have been made publicly available to the research community.

We now discuss in detail the results from the different lines of research of the consortium.

3.1 Towards a richer representation of the financial sector withinDSGE models
The modeling of the financial sector within DSGE models has been extended to include, inter alia, collateral constraints within a detailed formalization of the bond market, endogenous entry and risk-taking from home and foreign banks, the possibility of systemic banking crisis arising in the interbank market, the international dimension of bank portfolio problems, an explicit inclusion of the housing market subject to bubble-like expectation dynamics, the inclusion of public debt crisis and sovereign default, and imperfect information regarding shocks. This richer modeling of the financial sector constitutes a big step towards a better understanding of the functioning of the global banking system, its interconnection with the real economy and the spill-over effects from financial crises to economic recessions. This work has paved the way for the design of well-suited macroeconomic policy recommendations in the current context of the Euro area.

3.1.1 Modeling fragility, systemic risk and contagion
Network-based models have been employed to analyze how banking globalization and the rise in banking competition affect the risk sharing and risk taking behaviors of banks, and the implied fragility of the global banking system (D3.1). This type of models reflects key institutional features, such as heterogeneity of banks and bilateral linkages. It has been found that, in the long run, the effects of competition on risk taking are ambivalent, and depend whether competition takes place on the local deposit market or on the loan market. In the short run, events that improve growth prospects in the banking industry brings about higher loan rates and more risk taking by firms.

Taking into account the increasing internationalization of banks, the link between the commercial bank exposure to risk and their expansion abroad has been analyzed within a model of banking industry dynamics with endogenous entry and risk taking in a national market (D3.2 D3.6). The model shows that the bank’s riskiness decreases in case of more foreign expansion to destination countries with different business cycle co-movements and stricter regulations, or more intense competition than the home country.

This type of diffusion mechanisms has then been incorporated in a rigorous way into a standard macro framework. This result fills a gap in the existing literature where contagion and diffusion phenomena have been mainly studied in partial equilibrium models (D3.5). The tools of network analysis allow to assess the probability and size of global cascades in the financial system. Focusing specifically on the propagation of systemic risk in the interbank market of the Euro area, it has been found that the functioning of this market provides only limited risk sharing between countries within the monetary union (D3.3). In terms of policy guidance, liquidity injections that replace impaired interbank lending therefore have only a small effect on closing the growth gap between the core and the periphery.

Consequently, several policy recommendations concerning potential regulatory tools (namely the role of entry and monitoring costs) to reduce the fragility in the global banking system have been proposed (D3.2). The models developed offer some answers to several policy questions that are currently debated with regard to global banking. Specifically, a tightening of regulation of the type we are witnessing now may come along with an increase in regulatory burden and banks’ operative costs, which might severely impair their ability to extract business margins. This represents an unintended consequence of regulatory choices such as, in the case of European banks, increased capital and liquidity requirements (due to Basel III regulation and to the more recent update of the CRR). It is also highly recommended to monitor how global banks’ liquidity (short term liabilities) moves across borders. The balance for global banking groups between risk contagion and risk-sharing is of major concern as well. The important policy question of the optimal degree of public bailout in the context of banking crises is also discussed (D3.4).

An important source of financial frictions is also the monopolistic feature of the banking market that generates a stronger propagation mechanism with respect to a model with efficient banks. In combination with industrial dynamics, the DSGE models have been augmented with an explicit dynamics of the population of firms, featuring endogenous business creation and destruction (D5.1 D5.2). This extension can replicate the changes observed in the patterns of the capital and investment distributions over the recent financial crisis and replicate the associated business cycle fluctuations. As for the policy conclusions, in these frameworks, monetary policy has been found to have an important role in dampening these changes, and regulations such as Basel III capital requirements can mitigate the effects of negative financial shocks.

3.1.2 Constraints and risks on policy
In face of the Great Recession and an unprecedented fall in output, private consumption and investment spending, all advanced economies responded with a range of fiscal and monetary policy measures that have included increases in government spending, tax cuts, and various type of unconventional monetary policy measures, once monetary policy became constrained by the zero lower bound. As a result, the Great Recession deteriorated the fiscal positions of most advanced economies.

Endogenous crises To account for those dynamics, the consortium has developed DSGE models with non-linearities and binding constraints stemming from, inter alia, rising financial fragility and public indebtedness, the resulting endogenous borrowing constraints and the potential sovereign risk, the benefit from investing in a common fund (D6.1) and the zero-lower bound of the interest rate.

One challenge is to model endogenous crisis episodes, i.e. not triggered by shocks outside the model. Exploiting occasionally binding financial constraints have been proven to be a successful way to do so (D5.6). Those endogenous crisis episodes feature permanent falls in productivity, without generating implausibly large amplification in normal times. This is a novel method that is able to generate persistent medium-frequency cycles, meaning it matches well the data from short to long horizons.

In a similar vein, integrating an inter bank market which operates under moral hazard and asymmetric information can produce endogenous sudden market freezes, banking crises, credit crunches, and severe “financial” recessions (D5.5/D5.8). Those recessions follow credit booms and are not necessarily triggered by large exogenous adverse shocks, which is a clear improvement upon most existing models. State-dependent fiscal multipliers A large part of that strand of the research has then been devoted to the estimation of the so-called fiscal multiplier, and the discussion of the rationale for fiscal stimulus or consolidation in a recession once interest rates are zero (D6.2 D6.3 D6.4 D6.5 D6.6). This research has been mainly carried out with WPs 6 and 10. Importantly, the research has shown how to introduce debt crises, debt default and partial information into DSGE models. The resulting models address crucial issues such as the desirability of a certain level of debt, the timing of fiscal responses and, more generally, the conduct of fiscal policy under uncertainty. It is a very robust results in economic models that the multiplier of government spending is higher when the spending is deficit-financed rather than tax-financed. Yet, given the spectacular rise in public debt in the aftermath of the Great recession in the Euro area, it seems of primary importance to calculate the level of the output multiplier of a (tax-financed) increase in public debt.

In a period of mild recession in which nominal rates remain positive, the multiplier is generally quite small. The research conducted here has found that the debt multiplier is much larger, however, when the recession is severe and interest rates drop to zero. In this respect, the calls for fiscal consolidation in recession times seem ill-advised. Temporary shocks to the government budget should be largely absorbed by deficit adjustments. This policy is desirable because it avoids drastic tax increases or government spending cuts. If available, consumption taxes instead of labor income taxes may greatly limit the distortions caused by taxation when the government’s financing needs are high (D10.5). However, it has also been found that fiscal consolidation at the zero lower bound could succeed in reducing debt-to-GDP and providing stimulus, but only if combined with a decrease in labor tax rates. The results appear to depend also on the level of sovereign debt to GDP ratio, and may revert in case of low fiscal space.

In most of these models, the possibility of a debt default is ruled out by construction and debt crises are only envisioned as liquidity crises that are due solely to the inability to roll over debt, which is an obvious limitation. In that respect, the consortium has also developed a theoretical framework for analyzing government behavior in face of fiscal pressures (D10.1). In this framework, not running down debt, or running it up until default is unavoidable, can be part of the optimal strategy of the government whenever the recession is very severe and there is a nontrivial probability of a recovery. This framework provides an interesting tool for testing the implications of alternative policy responses. Information asymmetry between the government and its lenders can exerted an effect on the sovereign default risk, so that less transparent governments tend to have higher levels of debt, due to consumption smoothing behavior and low borrowing costs (D5.4 D5.7). The framework developed allows for the introduction of market sentiments that, if misaligned, can generate business cycle features that are close to the ones observed in the data.

Focusing specifically on the sovereign risk with the Euro area, the consortium discussed the policy implications of the European sovereign debt crisis. It is shown that, while EU countries have been isolated from exchange rate risk by the monetary union, they have suffered from the risk of banking crises. Moreover, while sovereign risk always appears during downturns, it has developed during recent years, because of the gap between GDP and export growth. The key challenge of sovereign risk management is to engineer debt consolidation in good times, rather than too late, when the downturn starts (D10.3). Several arrangements are then contrasted to face that risk (D10.5). It seems that in a monetary union, there is a positive risk of default, which is the outcome of the fact that default is not too costly : the reason has to do with the fact that the country regains its monetary policy while keeping the stability brought by the fixed regime. A higher degree of consumption habit simultaneously raises the risk of default. With weak fiscal instruments, the risk of default is larger for large habit formation and may explain why tougher fiscal rules are here needed.

In case of an exit from the Euro area of Southern European countries, the risk of hyperinflation can be significant if they leave the Euro with positive and persistent deficits, large value of debt-to-output ratios (D10.6). The analysis focuses on a case when agents in the economy have incomplete information about the economy, using the recently developed framework of “internal rationality”, thus departing from rational expectations. This paper serves as a methodological exploration of how to perform policy analysis under internal rationality. The inclusion of non-rational expectations is now discussed.

3.2 Beyond the rational and representative agent assumption within DSGE models The representation of learning dynamics within DSGE models with financial frictions has been improved along a number of dimensions. This work has been mainly carried out in work packages WP1 and WP2.

3.2.1 (Non-)Rational Expectations and Implications
A natural direction towards a better explanation of the expectation formation and resulting aggregate processes is to generalize the rational expectation framework, and allow possible temporary unstable paths (D4.1). This approach could identify the inflation rates of the 1970s with unstable paths of rational expectation models. Alternative behavioral models from the microeconomic literature, such as internal rationality (see above) or the prospect theory à la Kahneman & Tversky, have been used within parsimonious macroeconomic frameworks (see D1.1). Once accounting for behaviorally realistic assumptions regarding loss aversion in consumption, a fact that has found a wide support in the empirical and experimental literature, output displays more sensitivity to fiscal policy shocks in contractions than in expansions. This behavioral feature of the model explains why government spending shocks have more persistent and larger effects during recessions than during expansions, which is a widely documented stylized fact in the empirical literature.

The theory of diverse rational beliefs has also been integrated into a macroeconomic model with financial markets (see D2.3 D.2.4 and D2.5). Waves of optimism waves about the future are found to be able to increase aggregate output, but also cause fluctuations in financial markets. The Great Recession of 2008/2009 that followed the global financial crisis is found to be better captured by a business cycle model with diverse higher-order beliefs than in a corresponding business cycle model without belief diversity as a propagation mechanism.

As policy implications, among others, once diverse beliefs are taken into account, a central bank should target exuberance on financial markets directly. The conclusion that monetary policy authorities may face a trade-off when implementing policies between accommodating policies that have a positive effect on output but only at the price of a greater instability on financial markets, is a robust finding across different frameworks of the research conducted by the consortium (see for instance D3.5). Experimental evidence on behaviors and expectation formation in highly uncertain environments have been collected to empirically explore the features of belief formation and heterogeneity that lead to occasionally unstable dynamics, such as bubbles in financial or housing markets. Forecasting heuristics that are able to fit simultaneously several different economic environments have been identified. Those laboratory experiments with human subjects make a substantial contribution to the empirical assessment of expectation formation models, and the resulting market failures such as mispricing or coordination on bubbles that is often a feature of real financial markets. On the theory front, several behavioral models have been proposed to reproduce and explain these market failures, especially the use of simple rules, such as adaptive anchor or trend-extrapolating forecasting rule (see e.g. D1.2 D1.4 D2.1 D2.6). This research has developed and extended heterogeneous expectation models in environments with large fluctuations, with a parsimonious experimental foundation of individual forecasting rules, that is robust across different experimental settings.

This experimental approach has been applied to topical monetary policy proposals about price level targeting or explicit guidance (see D1.6). While subjects have been found to coordinate on similar behaviors, namely simple rules, large differences between different policy treatments prevail. For instance, guidance is found to have a negligible effect, whereas the aggressive Taylor rule specification turns out to be crucial in the presence of expectations formed by human subjects. Those behavioral patterns have been transported to larger-scale and policy-focused macro-financial models, resulting in so-called hybrid models, featuring both a DSGE skeleton and an agent-based behavioral block. This work is described in detail in the next section.

Different learning schemes and estimation procedures, especially but not only among the adaptive learning models, have been compared and assessed on the basis of their respective fit to empirical data of the Euro area and related forecasting performances. Those adaptive learning models have been combined with survey data of professional forecasters and estimated DSGE models, and have proven to provide a richer account of the joint dynamics in realized and expected inflation than the rational expectation model, even if augmented with survey data (D2.2 D4.2 D4.5).

3.2.2 Heterogeneity and Implications
Building on the heterogeneous agent literature in which labor supply decisions are endogenous, part of the consortium has investigated the effects of heterogeneity on unemployment, growth and technology adoption, and provided new insights into the widely debated skill obsolescence hypothesis. This hypothesis states that technical progress both destroys and creates skills, reducing employment and wages for middle- skill occupations but raising them for high-skill occupations. The researchers have offered both data-based evidence of that phenomenon, and a model that can account for it (D1.3).

Within this class of models, important insights have been gained into taxation and migration policies, and optimal taxation in the presence of technological changes (D1.5 D1.6). For instance, a more progressive tax scheme does not necessarily result in significantly higher government revenues, as the adverse effect on labor, savings and output may dominate the increase in tax revenues. Another example is the insight into migration policy and a discussion of selective immigration policies based on education. Solving this type of models with endogenous heterogeneity has been a challenge and the consortium provides a number of tools and methodologies to do so (see e.g. D1.6).

3.3 Towards robust policy analysis and recommendations
Once augmented by these two lines of research, the resulting macro-financial models account for features that are essential to understand the current economic situation, especially in the Euro area, and provide policy recommendations.

3.3.1 Towards a better understanding of the crisis dynamics
Those augmented DSGE models then serve to better understand the dynamics that have led to the crisis, and those underlying the recovery, compared to the models with exogenous (ad-hoc) adjustment mechanisms that used to prevail.

In that respect, those models can be estimated on empirical data, and hence allow us to identify the importance of financial shocks in triggering the Great Recession. Those shocks have been found to be more persistent in the Euro area than in the US, explaining the divergent post-crisis dynamics (D5.9/D5.10). Regarding the pre-crisis dynamics taking the example of Spain, falling risk premia on the housing and non-residential capital, a loosening of collateral constraints for households and firms, as well as a fall in the interest rate spread between Spain and the Euro area have fueled the output boom and the persistent rise in foreign capital flows to Spain, before the global financial crisis. During and after the global financial crisis, falling house prices, and a tightening of collateral constraints for Spanish borrowers have contributed to a sharp reduction in capital inflows, and to the persistent slump in Spanish real activity. The credit crunch was especially pronounced for Spanish households ; firm credit constraints tightened later and more gradually, and contributed much less to the slump.

Based on a collection of empirical evidence, it has also been shown that slow recoveries are associated with a significant drop in the growth rates of investment and bank loans, and with a surge in the growth rates of corporate bonds and only a minor role for non-financial corporate indebtedness (D5.11).

3.3.2 Learning as an amplification mechanism of financial frictions in DSGE models
The most recent studies that analyze DSGE models with financial frictions have documented problems in replicating the observed dynamics and explaining the “surprising” origin of the crisis and its propagation channels. Therefore, there has been a need for further research in the field of macro-financial modeling to consider additional features of the transmission mechanism that could potentially augment the interactions between the financial and real sectors, thus making models more successful in describing the business cycle dynamics.

Interestingly, once combining the two lines of research highlighted in the previous sections, the effect of financial frictions for the business cycle has been found to vary depending on the approach to modeling the expectations (D5.3). The results suggest that deviations from rational expectation strengthen the impact of financial frictions compared to the model with fully rational agents. In particular, some learning schemes based on small forecasting functions are able to amplify the effects of financial frictions relative to the model with rational expectations. Learning may create additional business cycle fluctuations in line with the real data by influencing the degree of economic persistence in asset prices and investments. These results imply that learning represents an alternative source of endogenous inertia and is helpful in explaining inflation, wage and investment dynamics in a way that outperforms rational expectations models. This seems to be a robust result, at least for the U.S. and European data.

In an open economy, the interaction between learning and financial frictions leads to more asymmetric fluctuations and lower business cycle synchronization (D9.5/D9.6). In this context, learning entails a greater need for both output and inflation stabilization, thereby requiring a more aggressive action from central banks. However, in any case, monetary policy is not able to fully anchor expectations in response to shocks hitting foreign countries.

3.3.3 Merging agent-based foundations with DSGE models
Models have been developed that combine an equilibrium skeleton with expectations formed within an agent-based component that encompasses several types of heterogeneous agents. Applied to the housing market on which fundamentalist and chartist forecasters operate (D7.1) it has been shown that the model can replicate the recent US house price dynamics, and account for interaction between the housing market and the business cycles. The model can also shed light on the causes of such bubbles : it seems that narratives based on "fundamentals", as the interest rate behavior or the credit market liberalization, appear to be less important. On the contrary, preference shocks seem to be the main explanation of the house price movements. From a policy standpoint, the model supports the view that an interest rate policy reacting to the house price could stabilize the house price dynamics.

In similar models of heterogeneous expectations consistent with the behaviors found in lab experiments or survey data, monetary policy and the related management of expectations have been analyzed at the zero-lower bound (D7.5). This type of models leaves the possibility for self-fulfilling pessimism that cannot be offset by monetary policy once the zero-lower bound hits and, therefore, there is a large danger of a long lasting self-reinforcing liquidity trap. Its occurrence crucially depends on the credibility of the central bank (D9.4). It seems that a higher inflation target, aggressive monetary easing, or a more aggressive response to inflation reduce the probability that self-reinforcing liquidity traps arise.

Further away from equilibrium modeling, so-called hybrid agent-based models have been developed within the consortium. The starting point is an agent-based macroeconomic model with a focus on borrowing/lending relationships within a banking network as well as between a banking network and the real sector (D7.2). This type of framework can reproduce the effects of adverse shocks to the credit network that trigger an avalanche of bankruptcies, culminating in a real economic crisis.

In this framework, monetary and macro-prudential policies have been analyzed, especially a modification of the maximum allowable leverage ratio and the required liquidity ratio of the banking sector. The former seems to be more effective than the latter in terms of frequency of crises, but the duration of those remains unaffected. By contrast, policy acting on the required liquidity ratio displays no significant stabilizing effect. This strand of the research also gives a methodological contribution to the literature. The agent-based approach has been employed to analyze a model with micro-foundations that are typical of the DSGE literature but incorporating also dispersed information and learning dynamics (D7.3). Policy can have an additional signaling role in such frameworks.

3.4 New policy evaluation tools

3.4.1 Model validation, simulation and comparison
The work performed mainly in WP8 has been devoted to the development of new software tools to improve estimations of DSGE models in presence of non-linearities, such as occasionally binding constraints or switching between multiple regimes (D8.1 D8.2 D8.3 D8.4 D8.9). This development makes significant computational progress despite the complexity of the task towards the resolution and the simulation of dynamic non-linear models that are necessary to capture the range of economic phenomena that have been discussed in the project.

One angle of improvement, for instance, is a faster and more robust extension of the so-called “path approach”, which was already implemented in Dynare, the main software used by the community of researchers to work with DSGE models. Another dimension concerns the abandoning of the strong assumption about the absence of future shocks : the modeler can now go beyond deterministic nonlinearities and partially recover the effects of nonlinearities on agents’ behavior with respect to risk. The main advantage of this global simulation strategy is its simplicity : its complexity does not grow exponentially with the size of the model, but only with the number of shocks. Also, it allows to take into account any kind of occasionally binding constraints.

The use of nonlinear methods, such as Sequential Monte Carlo (also known as particle filtering), allows the modeler to make an approximation of the state variable distribution at each date without formulating assumptions on the specific shape of the distribution. Two cases have been considered : i) the common approach that only uses the past information on the state variables and ii) more recent ones that include current observable information. The consortium has developed implementations in that respect in Dynare, focusing on the main difficulties found in the literature and discussing the alternative solutions proposed in the literature. For illustration purposes, they use the simple DSGE model that is generally studied in the literature on these methods and discuss how it can be estimated in Dynare with non-linearities. They also provide a numerical illustration in which the different methods are implemented and compared.

Applying efficient methodologies and computational techniques forms the basis of well-performing structural models and is key to provide policy analysis. Hence, further significant computational progress towards the evaluation process of a model and the comparison between macroeconomic models have been made (D8.5). This work provides the major extensions for comparing impulse response functions or model moments of different models thanks to a toolbox that encompasses a wide set of diagnostic tools to analyze and inspect global sensitivity patterns.

The applications of those key methodologies and computational techniques concern a wide range of policy questions and modeling features that have been discussed throughout the project, namely open/closed economy models, financial frictions, fiscal multipliers, endogenous growth and interaction with the business cycle, as well as the behavior of financial institutions with time-varying shocks (D8.11).

Those applications shed light on important policy questions. With non-linearities and occasionally binding financial constraints, it has been found that the financial accelerator mechanism, which enhances the effects of monetary policy on real outcomes, is only activated during times of financial distress ; at other times, monetary policy transmission is analogous to a standard existing linear models and financial intermediation is efficient. However, moderately large shocks drive the model into a region in which the financial constraint binds, generating endogenous financial crises.

3.4.2 Predictions with models
Important contributions have also been made in order to compare the predictive powers of alternative macroeconomic models, which is an important step for policy makers. A model platform has been developed (D8.6 D8.7). It contains a newly constructed real-time data set for the U.S. and the euro area, a collection of DSGE and Bayesian vector autoregression models and a graphical user interface that brings data and models together. Based on this platform, the authors conduct a comparative forecasting exercise to study the relative performance of DSGE models, VAR models and professional forecasters using real-time data vintages for the U.S. economy.

For the entire sample, professional forecasters outperform macroeconomic models at short horizons, whereas at medium horizons the performance is often similar. The Great Recession episode is then studied in more detail. During the periods leading to the recession, a financial accelerator model predicts the decline in GDP growth earlier than models without financial frictions. In subsequent periods, the models without financial accelerator but with relatively strong endogenous propagation perform better.

The literature on model comparison is then reviewed and the work presents a new approach for comparative analysis. Its computational implementation enables individual researchers to conduct systematic model comparisons and policy evaluations easily and at low cost. This approach also contributes to improve reproducibility of computational research in macroeconomic modeling. Several applications serve to illustrate the usefulness of model comparison and the new tools in the area of monetary and fiscal policy. They include an analysis of the impact of parameter shifts on the effects of fiscal policy, a comparison of monetary policy transmission across model generations and a cross-country comparison of the impact of changes in central bank rates in the United States and the euro area.

Finally, a large-scale comparison of the dynamics and policy implications of different macro-financial models is also included. The models considered account for financial accelerator effects in investment financing, credit and house price booms and a role for bank capital. An example shows how these models can be used to assess the benefits of leaning against credit growth in monetary policy.

3.4.3 Policy coordination
Once financial frictions have been taken into account, the interaction between the design of monetary policy and the macroprudential tool has to be reassessed. If the macroprudential tool imposes a too weak positive response of the bank capital ratio to private lending, the so-called Taylor Principle is violated leading to excessive fluctuations and macro-prudential policy is ineffective in stabilising debt (D11.1). Under a constant bank capital requirement, a strong reaction of the interest rate to inflation also destabilizes the financial sector.

Within a monetary union, the design of monetary and macro-prudential policies can be envisioned as a coordination game. Once balance-sheet effects are introduced in such a game and we allow for reserve ratios similar to those of the ECB, there are distributional effects from choosing optimal simple macroprudential rules (in harness with optimal Taylor rules) that incorporate responses not merely to credit growth, but also to GDP growth (D9.1). Similar effects are also seen when optimal simple stabilization rules are used.

The consortium has also tackled the criticisms of the European Central Bank’s asset purchase programs. They find both in the euro area and in Germany a positive effect on output, while prices do not respond significantly. Asset purchases reduce financial stress, but this beneficial effect is overturned in the medium run. The German financial market, however, shows signs of increasing imbalances resulting from central bank balance sheet policies.

A model where firms issue assets backed by projects with differentiated productivity has been developed to offer a rationale for the policy of quantitative and credit easing (D9.2). This model is able to explain the observed low risk premia and high volumes on markets for securitized assets prior to the recent financial crisis, and the sudden drying up of this market at the onset of the crisis. Indeed, a so-called adverse selection problem arises in a recession and is especially severe when the recession is preceded by a prolonged boom period, because financial firms tend to accumulate larger amount of low quality assets. In the case of a deep recession, the government policy of asset purchases, inspired by the quantitative and credit easing, can limit the negative effects of adverse selection on the real economy. This prevents the effect of large amounts of securities with low cash flow on the financial firms’ balance sheet, and thus mitigates the effects of deep recession as regards investment and capital stock. However, by supporting the price of low-quality-project-backed securities on the secondary market, this intervention increases incentives for financial firms to create such securities and conceal their intrinsic quality by offering implicit guarantee of the return, thus creating a moral hazard problem.

3.4.4 Robustness of policy recommendations
Robust rules across model uncertainty has become a growing concern for policy making given the increasing complexity of macroeconomic models used for policy design. Based on the multiple new models with richer financial sectors and expectation formation processes that have been developed and integrated into the Macroeconomic Model Database of the consortium, the form of robust monetary and fiscal rules for the Euro Area have been revisited – in particular, relative to earlier generations of macroeconomic models (D11.2 D11.3 D11.4). This exercise has provided a range of policy conclusions. The authors document that for the models with financial frictions studied, robust monetary policies feature a weaker response the output gap (D11.4) and sometimes a stronger response also to inflation (D11.2) than for their standard New Keynesian counterparts. They also document that, for this class of models, a systematic and direct response of the monetary policy rate to financial sector variables, i.e. a leaning-against-the-wind-type monetary policy, does not appear advisable (D11.5). Lastly, they consider an active rules-based role for fiscal policy in macroeconomic and debt stabilization.

As for the interaction between monetary and macroprudential policy, it seems that monetary authority should play the role of leader with respect to the macroprudential authority, that should be a follower. This arrangement delivers less volatile inflation dynamics without necessarily increasing the standard deviation of the credit imbalances, compared to an arrangement where both authorities perfectly coordinate (D11.5). This arrangement also leads to smaller potential costs due to model uncertainty of model-specific optimal policies. The best-performing robust macroprudential policies are modestly countercyclical as they yield the best performances on the credit market while achieving stable dynamics in inflation and the output gap.

Looking forward, the authors argue that model-averaging and embracing alternative modeling paradigms is likely to deliver a more robust framework for the conduct of monetary and macroprudential policies. Those conclusions, as well as their empirical relevance for policy making and implications are all discussed in a policy document (D11.6).
Potential Impact:
The target audience of the project comprises several groups : (i) researchers in academia and policy making institutions, (ii) professionals that are tasked with preparing and communicating policy advice at policy institutions, (iii) decision makers at policy institutions and the (iv) public media that report on these policy areas. In order to reach these diverse audiences, the consortium has been developing a dissemination strategy that accounts for the diverse information needs and accessibility of these groups.

4.1 Dissemination and Communication
The project has pursued this strategy by setting up a Communication and Dissemination team and prepare a Stakeholder analysis. Scientific Workshops and annual Conferences have been organized at different locations, particularly nearby policy making institutions. The involvement of Advisory Board Members in scientific workshops and annual conferences has been planned in order to use their visibility to also attract key European policy makers to these events.

To ensure high visibility, the progress made has been presented via a professional website and policy briefs. This website, has been a very important tool for communication, and has been updated constantly with all deliverables and all dissemination events, organized within the Consortium. To ensure that information will not be lost after the project has come to an end, the website will be available for two more years after the end of the project (until May 2019).

A functional calender was updated regularly during the whole duration of the project. Additionally, a stakeholder analysis has aimed to identify institutions potentially interested in the scientific output of the project. Stakeholders share a common focus on policy making in the subjects related to MACFINROBODS and/or a geographic proximity to the institutional members of the project. The stakeholder analysis has been reconsidered and fine-tuned on a regular basis in order to identify the right stakeholders to enlarge the possible impact of the project outputs.

In order to make the project output accessible to stakeholders and a larger group of academics, three annual conferences were organized, where consortium researchers and other scholars joined outstanding international personalities (including Advisory Board Members) to present and discuss the main findings of the research.

The First Annual Dissemination Conference was held at the Banque de France, Paris, on 15-16 June 2015. It was a very successful two-day event with over 60 participants. The conference started with a welcome speech given by Alain Duchateau, the general deputy director of Banque de France.There were 11 presentations of advanced-stage papers from all teams and a poster session, including 9 presentations of early-stage works. We also had 3 inspiring keynote lectures by Thomas J. Sargent, professor of economics at New York University and the 2011 Nobel Prize in Economics, by Tobias Adrian, Senior Vice President and Function Head Capital Markets Function at the Federal Reserve Bank of New York, and by Vítor Gaspar, Director of the Fiscal Affairs Department at the International Monetary Fund.

The Second Annual Dissemination Conference was held at the National Bank of Belgium in Brussels, on 20-21 June, 2016 (D12.5). It was a very successful two-day event with over 30 participants. There were 9 presentations of advanced-stage papers from all teams and a poster session including 5 presentations of early-stage works. We also had four inspiring keynote lectures by renown and leading researchers : Wouter den Haan, professor of Economics at London School of Economics, Stephanie Schmitt-Grohé, professor of Economics at Columbia University, Martin Uribe, professor of Economics at Columbia University and Chris Sims, professor of Economics at Princeton University, Nobel Laureate Economics, 2011 and member of the Advisory Board. All keynote lectures have been followed by a very active general discussion.

The Final Conference was held in Frankfurt (D12.6) and organized by Volker Wieland and Michael Binder at Goethe University on 4-5 April, 2017, followed by the ECB Watchers Event on Thursday 6 April. The conference was a very successful two-day event with over 50 participants, with two inspiring keynote lectures, Frank Schorfheide, professor of Economics at the University of Pennsylvania. and by John B. Taylor, professor of Economics at Stanford University and member of the MACFINROBODS Advisory Board. On the third day, “The ECB and Its Watchers XVIII” took place, with over 250 participants and more than 100 media representatives from all over the world, organized by Goethe University and starting with a Presidential Address by Mario Draghi, President of the European Central Bank. This event was a very stimulating meeting with several panel discussions between academic researchers and policy makers addressing many relevant themes from the MACFINROBODS project. Among the panel members were Volker Wieland (GU), and John Taylor and Vitor Gaspar, both members of the MACFINROBODS advisory board. All information on the Conferences can be found on the Macfinrobods website at the link

On top of these conferences, several other main dissemination events of the MACFINROBODS research were organized.
At the Dutch Central Bank, the Learning Conference “Expectations in Dynamic Macroeconomic Models” from 6-8 September, 2017 was organized by Cars Hommes, coordinator of the MACFINROBODS project. This is a highly visible policy-oriented conference that focuses on the important role of expectations and learning for macroeconomic dynamics, macrofinancial interactions and monetary and fiscal policies. All those topics are fully in line with the work performed within the consortium, especially one of the main lines of research regarding behavioral aspects of agent modeling. The Keynote speakers of the conference were James Bullard, President and CEO of the Federal Reserve Bank of St. Louis, whose talk focused on non-conventional monetary and fiscal Policies ; Seppo Honkapohja, Bank of Finland and member of the MACFINROBODS Advisory Board, and Bruce Preston, (Monash University, Australia), who presented his work “Optimal Monetary Policy under Imperfect Knowledge”. A poster session with several contributions of MACFINROBODS partners was also part of the conference. This conference provided a unique opportunity to discuss and disseminate MACFINROBODS research with international top academics and policy makers.

At the University of Surrey, UK, on the 20th of November 2015, the CIMS Workshop was organized by Paul Levine, node leader of the University of Surrey. The theme of the conference was “Agent-Based and DSGE Macroeconomic Modelling : Bridging the Gap”, a theme that is closely related to WP7 “Bridging agent-based and DSGE modelling approaches”. During this conference, presentations of research output from three different projects, MACFINROBODS, FINMAP (EU-FP7) and ESRC were given. MACFINROBODS node members from UvA, UCSC and SURREY presented their work performed within the project, and discussed with experts how to improve the work and how to use the work for policy analysis.

At Consejo Superior de Investigaciones Cientificas - Institute for Economic Analysis (CSIC) a dissemination workshop on Macro & Micro Perspectives on Taxation was organized on 19 and 20 June 2014. This workshop brought together some of the best experts on taxation (13 speakers).

In order to coordinate research activities and ensure interaction three Consortium Scientific Workshops were organized, in London on 15 and 16 January 2015, in Barcelona on 30 November and 1 December 2015 and in Como on 17 and 18 November 2016. All three Workshops were very successful two-day events, with presentations by senior- and young researchers from all teams, poster presentations of earlier-stage work by junior faculty mainly and very inspiring keynote lectures by Seppo Honkapohja (Bank of Finland and MACFINROBODS Advisory Board member), Jordi Galí (CREI and Universitat Pompeu Fabra, Juan Pablo Nicolini (Federal Reserve Bank of Minneapolis), Paul Baudry (Univeristy of British Columbia and Oxford) and Marco Del Negro (Federal Reserve Bank of New York). All information on the Workshops can be found on the MACFINROBODS website at the link

4.2 Potential Impact and Exploitation
Regarding the accessibility of the work performed by the consortium, it is worth pointing out that all the academic articles have followed the so-called “green route to open access”, i.e. the working papers are freely available on line to the public, systematically on the project’s website (see below for the address), and most of the time on the websites of the authors (The project did not include any budget to provide the so-called “golden route”, which gives a free access to the published versions in journals, often at the cost of several thousands of euros). The software toolboxes are all open-access, and offer freely to the community of researchers a wide range of state-of-the-art numerical approximation and solution techniques.

Numerous academic papers have been produced by the consortium teams, some have been already published and some are currently in the reviewing process of highly-ranked international academic peer-review journals. Outcomes for some of the deliverables already include The Journal of Political Economy, the Economic Journal, The European Economic Review, The Journal of Economic Behavior and Organization, The Journal of Economic Dynamics and Control, The Journal of International Money and Finance, Macroeconomic Dynamics, Computational Economics, Economic Policy.

Those papers have also been presented at key international conferences, such as the ASSA meetings, the EEA-ESEM meetings or specialized, highly visible conferences.

The consortium has achieved a number of improvements upon existing tools in macro-finance modeling. These improvements concern a better description of the financial and the banking sectors and a more realistic representation of the behavior and expectation formation of the actors in the economic system. These improvements represent quite a big step forward in the understanding and modeling of a wide range of economic and financial phenomena that have turned out to be of critical importance in the economic dynamics that have ensued the financial crisis and the Great Recession, especially in the Euro area. These dynamics include the contagion of shocks in the financial sector into the real economy, bankruptcies cascades and spill-over effects within an interconnected and globalized banking system, bubble-like asset and housing price dynamics, the challenge raised by the zero-lower bound and the constrained reaction of the monetary authorities, the effect of fiscal shocks in a context of high public debt levels and rising sovereign solvency concerns, especially within a monetary union.

These improvements constitutes no doubt a solid ground for understanding and acting upon phenomena that have caused the dramatic economic and social costs that the Euro area has been dealing with in the wake of the financial crisis. A better understanding of those mechanisms is a necessary condition towards the design of more effective policies to prevent and counteract those costs.

In line with this eventual objective, the ultimate achievement of the consortium has been to design robust monetary and fiscal policies and robust macro-prudential and regulatory policies under model uncertainty, while accounting for those phenomena. These policy recommendations are designed as simple rules that minimize the welfare loss compared to fully optimal, but complicated and model-specific policies.

The range of policy questions that have been addressed in a systematic and rigorous way includes : the design of monetary policy reaction in the presence of asset price and housing price excess volatility, the benefits in terms of risk-sharing from a cross-country common investment fund in a monetary union, the maximum level of indebtedness that is desirable in face of the possibility of a fiscal crisis and the engineering of debt consolidation programs in good times to accommodate fiscal crisis when a downturn starts.

Undoubtedly, those policy issues are of key relevance for the recent and current economic experience in the Euro area. These policy recommendations are extremely useful to policy makers at major economic and financial institutions, as the consortium clearly addresses the type of issues that they have been facing ever since the beginning of the financial crisis, while they did not have the appropriate tools to guide them back at that time. The hope is that the work performed under the MACFINROBODS project can help avoid such a situation in the future.

The achievements of the consortium will also without doubt push further the macro-financial modeling line of research within the profession, and academic research. In order to facilitate this research advancements, the consortium has also developed new computing techniques and corresponding software packages to solve and estimate the proposed models while dealing with the non-linearities and constraints that they imply. The Macroeconomic Model Database will also be made public, in order for other researchers to be able to use these models, and build upon them their own research program. In the near future, those models and tools should allow the professions to investigate an even wider range of policy questions than the ones that this project has been discussing, and in a robust way.

More generally, improving our understanding of the causes and consequences of financial and economic crises should help us develop better analysis and policy tools that can be used with a fair degree of confidence by policy makers, despite the fact that the true model of the economy will probably remain unknown. More reliable and efficient policies should be able to limit the economic and social effects of such crisis episodes, and even, ultimately, help prevent their occurrence. More accessible research should also work towards more understanding and acceptance of policy options and choices made by a larger public, therefore improving the accountability of policy institutions such as the central banks. This research has strongly contributed to those developments and has offered some promising findings and directions for the future.

List of Websites:
Contact details:
Cars Hommes, project coordinator (
Karin Breen, project manager (

Amsterdam School of Economics
University of Amsterdam
Roetersstraat 11
1018 WB Amsterdam
The Netherlands

Project public website address:

List of all beneficiaries/partners:
1. Universiteit van Amsterdam (UvA), Professor Cars Hommes (node leader and coordinator)
2. Centre Pour la Recherche Economique et ses Applications (CEP), Professor Xavier Ragot and Dr. Eleni Iliopulos (node leaders and vice-coordinators)
3. Joint Research Centre - European Commission (JRC), Professor Marco Ratto (node leader)
4. Universita Degli Studi di Pavia (UNIPV), Professor Guido Ascari (node leader)
5. University of Surrey (USU), Professor Paul Levine (node leader)
6. Katholieke Universiteit Leuven (KUL), Professor Vivien Lewis (node leader)
7. Universita Karlova v. Praze (CERGE), Dr. Sergey Slobodyan (node leader)
8. Johann Wolfgang Goethe Universität (GU), Professor Michael Binder (node leader)
9. Universita Catholica del Sacro Cuore (UCSC), Professor Domenico Delli Gatti (node leader)
10. Agencia Estatal Consejo Superior de Investigaciones (CSIC), Professor Albert Marcet (node leader)
11. The City University (CITY), Professor Joseph Pearlman (node leader)
12. Universite Libre de Bruxelles (ULB), Professor Robert Kollmann (node leader)