Streetwise Professor

In the midst of the credit crisis, it has become conventional wisdom that OTC derivatives, especially credit derivatives, should be centrally cleared. Like Dudley Dooright, the central counterparty (CCP) will come in and save the day. Given that this idea is supported by many people who had probably never heard of clearing until quite recently, there is room for doubt that these high expectations are grounded in a firm understanding of what clearing is and what it entails.

I will explore various aspects of OTC clearing in several posts, of which this is the first.

The motivation behind these posts is a simple question: If clearing is so obviously wonderful,why has it not been adopted before this?

I think there are two answers to this question, answers which are not mutually exclusive. The first is that clearing would be welfare enhancing, but as a result of some collective action problem it has not been implemented. The second is that there are costs and benefits to clearing, that these costs and benefits vary by product, and for some products, the costs exceed the benefits.

The public policy implications of the two alternatives are obviously quite different. The first provides a justification for robust government intervention: the second does not. I think that both factors have an element of truth. This implies that there is a basis for government involvement, but that considerable care should be exercised.

I have explored both questions in my academic work, and will draw upon that work in drafting these posts. I have recently done more work on the collective action issue, and will report the results of that analysis for the first time in the blog; I will write up the results of the analysis more formally in the coming weeks.

This post focuses on the collective action issue. If clearing is welfare enhancing, but it has not been adopted due to transactions costs/collective action problems, then arguably a case can be made for regulatory or legislative initiatives to encourage–or even force–its creation. Indeed, one can argue that the NY Fed’s frantic efforts in recent weeks to push market participants towards creation of a CCP reflects such a belief.

I first mooted the possibility that CCPs were not adopted due to some sort of collective action problem in a working paper in 1997. I also mentioned this idea in a more recent paper. The basic idea was that clearing could have distributive effects, most particularly, redistributing wealth from well capitalized firms to less well capitalized ones. This could occur if clearing leveled the competitive playing field by eroding the competitive advantage held by well-capitalized OTC dealers.

In my earlier work, this argument was made verbally. It is not an easy thing to formalize it, but I have made some progress in that direction. The model is a relatively simple one. There is a representative agent hedger, with CARA utility. The hedger has an endowment of a risky asset. There are two risk neutral dealer firms. These dealers are risk neutral, and have no endowment of the risky asset, but they have limited capital. Their capital is also risky. Therefore, if they enter into derivatives contracts, they may default as their losses on their derivatives positions may exceed their capital.

The two dealers have different starting capitals. There is a large dealer and a small one. The volatilities of the firms’ capitals can also differ. The dealer capitals are imperfectly correlated.

The dealers incur operational costs to execute derivatives transactions. For simplicity, these costs are quadratic in the quantity of contracts executed. Although the dealers have different capitals, I assume that they are equally operationally efficient. That is, their costs of execution are the same for a given number of transactions.

It is well known that a defaultable derivatives contract, such as a forward or swap, is effectively an option. In this model, the hedger that buys a forward contract from a dealer actually enters into contract with a payoff equal to the minimum of the payoff on a non-defaultable contract, and the counterparty-dealer’s capital.

Due to this optionality, the problem is inherently non-linear, and even assuming CARA utility, must be solved numerically. Therefore, since that nettle has already been grasped, there is no need to make the drunk-looking-for-his-wallet-under-the-lamppost assumption of normality in the distribution of the price of the risky asset, or the values of the dealers’ capitals. Normality facilitates the determination of closed form solutions in linear problems, but since this problem is non-linear that is of no help. I therefore utilize more realistic assumptions. I allow for heavy-tailed and skewed price distributions, by assuming that returns on the risky asset are characterized by a negatively skewed normal inverse Gaussian distribution. (This means that crashes are more likely that upward spikes in prices, as is the case for equity indices.) I assume that dealer capitals are characterized by fat-tailed Student-t distributions. Skewness and kurtosis are realistic features of real markets, and may provide an important reason to hedge and affect the likelihood of default. The world is not normal, and non-normality is especially important in this context.

I first solve for equilibrium in a bilateral OTC market. I assume that all parties are price takers. The resulting equilibrium is quite sensible. The hedger trades more with the firm with the greater capital, and sells at a higher price to the less-well capitalized firm. That is, since the deals are negotiated individually, and since the counterparties are heterogeneous, the terms of trade and quantities of trade differ. The hedger sells more at a lower price to the advantaged well-capitalized firm.

I next solve for equilibrium in a cleared market. In the cleared market, the two dealers form a CCP and share default risk. If one firm defaults, the other party covers the loss, up to the amount of its capital. The hedger suffers from a default only if the dealers’ collective losses on their trades is larger than the sum of their capitals.

The details of possible sharing arrangements under a CCP are complex, and potentially varied, so I simplify matters by assuming that the CCP allocates trades among the two so as to maximize joint profits. This would require some sort of pricing mechanism, or some other means of allocation, which is not modeled specifically. I do turn attention to the possible pricing issues later.

Due to the equal operational efficiency of the two firms, the efficient outcome is for the output of the two dealers to be equal. In the clearing equilibrium, (a) the hedger trades more, (b) the output of the small capital dealer goes up, and (c) the output of the large capital dealer goes down. Moreover, the price at which the large dealer buys goes up slightly, and the price at which the small dealer sells goes down. The hedger’s expected utility rises, and joint profits of the dealers are higher because (a) total output is higher, and (b) total costs go down due to a more efficient allocation of output across the dealers. There is also a more efficient allocation of risk, and the incidence of default falls, even though the scale of trading rises, because of the diversification of the dealers’ capital risks.

In brief, the formal model, solved numerically, generates results along the lines of what I laid out informally in the earlier working papers.

The implications are quite striking, and important. Implementation of a first-best risk sharing mechanism improves welfare (obviously) but redistributes wealth from the large dealer to the small dealer. The small dealer’s smaller capital makes him a less desirable trading party for the hedger, even though from a social perspective diversification and operational costs make it efficient for the hedger to split his trades between the two dealers. That is, from a welfare perspective, pooling of the dealers’ capitals is efficient as it facilitates efficient allocation of output across dealers. (This raises the question of why don’t the dealers just merge? Well, merger of all the dealers into one would effectively create a clearinghouse. Moreover, there may be sources of diseconomies of scale and scope not captured in the simple, formal model that mean that sharing of risk via a clearinghouse is a more effective way of pooling capital for default mitigation purposes than a formal merger.)

Since the joint profits of the dealers are higher under clearing, the small dealer’s profits rise more than the large dealer’s profits fall. In the absence of transactions costs, the dealers would execute a Coasean bargain that would make both better off, and one would expect the CCP to form.

However, it is well known (see particularly Libecap’s work–some done with Steve Wiggins–on the creation of property rights in natural resources), that distributive effects can interfere with the creation of efficient property rights. The common pool problem in oil (which I discussed in earlier SWP posts) is a classic example of this. Even though unitization of oil fields–the efficient allocation of output across potential producers–maximizes joint profits, it has proved devilishly complicated in practice for competing drillers to consummate this wealth-enhancing bargain, due to the tussle over rents.

Moreover, due to the complexities in bargaining, and the potential costs of enforcing any arrangement among the firms, the dealers may arrive at a pricing or trade allocation structure that limits the distributive impact of the formation of a CCP. In the stark model, however, any such deals reduce welfare. Thus, even if a CCP is formed, it may not achieve the efficient outcome. Alternatively, the feasible bargaining outcomes may not include the efficient outcome. If there are fixed costs to implementing the CCP, it may not form.

In particular, as I have shown in my work on exchanges (JSTOR access privileges required), coalitions of heterogeneous firms create the potential for redistributive rent seeking. As a result, organizational form and governance structures for coalitions of heterogeneous firms are usually crafted to mitigate the potential for such redistribution. The resulting structures and pricing mechanisms may not be first best.

It should also be noted that even though CCPs are typically formed as coalitions of dealers, some of the benefits flow to (in the model) hedgers. That is, CCPs are usually coalitions of sell-side firms, but the benefits of CCPs spillover to buy side firms. Again, since the CCP is first best (in the model), there is a potential grand bargain that makes the buy side and the sell side firms better off. This only expands the bargaining complications, however.

Thus, one explanation for the failure to adopt a CCP in OTC derivatives is that even though it is efficient, collective action and bargaining problems have impeded, and in some cases precluded, its creation. Distributive effects can preclude the adoption of efficient organizations and institutions. The model shows that the adoption of clearing can have pronounced distributive effects. Therefore, the failure to adopt clearing may reflect the effects of distributive effects and rent seeking behavior that impede collective action.

In this situation, government action (and essentially, government can serve as a mechanism to reduce the costs of collective action) can improve welfare to the extent that it reduces transactions/coordination costs.

In essence, government is just another mechanism through which parties can act and bargain to achieve a given outcome. Government possesses coercive powers. Moreover, especially in democratic/republican polities, a broad array of interests can influence/bargain through the government. This can be important, and sometimes result in enhanced efficiency–though it can reduce efficiency at other times. In the present instance, for example, buy side firms that would benefit from the formation of a CCP can exert influence on various government entities (the Fed, Treasury, the legislature), and beseech the government to employ its coercive powers to advance the interests of these firms. It may be cheaper for these firms to bargain through the government, than directly with the sell side firms.

This analysis reflects the strongest case for government intervention to, ahem, encourage the formation of a CCP for credit derivatives, and other products traded OTC. But one must be careful. Clearing is a risk sharing arrangement, and it is well known that risk sharing arrangements are often plagued by adverse selection and moral hazard costs. These costs can vary depending on the type of risk being insured. These costs reduce the amount of risk sharing, and in some cases, can make it better not to share risks at all. Thus, the failure to form a CCP that shares default risks for some products may also be due to the fact that moral hazard and adverse selection costs are so large as to make risk sharing inefficient. I will consider that possibility in the next post, and argue that this possibility is particularly pertinent for products like credit derivatives. Thus, the benefits of the formation of a CCP for these types of products may be substantially smaller, and the case for government intervention weaker, than the collective action problem-focused analysis of this post suggests.

Dr. Roger Pielke has resumed his excellent Climate Science blog. It is particularly refreshing in the aftermath of all the Bali-hooey to read succinct commentary from someone who recognizes that the science is almost never settled, and hence keeps his mind open to all the evidence. His recent posts on the lack of evidence that water vapor in the atmosphere is increasing (as the climate models predict it should) are particularly illuminating. The positive feedback from water vapor is where the real temperature increasing action in the climate models comes from; the absence of parallel increases in water vapor and temperature is inconsistent with the predictions of these models. In real science, this is called “rejecting the hypothesis.” In global warming “science” this is called an inconvenient truth that shall not be spoken of.

Anyways, it is good to see that Dr. Pielke recognized the importance of his voice and his viewpoint, and is using the forum of his blog to continue to disseminate his science-grounded approach to a broad audience.

The Nobel Peace Prize is supposed to be about, you know, peace. In awarding this year’s Nobel to Al Gore and the IPCC the prize committee had to pile one mass of tenuous speculations (about how climate change would lead to widespread violence and strife) on top of another (whether climate change will occur and how it will occur) in order to justify its choice. This is a travesty in a world filled with real violence where one needn’t speculate about its proximate causes, and where men and women are risking their lives every day in order to stop this violence or ameliorate its consequences. All Al Gore risks in his jetset Jeremiad is jet lag. And perhaps death by exploding ego.

I read with regret that Dr. Roger Pielke is ending his excellent Climate Science blog. I have learned a great deal reading it over the past couple of years. Dr. Pielke’s last post provides an excellent summary of his main conclusions, which are quite reasonable and persuasive. Most notably, Dr. Pielke argues very cogently that a monomaniacal focus on CO2 as the only important anthropogenic climate forcing is counterproductive. Moreover, Dr. Pielke’s criticism of average global temperature as a climate metric is tremendously important and deserves more attention. Temperature is not a conserved quantity, and temperature measurement is rife with errors arising from poorly sited instruments; Dr. Pielke has shown very convincingly that daily minimum temperatures are a particularly flawed measure of climate trends.

Dr. Pielke is also very measured and reasoned in his approach, and is the quintessential scientist. His sober and scientific tone couldn’t be more different from the hysterical millenarianism of GISS’s James “Jeremiah” Hansen. Hansen has been over the top for literally decades, but his stridency and evangelical fervor have reached new extremes in recent months. Rather than honestly acknowledging errors in his data, Dr. Hansen viciously attacks those with the temerity to point them out, and accuses those who disagree with his True Belief with callous indifference to the fate of the planet. [As an aside, Hansen’s dismissal of the irrelevance in the correction of his US data is incredibly lame. If there is such a thing as “global temperature,” temperature one large region (e.g., the US) should be a good proxy for the global temp, so a lowering of temperature in recent years in the US is a bigger deal than Hansen is willing to admit–if Hansen is right about global mean temperature as the best metric of climate change.]

The blogosphere will certainly miss Dr. Pielke’s reasoned voice. He will certainly continue making his prolific contributions to the scientific literature, but more popular media need to hear his views too. We know that Jeremiah Hansen will continue to use the popular media megaphone. Who will make the more reasoned scientific view more accessible to the public? Dr. Pielke, you will be missed.

Roger Pielke Sr. has a very illuminating blog entry on climate modeling vs. weather modeling. A common objection to long range climate prediction is “If we can’t predict next week’s weather, how can we predict the next century’s climate?” The stock climate modeler response is that climate is stationary, so we can predict future averages (where averages are taken over relatively long periods) with considerable accuracy. The short run vicissitudes of weather around long run means are effectively averaged out when forecasting climate.

Pielke notes, however, that this is only part of the story, and arguably not a very important part. The “stationary” part of climate models is in the basic physics, but this represents only a fraction of the forces that affect climate. Climate also depends on numerous other processes and feedbacks. Climate models “parameterize” these processes. Unfortunately, some of these processes are very poorly understood (e.g., clouds, just to name one), so the parameterizations are crude. Moreover, these factors can interact in complex–and poorly understood–ways. Even if the climate system is indeed stationary, the relevant model may be badly misspecified due to incorrect parameterizations. This can lead to biased projections. Even if the vicissitudes in weather indeed average out, the models may be waaaaay off, because the estimate of the mean can be biased due to biased parameterizations. That is, in the presence of potentially large uncertainty in the parameterizations, it is a non sequitur to assert that stationarity implies that climate forecasts are more accurate than weather forecasts. Large model errors can swamp the effects of stationarity.

Dr. Pielke deserves plaudits for pointing this out. He does so in an evenhanded, thoughtful, and fairminded way. (In stark contrast to many in the climate sciences–at least many of those who have a high public profile.) He seems to be a serious scientist–one who acknowledges anthropomorphic effects on climate, but who skeptically evaluates competing claims and evidence. Would that more were like him.

My post on the Antarctic temperature research neglected to mention a very interesting book by Marcel Leroux titled Global Warming: Myth or Reality. Leroux’s theory of Mobile Polar Highs can perhaps explain some of the discordant data from the polar regions. Specifically, whereas the Antarctic Peninsula is warming rapidly, the rest of the continent is not. Leroux argues that colder polar temperatures (especially in the summers as a result of lower insolation) cause more intense MPHs. As they move towards the equator, the MPHs cause cyclonic movement at their leading edges. These cyclones transport warm, wet air back towards the poles. More intense MPHs result in more intense cyclones. This could explain the more intense winds mentioned in the post. It could also explain warming in the Peninsula occurring simultaneously with cooling in the interior of the continent.

This is an interesting theory that reconciles the varying temperature trajectories in different parts of the Antarctic that the general circulation models cannot readily explain. Overall, Leroux’s book is an entertaining and iconoclastic (and very Gallic) take on the climate change debate. Many of his points are well taken and deserve more attention.

File this under “looking for the cloud surrounding the silver lining.” Un-freaking-believable. A sad commentary on the state of climate science and its dominance by the modeling priesthood.

On 15 February, David Bromwich of Ohio State University released results of an important study showing that “[a] new report on climate over the world’s southernmost continent shows that temperatures during the late 20th century did not climb as had been predicted by many global climate models. . . . It also follows a similar finding from last summer by the same research group that showed no increase in precipitation over Antarctica in the last 50 years. Most models predict that both precipitation and temperature will increase over Antarctica with a warming of the planet.” Bromwich states: “What we see now is that the temperature regime is broadly similar to what we saw before with snowfall. In the last decade or so, both have gone down.”

So far, so good. Solid empirical research on a crucial geographic region. Indeed, inasmuch as the most alarming (and probably most alarmist) prediction regarding global warming is that heating in the polar regions will result in a massive release of water that will raise sea levels by catastrophic amounts, these findings represent very good news. Or at least, you’d think they would. But you’d be wrong. And that’s where the real interest in this story lies.

These results regarding snowfall and temperature are starkly at odds with the predictions of global climate models. These models predict that warming should be greatest at polar latitudes. Hence, these empirical results represent a major challenge to these models—and hence to the entire edifice of climate change theorizing and policy that rests on the models.

But rather than point this out in the emphatic way it deserves, Dr. Bromwich does everything he can to minimize the importance of his own research! (What’s up with that? The usual academic sin is to over-hype one’s own results.) Bromwich says “It isn’t surprising that these models are not doing as well in these remote parts of the world. These are global models and shouldn’t be expected to be equally exact for all locations.” [Query: Why should remoteness from civilization matter a rat’s rump in determining the implications of a region’s climate for global climate change? Isn’t the global climate a collection of largely remote regional climates? Or if it does matter, shouldn’t this cut the other way? That is, if economic development corrupts the surface temperature record, as has been plausibly argued and reasonably documented, wouldn’t you want to look at remote regions less affected by such development and place more emphasis on the results derived therefrom? Is Dr. Bromwich a reverse Bayesian who wants to put less weight on less biased observations and more weight on more biased ones? For that matter, if remote regions don’t matter, why is Dr. Bromwich wasting his time—and presumably freezing his tukkus—studying them? Masochism?]

In some ways the best response to Bromwich’s denigration of the poor regional performance of these models is “Duh”: it is well known that the skill of models in predicting the regional climate is pitiful. But the problem is much deeper than that. If the models are right, the regional impact of greenhouse gases should be greatest in the polar regions because the models predict that these regions are most sensitive to CO2 emissions. That is, if the models are right, they should do best in the Arctic and Antarctic, rather than failing miserably.

Dr. Bromwich attempts mightily to spin away from this obvious conclusion. He avers that “[i]t’s hard to see a global warming signal from the mainland of Antarctica right now. It’s very hard in these polar latitudes to demonstrate a global warming signal. . . . Part of the reason is that there is a lot of variability there.” However, the models predict that the climate change signal should be STRONGEST in the polar regions! If we can’t find it there, where the models predict it should be strongest, how can we have any confidence that we can detect it anywhere else? That is, if natural variability makes it “very hard in these polar latitudes to demonstrate a global warming signal” where the signal should be strongest how in God’s name can we detect it where the models predict it should be weaker? [Further my last aside: shouldn’t this enhance suspicion that the purported detection of a global climate signal from the instrumental record reflects in part—and perhaps in large part—biases in that record?]

There are other interesting tidbits. For instance, the research documents increasing strength in westerly winds in Antarctica. Well, an increase in the temperature gradient between polar and temperate regions is the most likely cause of an increase in winds. Standard models predict that GHG induced global warming should reduce the gradient (i.e., the poles should warm relative to temperate regions) and hence the winds should decrease in intensity.

Strike Three!: temperature predictions (wrong); snowfall predictions (wrong); wind predictions (wrong). Not even a foul ball in the whole at bat. Three cuts, three bad whiffs.

Bromwich notes that the increase in westerlies may be contributing to the warming of, and the consequent breakup of ice on, the Antarctic Peninsula—another iconic image in the Church of Global Warming. But if the winds result from an increase in the temperature gradient, this warming of the Peninsula is not a signal of global warming. Indeed, it could actually be a signal of global cooling—or at least polar cooling, which is flatly inconsistent with the anthropogenic global warming hypothesis which unambiguously predicts that the warming should occur disproportionately at the poles.

Like the recent research on the decline of heat content in the ocean in recent years, this study presents a major challenge to the model-driven “consensus.” (Ugh—how can self-respecting scientists worship consensus?) Rather than shout this from the rooftops, Dr. Bromwich cringingly attempts to shield the models from criticism. It’s not quite an auto da fe, but I suspect that Dr. Bromwich is striking such an attitude precisely because he wants to avoid offending the modeling establishment, thereby deflecting criticism and funding problems, or the necessity of walking through the public square recounting and recanting his heresy.

This is an inversion of the scientific method. Models are wrong. Always. Good modelers know they are wrong when they create them, but at the outset we don’t know how or why they are wrong. They are fleeting matches we light in a dark cave, not the Eternal Light of Truth. We are ignorant of the particular ways in which they are wrong, and learn by discovering their failures and adjusting our models and our understanding accordingly. Models push empirical work, and when empirical work documents failings in models the appropriate response is to revisit and revise—and sometimes junk—the received models. Scientific progress should be built on the rubble of discarded models.

Acting like the sycophants in “The Emperor’s New Clothes” is exactly the wrong way to respond to anomalous empirical results. The models are disposable, rapidly depreciating assets. The empirical evidence is not. The latter should take precedence. Spinning the empirical results so as not to discredit the models is antithetical to the advancement of our understanding of complex processes.

This episode and the ocean heat content one that preceded it (explained away as a “speed bump” as described in an earlier post) are symptomatic of a divergence between how science should work and how it works in practice. This should not be surprising. Kuhn pointed out years ago that there was a gulf between how science really works and the ideal scientific method. In practice, the tendency is to suppress empirical anomalies that challenge the conventional wisdom, rather than to pursue them doggedly in order to improve on—and perhaps jettison—the conventional wisdom.

So, there is little new under the sun. However, given the immense stakes in the climate change debate, these methodological deficiencies are much more costly than when the dispute involved phlogiston theory or the theory of the ether.

So please, Dr. Bromwich: revel in your results. Point out with relish how they conflict with the predictions of the models. Relentlessly needle the modelers into improving their products. Exult if your research consigns current models to Ozymandias-like oblivion. Go out and collect more data, and let the empirical chips fall where they may. Wear apostasy like a badge. Disdain consensus. In other words, be a scientist.

No, it’s not what you’re thinking. And it’s not about the CME-CBOT merger either. It’s about models–climate models, in fact.

I have been reading a lot about these models in recent months. They embody a convergence of academic interests–in energy (because global warming and the possible policy responses thereto will have a major impact on energy markets, and are leading to the creation of new commodity markets for carbon), and in numerical methods (I research and teach on numerical methods in finance).

I must say that the models are incredible achievements. They are extremely complex. Indeed, they are far more complex than even the most advanced derivatives pricing models. The typical derivatives valuation model is a 1 dimensional or 2 dimenional single-equation linear PDE. Climate models involve systems of non-linear PDEs. Moreover, climate models must characterize the behavior of complex weather phenomena (such as clouds and precipitation) that cannot be resolved due to their small scale (relative to the scale of the discretized grid on which the equations are solved). Due to their complexity, and the numerical constraints that must be satisfied to ensure stability of the solution, these models must be solved on state-of-the-art supercomputers (whereas even fairly sophisticated derivatives models can be solved on desktop machines.)

That said, there is room for skepticism about these models. The characterizations of small scale phenomena–the “parameterizations”–are often problematic. Important forcings (e.g., solar forcings not related to the intensity of the sun, aerosols) are poorly understood, or missing altogether. Similarly, important feedbacks may be missing or mischaracterized. If there are non-linearities in climate processes (a proposition for which there is considerable theoretical support) observational errors in initial conditions can lead to forecasting errors. Coupled ocean and atmospheric models require computational “fudges” to permit solution, and to give even passably realistic results.

These problems manifest themselves in some serious, but too seldom acknowledged, problems. These models have very little skill in predicting regional climate variations. Their predictions about tropical climate are particularly bad. Moreover, their predictions about tropospheric and surface warming differ substantially from observed data (which either indicates a problem in the models, or in the surface observations, or both). Predictions about precipitation and soil moisture are especially dubious. These empirical rejections of the models’ predictions should give everyone pause when interpreting the dire model-based forecasts of impending climate doom–but this is too seldom the case. Day after day I read in the press of studies predicting dire changes in regional climate–notably regional precipitation–based on models alone. The models’ poor track record in predicting regional climate variations and precipitation is seldom noted in these reports. The modelers have become like the Wizard of Oz–imposing, disembodied authorities speaking down to the little people, shrouded in smoke, illuminated by flashing lights, raging at those who dare question them. Pay no attention to the little man behind the curtain.

The analogy is not exact, but the situation reminds me of macroeconomics in the 1970s. That was the era of the big model with hundreds of equations. Those models were also tours de force, but it eventually became evident that bigger wasn’t better. The big models did not perform well. The real economy was too complex. The Lucas critique also showed that these models could not capture important feedback effects–notably, the maximizing responses of economic agents. The models became progressively more complex, but their performance did not improve dramatically, and as Lucas noted, they faced inherent limitations; you could make them bigger and bigger, but that didn’t mean they got that much better. There are still big models around, but their day has passed.

I wonder if the same might not happen in climate science.

One alternative to these structural models is reduced form time series models, like those described in Dobrovolski’s book on stochastic climate models. These models clearly have their own deficiencies. Notably, they do not specify the causal relations and feedback mechanisms in the way that structural climate models do. However, they do allow parsimonious characterizations of the data. Moreover, they exhibit different behaviors at different levels of temporal and spatial aggregation. These can provide benchmarks against which the structural models can be tested: can these models generate the same time series patterns as observed in the data, and the diversity of behaviors at different levels of aggregation? Through use of Kalman filtering and MCMC methods, these time series models can also handle the observational errors that are inherent in climate data. Thus, although the structural and time series approaches have largely evolved along separate paths, there seems to be considerable benefit to integrating them in some fashion.

Recent research shows that the ocean’s heat content declined substantially in 2003-2005, losing in these two years 20 percent of the heat that built up in the 1955-2003 period. This is embarrassing, to say the least, to advocates of the anthropogenic global warming hypothesis (especially NASA’s James Hansen, who has argued that “missing” atmospheric warming he’s been predicting for nigh on to 20 years is due to the ocean soaking up huge quantities of heat.) So, NASA has to spin mightily to explain away these results:

The average temperature of the water near the top of the Earth’s oceans has cooled significantly since 2003. The new research suggests that global warming trends are not always steady in their effects on ocean temperatures. (Click NOAA image for larger view of Sea Surface Temperatures for the Western Hemisphere for Sept. 20, 2006, taken at 10:59 a.m. EDT. Click here for high resolution version. Please credit “NOAA.”)

Although the average temperature of the upper oceans has cooled significantly since 2003, the decline is a fraction of the total ocean warming seen over the previous 48 years.

“This research suggests global warming isn’t always steady but happens with occasional ‘speed bumps’,” said Josh Willis, a co-author of the study at NASA’s Jet Propulsion Laboratory, Pasadena, Calif. “This cooling is probably natural climate variability. The oceans today are still warmer than they were during the 1980s, and most scientists expect the oceans will eventually continue to warm in response to human-induced climate change.”

Note the assertion that the cooling is due to natural climate variability–but the previous warming is (implicitly) assumed to be man-made. And it is only an assertion on top of an assumption. Where is the evidence that rejects the null hypothesis that the 1955-2003 warming was due to natural climate variability, in favor of the alternative that it was anthropogenic in origin? If the impact of anthropogenic greenhouse gases is so dominant relative to the natural variability of the climate, how could the latter overwhelm the former? This is a lame rationalization of results that contradict the dominant paradigm–something that should be familiar to anyone who has read Kuhn.

I also note the different standards applied to evidence that contradicts the AGW paradigm and the evidence that supports it. I have read in major media sources reports of numerous studies which document very short-lived trends, such as increases in temperatures or decreases ice over two or three year periods, that are trumpeted as evidence of global warming. They are never explained away as naturally-occurring “speed bumps.”

One last point. In his book Stochastic Climate Theory, Russian scientist S. G. Dobrovolski notes that many climate variables that are averaged over large geographic areas (like ocean heat content) are indistinguishable from random walks. That is, they are integrated time series. Equivalently, they have unit roots. This means that these series tend to exhibit large swings that appear trend-like. They increase for many years, and then may decrease for many years. (It is sometimes said that integrated time series exhibit “stochastic trends.”) That is, long periods of increases in one of these time series may merely reflect their natural time series dynamics, and may not be due to human forcing.

Carbon dioxide has been trending up for years. So has ocean heat content. Both series are essentially integrated. Regressing one on the other will almost certainly result in a highly statistically significant relationship between the two variables–but this regression is spurious. It says nothing about causation.

I recognize the virtues of using ocean heat content as a measure of global climate. As Roger Pielke, Sr. has noted, it is not contaminated by the effect of land use changes or urban heat islands or other factors that can distort temperature readings from land-based weather stations. However, as a globally averaged variable that per Dobrovolski is likely to be integrated, it is also very difficult to use as a measure of the impact of greenhouse gases on climate trends.