Background


The ACE revolution is based on two pillars:

 

  • advances in mathematics and theoretical developments in the understanding of complex systems
  •  

  • technological breakthroughs in computer software and hardware that has made the creation of dynamically evolvable artificial environments possible
  •  

    Examples of the latter at a high level of sophistication are internet computer games with human user interactive facility. In the fully fledged ACE methodology there is also an automatic and continuous time probe that monitors and collects the data from the underlying system. This new technology is increasingly being adopted for economic modelling, market and policy design. The power of ACE tools for market and policy design to a large extent arises in the first instance in the way in which the underlying systems are modelled in ACE as complex adaptive systems. This often also includes digitized reproductions or “model vérité” that replicates structural features in an as is mode with few simplifying assumptions and which is capable of using real time data feed to perform “wind tunnel” tests of proposed policy intervention or market protocol.

    See Example I of a fully digitized study of congestion formation in a real cityscape which forms the basis of the market solution being proposed for congestion. A fundamental new avenue for the use of ACE in its pragmatic aspects has arisen from internet commerce. In pre-internet economy – markets were a given. In the post-internet era the design and implementation of electronic markets/trading platforms has become not only a commercial activity of the private sector but is also an effective way of regulating negative externalities such as pollution. The use of such e-markets, internal to large conglomerates and organizations to over come the rigidity of command and control management, is also being pioneered (See Example II)

     

    The computational testing out of specific auction protocols or policy interventions prior to implementation which can be done in an interactive mode is a vast advance to the way in which traditional methods dealt with the process of institutional design.

     

    ACE vs Traditional Economic Modelling and Policy Analysis

    ACE applications can overcome the limitations of deductive methodology of traditional economic analysis. That is, problems that are NP-hard or non-computable and those that can only self- organize as a result of large number of socio-economic interactions of people. Self-organization, by and large, refers to the outcomes of systems that defy command and control techniques of management. This includes growth of cities, transport, innovation, socio-economic networks etc. Analytical economic models often have to make simplifying assumptions for purposes of tractability and this may assume away the crucial features of the problem in question. The CCFEA/Bank of England ACE model in Example III below shows how ACE modelling can highlight the systemic risk implications of a highly non-symmetrical interbank network system which would be difficult to understand with traditional models. Econometric models that have been used traditionally for quantifying the performance of systems via forecasts or to evaluate impact of policy interventions run into a fundamental problem that the structure is rigid while in fact it cannot be assumed to be invariant to policy change. There is no scope to follow unintended consequences of policy.  Further, the econometric models are no good for monitoring on going phenomena (example of this was the monitoring of foreign exchange crisis, see Example IV below) as there is a lag in the updating of data for the equations of the econometric model to be reestimated.  By then the crisis will be over! Thus, a major strength of the electronic agent technology is its capacity for monitoring and responding in real time – a direction in relation to financial and banking systems is outlined in the final project in the Appendix.

     

    Way Forward


    There has been little or no interest from the bulk of UK academic economists in this potential paradigm shifting mode of analysis. (Ironically, it is the RAE which has been detrimental to ACE on the grounds that as this work veers off mainstream economics and it is not highly RAE rateable.) However, the response of policy oriented institutions to the potential inherent to having an interactive tool to understand systems for which policy interventions may be necessary – has been phenomenal.

     

    In summary, successful ACE modelling for purposes of market and policy design requires:

     

    (i) In depth domain and also multi-disciplinary knowledge and collaboration.

     

    (ii) The problem should be amenable to digitized mapping of the key structural features.

     

    (iii) Statistical data or live feed from real time data is needed to calibrate and validate the ACE simulator, and

     

    (iv) Modelling and programming the artificial environment are key inputs.

     

    (v) In many models where market/auction based robustness tests are required – either human user input via remote networked interactive computer system is required as in internet games or a general purpose lab for real time interface with a networked computer system for human subject  submission of experimental data into the ACE model is needed.  This involves substantial overhead costs over and above personnel costs.

     

    (vi) Last but not least, while it is clear that there is an urgent need for an ACE modelling platform for policy design – this clearly cannot be done as a pure theoretical exercise independent of the perceived needs of regulatory entity and industry.  Funding and the commitment to developing expertise in ACE models at an academic research centre and at the level of government economists – requires some concerted effort at the governmental level.  As it is the Foresight involvement in pushing the frontiers of science and technology forward that has brought these new modelling techniques to light, I commend those involved there.  CCFEA is unique in the UK for the strides that have already made in ACE modelling for policy and should be used as the hub for future developments.

     

    Appendix


    Five examples of work in ACE for policy or market design associated with CCFEA staff as lead researchers is given below

    1. Foresight (DTI and Directorate of Science and Technology) has commissioned a piece of work from 3 leading academics, Dr Sheri Markose, Director: Centre For Computational Finance and Economic Agents (CCFEA), Essex University; Professor Peter Allen, School of Management, Cranfield University, and Professor Phil Blythe, Director: Transport Operations Research Group (TORG), Newcastle University, to investigate how agent-based and transport models may be combined to allow us to explore Smart Market Protocols for Road Transport.
      This may be one of the first ACE models on this scale being developed for purposes of policy design.

    2. BT Market-based Workforce Management Project
      This is a joint project with BT Research Labs (Martlesham). At University of Essex, it is led by Professor Edward Tsang. The aim is to tackle BT's work force scheduling problem through a market-based approach. Preliminary work was presented in MISTA 2005 (click here for paper). http://cswww.essex.ac.uk/CSP/bargain/
    1. CCFEA joint project involving Sheri Markose and  Amadeo Alentorn  with Bank of England researchers Steven Millard and Ying Yang has developed  the Interbank Large Value Payments Simulator (IPSS).  IPSS is capable of running real time data of the interbank payments which is approximately equally to ¼ of UK GDP – on a daily basis.  The properties of the interbank system can be well understood and quantified. These include relative size of failed payments and loss of liquidity and the possibility of contagion. The impact of a change of regime which may alter bank behaviour is more problematic and needs industry/practitioner input to model the appropriate strategic behaviour of banks.

    2. CCFEA ACE Modelling of a classic example of poor policy design leading to collapse of a system: Black Wednesday and Collapse of the ERM Currency peg or 19 September 2002. George Soros made £2bn taking a short position against the Sterling and the Bank of England. He is alleged to have used the Liar or Contrarian Strategy. Why did Soros win: or why did all currency pegs collapse (from Mexico to the Asian ones) at great cost to the tax payer? Firstly, the ACE model shows that what provokes the attacks is the transparent defence: Speculator Sells forward after the central bank raises the exchange rate to above the lower bound. Thirty different simulations using a ACE wind tunnel test of the currency peg with a central bank intervening to raise the exchange and speculator taking a short position shows that the bank cannot win even once viz. ran out of reserves. Flawed macroeconomic literature on precommitment to transparent strategy caused IMF to support currency pegs and led to the worst policy induced failures of our time. ACE testing may have averted such policy disasters.

    3. ACE Modelling Based on Real Time Banking and Financial Liquidity Data:  Future Project of Great Significance.

      The important project here is to see that the data from the 3 main real-time systems CREST, Chaps and the Forex Markets can be integrated and monitored – so that not only regulators and market participants can view it – but also data can be used for analysis and research. Currently, nobody in the financial system – least of all the regulators have the capacity to respond in real time based on real time feeds from these systems. As Essex has a great reputation in data management with the ESRC Data Archive and ISER, the processing and analysis of live feeds from real time financial systems by CCFEA is a project of great significance. Funding from a consortium of interested City firms with that of the DTI to support research and lab facilities in this area is being sought.