3. About ActiveState
Founded 1997
2 million developers, 97% of Fortune 1000 rely on
ActiveState
Development, management, distribution solutions
for dynamic languages
Core languages: Python, Perl, Tcl
Other: PHP, Ruby, Javascript
5. About R&R Consulting
• R&R Consulting is a 10-year-old leader in cybernetic finance
• A standards firm for debt capital markets
• Balanced emphasis on risk-responsibility and value creation at each
link in the global supply chain of debt capital
6. What we will cover
• eGov & the Appearance of Transparency
• Case Study: Financial Content at US SEC
• Why Reg AB was not enough
• Python & Financial Models – the next Wave
7. Premise
• “Open” Gov’t Data needs “Open” Tools
• Tool Vendors need Open Source technologies
– Expedite delivery of consumption tools to market
– Ensure consistency in interpretation of the data
– Ensure equal access to tools across the supply
chain
8. So what is eGov?
• The way in which government has to adapt itself to a world in
which most people regularly use the internet
– http://poit.cabinetoffice.gov.uk/poit/
9. Motivation
• Transparency and engagement
– holding government accountable and promoting choice by
informing citizens
• Efficiency and enhanced public services
– enabling re-use of information within the public sector
• Innovation and economic growth
– encouraging and supporting data-driven innovation
10. eGov in US: Open Government Directive
January 2009
• President Obama issued a
memo on transparency
directing his top officials to
develop plans for an Open
Government Directive to
promote transparency,
participation, and
collaboration.
11. Growing availability of Open Gov Data
• US, UK, Australia,
Netherlands, Denmark,
Sweden, Spain ...
• Washington, Madrid, London,
Vancouver, ...
12. Publishing Data on Data.Gov
• XML
– Industry, agency, project-
specific
• PDF
• RDFa
• Shape
• Excel
13. Unraveling Open Data
• It’s more than just documents for people to read
• Need to enable machines to
– traverse, aggregate, analyze, answer
• Tim Berners-Lee's vision of a “Semantic Web”
– "Semantics", "Ontologies”
– RDFa, Linked Data
14. More Context
• Open government data is a reality
– “Open” doesn’t necessarily mean:
• “Equal Access”
• “Free”
• “Transparent”
• Semantic Web is a vision
– Should be nurtured, supported and encouraged
• But we’re in the middle of a Financial Crisis
– We need to work with the tools we have now
15. Financial Content @ US SEC
• A Case Study
• Lessons Learned the hard way
Part One: The Data
http://www.xbrl.org
http://www.xbrlnetwork.com
16. XBRL @ SEC
• A standard format in which to prepare financial reports that can
subsequently be presented in a variety of ways
• A standard format in which information can be exchanged between
different software applications
• Permits the automated, efficient and reliable extraction of information by
software applications
• Facilitates the automated comparison of financial and other business
information, accounting policies, notes to financial statements between
companies, and other items about which users may wish make
comparisons that today are performed manually
17. Why did XBRL make sense for US SEC?
• *Responsible* Publishing of data
• *Easier* for people to consume financial data
• *Solve* some snags other approaches miss
19. It’s still a black box world
• Processing Financial Content
– requires specialized tools
– proprietary processors
• Business Decisions are all about timing
– access to data to make decision that happen in *nano*
seconds
– driven by algorithmic, computerized trading
20. Lesson Learned in First Phase:
Open Data at US SEC
• Appearance of transparency
• Granting of Access is insufficient
• No Open Standard API for XBRL available (yet)
• Asynchronous access to data (latency)
21. But what if the data isn’t enough
• A little structure goes a long way if you combine it with
– A human being with a lot of intelligence/domain
knowledge
– (tools, protocols, means of communication)
– (browser, http, share)
• Complex legalese in Filings & Prospectuses
– Logic embedded in them needs to be interpreted
22. http://en.wikipedia.org/wiki/Subprime
_crisis_impact_timeline
Off to the Next Crisis
Enter the
Financial Ninjas
http://benbittrolff.blogspot.com/2008_12_07_archive.html
23. Seeds of the Latest Crisis
• Industry participants have real problems interpreting how the
cash flow is allocated
• The lack of certainty creates serious valuation problems for
holders of the securities
• Investors and their custodians also have very real settlement
problems
Regulators/Watchdogs
24. ABS is the new black box
s
Ra ting
Assets it Liabilities/Capital Structure
C red
• Corporate credit is stochastic: • WYSIWYG/Not WYSIWYG.
YES/NO. ABS is statistical, not • Deal waterfalls defines value.
stochastic: YIELD/% OF PAR?
• To implement the standard form or
• Corporate data disclosures are defined amortizations, modeling
public. ABS data disclosures are skills and knowledge of finance are
private. required.
• Corporate data are a moving • Waterfalls are not always
target—ratios change with business unidirectional (as we intuit them).
flows. ABS credit-sensitive data is Some are complex, with recursions
taken on a static pool basis. and nonlinearities.
• Corporate disclosures are a snapshot • At origination, Sellers and Buyers are
of performance, easily “doctored.” supposed to be risk-neutral.
ABS data should be updated: when Modeling mistakes totally change
the pool amortizes, so does risk. the game.
25. How a Deal Waterfall Works, Conceptually
• The whole thing is a state
machine that needs to be
precise in every detail; as
generally once the deal is
set up, the trustees and
managing agents have no
discretion
• They have to follow the
script precisely as set out in
the deal prospectus
26. What Real Waterfalls Look Like
(This One Has Loops)
Sample Waterfall (AmeriCredit 2001)
Copyright 2010
26
27. Here, the Master Waterfall Has A Trigger
Distribution Structure:
Calculate Available Funds
using the SWAP Payment
and Yield Maintenance Cap.
The result will be distributed
across all tranches in the
liability structure.
Allocation Structure:
Use a Trigger to determine,
“Which Allocation Plan do I
call, A or B?”
Sample Waterfall (Bear Stearns 2005)
Trigger
Copyright 2010
28. …If Tripped, Pro Rata Becomes Sequential
Sample Waterfall (Bear Stearns 2005)
Distribution
28
29. …Mezzanine Class Credit Enhancement Is
Linked to 60+ Delinquencies
Sample Waterfall (Bear Stearns 2005)
A simple difference linking two
Asset Trigger Event
complex calculations: 60+
days’ delinquent accounts as a
percentage of the current pool
balance and a run-rate target
enhancement percentage.
30. Enter Regulation (“Reg”) AB
The necessary, robust disclosure framework developed by the
SEC for structured securities.
• Necessary because:
– ABS use different (private data) and different measures (static pool
delinquencies, defaults, losses)
– ABS have a different risk/return profile than corporate debt
securities—the risk amortizes as the pool amortizes
• Robust because:
– Reg AB incorporated all key disclosure items as recommended by the
market
– It was strongly endorsed by large pension funds and institutional
investors
32. Regulation AB: Evolving, Closing
Loopholes
,
ure
2005 Version clos ck,
Dis dba 2010 Proposal
t
Fee cemen
• Requires disclosure of static pool or • Requires updated disclosures.
Enf
data in prospectus, website • Lowers the threshold of materiality
disclosure of performance data to 1%
– Delinquency • No shelf registration without risk
– Loss retention
• Sets a 5% standard of materiality in • Waterfalls must be published, as
misleading disclosures part of the data set
33. How Access to Financial Models
helps..
• Today, it’s hard to Price Securities
– complex prospectus documentation
– write your own waterfall program
– data must be mined from investor reports
– determine the cash flow every month
• Proprietary analytic packages that do this, but it’s expensive
– pay 100K USD for the model to find out the securities are fairly priced
• Having the model & updated data freely means:
– Securities can be valued in a far more cost effective way
– Leaves less up to interpretation
– Run scenario’s on the portfolio and access the impact on the underlying
securities in real time
35. What is the SEC asking for?
• Python code to document the waterfall model
– decision logic which determines the cash payouts made to
all the securities attached to the a specific mortgage pool.
– depending on specific events the cash flows distributed to
securities will change over time based on events occurring
in the underlying portfolio
• This is documented in the prospectus which can be > 600
pages of detailed legal language
http://www.pylaw.org/demonstration.txt
36. Why Python makes sense for US SEC
• Open Freely Available now
• Available libraries for working with
complex algorithms
– Numpy, matplotlib, Rpy
• Bindings exist for other proprietary
numerical Financial libraries
• Already widely used for Financial
Modeling
• Python is a very readable language when
coupled with the other proposal that
detailed asset level data be also
provided in machine readable (XML)
format
http://jrvarma.wordpress.com/2010/04/16/the-sec-and-the-python/
37. Next Steps
• Key: Give Feedback to SEC on it’s Proposal
– Depends upon economics, politics, culture and
technology
– Which could easily change in radical ways..
• Through an invention
• Through a insight into a practical application of an existing
technology
• Fostered by Open Collaboration, Open APIs
38. It’s still a black box world
• Collaboration across constituencies..
– “Open” Gov’t Data needs “Open” Tools
• Open Source technologies
– Expedite delivery of consumption tools to market
– Ensure consistency in interpretation of the data
– Ensure equal access to tools across the supply chain