"Challenges that the Recent Financial Market Turmoil Places on our Macroeconomic Modeling Toolkit"

Many years ago I was at an NBER conference. A lot of the time is spent in sessions, but there's also quite a bit of time when it's more like a social event. However, I'm not the best mingler in the world, far from it, but one person who was there, Charles Evans, went out of his way to make me feel welcome and I haven't forgotten. So I don't expect me to ask him any really tough questions in this simulated interview on the usefulness of modern macroeconomic models in the present crisis:

Challenges that the Recent Financial Market Turmoil Places on our Macroeconomic Modeling Toolkit, Charles L. Evans President, Federal Reserve Bank of Chicago ["Pseudo-interview" based upon the linked speech. Edited slightly, deletions only]:

Q: Thanks for agreeing to do this. I said I'd let you pick the topic, so what do you want to talk about?

I would like to discuss some of the challenges that the recent turmoil in financial markets poses for research in both financial markets and monetary economics.

Q: Sounds good. What challenges? Aren't we, more or less,  converging on a particular model for policy analysis?

Over the years, the research community and policymakers have developed a good framework for thinking about monetary policy reactions to real–side shocks. For instance, if we think about the appropriate policy response to an oil shock, there is widespread agreement on the broad outline of models that are useful to answer the question.

Q: What kind of models are used?

Typically, we would turn to a dynamic stochastic general equilibrium model, (DSGE) model, with both nominal and real rigidities; estimate or calibrate the parameters of the model; and then trace out the responses to the real shock under different policy actions—and perhaps under different policy responses.

Q: Has the research community settled on a particular model? I'd be surprised to hear that it had.

Certainly many details remain disputed, such as the extent of nominal versus real rigidities, or the appropriate form of adjustment costs; these details are very important, but they do not challenge the basic tools we empirical macroeconomists use for policy analysis.

Q: Yes, I think Blanchard recently made a similar point in "The State of Macro." So have those tools served you well in the present crisis?

Unlike real shocks, we don't, as yet, have such a well-developed toolkit with which to analyze financial shocks. The events of the past year certainly have reminded us of this point. So I would like to discuss some of the difficulties we face in modeling financial shocks, and do so through the prism of the recent turmoil in financial markets.

Q: Okay. As I said at the outset, I'll let you set the roadmap, so what particular questions do you want to address?

Let's think about some of the tough questions I am sure all of us have been grappling with over the past year. Can markets easily work around the disruptions to the credit intermediation process that channels funds from savers to borrowers? Or have we experienced a permanent destruction in something we might want to think about as the financial sector capital stock? If so, does the economy need to develop entirely new infrastructures for some types of intermediation, or perhaps will reviving earlier more traditional approaches suffice once the turmoil has abated? And how will we know when markets settle out to the "new normal steady state?" With regard to policy, what—if any—public sector actions can assist this transition? Finally, and most importantly, what are the consequences of this turmoil for the achievement of the Fed's ultimate policy goals—maximum sustainable growth and price stability?

Q: You weren't kidding when you said the crisis raised a few questions. What is missing in existing models?

We need to be able to model how various disruptions in financial markets affect credit provision, and how these effects, in turn, impact the real economy. Ideally, we would want to analyze these financial developments using a well-articulated and empirically implementable modeling framework. Executing this program is tricky business. And we still have some work left to do—first in better understanding the fundamentals underlying the current financial shocks and, once we do, in figuring out how to best adjust our empirical models to measure their influence on the macroeconomy.

Q: I assume you tried inserting some measure of financial market turmoil, TED spreads, something like that, into the standard framework?   

At the Chicago Fed, our initial response to the financial turmoil was to try to classify the financial crisis as a shock that could easily be inserted into a standard DSGE model.

However, deciding on the appropriate characterization of the shock has turned out to be a frustrating exercise. Modern macro models have identified and estimated a number of "standard" shocks—shocks to total factor productivity (TFP), marginal rates of substitutions, monetary policy, and so on. But less progress has been made with regard to modeling a "financial infrastructure shock."

Q: How did you try to overcome this?

The first step is to look at abstract theoretical models. These have shown that asymmetric information, moral hazard, and agency issues can generate an important role for liquid assets in channeling credit between lenders and borrowers.

One natural starting point is the model of Diamond and Dybvig and the large literature that followed it. These papers consider the problems that information asymmetries pose for the functioning of primary credit markets, the role of liquidity in addressing such problems, and the possibility of speculative runs in such environments. Accordingly, these models are helpful for thinking about the current situation, in which liquidity shortages have caused a number of markets to seize up and have resulted in run-like activity on some non-bank lenders.

Q: You say "starting point". That implies these models aren't sufficient straight out of the box.

There is a good deal that these models leave unexplained. They simply take as given the illiquidity of long-term investments that are traded in primary capital markets in which borrowers and lenders interact directly. They do not explain illiquidity per se. As such, they fall short in explaining the current turmoil, which in large part reflects the sudden illiquidity of some assets that used to be re-traded in very deep secondary markets.

Q: Is that the only model that can serve as a foundation for thinking about these issues?

Recently, Kiyotaki and Moore developed a model in which credit frictions operate not only in the primary market but also in secondary markets. They trace out some implications for consumption when liquidity seizes up, and have some interesting policy prescriptions. That said, they too take as given the degree of liquidity: The shock to liquidity is a primitive in their model, with no explanation why some assets that previously were very liquid suddenly stop being traded. Without knowing this, it is difficult to think about the full effects of the shock on economic activity or the way central bank policy ought to react.

Q: So existing models don't help much?

I think there are some models that give plausible explanations for certain aspects of the current shock to liquidity. One stems from Akerlof's "lemons" paper, which studies markets under adverse selection.

Q: The adverse selection is, I presume, from asymmetric information on security values. How does this relate to the present crisis?

As Akerlof showed, adverse selection can generate a vicious cycle that could even lead to a complete market shut down: This is because sellers of high-quality securities drop out, and this depresses the market price and leads more sellers to drop out, thus setting off a continuing cycle of price declines and exit of high-quality assets.

I think this story begins to approximate some of the malfunctions in the mortgage-backed securities market. But it leaves open the question of why liquidity dried up in markets that were not tied to housing.

Q: Do you have any ideas about why liquidity dried up in all markets rather than just in housing?

One reason may be that the securitization process was similar across a wide spectrum of financial liabilities. This may have led market participants to worry about the potential for similar lemons problems to emerge among a much broader class of asset-backed paper, generating contagion across financial markets.

Q: So how would you summarize this class of models?

While very useful for organizing our thinking, these models provide only highly stylized characterizations of how liquidity problems arise, and they fall short on modeling how they might feed back onto the consumption and investment decisions of households and businesses.

Q: Can we overcome these problems?

The first question to ask is whether we can make some straightforward modifications to an otherwise standard-looking DSGE model and obtain useful policy prescriptions. There have been some important efforts to do so in the literature.

Q: Such as?

A financial market shock may have some characteristics of a DSGE TFP shock: The cost of producing an intermediate input—credit intermediation—has become more expensive. This is how the shock is modeled in recent research by Curdia and Woodford. In their analysis, a negative financial shock increases the costs of intermediation between savers and borrowers, and increases the spread between the interest rates faced by the two groups.

Q: Are there any other approaches to overcoming these problems?

Alternatively, there is a long history of empirical studies of the effects of credit frictions on output and inflation. Here I am thinking about the VAR and other reduced form models meant to capture the influence of the financial accelerator and the credit channel for monetary policy. Subsequently, some attempts have been made to incorporate these effects into more structural models.

Q: There is quite a bit of work along these lines. Is it helpful?

These studies are very important and provide us with a very useful perspective. Yet they are identifying financial shocks largely in a reduced-form manner—either as an exogenous increase in interest rate spreads or net worth or as a shift in the demand for capital due to tightening of credit market constraints. To capture the current situation, we need to have a better idea of how the events in financial markets map into the parameters and shocks identified in these models.

Q: What's your view on this? Can that be done?

My feeling is that this probably requires embedding a richer specification of the financial intermediation process into these empirical frameworks.

Q: What do you mean?

There are dimensions along which financial capital is not like the typical intermediate input in a DSGE model. In a number of theoretical models, financial inputs can be subject to discontinuities: Seemingly small changes in the borrowers' net worth or perceived risk can bring about radical consequences in market performance. Perhaps the recent events in the mortgage-backed and related structured security markets are an example of such a breakdown—indeed, one that is of macroeconomic importance.

Q: If discontinuities are a key element in theoretical models, how do you model the jumps empirically?

Such discontinuities are a challenge for our standard empirical models. These models are typically solved by taking a local approximation around a balanced growth path. This approach is not designed to capture discrete jumps in equilibria. In addition, some of the estimated coefficients in our standard models may not be invariant to the onset of a liquidity crisis.

Q: Does that point us in a particular direction?

This suggests that a regime-shifting model could be helpful, with a "normal" regime alternating with a "liquidity crisis" regime. But, still, how should we parameterize the crisis regime? If we observed liquidity crises often, we could estimate these parameters. But we don't experience crises very often, so it's asking a lot of any empirical model to accomplish this identification.

Q: What's your view generally. Are the models useful or not?

So my guess is that we cannot gain a good deal of insight into when markets will "return to normal," and what "normal" will look like from these models' estimated impulse response functions. Indeed, the rarity of crisis events and changes in the structure of markets makes it difficult to use even simple statistics, such as risk spreads and volatility measures, to gauge the path to the "new normal."

Q:  So both the theoretical and empirical models have yet to be fully adapted to incorporate the type of financial crisis we are experiencing. Without the guidance those tools provide, how do you do policy? What has driven policy prescriptions?

Since last August, the FOMC's policy decisions have been calibrated in part to avoid a "adverse feedback loop" between disruptions to financial market stability and the real economy. This focus has influenced both the setting of the funds rate and the implementation of a number of new policies.

In this context, the Fed has added a number of new lending facilities: the Term Auction Facility, the Term Securities Lending Facility, and the Primary Dealer Credit Facility.

In addition, on September 7 the Department of Treasury added a new lending facility for Government Sponsored Enterprises (GSEs). This is aimed at supporting the flow of credit to mortgage markets and the associated liquidity of mortgage-backed securities.

Q: Is there any theoretical basis at all for these policies?

Such policies do have counterparts in some of the work I cited earlier. As an example, in Kiyotaki and Moore's analysis, efficiency may be enhanced if the central bank offers more liquid assets to the private market in exchange for less-liquid ones. This strategy is potentially even more important if adverse selection is in play because here the intervention does affect liquidity. By allowing newly illiquid assets to be used as collateral, the central bank may be setting a floor on their value, which may undo—or at least limit—the vicious cycle that I described earlier in the Akerlof model. This could buy time for information to spread among market players and reinvigorate the intermediation process—perhaps in a different form, though, than how it operated in the pre-turmoil period.

Interestingly, our Term Securities Lending Facility, which the Fed implemented in mid-March, roughly corresponds to the Kiyotaki and Moore recommendations.

Q: It sounds like there is a lot of work to be done before we have good policy guidance for crisis like the one we are in.

We are left with many questions. How long should the central bank intervention last and how large should it be? What are its effects on the incentives for market participants to acquire information before a crisis unfolds? What markets deserve our attention? After all, Akerlof's original example was about the used car market; I certainly would not want the Fed to start accepting used cars as collateral!

Even once we answer these questions, we have to address the implications of the new policy tools for our ultimate policy mandates of promoting maximum sustainable growth and price stability. This includes accounting for the interaction between the new tools and our traditional instrument of monetary policy—the short-term interest rate. An important component of this analysis is thinking about the degree to which the impaired assets used to be close substitutes for what some may call "traditional money." If they were good substitutes, does a market breakdown that renders these assets illiquid create a potentially deflationary shock? If not, are there inflationary implications from swapping liquid Treasury securities for less-liquid assets or expanding the direct lending facilities that use such assets as collateral?

This question brings us back to the notoriously difficult issue of identifying the channels through which liquid assets generate price pressures, especially in the short run. And, as an empirical question, we are faced with identifying which credit aggregates or interest spreads best capture the relevant degree of substitutability.

Q: Looks like we are just about at the end of the time, I know you have a meeting soon, so maybe we could end with a general summary of where research effort needs to be directed.

To answer all of these challenging questions, it will be extremely useful to seek a more unified perspective of the role of financial frictions in modern macroeconomic models. In this way, we will be able to identify the independent roles of our traditional monetary policy tools and our new lending facilities, as well as the interaction between the old and new tools.

Q: Who should do this research?

Developing these models is up to the research community—from the research shops at central banks to academic departments to independent think tanks. And we policymakers are eagerly awaiting the results.

Q: Thank you.