Projecting Portfolio Performance: Improved Diagnostics Become
Critical in the 21st Century
John R. Davies; Signals, Spring
2003
Consumer loan portfolios, both secured and unsecured, comprise
trillions of dollars of receivables globally. The accumulation
of credit at this scale is a relatively new phenomenon and the
annual losses from these huge pools of assets run in the billions
of dollars. For the large portfolios, losses can exceed one billion
dollars even in a normal year.
Despite these conditions, the consumer lending business is still
very profitable. But history has shown that when the economy cycles
downward, the risks to portfolios can become very large. For example,
in the early 1990s when the economy and many financial institutions
were in trouble, the unsecured Citibank Card portfolio began to
move incrementally in the wrong direction at the same time the
mortgage business was taking a large loss. Card profit was notching
down from its typical $1B+ towards $500m, and the write-offs were
moving from their typical $1B+ towards $2B. Had the economy not
turned when it did, it is not clear how long Citicorp would have
survived. With John Reed's leadership, Citicorp made it
through that difficult time by significantly cutting expenses
and taking other aggressive measures such as selling businesses.
The ability to understand what is going on within portfolios
and what that portends for the future has become essential. Just
as portfolio effects in bonds or equities have been harnessed
by modern finance, the time has come to better understand and
predict the future dynamics of large consumer portfolios.Yet today
there are few comprehensive tools that address this area due to
modeling complexities. Portfolio risk modeling for commercial
loans and corporate securities has seen several breakthroughs:
Risk Metrics, the McKinsey model, and KMV come to mind. Meanwhile
portfolio modeling for consumer portfolios remains a challenge,
just as the consumer loan business has reached a point of maturity
and scale that impacts the entire global economy. Experience tells
us that portfolio performance is affected by:
- Economic cycles
- Seasonality
- Actions taken by portfolio managers,marketers
and system operators
- Customer response to opportunities, requirements
and limitations of their credit relationships
- Domestic and international fraud
- Competitive product offerings.
Today new effects have entered the scene and have expanded this
list. Technology advances and global politics make consumer credit
portfolios and customers associated with them vulnerable to dangerous
attacks, which are costly whether mischievous, criminal or terrorist
in nature. Privacy and information security for consumers is being
seriously challenged. The attacks, in turn, have created increased
volatility in fear and greed, further affecting consumer behavior.
Over the years there have been several "quantum leaps"
in the level of analytical sophistication brought to bear upon
the management of consumer loan portfolios. Although Henry Wells
of Spiegel is typically cited as having deployed the first credit
model in the late 1940s, scoring did not gain widespread use until
Bill Fair and Earl Isaac entered the field in the early 1950s.
Even then it was slow going. By the 1970s behavior scores began
to catch on, accompanied in the '80s and '90s by the
widespread use of account management software to enable a more
effective use of scores and an unbiased assessment of the impact
of account management and acquisition strategies on portfolio
performance.
In the late 1980s and early 1990s non-linear algorithmic techniques,
and in particular neural networks, nearest neighbor algorithms,
and genetic algorithms were being considered for operational use
for fraud and bankruptcy detection, as well as other areas of
behavior prediction and forecasting previously considered intractable.
At Citicorp/Citibank we were working to deploy the first neural
networks for credit card fraud detection. Given the economic conditions
at the time, card managers were not of an early-adopter mind set.
Despite very strong early R&D results, we could not move forward.
During that same period Citi's Advanced Technology Group
worked with leading figures of academia and business including
Brian Arthur and John Holland of the Santa Fe Institute, Peter
Fry of Northwestern, and Alan Jost of HNC. Their early efforts
led to new teams at Citibank focused on using emerging non-linear
behavior analytic techniques. As the industry matured in the use
of these non-linear science areas, commercially viable tools became
available. Today, it is unthinkable for an organization to be
engaged in consumer financial management, especially in a market
as sophisticated as the US, without deploying these tools.
Consumer loan portfolios, in the context of the emerging global
economy and related politics, require a thorough understanding
of the underlying fundamentals for their effective management.
It is critical that people responsible for these consumer portfolios
develop and implement the ability to dynamically forecast every
possible determinant of portfolio risk going forward into the
21st Century.
A timely understanding of value enhancing and value destroying
strategies in the markets where an institution operates is critical.
For example, strong capabilities are required to:
- Identify accurately high-risk situations
while they are emerging
- Recognize unusual underlying market characteristic
dynamics
- Evaluate the true potential and risk of
a business.
- Where these capabilities are available, key proactive actions
can be taken to:
- Uncover situations where reported results
are being "managed"
- Proactively address situations where managers
may be either too cautious or too optimistic in their forecasting
- Bring rigor to the budgeting process
- Detect emerging illegal activities, e.
g., fraud and mischief
- Avoid embarrassing public performance
announcements.
Strategic Analytics has developed an operational, unbiased assessment
of exogenous impacts upon portfolio performance that is cleanly
separated from maturation effects or changing originations quality.
Using this technology each determinant of portfolio dynamics is
continuously evaluated looking for patterns and emerging characteristics.
Accurate forecasts are now possible by using a portfolio's unique
experience and behavior in accurate simulations taking the portfolio
through various scenarios. These scenarios use the best practice
and expertise of key management, marketing, operations, and treasury
people responsible for the portfolios.
Analysis of this kind has not been possible historically. Only
over the past decade have combined efforts in technology and business
resulted in such innovative approaches yielding operational solutions.
The timing could not be better given the current state of both
the value of the global consumer loan business and the state of
global security.
John Davies works with technology companies
(including Strategic Analytics and Metacerebra), venture firms,
and scientists to address critical emerging business needs and
opportunities. John co-founded Center for Adaptive Systems Applications
(CASA),a for-profit research and development corporation dedicated
to solving practical industry needs using complex adaptive systems.He
has recently served on the Los Alamos National Laboratory External
Advisory Board,a group dedicated to facilitating the commercialization
of New Mexico's extensive science and technology base. He
is also a board member of the Los Alamos Commerce & Development
Corporation. John resides in New Mexico, spending much of his
time researching family history and thinking about how advanced
simulation techniques might be applied to this field.
|
Top of page