Brad Setser

Brad Setser: Follow the Money

Print Print Email Email Share Share Cite Cite
Style: MLA APA Chicago Close


Nouriel Roubini gets a medal …

by Brad Setser
August 17, 2008

A well-deserved one too. Nouriel stuck to his core views — housing was massively over-valued, the financial system was heavily exposed to a fall in home prices and the fall out from a fall in US home prices wouldn’t be contained either nationally or globally – when those views were decidedly unpopular.

Back in early 2007, there was a great deal of complacency among America’s financial leadership. Many thought macroeconomic volatility had been vanquished, and as a result financial volatility was justly low. High levels of leverage consequently made sense — and a range of asset market prices reflected this. In the language of the time: credit markets weren’t over-valued, equity markets were under-valued. Recessions – or at least severe recessions and financial crises – were things that happened to other countries, not the US. The US had survived the .com bubble with only a shallow downturn. The 2003-2006 rise oil prices hadn’t put a big dent in the US economy. The large US current account deficit reflected high savings abroad and the attractiveness of the US financial assets; the US, after all, had a comparative advantage in financial-engineering. The IMF wrote that “innovative US fixed income markets [provided] many assets which simply aren’t available elsewhere” (see p. 12). There wasn’t much too worry about.

Read Michael Lewis’ argument that Davos man spent too much time worrying. He wrote in 2007:

Oil prices double, the U.S. housing market tanks — no matter what happens, financial markets adjust quickly and without hysteria. There are obviously a few things to worry about just now in the world, but the inability of traders to find a sensible price for the spread between European junk and European Treasuries isn’t one of them. So why do these people waste so much of their breath and, presumably, thought, with their elaborate expressions of concern?

Even the IMF – which is paid to worry – was tired of worrying. In late January of 2007, Chris Giles of the FT ran an article, based on an interview with the IMF’s Deputy Managing Director, that was titled “Big risks to global economy receding.” I thought that captured the mood of those times well.

Nouriel didn’t waver then. Others (myself included) did. Standing apart from the herd can be hard.

Over time, the focus of Nouriel’s concerns has shifted over time from the United States’ external deficit to the housing market and the financial system. But there has been a core consistency to his views: he never thought that it was healthy for the US to borrow heavily from the rest of the world to finance large fiscal deficits, high levels of consumption and lots of investment in suburban housing. And he thought this borrowing binge would end badly. Very badly.

Stephen Mihm notes that Nouriel’s book – which I co-wrote– doesn’t contain a single equation. That is true. I also think it is a little unfair. It was a book about responding to financial crises in emerging economies in the 1990s and the first few years of this decade. There were only a limited number of cases that could be examined. It didn’t take a huge regression to show that the IMF has extended large sums of money to countries with different levels of external debt and that in general, those countries with less debt paid the IMF back more quickly.

And I am not sure that it is possible build a formal model that captures all the dynamics of IMF lending to an emerging country. There is uncertainty about the amount of financing that will be made available. There is uncertainty about the policies that the country will actually adopt. There is uncertainty about the impact of those policies (a lot more uncertainty than there is over the impact of the US stimulus) that are adopted. There is uncertainty over how external bond holders will respond both to IMF financing – and to steps to limit the pressure private creditors can place on a country’s reserves through various rollover agreements and more or less market friendly debt restructurings. There is uncertainty over how domestic bank depositors will respond both to IMF financing, IMF conditionality and any debt restructuring. And all the developments in the crisis country can be overwhelmed by external shocks – like a big move in commodity prices or a major change in market conditions in the US or Europe.

It is possible to model IMF lending to an emerging economy. Nouriel co-authored a paper (with Corsetti and Guimaraes) that tried to model the IMF as a large player that can shape the dynamics of a small market. But the outcome of most such models seem determined by the assumptions used to create the model more than anything else.

I am biased, but I think it is possible to be analytically rigorous (and to use real data to inform your conclusions) even in the absence of a formal model.

Yale’s Shiller notes that Nouriel’s greatest strength is his capacity to synthesize an enormous amount of information:“Nouriel has a different way of seeing things than most economists: he gets into everything.” I wrote Bailouts and Bail-ins with Nouriel and I then worked for Nouriel at RGEMonitor – and I fully agree. The breadth of Nouriel’s interests — and his ability to synthesize information from multiple sources — is extraordinary.

I wouldn’t mind if Dr. Roubini was proved to be a bit too pessimistic, and not all the near-term risks he sees come to pass. But I also think it would be a mistake to base policy on the assumption that the worst of the credit crisis is over.


  • Posted by Tony Buzan

    Perhaps one reason that Roubini does not use equations is because he understands that not all of this pricing phenomena can be characterized by Gaussian distributions.

    That failure to account for Cauchy-type curves and their fatter tails is at the very core of what went wrong.


  • Posted by Felix Salmon

    And here’s me thinking he was just getting more media exposure. But I guess that’s the power of the NYT: going on the telly is par for the course, but getting your picture in the NYT Magazine is a medal.

  • Posted by J Mitchem

    Think of, say, 2005 — a scant 5 years after a brutal stock market crash and on the cusp of the entry of baby boomers into retirement. We saw the cancerous proliferation of homebuilding and the insane double digit skipping up of home prices, grotesquely large SUVs with 100% financing standard kit for people with mid-5 figure incomes, the mountainous government debt, the trillion dollar war, etc. etc,

    Well, statistics and formulas could actually obfuscate your understanding.

    The United States has been gorging on debt to spend on non-productive assets – oversized homes and automobiles, going out to dinner, brand-name socks.

    The flood of global capital, well documented in the statistics – allowed for all of this and actually created the illusion that it could go on.

    You know what you know – and macroeconomic/financial analysis actually was the facilitator of both the TMT bubble in the late 90s and the housing/credit bubble 10 years later.

    Tens of millions of Americans are heading into retirement with no financial assets and no health insurance.


  • Posted by Michael

    There’s only one formula needed to predict the course of our economy: aggregate debt/aggregate gdp. Ours has been rising for decades, as inflation of credit permitted the government, every business, and every household to expand its “wealth” and power, without inflating the money supply or raising real wages. And the current number provided by that formula shows how fragile we have let our economy become and how deep the fall can possibly be: In 2008 aggregate debt/aggregate GDP = 3.5 (350%).

  • Posted by Rien Huizer

    Any long time reader of the AER will agree that the formal (“rigorous”) aspect of economics seems to grow, topics get narrower and the real world seems farther and farther away.

    Formal approaches are nice, indispensable for describing complex mechanisms and linking theory to data. But economics is not physics (although that appears to be a favorite undergrad basis for economics), it is merely a social “science”, something to help us think about political and business problems that are far too complex to be captured by “tractable” constructs. Economics as a branch of applied mathematics is interesting, but so is the world of Bridge Conventions.

    Formal approaches are also quite dangerous (thanks, Tony). I do think that the naive use of models (or worse, cynical use of things with possible, dangerous defects, by opportunistic agents with poorly designed incentive structures) is an important contributor to the current credit market problems. Things like illiquidity, market breakdown, shifting correlations, etc were all well known to everyone with market experience, but perhaps not to the quants who designed these models.

    So, I applaud books that can be read intuitively. They may not be rigorous but can be important. And if it takes a little longer to analyze the argument, one can alwyas plug the intuition in one’s modeling package and look for insconsistencies. In this case, that may not be the easiest way to find those. Thinking without a keyboard and screen. Refreshing.

  • Posted by Bernardo A

    “…if one keeps predicting a market decline (or rise) one is bound to be correct at some time” -Malkiel, A Random Walk Down Wall Street.

    I am really looking forward to see what dr. Roubini (who is also in part a glorious representative of the great Italian school) will be writing about in the future. Will he be able to spot out the next Wall Street’s great con as soon as the market will start recovering again?

    What Malkiel wished to conceive in his book was that being a chartist (i.e. technician) is a trivial job to the extent that you keep telling people to always sell or always buy (at the end, you will be right sometime). But this is not what economists do. Some economists believe in the presence of long-run fundamentals. And one of the most challenging jobs deriving from this conviction is to spot out what is behind temporary shifts from the long run trend. Ideally economists should do so while the bubbles are occurring. This is not always easy, but I in my view this round dr. Roubini has managed brilliantly.

  • Posted by C. Michael Reilly

    Brad Setser’s post on Nouriel Roubini getting a medal ( the NY Times Magazine article being ‘the medal’ referred to I assume ) raises some issues that need to be treated carefully.

    Brad defends Stephen Mihm’s observations about his book not containing a single equation by pointing out the lack of data and then goes on to imply that ‘equations’ are associated with a ‘huge regression’. What needs to be treated carefully here is the implied assumption that an ‘equation’ is something that needs data to be useful. E= MC^2 is an equation Einstein’s amazing mind came up with that combined with Monte Carlo simulation experiments led to the design criteria ( design assumptions ) needed to create an atomic bomb. Could this dubious yet amazing achievement have been accomplished without the equation? If the equation did not exist, why would anyone have done the Monte Carlo experiments? Einstein’s equation was an assumption about behavioral relationships in the physical world. What needs to be treated carefully here is to not dismiss ‘equations’ as data dependent notations that are often thought of by many as not intuitive when what they are is a way to represent assumptions about behavior in the physical AND social science realms. People often say something is not intuitive when it represents a concept they do not readily grasp, when it represents a way of thinking divergent from their own views or when it presents a concept that is not aligned with their agenda. Yes, equations need data to be verified either with repeatable experiment outcomes or with historic time series but that misses the point that the data are only given relevance and substance with equations that represent relational/behavioral assumptions. Relating ‘equations’ with a ‘huge regression’ is perjurious to the notion of an equation to anyone who has studied statistics and has tried to profitably use them to help make decisions. Yes lots of people study statistics and use ‘huge regressions’ but only to sell them to generate revenues by serving as guides to people so desperate for views on the future that they don’t bother to seriously consider the foundations and substance of those guides.

    Brad continues with the following sentence: “And I am not sure that it is possible build a formal model that captures all the dynamics of IMF lending to an emerging country.” Brad then highlights a number of uncertainties that need to be addressed when building a “formal model” but then follows with: “It is possible to model IMF lending to an emerging economy. Nouriel co-authored a paper (with Corsetti and Guimaraes) that tried to model the IMF as a large player that can shape the dynamics of a small market. But the outcome of most such models seem determined by the assumptions used to create the model more than anything else. “ While I am not sure what the distinction between an emerging country and an emerging economy is, what concerns me is the last part of what he says: ”But the outcome of most such models seem determined by the assumptions used to create the model more than anything else.“ What needs to be treated carefully here is a dismissive attitude with respect to the model assumptions. What can a model’s relevance, substance and usefulness be based on if not its assumptions ! Recall E = MC^2 ! What are data without assumptions ?

    Brad concludes with :” I am biased, but I think it is possible to be analytically rigorous (and to use real data to inform your conclusions) even in the absence of a formal model.” Brad then observes that Nouriel’s strength is his capacity to synthesize enormous amounts of information. So how does Nouriel’s capacity to synthesize an enormous amount of information manifest itself in his writings ?He lists concerns, discusses their relationship with potential outcomes and discusses behavior that links his concerns and outcomes. Isn’t the most unambiguous way to accomplish this done via describing each of the concerns and outcomes as behavioral relationships in equations? Otherwise, how can one assess all the uncertainties Brad cites about formal models? What needs to be treated carefully here are the risks attendant in using ‘analytically rigorous’ analysis that uses ‘real data’ to inform your conclusions when the underlying equations are left in written language form, hence potentially ambiguous and incapable of the type of uncertainty analysis Brad himself cites.

    I wouldn’t want to be on an airplane, bridge …, or nuclear submarine that wasn’t designed using equations. Yes, the social sciences are not physics so the social sciences equations will need to factor in lots of uncertainty … that might make selling services to people desperate for views on the future tougher but we’d all have a smoother ride.

  • Posted by Hal

    It required mere common sense to know that excessive debt is dangerous. But that lesson was blithely ignored. One doesn’t need equations to know that. Roubini was unique in applying common sense to economics amidst all the reliance on math, etc., to do foolish things. (And of course Cheney’s “Deficits don’t matter” will be another stupidity that will in time come back to bite the USA).

  • Posted by bsetser

    C. Michael Reilly.

    interesting points. i wasn’t precise enough — some equations (as you note) can be very valuable. especially equations that yield testable predictions.

    my sense tho is that there is a lot of economics that dispenses with a lot of uncertainty that is intrinsic to complex systems in order to yield cleaner theoretical results. and those results are only as good as the assumptions underlying those results.

    and there can be a tendency to rely a bit too much on past patterns to infer the future. housing prices never fell nationally for example. that seems to have been true. but it also was true that until recently most mortgages were financed by a local s and l and there wasn’t a national housing market in the way there is now.

    one of the things that I find interesting is that one of the more robust empirical relationships is that a fall in the real exchange rate boosts exports and reduces the trade deficit. yet there are loud voices arguing that exchange rates don’t matter –

    whether b/c they don’t like exchange rate adjustment
    or b/c they claim that the trade balance is driven by the savings and investment imbalance and thus independent of exchange rates.

    and this is a case where the bridge designers, so to speak, would look at an equation and conclude that exchange rates do have an impact.

  • Posted by Rien Huizer

    MIchael, your engineering analogy rules out most applications of prescriptive economics. A lot may have been developed using formulae, but there are (fortunately) no nice, consistently right economic prescription robots, as there are off the shelf solutions for bridge design..Too many conflicting theories all with some plasusibility, lack of knowledge about factor relevancy for intended and unintended outcomes and so on. How would you structure uncertainty in a prescriptive model for say, Sudan, in a state of civil war, or Afghanistan? There is no way to model the strategies of the actors you think may be relevant, there is no way to test therories. In other words you might end up with a signal to noise ratio of close to zero after years of work. What prescriptive economists do is (a) use their preferences -probably ideoplogically biased), (b) see if the situation shows similarity to well documented precedents (c) talk to the people involved (d) if nothing else seems to be better, ram a lawyer-approved cookie cutter down the victim’s throat, because that will not get you fired. Apologies to prescriptive economists..

  • Posted by Andy

    Roubini has no credibility among those who take risks and positions, because he cried Wolf throughout a monumental boomtime. Sure, the party eventually ends, 3-4 years after he says it will, and he is a soothsayer. But Roubini Capital Partners would have closed up shop sometime in 2005/06. He wins plaudits as a pundit/economist, but those who successfully timed the market, like John Paulson, are far more impressive than someone who was dead wrong for three years.

  • Posted by C. Michael Reilly

    To; bsetser;

    Thank you for sharing your thinking on my ‘defending equations’ post.
    I am a student of the financial markets and use equations to try to understand
    how all the forces acting on markets cause it to evolve the way it evolves.
    As such, I am sensitive to those that put a negative spin on
    using simulation models constructed from equations to understand
    financial market behavior better. Now that I have reflected on yesterday’s post
    I realize more than anything it was Tony’s post that justified avoiding equations
    because of an awareness that normal distributions don’t reflect lots of observed
    phenomenon that tipped me over into spending the time to write and submit the
    ‘defending equations’ post. The fact is that the assumptions underlying ANY probability
    distribution based only on data and not based on one or more behavioral equations is
    to me not really telling you anything about the forces that brought that particular
    distribution about. So what really bothered me about Tony’s post was that he was condoning not using equations because probability distributions derived from data don’t rely on underlying relational/behavioral equations; this to me is the key problem with how risk management is done on Wall St. Just because many
    derive conclusions about risk on data alone without attempting to model any underlying structural relationships in an attempt to model behavior is to me not a reason to condone not using equations.

    I think both you and Nouriel model economic problems in your own way better than most; otherwise I wouldn’t spend lots of time reading your blog posts that shed light on the most challenging economic issues of our time. I am glad you got something from the ‘defending equations’ post. Good day.

    To Rien Huizer;

    Rien asks, “How would you structure uncertainty in a prescriptive model for say, Sudan, in a state of civil war, or Afghanistan? There is no way to model the strategies of the actors you think may be relevant, there is no way to test theories. In other words you might end up with a signal to noise ratio of close to zero after years of work.”

    A fair question; I’d have to understand the institutions that exert power in the region, their goals, their
    time horizons for getting results to the pressures that are making them act the way they are acting. I would have to understand what people make up these institutions, how they came to power, how they maintain their power, what the rules of these institutions are. They may not have a Constitution but they have cultural norms and a perspective about themselves and others that could help shed light on why they act the way they act. Is it a fight for petrodollars in various regions where various factions cannot resolve the resource allocations non-violently? Given this environment, what prescriptive actions could lessen the violence?
    If I took on this problem and tried to model it using equations I would very likely as you say end up with a signal to noise ratio of zero for any modeled policy responses (prescriptive actions.). You cite a problem that needs bold assumptions, bold assumptions about the actors’ behavior and WHY they behave the way, they do. It has been said that if ancient Greeks could see the World today that they would be amazed at all the technology and would not understand it, but they would completely understand what’s going on in a session of Congress in the USA or in the dreadful situation in Darfur. To say, “There is no way to model the strategies of the actors you think may be relevant, there is no way to test theories.” underestimates your and humankind’s potential to address pressing problems. If some group of people think they can address such problems using equations and simulations that try to model the institutional and involved individuals’ behavior, then the approach and any prescriptions should be scrutinized and vetted but they should not be dismissed out of hand.

  • Posted by Martin Margolis

    Much of the NYT article builds on the Roubini IMF presentation where he was viewed as a “permabear.”

    Reading the actual IMF transcript of what happened at the presentation offers a different story:

  • Posted by Rien Huizer


    Thanks, we seen less far apart than I thought. My grudge is mainly with people who use models opportunistically, and with people who think economics is an arena for showing off mathematical skills to potential competitors for tenured positions. I think a good economist should certainly be able to present an argument in a formal manner and where possible strive for empirical validation, requiring models with a manageable quantum of components.
    Just take a look at the March 2008 issue of the earlier mentioned AER ( 2008, 98:1, 267-293). Something every successful third world politician knows (or else he would not be around anymore) thugs beat rules when they get the chance. It is interesting, not easy to read and unsurprising, except for the naive. Good economics means being able to present economic thought in a way the public will understand, because it is not a bloody bridge, it is changing, fickle and ingenious human beings who make an economy and who need to be educated to be responsible citizens.

  • Posted by C. Michael Reilly

    re: reply from Rien Huizer;

    I hear you, your points are well made.
    Good day.

  • Posted by bsetser

    C. Michael Reilly

    Thanks for your reply; my initial comment (and yes, you were responding to Tony, but also to me) was motivated by a sense that:

    a) a lot of formal modeling is done without ever really looking at data. I assume X, Y and Z. I think those assumptions are relevant to the world (an often unstated assumption). Those assumptions and a bit of math yield an interesting theoretical conclusion. For example, IMF lending has no impact …

    b) on the other hand, a lot of analysis of past data is done with limited amount of thinking about what might have produced the outcomes that emerge from the data.

    it is for example possible to construct a national housing market index. but that doesn’t mean that the housing market in 1970 is like the housing market in 2005. I remember well that missouri banks weren’t allowed to operate in Kansas back then. The big issue bank regulators worried about was regional concentration. Texas banks had a lot of energy exposure, and suffered when prices fell in the 80s. National banks with more diversified balance sheets were the answer — but they also helped to nationalize (together with mortgate brokers) a change in the lending standards nationally that helped create the conditions for a more correlated national housing market.

    another example is that the losses on subprime loans were lower than expected in the last recession, largely b/c housing prices didn’t fall during the recession to the underlying collateral remained good. and since subprime was a fairly recent innovation, there wasn’t a data series that stretched back far enough to capture a housing market recession. net result: I think some risk models concluded that the risk on subprime had been over-estimated, and that it was a better bet than people thought …

    but then behavior started to change based on that assumption, and the conditions that produced the lower than expected losses dissipated.

    my guess is that we do not fundamentally disagree — using data to make judgments about financial risk and reward is hard.

  • Posted by bsetser

    Andy –

    Paulson might have just gotten lucky. Getting the timing exactly right is hard. But yes, he called it right, and Nouriel was early.

    I know this well. I used to argue that RGE shouldn’t send out an email every week about nouriel’s latest prediction that the recesion was coming.

    I still think Nouriel deserves credit for two things:

    a) he didn’t waver; the imbalances that he believed would drag the economy down didn’t correct, so he got more worried over time not less even tho his initial forecast was early
    b) he was almost alone (and right) in arguing that the financial sector would be hard hit by the combination of falling housing prices and a slowdown, and that all the new fangled securitization technology had dispersed risk to those able to bear it and the us financial system wasn’t well capitalized. on that he has been proved right — and that is big. sure he was early. but, to be a bit crude, a lot of his critics would be bust now if not for unprecedented US government intervention to support the financial sector. and i do think the regulators should have spent less time talking about the wonders of risk dispersion and a bit more time trying to find out who exactly was holding housing market risk.

    note the points i made about complacency in the post. Senior folks in big banks were criticizing roubini as a scare mongerer who didn’t understand the resilience of the us economy and us financial system. the IMF was going to reorient itself towards helping with growth rather than worrying about risks. and so on …

  • Posted by Cedric Regula

    Hi, I’m an engineer by training and we do lots of math stuff in my field, but @$@$^^%$#@#%^&**((&^%%%^^^^&&**(($$#$#@

    sorry, my keyboard just blew up…had to plug in a new one.

    But as I was saying, I find it tremendously funny that economists think they can use math in their field.

    Not to say that a competent understanding of simple arithmetic wouldn’t be useful…like when calculating government budgets, and government statistics could use some dumbing down, and structured finance would work better if 20% down mortgages to people who have had their incomes verified by the lender was what we are buying to get an inflation matching yield.

    But I’m perfectly happy with Professor Roubini assimilating large amounts of information and then giving us his “gut feel”.

    Also, you can’t get the timing right on macro predictions when you don’t know how many trillions in external financing you can get !

  • Posted by Ken

    Re: Andy’s remarks. Yes, perhaps *investors* wouldn’t have profited significantly from NR’s predictions the past several years. But, just as Reagan famously introduced the question, “Are you better off now than you were 4 years ago?,” we can ask whether the *country* and the *world* wouldn’t have been better off today had it taken account of NR’s warnings then. This is an issue of far greater urgency than mere market timing (when’s the next recession?): this was accurately predicting potentially catastrophic conditions. (Mohamed El-Arian comes very close, in a recent FT article, to suggesting that there is now some (small) risk of world wide depression if countries and SWF’s do not act rationally and in concert.) Macro economists have a loyalty to things greater than Wall Street.