Engineering Failure Analysis 18 (2011) 550–556

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

The 2008 financial collapse: Lessons for engineering failure

David Fisk
Department Civil & Environmental Engineering, Imperial College London, London SW7 2AZ, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: The banking system, just like engineering systems, fails from time to time. As a conse-
Available online 26 August 2010 quence, again like engineering systems, it has imposed on it a risk regulatory system. While
the ‘technology’ being regulated is clearly very different, it is shown that the wider incen-
Keywords: tives to take risks and the consequent problems for risk regulation are very similar to real
Financial regulation world engineering. The evolution of the crisis is described drawing on published inquiries
Finance and the parallels at each stage of the crisis to engineering systems failures are identified.
Regulation
The conclusion is drawn that the real failure was not the regulators but their oversight sys-
Risk
tem – the meta-regulation framework. Conclusions are drawn for the function of risk reg-
ulation in engineering especially as applied to complex systems.
Ó 2010 Published by Elsevier Ltd.

1. Introduction

While engineering technology has little contact with the banking system, finance often plays a major role in the engineer-
ing of real products. Engineers will be aware of the impressive toolkit of regulation that is intended to protect those who use
the banking system from its occasional failures. Banks have always failed just as engineering devices have always failed. The
banking risk regulation system is better endowed in banking than in engineering because the former is much more transpar-
ent, independently audited and has to hand an enormous statistical database that would be the envy of any forensic engineer.
But if this is the case how did the global banking system collapse in 2008? Does it imply that if such a sophisticated system can
fail so catastrophically that engineering systems, particularly large and sophisticated systems can and will suffer the same
fate? To answer that question and explore lessons to be learnt for engineering failure analysis this paper reviews the causes.
The paper is in four parts. First the nature of banking is reviewed to highlight the role of risk management and risk reg-
ulation. In particular the argument seeks to draw the parallel between the bank and the engineered product as a mediator
between service provider and service seeker. Second the collateral debt obligation is introduced and presented as a new
‘product’ that became available at the end of the 1990s and is argued was the cause or vector of the collapse. Third the failure
of the supervisory system is reviewed as use of the CDO unwound the system in apparently unexpected ways. Finally general
conclusions are drawn and applied to engineering, especially engineering networks.

2. Banking and banking products

2.1. Banking

At its most generalised, the bank is a brokerage between those with a call on resources that they can be tempted not to
exercise immediately and those who would use those resources now on promise of paying back to the first party before the
resource is required by them. Although economics textbooks seldom pay much attention to ‘financial’ issues the banking

E-mail address: d.fisk@imperial.ac.uk

1350-6307/$ - see front matter Ó 2010 Published by Elsevier Ltd.

doi:10.1016/j.engfailanal.2010.08.012
 D. Fisk / Engineering Failure Analysis 18 (2011) 550–556 551

system is central to the modern economy. Given that the future is uncertain and debts may not be repaid managing risk be-
tween lender and borrower is the essential part of a bank’s function and its skill a source of its profit. Engineers are familiar
with ‘retail’ banking at a personal level. Retail banks take deposits from lenders who have been tempted by short withdrawal
periods and a small reward in interest. The banks then lend to borrowers on longer term and seek a higher reward in return.
As long as the occasional withdrawals of lenders are not correlated, this enables more to be lent than would ever occur
through bilateral agreement between a lender and borrower given their different time perspectives. Banks have to keep a
reserve of low risk securities so that random clusters in time of withdrawals can be paid and do not create an illusion of
a ‘run’ on the bank. This reserve on the balance sheet is essentially the ‘resilience’ in the banking product.
‘Investment banks’ will be familiar to engineers involved in substantial projects. These do not take deposits but sell bonds
and use the proceeds to lend, using the loan interest to pay bond holders. Unlike retail products the value of bonds depends
on the market. Up until the end of the 1990s investment banks were required to be distinct from retail banking. The idea had
been to partition the higher risk world of the investment bank from the less volatile world of main street banking. Dereg-
ulation presumably reflected a confidence that risk management was now so sophisticated that the distinction was irrele-
vant. A man carrying a red flag no longer walks in front of trains.
The banking model is very similar to engineering networks where services are provided more economically by aggregat-
ing demand and link or node failure rates can be tolerated because of built in redundancy in the system. In fact it is not that
uncommon in the finance literature to use engineering failures as the metaphor for financial catastrophes [1] (e.g. Miller,
2002). The object of this paper is to play the analogy in reverse.
The next section sketches at an elementary level the essential features of the banking collapse.

2.2. The disaster

The world’s finance system in 2000 was almost certainly the most elaborate and sophisticated system for both managing
risk and protecting third party interests ever devised yet it failed on a cataclysmic scale within the decade. It appeared to
take the rest of the global economy by surprise, as the hubris of the World Economic Forum meeting of 2006 testifies [2].
But then the first Tay Bridge was being applauded as a wonder of engineering before it too collapsed.
Just as one might re-run the flight recorder, the story of the banking collapse told in reverse chronology (in this case by
the Wikipedia community accessed 1 July 2010) runs as follows:
2008.

1. October. The balance sheets of banks all over the world had become meaningless. Unable to be sure of the viability of
trades, inter-bank trading ceased and the global financial system froze completely. In a matter of weeks it threatened
to halt all commercial activity in the global economy.
2. September. Investment bank Lehman brothers collapses. Half of US mortgage market effectively nationalized.
3. June. FBI makes arrests for fraudulently misrepresenting mortgage risks to investors.
4. January. Economic downtown increases mortgage defaults.

2007.

5. November. Financial Accounting Standards board introduces ‘mark-to-market’ to increase transparency but with con-
sequent greater sensitivity to market fluctuations.
6. October. US housing price bubble near bursting.
7. August. Credit crunch as subprime mortgages ‘found’ in bank balance sheets around the world.
8. May. Mortgage lenders seen as possible systemic risk by Fed Reserve.

2006.

9. October. US construction index drops 40%.

2005.

10. SEC suspends net capital rules on investment banks leaving them free to create derivatives. HUD raises affordable hous-
ing target for mortgage lenders. In fact 40% of real estate loans not for personal homes but ‘for investment purposes’.

2001.

11. ‘Derivatives’ trading applauded as means of stabilising risks to banks.

1999.

12. Banking deregulation.

552 D. Fisk / Engineering Failure Analysis 18 (2011) 550–556

As in engineering, how a system failed is not the same question as why it failed. The principle resource to answer this
second question is the Turner Review [3], prepared by the UK for the 2008 G20 meeting. In some respects the review does
not meet the forensic standards expected in engineering. For example it was conducted by a party to the catastrophe for pre-
sentation to a context where Governments would be seeking to distance themselves from blame. But it is a useful and re-
spected resource nevertheless.
It is important to emphasise that this is not a story of criminal activity. While dishonest selling in the mortgage market of
a few US States arguably primed the collapse, the ‘downturn’ in the US economy played a much larger part in initiating the
failure. Most players did not actually seek to break the letter of the law. The parties did not actually seek to go bankrupt
either and had put in place elaborate internal controls to ensure they remained solvent. For most players the ‘law’ with
which they sought to comply was not actually for their protection but for the protection of third parties. The players were
in business to maximise profit, and that meant that loopholes in the regulatory framework were as much (legal) fair game as
other more conventional business opportunities. None of this was new, either in banking or for that matter engineering.
Confident in the force of banks own self-interest in remaining solvent and the continued sophistication of risk systems,
regulators acquiesced to political demands to de-regulate. And at every removal of regulatory barriers the banks responded
positively by offering more elaborate and ingenious products for customers, to much applause at their ‘innovation’ and its
contribution to ‘economic growth’. Central to these new products was the ‘collateral debt obligation’. Its intention was be-
nign, but its effect was to prove malignant.

3. The collateral debt obligation

The collateral debt obligation (CDO) that came into being at the end of the 20th century is as close as one is likely to
get in finance to the analogy of a new engineering product. Ironically the CDO relied heavily in its construct on concepts
in engineering risk and statistical physics. The bank creates a new corporate entity. That entity receives income from a
group of assets, usually mortgages that it has bought. The entity issues bonds to pay for buying the underlying assets.
The bonds are sold to investors reflecting tranches of risk estimated by statistical models applied to the class of loans in
the pool. So in principle the CDO vehicle could better match the differing risk profiles of the spectrum of lenders and
borrowers and so bring more deals onto the market. Banks got income from the fee for assembling the CDO packages.
In the original model it was sophisticated buyers not banks who were supposed to hold the bonds. This is an analogous
role to the engineering product where the manufacturer plays no part in merchandising or buying it. Obviously the sta-
tistical model (essentially expected number of mortgage defaults) is the key. The CDO was a product success story by
any measure. From effectively zero in 2000 [12], CDO issuance had reached $1.7 trillion by 2007! To make CDO’s attrac-
tive to lenders seeking a low risk home for their money some CDO tranches were submitted to ‘rating agencies’ to be
accorded the safest AAA status. Indeed since the banks and the agencies were using essentially the same models the
products could be ‘engineered’ to be just AAA.
To the extent that regulators believed their role was to protect lenders, CDO’s were viewed as ‘clever rich boys’ toys’
where caveat emptor could legitimately be believed to hold. ‘Adverse asset selection’ was essentially watering down the
security portfolio with bad risks to the point of just making AAA. But all this was obscured from the bond purchasers’ view-
point. The nearer to the AAA boundary, the higher the premium the CDO could transfer from high risk borrowers paying high
interest rates to AAA lenders.
Engineers ought to recognise this ‘indicator’ problem. Where regulation or standards offer a precipitous ‘go/no-go’ deci-
sion (as for example in standardised performance based regulation using surrogate test conditions) engineers are incentiv-
ized to get as close to the boundary as possible. Any error in that boundary’s location and the real world systems start to fail.
Engines engineered to just get through mpg driving cycles may disappoint consumers in on-the-road use. If this was all that
was at risk, the system that unwound would have been bad news for the super rich who bought the bonds (much in the way
of the collapse of Lloyds Names) and not such bad news for the poor who occasionally paid off a mortgage they would not
normally have been granted. But in all the early discussions of the CDO as a special investment vehicle [4] no one seems to
have guessed (except perhaps bankers themselves) the use to which they would largely be put.
Of course it is remarkable that very few saw the problem posed by the critical AAA boundary. Indeed one of the
problems with finance in general is that those who see the problem can do better to sell short against it than publicly
inform the market. Dealing only in money it is a largely socially autistic enterprise. The investment banker Goldman
Sachs infamously advised the Greek Government on how to present its debt in the most favourable way and then short
sold the Euro.
Unfortunately CDO bonds started to appear on the books of banks themselves. They were after all offering substantially
better interest than the conventional Treasury Bonds used for ‘resilience’. The point that was missed was that the statistical
calculations were based on defaults in a banking system that did not have CDO’s on the balance sheet. But if random fluc-
tuations in the economy meant that some CDO’s lost value, then bank balance sheets would have to be rebuilt and credit
would be less easily available. That now increased the rate of default and the cascade failure begins.
Once the statistical model proved invalid the value of bank balance sheets became indeterminate. Once the large invest-
ment bank Lehman’s failed it appeared possible that any bank could fail and the banking system ceased to trade with itself or
anyone else. The train had crashed. Where were the bank regulators?
 D. Fisk / Engineering Failure Analysis 18 (2011) 550–556 553

4. Analysis

This is as far as this paper proposes to go into the details of financial crisis. In the annex we use a summary of the Turner
Review to construct the kind of Formal Systems Model that has been advocated for analysing engineering failures [5]. The
Turner narrative fits easily into the formalism. The methodology’s attraction is that it presents the failure as a systems effect
not a single cause. The system is redrawn with the standard three system domains in Fig. 1.

4.1. The regulatory system

This contains two sub-systems. The first defines the liability that is being managed offline from the regulator (i.e. the bal-
ance sheet). The characteristic is that the cognitive responsibility for the safety of the product is diffuse. Each player is
assuming that another player has assessed the risk correctly. The second subsystem is the oversight regulator, instructed
to be principle rather than prescription based and consequently also not aware how the product really worked.

4.2. The wider system boundary

This contains the meta-regulatory system which is the Government and legislature and which sets the terms for the reg-
ulator in the inner system. It also includes accountancy standards bodies, academia and the press. Over all the meta-regu-
latory system is misreading the situation and sending the wrong instructions to the inner system. The strongest message it is
receiving is from the pressure of multi-nationals and mobile capital to retain ‘competitiveness’. The meta-regulatory system

Fig. 1. Formal Systems Model of pre-collapse system.

554 D. Fisk / Engineering Failure Analysis 18 (2011) 550–556

does not have a clear view of events in the risk subsystem, nor has it the capacity to comprehend what wider risks might be
associated with the product.

4.3. The environment

The wider environment is dominated by a shared understanding of economic expectations that proved to be false. The
theory of economic growth implies expected norms of GDP increase and beliefs that faster than norm rates can only be
‘good’. It also implies that lower than norm in some sense indicates underperforming and talk of ‘need for recovery’. In either
case the regulator is inhibited from interventions that reduce GDP. But GDP is not ‘wealth’ but income from wealth and so
quite misleading as to the nature of the real economy.

5. Lessons

We conclude that there was nothing that remarkable about the 2008 crash. It reflected very familiar problems in risk reg-
ulation of engineering products. There is a diffuse responsibility for risk management that is not pinned down by a regulator.
The regulator is looking for exceptions before deviating from existing principles, goaded on by a meta-regulatory system
with little or no knowledge of what is happening but easily influenced by supposed economic threats. The alarming lesson
for engineering is that we are not addressing errors by relatively unskilled operatives but errors by a technical elite in an
industry infamous for creaming off a country’s best graduates from science and engineering.
In a sense the consequences of the crash are more severe because the elite, of all the players, were thought to be able to
look to their own self-interest. The idea that the crash was not foreseeable in principle because such foresight if it existed
would have then led to its effective management [13] indicates the kind of intellectual knots in which main players had
entangled themselves. Rather than cleverness giving complete mastery of the whole system by every player it had created
a complexity that fewer and fewer could master. The regulator and the meta-regulator were left far behind. The complexity
of the product was clearly a contributory factor and this is beginning to occur in engineering as well. Few power engineers
would have known to look for the IT programming errors in communication networks that played such an important part in
the 2003 NE US blackout. The risk assessment situation presumably only gets worse with smart grid technology! Similar reg-
ulator issues have arisen since the financial crash on ‘fly by wire’ vehicle control and Deep Ocean drilling.
In this sense we can ‘learn’ for analogous engineering solutions that:

 Technical capability may be incentivised to make matters too complex

 Before relying on self-regulation we should check that risk cannot be more easily transferred rather than managed
 The meta-regulation system may not be capable of assessing weaknesses in regulator competence or scope and the reg-
ulated may be incentivized to exploit those weaknesses
 Where the risk is multi-national and regulators are organized as a confederacy rather than federally, the regulatory sys-
tem is weak against multi-national organisations’ strategies
 Well intentioned adherence to an economic growth model can be dangerous when managing regulators if there is no
information about the underlying real economy

5.1. Designing interventions

It is tempting to push the analysis further because the model raises issues in precautionary risk management. The meta-
regulation might position regulators so that nothing happens that the regulators do not fully understand. From time to time
this has been the position in environmental regulation where the industry did not expect to have the relevant expertise. But
engineers and financiers would argue that society would not get a good service if the pace of technical innovation was set by
the regulator’s expertise in innovation. Technical innovators are in industry not regulators’ offices. Government in any case
would probably be more comfortable regulating how risk is managed by third parties rather than prescribing the risk man-
agement itself.
The Federal Reserve had noted the growth in CDO’s as an issue [6]. The appropriate action with hindsight would have
been to require banks to increase their reserves. The regulatory impact was certain (reduced bank profits and less credit
available) but the hazard indeterminate. So while early action would have been applauded by hindsight it is not difficult
to emphasise with the Federal Reserve’s judgement that it was politically untenable. The resultant position was not unlike
‘no cause for concern but we will continue to monitor’. Where the system seems to have failed is that it had no estimate of its
response time once the monitoring flagged up a problem. As we have seen at some undetermined critical point requiring the
banks to rebuild their balance sheet would cause a credit crunch that simply increased the defaults on their remaining loans.
Regulators sometimes find that, when they intervene late to make a product ‘doubly safe’ and its industry secure, the height-
ened public suspicion destroys demand altogether and the industry collapses. The meta-regulatory system ought to be the
component charged with asking the ‘but have you time to act?’ question.
The main failure seems to have been in meta-regulation. This system neither probed the competence of regulators nor
demanded modelling of regulatory response. Unfortunately it is the meta-regulatory system that sets the framework for post
 D. Fisk / Engineering Failure Analysis 18 (2011) 550–556 555

collapse ‘reform’. Evidence from finance would concur with engineering experience. The reforms have largely patched up the
inner system faults of the last collapse and not addressed the faults in the wider system at all.

6. Conclusions

The banking collapse of 2008 was an extraordinary event. It focussed around a new banking product and has proved to
show similar systemic causes to engineering failures. Lessons can therefore be learnt. Its major lesson is that clever systems
are no more secure than simple ones. Indeed the very cleverness makes them less transparent and more dangerous. This
seems to be an emerging issue in complex engineering systems [14].

Appendix A

A.1. Introduction

This annex gives a fuller rationalisation of Fig. 1. To give economy in referencing it draws specifically on evidence given
by the past head of the Federal Reserve, Alan Greenspan, to the House Committee on Oversight and Government Reform
(Greenspan 2008), and the UK Financial Services Authority’s (FSA) analysis (‘the Turner Report’) published ahead of the G20
Summit in March 2009 [3]. The latter tends to add explanations rather than critically appraise them but that is to be ex-
pected in a document seeking to lay the foundations for a consensus. The G20 meeting is not the final word in peer review
but its broad endorsement of the analysis serves the purpose here. The following paragraphs summarise. Summaries in
bold link to Fig. 1.

A.2. Wider environment

Turner [7] summarises the world context before the crash. One of the notable qualities of finance is that since it is about
money and nothing else it has no physical substance in its discourse and is dominated by narrative and sentiment (‘credit
crunch’, ‘double dip’, etc.). The dominant narrative at the New Millennium was the ‘End of History’. Markets had proved tri-
umphant. Only China represented an inconsistency but that was only a matter of time and reform. Across the globe econo-
mies GDP was growing. It appeared that the expectations of growth embedded in liberal economic growth theory were being
fulfilled.
If there was a problem it was the growing global financial imbalances [11]. China, the workshop of the world was putting
dollars on deposit rather than spending them in domestic consumption. The Middle East States were also accumulating dol-
lars as oil prices were running four to five times higher than production costs. So there was a good deal of global money look-
ing for a safe home rather than a quick gamble. Global economic expansion created an appetite for innovative financial
products.

A.3. Wider regulatory system

The US Government and consumers were happy to take the credit, not least to buy new houses and as a consequence
house prices climbed. But this was then largely seen as a growth pains problem for the US economy, with little linking to
health of the global banking system or the emergence of investment vehicles. The fulfilment of aggregate growth expecta-
tions hid the mounting problems [6].
In any case regulation by Congress to address the earlier failure of auditors of the highest international standing (and fee
rates) to see that Enron and WorldCom at a senior level were hiding liabilities off the balance sheet had not been well re-
ceived. Legislative complexity was beginning to be counter-productive. The Sarbanes-Oxhey Act (SABOX) required even
more onerous declarations and processes. Now since most firms were not trying to defraud their shareholders these condi-
tions were burden with little gain and the Act seemed generally unpopular. The idea that industry knows best self-regulation
with light touch regulation would be preferable to clumsy ham fisted legislation had some force given the apparent stable
economic growth.
New York was claiming that the Act was losing listing business to London. Indeed even after the crash the FSA’s own pro-
posals for banks to retain assets in each country where they operated were still be lobbied against in 2009 as ‘denting the UK
competitiveness if the regulator forged ahead without waiting for other regulators to agree to international action’ [8]. In
developing their innovative products multi-national banks were able to play one regulatory regime off against another to
argue for lighter not stronger regulatory touch.
Competition between international banking firms in different countries for business was incentivizing weakening of reg-
ulatory overhead without there being any relaxation in its need. Congress was in no mood to repeat SABOX interventions
with precautionary action. Advancing globalization without appropriate structures for global governance left a good deal
of combustible material hanging around (A point made by Stiglitz [9] in a wider context).
556 D. Fisk / Engineering Failure Analysis 18 (2011) 550–556

A.4. The regulated system

Neither the US Federal Reserve nor the UK Financial Services Authority (FSA) had much appetite to seek legislative ap-
proval to step ahead of the other with interventionist regulation. The London Financial regulator made a virtue of its ‘prin-
ciples-based’ regulation over ‘rules-based’ approaches [7]. ‘Principles-based’ safety regulation is common in engineering
regulation. It clearly relieves industry of a load of legal clutter. But, as with performance based standards, enables the reg-
ulator to shift the public risk. Light touch regulation is not precautionary by nature (‘if it ain’t broke don’t fix it’).
The regulators, especially the ‘light touch’ FSA had limited understanding of collateral debt obligations [10]. The Federal
Reserve had noted the emergence of CDO business but had no way of knowing its global extent [6]. As a consequence of an ex
post attitude to risk, regulators had a poor understanding of the new products and they appeared on bank balance sheets
unimpeded.
Competitive pressures ensured a focus on offering the highest rate of return on the securest products. But banks inferred a
transfer of risk assessment to rating agencies. Rating agencies did not profess to guarantee risk only rate it. Banks senior
supervisory systems (i.e. the board) oversaw but did not always understand the process. Banks themselves become custom-
ers for the high rate of return novel products. Unlike the Enron affair, no one in this system was acting illegally beyond the
principle of caveat emptor.

A.5. What is not in the model

This completes the Formal Model exposition from the Turner Review. But it is almost impossible in a failure analysis to
halt the event becoming a Christmas tree for other anxieties. The banking collapse is no exception and for completeness two
issues that have received prominence but have become excluded by the Formal Analysis process are aired below.

A.5.1. Greed (or just envy?)

Bankers earned spectacular salaries and bonuses and it was perhaps natural for popular debate to assume that greed must
have spurred on such reckless behaviour. But it is hard to justify accommodating this in the formal model because the reg-
ulatory system is there because behaviour is sometimes reckless. The incentives to promote CDO’s and devise means to get
them onto balance sheets certainly existed and may have contributed to the speed of events, but ‘greed’ is hardly a novel
external factor.

A.5.2. Too big to fail?

Debate also focussed on whether banks had become too big to fail. This has a bearing on how Governments become em-
broiled in rescue operations. Indeed rejecting a precautionary regulatory regime while at the same time allowing failure to
become catastrophic was a growing inconsistency in the meta-regulatory regime. But it is not a cause of the crash, only a
parameter that defined the magnitude of its consequences. Suggestions that ‘casino’ banking should be separated from ‘Main
Street’ banking usually missed the point. The two were not embroiled because both services were sometimes provided by
the same banking enterprise. Rather it was because ‘casino’ products had found their way into retail banking. Indeed it
was ‘narrow’ banks that failed most frequently after the crisis. The international scope of banking is clearly an issue when
some banks are very large. Ironically the problem of rescuing a bank operating in many jurisdictions was a conclusion from a
‘war gaming’ exercise conducted by the UK Treasury in the quiescent period before the crash [12].

References

[1] Miller M. Sovereign default by Argentina: ‘slow motion train crash’ or self-fulfilling crisis? CEPR discussion paper, No. 3399; May 2002..
[2] WEC. Annual report 2005/6 Davos; 2006.
[3] Turner A. The Turner review. London: FSA; 2009.
[4] Rule D. Risk transfer between banks, insurance companies and capital markets: an overview. Financial stability review: December Bank of England;
2001.
[5] Fortune J, Peters G. Failure from learning. Wiley; 1995. p. 111.
[6] Greenspan A. Evidence to house committee oversight and Government reform, 23 October 2008.
[7] Turner A. The turner review. London: FSA; 2009. p. 12.
[8] Financial Times. Lehman team warns banks to hold more assets in local units, April 15 2009.
[9] Stiglitz J. Globalization and its discontents. W.W. Norton & Co; 2003.
[10] Turner A. The Turner review. London: FSA; 2009. p. 89.
[11] Turner A. The Turner review. London: FSA; 2009. p. 14.
[12] Financial Times. Bankers ‘should rehearse for global crisis’, Nov 17 2006.
[13] Greenspan A. We need a better cushion against risk. Financial Times, March 27 2009. p. 11.
[14] Fisk DJ, Kerherve J. Complexity as a cause of unsustainability. Ecological complexity 2006;3:336–43.

David Fisk holds the Royal Academy of Engineering Chair in Engineering for Sustainable Development. He was previously Chief Scientist in a number of UK
Government Departments and was actively involved in negotiations of the main international environmental treaties of the 1990s.