Economics current debates from a policy perspective Jonathan
Economics – current debates from a policy perspective Jonathan Cave
Outline • The role of economics in policy • Challenges to conventional economics • Some policy challenges and debates
Economics and policy “Practical men who believe themselves to be quite exempt from any intellectual influence, are usually the slaves of some defunct economist. Madmen in authority, who hear voices in the air, are distilling their frenzy from some academic scribbler of a few years back”
How economics drives policy Creating policy imperatives – “it’s the economy, stupid” Framing issues and options – Incentives – why do people do what they do? – A common basis of comparison – adding Apples ™ and Oranges ™ – The animating force of the market – societal infrastructure, primordial network, measuring stick and channel of influence – Evidence of what works – empirical and experimental – Regulatory and other policy instruments – Impact assessment and evaluation Economic Theory Administrations Research Policy Business, Market Models Business Citizens
Building blocks Levels: – Micro – Rationality (and psychology and sociology) – Meso – Firms, households, sectors (and networks) – Macro – Countries, blocs and aggregate measures – ‘Scale-free’ results – each level looks like the others Interactions: – Competition, collusion and conflict – Game theory • Players, strategies, preferences (payoffs), information • Non-cooperative, bargaining, cooperative • Mechanism design – from contracts to auctions
Policy Basis – market failure (to do what? ): – – – Allocational efficiency – can everyone be made better off? Technical efficiency – can we produce more of everything? Dynamic efficiency – including growth, recovery and innovation Equity Delivery of external and public societal benefits Mechanism: – selection (who plays) – incentives (what they do) Tools – Property rights (to allow trade and investment) – Ex ante restrictions (licenses, standards) and Ex post rules (conduct or outcome-based) – Taxes and subsidies – Contests
Internet challenges to the standard model Economic policy assumes that things are done for money Rationality and meaningful consent may not be reliable Non-human actors Herding and contagion Information goods involve access as much as ownership Two-sided (platform) markets Network externalities - tipping, excess volatility or inertia Complex systems behaviour New stuff: Io. T, Cloud, Big Data
Public goods and the value of information Standard theory assumes exclusive and transferrable property rights A few exceptions are recognised (in red): – Externality determines how we aggregate costs and benefits to decide what is efficient – Permission determines whether we can use markets to determine efficient outcome, organise production and access, and pay the costs Permission to access: Externality Owner User Perfectly negative (rivalrous) Zero (non-rival) Positive (‘network’) Y Y Ordinary good Club good Voluntary provision N Y Congestion good Pure public good Commons Y N Liability Fees Charity N N Environmental goods
Value of information Common sense assures information is valuable – we can always ignore it, right? Consider the game of Prisoners’ Dilemma – – Each person can choose between selfish and cooperative behaviour e. g. selfishness is traffic shaping (S) and cooperation is net neutrality (N) Selfishness helps the person less than it hurts the other, regardless of what the other does There is a unique individually rational equilibrium – selfishness all round But all the other outcomes are collectively rational (Pareto optimal) – it is not possible to make anyone better off without hurting someone else Now suppose that either strategy could be cooperative (shaping may help traffic types to cluster together; net neutrality may treat unequal parties equally) – – Suppose that player 1 (row) knows which is which, and player 2 observes 1’s move If the informed player (1) uses this information, the result is always the inefficient equilibrium If the informed player randomises, so does the other; not perfect, but better than before So the information has negative value to both players! S N S 3, 3 0, 4 N 4, 0 1, 1 Shaping is cooperative S N S 1, 1 4, 0 N 0, 4 3, 3 Neutrality is cooperative S N S 2, 2 N 2, 2 Random
Policy 1: Digital Single Market Policy context: Europe 2020 strategy – A policy chapeau – many initiatives, e. g. • • Future Internet PPP Horizon 2020 Flagships EFSI – Digital Agenda for Europe – now Digital Single Market The challenge of the Internet – – Jurisdictional problems – regional, national, EU, global Market definition and regulatory traction Regulators not set up for Internet, with different cultures Conflicting economic objectives – competition, competitiveness, GDP, employment… – Tension between harmonisation and comparative advantage The challenge of Grands Projects
Economic case for the Digital Single Market* Massive potential for Europe – – – 315 million daily Internet users € 415 billion in additional GDP/year Substantial (unknown) contributions to employment) Europe’s digital market is not her own : – 42% of online economic activity lies within Member State borders; only 4% is cross-border. • 15% of consumers bought cross-border; 44% bought domestically; Cross-border competition could save them up to € 11. 7 Billion per year • 7% of SMEs managed significant cross-border sales – average extra cost of € 9000 per year; uniform rules would increase proportion to at least 57%; VAT compliance adds € 5000/country • Shipping costs are a barrier for 90% of e-shoppers and 62% of companies • 52% of attempted cross-borders within the EU are geo-blocked – 54% of online economic activity involves US-based services Framework conditions are also lagging – – Data protection reforms stalled, Safe Harbour is sunk, TTIP risks Patchy penetration of fast broadband (22. 5%) and 4 G 59%-15% (rural) New opportunities – – Cloud data storage (20%-40% in next 6 years) Data analytics could save top 100 mfgs € 415 billion, raise GDP growth by 1. 9% * Data from Eurostat and Digital Agenda Scorecard
Policy case for the Digital Single Market Starting points: Barriers: • Elaborate and mature Single Market vision • Good examples e. g. Estonian e. Government, German start-up scene, UK fintech ecosystem, HD broadband (e. g. Ie, Be, Sw, Nl) • Perceived inequalities and inequities • Consumer access to services • Market access by start-ups • Barriers to innovation (e. g. sharing economy, P 2 P services) • Lack of digital-by-default services and Internet-ready regulation • Small differences increasingly important Consumers: • Pan-European access to subscription services, FRAND IP protection • An end to locational discrimination in pricing and conditions • Unified or harmonised consumer rights • Effective, certain, innovation-friendly data protection framework • Improved access to public and private online services Entrepreneurs: • Single point of registration, once-only reporting • Competition rules that work for data-driven economy • Tax reform to address BEPS, VAT issues, fit digital business models • Sa. MBA default exemptions for growing/new businesses, REFIT • Interoperability standards, unified set of wholesale access products • Open Data Charter
Other Digital Agenda for Europe policies Innovation Union – – Targets human capital, finance, patenting costs, regulations and procedures, standards, strategic public procurement, fragmentation Large variations, faltering progress (scorecard leaders, followers) – Dimensions of innovation European Fund for Strategic Investments – – – Regulatory and structural reforms to improve investment climate European Investment Advisory Hub to channel finance to real economy Supporting higher-risk financing All owned by different players, and subject to internal and external shocks All rely on others’ participation, but create unique risks Each relies on uncertain (and potentially inconsistent) economic modelling 4
Policy 2: Net neutrality What could be wrong with neutrality? The separability of the transport layer – a bit is a bit? The race to zero (rating) Two-sided markets and walled gardens – Competition in the market or competition for the market – Is it all about content delivery? Dimensions of performance and discrimination – Congestion externalities and crowding types – who interferes with whom? – The indirect value of a subscriber base – cream-skimming and sludge-passing – Quality of experience – latency, jitter and relative speeds
Net Neutrality 2: the necessity and efficiency of discrimination Often, platforms or infrastructures have large fixed costs AC Price – Allocationally efficient marginal-cost pricing will not cover them – Any feasible single price regime will generate welfare loss – Have to price according to inverse elasticity of demand (Ramsey) How to ensure the right kind of differentiation? Secondary question: who should have market power? How to balance regional and global interests, value capture and creation? Future-proofing the rules In diagrams, cost = area under MC plus fixed cost (shaded rectangle); revenue = sum of unshaded rectangles Monopoly AC Monop D D D MC MC MC Quantity Inefficient single-price Efficient multi-price – includes subsidies Extreme case: market would not exist without discrimination
Policy 3: Competitiveness and growth Competitiveness is not the same as competition – Incubators vs. boot camps vs. Darwinian sandpits – The importance of failure – Building market share, IP, experience, organisational capital Creative destruction - incumbents vs. new entrants New forms of enterprise – Classifying and understanding them; guiding their development – Size matters – but indirectly – Networked and transitory affiliations; rules assume corporates and growth Finance – Not just the amount, but the modalities – Competition regulation and financial regulation Structure (SCP) – Understanding market networks – Macroprudential regulation – Ecosystem services New industrial policy and Better Regulation
Policy 4: Competition policy What is a market? Cooperation or collusion New forms of enterprise Standards and self-regulation From promoting competition to promoting efficiency Nudging the self-organisation of markets and products Avoiding capture and foreclosure Balancing public needs and private incentives
Policy 5: Privacy and security Economic or fundamental right Adequacy of legal roles Third-party monetisation and the two-way value of personal information The Right to be Forgotten – or remembered? Privacy of information or of action? The draft General Privacy Regulation Safe Harbour type arrangements Security concerns – and the effective placement of liability
Policy 6: Financial trading Non-human actors, quants and complex systems The rule of algorithms (e. g. Gaussian Copula) The interaction of technical and financial efficiency Behavioural responses and feedback loops Seeing what is happening “The regulators had all the data; the investment houses had all the brains”
Regulating the cloud: more, less or different regulation and competing agendas
Introduction The cloud is: – a fad; – a metaphor; – a critical phase in complex ICT system development; – A microcosm for issues of Internet regulation – dead [choose all that apply] Cloud-like things challenge regulation: – Unique issues – Existing issues made harder (or easier) It shares this characteristic with “Internet regulation” Not regulatory convergence, but a regulatory network Rewards new models and approaches
Operational definition of the cloud 5 -3 -4 … …You know the rest
The cloud is already regulated Technical: standards; interoperability; Qo. S; security… Economic: general competition � consumer protection; IPR Social: privacy; content; liability(? ) Sector-specific: finance, health, transport, services Up- and down-stream regulation: cloud, cloudbased and cloud-enhanced services
Why should it be regulated? Issues unique to the cloud (few) Issues made harder or easier by the cloud A convenient point of intervention – or at least discussion A natural platform for self- and co-regulation A ‘model’ for atomistic and dynamic competition (the cloud version of the app ecosystem)
Can it be regulated – and by whom? Indirect relationships and regulatory traction Implementation problems (e. g. jurisdiction) Conflation of regulatory and stakeholder agendas – a hard problem for regulatory design Limitations of existing statutory duties and powers Need to assemble networked governance ecosystem to parallel – or to invade - cloud
How to frame the problem Map regulators’ statutory remit and tools Assess problem: – Whose problem is it – Is it due to or changed by the cloud – Context: • Can it be fixed without damaging other things • Will it go away by itself Identify or develop instruments – Evidence and evaluation – finding the problem and fixing blame – Crafting and implementing a remedy – Changing behaviour Monitoring and enforcement
Selection framework: legitimate interests Citizen interests – Access to critical telecommunications services – Participation in society – Citizen protection Consumer interests – Benefits of competition – Consumer protection – Consumer empowerment
Selection framework: specific duties Ensuring the optimal use of the electro-magnetic spectrum Ensuring that a wide range of electronic communications services – including high speed data services – is available throughout the UK Ensuring a wide range of TV and radio services of high quality and wide appeal Maintaining plurality in the provision of broadcasting Applying adequate protection for audiences against offensive or harmful material Applying adequate protection for audiences against unfairness or the infringement of privacy Citizen and consumer interests
Screening test Is there citizen harm? Is there consumer harm? Is ‘the market’ likely to mitigate or eliminate this harm? Does it fit within regulator’s remit? Is new intervention/power required? Note: regulatory duties are not the same as policy objectives Not the same for all nations (esp. within EU)
Citizen and consumer issues Market solution? Existing regulation Consumer empowerment Consumer protection Benefits of competition Citizen harm Description Consumer switching and mobility Consumers may be tied to cloud service providers by limited portability of data and applications, restricted interoperability, lack of information N I Copyright and IPR i) Questions over whether Cloud Service Providers may be liable for actions of users and hosted service providers; ii) Questions over N ownership of user-generated content; iii) Concerns that content-matching services may serve to legitimise unauthorised content copying I Unfair and potentially anti-competitive Standard form SLAs are efficient for large numbers of consumers but prevent bargaining. This raises switching costs and can lead to lock-in, contract terms damage competition or promote potentially anticompetitive market segmentation. More serious for cloud because providers ‘have’ user data. ? Remit Issues Security, reliability, resilience capacity Concerns: a) transparency of cloud providers’ practices; b) security and reliability of services; c) data loss and unauthorised release - even when not illegal, this can create significant harm, especially when data owners are not aware; d) Data security includes e. g. infrastructure resilience (continuity), authentication. Crime Users of cloud-hosted services may face an increased risk of a range of criminal threats: Identity, Data theft; Fraud; Malicious system, processing or data interference; Data loss or unauthorised release. Privacy may be weakened by indirect relationships with cloud-hosted service providers who hold personal data, limited visibility, obsolete ? legal roles, privacy-invasive technologies and business models. I Communications as a Service (Caa. S) Integrated or converged video, voice and data communications and associated services that overlap with existing regulated communications but are not limited to communications service providers. Policy issue: whether/how to regulate them. D Advertising and marketing Consumers may not be fully informed of what they sign up to; may see repeat of problems with e. g. broadband or mobile. Providers may be unable to certify, deliver or even inform consumers about the services they expect and those they receive. Cloud as a utility (including risk of market Cloud computing share technical, economic and societal features with other utilities (e. g. scale economies, universal service potential). Policy foreclosure) issues: potential monopolistic foreclosure, social case for Universal Service (quality, affordability, open access) regulation ? ? N ? E E N E
Implementation issues Consumer empowerment Consumer protection Benefits of competition Locations are hard to verify and constantly changing; this raises consumer protection and jurisdictional issues. Locus of control Difficult to identify points of leverage for effective intervention; existing regulation could be undermined by cloud-hosted alternatives that D, I, ? operate outside the regulatory sphere. Also, control is different in different architectures and deployment models. E ? Remit Location and jurisdiction Market solution? Existing regulation Description Citizen harm Issues D Linked to mobility, but important to other interests that depend on consumer choice, e. g. identity management, which is central to consumer Consumer information, transparency of choice and protection; if individuals cannot know or verify the identities of those with whom they transact, it may be hard to enforce rights CSP practices Includes adverse impacts of unfair cloud contracts and advertising and marketing practices. ? I Complexity of the cloud The cloud's inherent complexity and adaptability challenges conventional regulation. D Certification and other self- and coregulation There are currently a variety of such ’market-provided' ways to address a range of concerns; may require national monitoring, and/or enforcement and/or be multiple, inconsistent, ineffective, costly, unmanageable, or anticompetitive. Y I Cloud neutrality Cloud should be OS, hardware, software. neutral; but this may be restricted for purposes of efficiency. As with net Neutrality, it is an empirical question whether non-neutrality is harmful and whether harmful non-neutrality can be countered. N I Consumer/SME similarities, regulatory heritage and convergence In the cloud, both SMEs and users are potential cloud-hosted service providers therefore consumer harms thus occur in both B 2 B and B 2 C. A Potentially applies D related issue is the 'fit' of communications, privacy. regulation with the cloud environment. across the board Trust Security (esp. for firms, including SMEs) and privacy issues may reduce trust in cloud-hosted services, affecting their uptake; alternatively they may lead to to “privacy/security as a service” innovations N E
Modelling challenges Complex adaptive system Protean N-sided market Salience vs. reality Monetisation and participation rights Technical issues: capacity management, privacy and security as a service, data access vs. processing Challenging sectors: computerised/HF financial trading, health diagnostics, shared innovation, content sharing, supply chain data repositories
Three scenarios: dimensions Sovereignty over cloud regulation: national, international, market/selfregulation Locus of power in cloud services: Telcos (current EU situation), Google/Amazon (current US situation), Hypervisors (Vmware, MS/Citrix) Balance of cloud deployment from regulatory perspective: public (consumer and citizen harms incl. privacy), private/enterprise (competition), hybrid Architecture of (future) Internet: cloud as niche/overlay on current architecture or cloud-centric architecture (as in NEBULA).
Summary of scenarios Cloud-max Nothing new Another critical infrastructure Sovereignty: Market/self-reg. National/international Power: Google/Amazon Telcos Hypervisors Deployment: Hybrid Public clouds Private/enterprise (provider responsibility) Architecture Cloud-centric Niche/overlay
Cloud centric Cloud develops under existing reg. framework Increasingly central to Internet architecture and governance Emergent issues tend to be managed through a combination of market solutions and self-regulation Market and governance power lie with dominant B 2 C providers of global Internet Dominant mode of computation, data storage and informationintensive communication and transaction Could drive re-examination of regulators’ duties and mandate; in short run, most issues are out of scope
Nothing new Continues trend as currently experienced in UK Regulatory sovereignty according to existing mandates (some ad-hoc additions) May create race to bottom amid contradictory requirements (especially at pan European level), as national regulatory agencies strive to attract profitable data centres. Dominant model is public (B 2 C, SME) cloud; dominant players are telcos Existing regimes continue in sub-optimal and fragmented fashion: – – – telecommunications regulators adopt consumer protection perspective data protection authorities looking into how cloud service providers address privacy and data protection obligations security and law enforcement agencies provide guidance on managing cloud risks
A new critical infrastructure Currently, unlikely due to perceived security weaknesses (confidentiality, availability and data integrity) and regulatory interest Private clouds are becoming more central, esp. in highly-regulated sectors subject to margin squeeze Another reason is expansion of Big Data Analytics Regulatory sovereignty will be increasingly joined up within and among nations Critical core primarily provided as private/enterprise clouds; with the regulatory protection afforded by their centrality, they can compete successfully for citizen and consumer business as well. This reinforces the status of cloud service providers.
Conclusions The cloud is primarily useful as a metaphor, and a means of raising challenges: – Definitions – Policy linkage (economic growth, finance, …) – Realigned roles and responsibilities – Sandbox for n-sided market, app ecosystem issues The technical issues may need prior resolution – Qo. S – Self-organised complexity – Demand smoothing…
Algoarchy – the rise of formulÆ and machines 06 and 13 March, 2015 39
I. Debt derivatives and the Gaussian Copula The Internet connection • Like the Internet itself, the simplicity of these formulae opened participation to all sorts of players • Lines of information and accountability blurred • Models that interpreted market data developed hidden bias and error • Trading happened over the Internet, using big data analytics to which fast and stupid models were applied • Systemic behaviour became harder to predict as individual elements became simpler. 06 and 13 March, 2015 40
Gaussian copulas – the formula that killed Wall Street? In the US, 2007/2008 marked the bursting of a Housing Bubble This triggered a major recession; people looked for scapegoats. Initially, blame was fixed on major financial institutions (Bear Sterns, Goldman Sachs, AIG, etc. ) Later, the finger was pointed at the formulas they used to assess investment risk Chief among these was David X. Li’s Gaussian Copula formula
Collateralized Debt Obligations A CDO is a structured asset-backed security (ABS) whose value and payments come from an underlying portfolio of fixed-income assets: bonds; loans; credit default swaps (CDSs) and mortgage-backed securities The first CDO was issued in 1987; they became steadily more popular starting in the late 1990 s until the mid 2000 s The same period saw the growth of the CDS The CDO offers different tranches of security – “Senior” tranche - paid first, most secure, most expensive – lowest (subordinate/equity) tranches are riskiest but cheapest – Investors have ultimate credit risk exposure to underlying entities, so banks used CDOs to transfer risk from themselves to investors
CDOs, 2 On each tranche, investor has “attachment percentage” and a “detachment percentage” - when the total percentage loss of the entities in the CDO reaches: – The attachment percentage, investors start to lose money (not get paid fully) – The detachment percentage, investors won’t get paid at all
CDO Example – – Tranche 1 (equity tranche) = 0% - 5% Tranche 2 = 5% - 15% Tranche 3 = 15% - 30% Tranche 4 (senior tranche) = 30% - 70% If CDO has 3% loss, Tranche 1 (the equity tranche) will absorb that loss; other investors unaffected. If CDO has 35% loss, Tranche 1 and 2 gets no payment, Tranche 3 loses most of its payment; Tranche 4 unaffected
Credit default swaps (CDSs) Like an insurance policy that pays off in the event of default Unlike an insurance party in that it does not (necessarily) involve the original debtor No limit to the amount of CDS that can be written on a single “underlying” credit. – Every underlying gets a certain amount of “basis points” (representing. 01%) – These depend on stability/riskiness of underlying – The riskier the underlying, the higher the basis points. • Reflects market perception of default risk over riskless rate; like percentage odds that underlying will default before maturity
Gaussian copula Purported to model correlation between default of two obligations (or the entities that control them) without using historical default data. Instead, formula used CDS pricing data (initially had less than 10 years’ of observations) Implicit assumption: CDS market was able correctly to price the default risk correctly on the underlying assets A copula is used in statistics to couple behaviour of two or more variables and determine if they are correlated With so many underlying entities in CDOs and portfolio/index CDSs, copula seemed ideal Li’s Gaussian formula was the only copula used in practice
The formula itself
Response Financial industry embraced it, and used it to create and sell unprecedented amounts of “AAA-rated” securities This was easy: no need to examine (or even identify) underlying entities, just use one number If underlying entities were believed to be uncorrelated, the perceived risk of a CDO built of these CDSs was near 0, especially in senior tranche Banks began combining all kinds of risky underlyings; if they did not appear correlated, CDO was highly rated Markets grew rapidly: – CDS – from $921 B end ‘ 01 to $62 T by end ’ 07 – CDO – from $275 B in ‘ 00 to $4. 7 T by end ‘ 06
Impacts Used to be good practice to diversify underlyings With copula, a group of (apparently) uncorrelated home loans (say) could be advertised as a safe asset, because you’d ‘never’ lose everything (in the senior tranche) Banks started to sell riskier CDOs; they also made riskier loans (because they could lay off the risk) – Exacerbated by government pressure to make more loans – Sound familiar? When the initial growth occurred, underlying (house prices) was increasing rapidly, meaning prices reinforced impression of low and uncorrelated default risk By the time the bubble burst, this misleading price record was ‘set in stone’ – By the time defaults showed up, it was too late: AAA CDOs became worthless
What was wrong? Underlying correlation assumptions defeated by derivative cross-linking Model intended for analysis, not decision-making Fundamentals not understood by model users Certainly not scalable Conspiracy of optimism lasted 6 -7 years – more ‘good history’ to reinforce belief in model Maybe we can do better now with – Network models – Big data to identify high-dimensional correlations – Avoids even possibility of ‘one formula to rule them all’
II. High-frequency and computerbased trading
Outline: High frequency and computer-based trading What are CBT and HFT? Financial stability and CBT Impact on liquidity, price efficiency/discovery and transaction costs The technology of CBT/HFT
What are Computer-based and high-frequency trading? • Look at stability as a source of confidence in capital markets (as a store of wealth, source in ‘real’ investments, etc. ) • Fluctuations are always expected, but large, unexpected/inexplicable changes can impair the investment mechanism, erode confidence, and undermine financial stability. • Example: 6/5/2010 “Flash Crash” • US equity market dropped 600 points in 5 minutes • Destroyed $800 bn of value • Regained almost all losses within 30 minutes but • Led to several months of outflows from retail mutual funds. • Mechanisms that may lead to instability: 1. Nonlinear sensitivities to change 2. incomplete information 3. “endogenous” risks based on feedback loops.
Computer-based/highfrequency trading • The internal chains of cause and effect producing endogenous risk create positive feedbacks that • Amplify detrimental interactions among management processes • Can even be worsened by risk-management systems • Can be driven by • Changes in market volume or volatility • Market news • Delays in distributing reference data. 4. Social instability - normalisation of deviance (unexpected and risky events come to be seen as ever more normal, until disaster) 5. Network topology determines the stability and the flow of information and trades, hence overall system stability
Why is CBT/HFT different? interactions take place at a pace where human intervention could not prevent them given this, computer based (mechanical) trading is almost obligatory, with all its system-wide uncertainties Information asymmetries become more acute (and different) than in the past the source of liquidity has changed to computer based and high-frequency trading, which has implications for its robustness under stress Latency and other technical characteristics matter enormously
Varieties of CBT/HFT Can trade on an agency or proprietary basis May adopt liquidity-consuming (aggressive) or liquidity-supplying (passive) trading styles May engage in uninformed or informed trading
Computer-based/highfrequency trading • The balance between “Computer Trading” and “Human Intervention” has yet to be determined • Further complexity arises from the divergent evolution of different markets • But the base drivers remain largely common • The pace of evolution means the answers to the questions arising must be in 2021 terms if they are to be delivered!
Balance between “Computer Trading” and “Human Intervention” has yet to be determined – black vs grey (vs white? ) “Computer Trading” = - Mode of trading stocks, bonds, forex, derivatives and commodities with the “use of machine” 1 Includes e-trading, programme trading, automated trading and their subsets algo-trading and HFT Trades on or off exchange (eg ECNs, MFTs). Currently no regulation specific to computer trading – but watch this space Growth has been significant - Overall volumes trebled between 2004 -2009 70% NASDAQ trades are HFT >30% UK equity trades are HFT Asia Pacific picking up speed , from 15% to 18% of total Percentage of institutional flow traded using low touch channels 63% 53% 31% 2004 2005 2006 *Percentage of Equity trading From HFT 6 61% 52% But levelled off in Europe and US in 2010 - Maturity phase? Pause for breath/technology? Calm before the (flash crash) storm? “Curate’s Egg” debate continues - Greater volume => liquidity => stability, price transparency? Greater price competition => added market depth => decreased volatility? 1987, 2010 Flash Crash 2007 38% 35% 20% 1% 2005 26% 6% 2006 56% 29% 21% 9% 2007 2008 Europe 2009 2010
Further complexity arises from the divergent evolution of different markets Factors: existing market structures, regulation, competition and trader needs have all affected the transition to electronic trading Equity markets US: many electronic trading venues, relatively few traditional exchanges Europe: electronic trading generally incorporated within its many traditional exchanges Both: fairly straightforward and cost-effective to introduce computer trading (liquidity, homogeneity) Foreign exchange markets 41% of interbank trading in major currencies (BIS triennial survey) Fixed income markets Slower migration vs. equities – strong legacy of telephone dealing Commodities markets Overall low penetration, perhaps owing to wide differences between traded contracts on different exchanges Derivatives markets Almost entirely electronic in US and Europe. Foreign exchange market turnover by execution 2010 Total Electronic Methods
But in the quest for speed and volume, the base drivers remain largely common Increase of Trading Volumes Decreasing Latency Co-location eliminates geographic and some system constraints Hardware acceleration further improves Timeframe Overall step changes in system speed (-> milli -> micro -> nano) on both exchange and trader side Direct Market Access Exchanges opening up beyond “traditional” approaches and numbers/types of participants Algo volumes and speeds => greater capacity (and reduced average trade size) Intense Regulatory Scrutiny New legislation, e. g. Dodd-Frank, Consumer Protection Act, is expected to: • increase exchange activity • drive competition /new business • “improve” technology and products SEC developing policies to address disparities between competing exchanges
Traders want first call Exchanges want volume Transparen cy? Technologi cal arms race Impact on Exchanges ? Geography ? Changing competitive dynamics/ barriers Tipping points quickly established i. e. must be in it to win it New products/ extended services? Prices? Counterparties? Business models? Price /cost efficiency? As a result of recent/pote ntial trends…. # players – bulge or breadth? Tightening /universal regulations ? Frog or prince? Liquidity always greater? Volatility always reduced? Increased “local” activity? Operational risk? .
First issue: Financial stability and computer based trading
No evidence that HFT/AT has increased volatility A number of empirical studies support this. For example: 1) Jovanovic & Menkveld (2011) compare the volatility of Dutch and Belgian stocks before and after the entry of one HFT firm in the Dutch stocks at Chi-X & Euronext. They find that the relative volatility of Dutch stocks declines slightly.
2) Linton (2011): “The period 2008/2009 was one of great macroeconomic uncertainty, which resulted in a big increase in volatility. This fundamental volatility has since decreased”
However, in specific circumstances, a key type of mechanism can lead to significant instability in financial markets with computer based trading (CBT): self-reinforcing feedback loops can amplify internal risks and lead to undesired interactions and outcomes.
Feedback loop 1 - hedge For example: the portfolio-insurance-led market decline of 1987. Hedge feedback loop
Feedback loop 2: risk May 6 2010 (I):
Feedback loop 3: volume May 6 2010 (II): Volume feedback loop
Feedback loop 4: shallowness
Feedback loop 5: quote delay
Feedback loop 6: Systemic divergence
Second issue: Impact on liquidity, price efficiency/discovery and transaction costs Oliver Linton, Cambridge University
Improved liquidity? 1) Hendershott, Jones, and Menkveld (2011) use the phased automation of the NYSE to measure the effect of algorithmic trading on liquidity. They found that algorithmic trading improves liquidity and enhances the informativeness of quotes.
2) Payne (2011): Bid ask spreads declined while book depth has increased. LSE order book best spreads in FTSE 100 stocks (Basis Points) LSE order book depth available at the best quotes, (GBP) FTSE 100 stocks
3) Menkveld (2011): general improvement in liquidity supply (bidask spread and depth at the best quotes)
4) Linton (2011) looks at time series evidence from FTSE All Share index and individual stocks. He finds that after 2008 illiquidity increased considerably (although still much lower than in 2000). Illiquidity has since come down from this short term high. Illiquidity
Better price efficiency and discovery Some empirical support e. g. – Castura et al. (2010) investigate trends in variance ratios on the Russell 1000 and 2000 stocks over the period 2006 to 2010. 50 They show that efficiency has improved over time.
• Hendershott, Jones, and Menkveld (2011): more algorithmic trading leads to more efficient price quotes. • Brogaard investigated 120 Nasdaq stocks. He estimates that in the absence of HFTs, a trade of 1, 000 shares would cause the price to move an additional $. 056. He argues HFT contributes more to price discovery than do non-HFT activity.
Linton (2011) provides evidence based on daily UK equity data (FTSE Allshare). He computes variance ratio tests and measures of linear predictability (inefficiency) for each year from 2000 -2010. He finds no trend in efficiency in the UK market, whether good or bad.
Falling transactions costs Angel et al. (2010) show that average retail commissions in the USA have decreased between 2003 and 2010, a period relevant for inferring the effects of computer trading.
Effect of entry the entry of Chi-X into the market for Dutch index stocks in 2007/2008 had an immediate and substantial effect on trading fees for investors through: – the lower fees that Chi-X charged – the consequent reduction in fees that Euronext offered. [Menkveld (2011)]
Technology
Three things about trading and technology (1) Speed has always mattered to traders – early birds, worms (2) Technology has always increased speeds & reduced delays – horses, pigeons, telegraphs, tickers, telephones, internet (3) In the past decade, trading technology has gone superhuman – We’re not in Kansas any more, Toto
Three major technology shifts that are happening now (1)Using custom silicon to “bypass the PC” – field programmable gate arrays (FPGAs), software defined silicon (2) From multi-core to many-core • my GPU trumps your CPU (3) Cloud computing & remotely hosted services (e. g. EC 3) – Pay-as-you-go access to supercomputers
Three major consequences: Depopulation of the trading floors – One computer, one man, one dog Lowering of barriers to entry, rise of new hubs – New global exchanges built from BRICs Extreme failures of risky technology via normalisation of deviance – “Challenger, go at throttle-up”
Failures in risky technology, normalization of deviance 1984 1997 2005
Why The Failures?
http: //www. nanex. net/Strange. Days/06082011. html
What the Dickens? ? ? ? ? ?
Unexplained mini-flash crashes – e. g. in single stocks despite volatility bands/circuit breakers EU indices: unexplained slump in index futures on 27/12/10 Euro Stoxx and DAX: only -2% thanks to circuits breakers CAC 40: -4% in less than 3 min, -3% in a few seconds
Rethinking the Financial Network, Redux… (1) Map Understand the system that we have, its network & dynamics (2) Manage Develop policies that are suited to the map, keep the map updated (3) Modify Alter the network to make it less risky, more resilient
Possible Scenarios S 4 • • Globalisation advances Competing world powers and economic models financial elite [Concentration of capital] • In Asia, retail trading of derivatives and other synthetics explodes – copied in the West [Democratisation] • New instruments and online exchanges for company Distributed technologies • Companies in much of Asia owned by state and S 1 [Global economic growth] [Geopolitics] financing • Tech-literate new trading generation [Education, Investor profile] • Competition leads to technology innovation, pressure on traditional trading margins [Competion/lower costs, Business model innovation] • Interlinked global exchanges, single exchange view [Market structure] • Competing trading (‘Trade. Station’) and clearing (‘Clear. Pal’) components [Asset classes, Technology] [Social media/networks, Settlement/execution] • Investment in systemic risk surveillance [Risk management] Closed systems Open systems S 3 • Economic systems worldwide retrench in the face of severe governing economic and political systems challenges [Politics/geopolitics] [Global economic growth] • Rebalancing of capital and trading volumes to • Pressure on exchanges, trading firms, leads to consolidation, emerging markets rebundling, monopolies [Emerging economy] [Market structure] • Regional exchanges dominate, interconnected in a [Greater use of synthetics, Financial engineering] • HFT grows. Churning. Copycat strategies [Competition, lower costs] • Endogenous feedback loops create risk [Risk management] Centralised technologies • Proliferation of synthetic products with opaque structures, responding to demand for ‘safe returns’ S 2 • International institutions play an increasing role in carefully regulated system [Market structure] • Responding to low beta margins, more macro strategies lead to correlation, lower volumes [Asset classes, information homogeneity]
Policy Options 1. Harmonised circuit breakers 2. Different types of circuit breakers 3. Minimum obligations for market makers 4. Tick sizes 5. Central Limit Order Book (CLOB) vs. Exchange order books 6. Real time surveillance 7. Market abuse surveillance 8. Maker-taker fees 9. Minimum resting times 10. Order preference 11. Continuous market vs. randomised stop auctions 12. Algorithmic regulation 13. Internalisation 14. Priority rules
- Slides: 93