Credit Management

[George Richardson]

On Thu, 1 Aug 1996 PADDY@cxa.ie wrote:
[...]
> SD teaches us that structure influences behaviour. I am interested
> in understanding how the structure of an individuals personal
> circumstances results in bahaviour patterns which lead to bad debts.
[...]

My reply is that dynamics is not particularly good for analyzing how external
situations drive internal system behavior, what might be called
"exogenous" theory. System dynamics models and closed-loop systems
thinking are most powerful investigating "endogenous" theory in which the
internal structure of some perceived system can be seen to cause or at least
exacerbate the problem behavior of interest.

That suggests that the problem statement needs a bit of revision. We
want to draw the system boundary widely enough so that "the individuals
personal circumstances" which are involved in a systems view of "bad debt
behavior" are both causes and effects of variables in the "person bad
debt system."

Can the poser of this problem expand on the problem definition in such a
way, and maybe include some dynamics to be explained more or less
endogenously by a systems view or system dynamics model?

...GPR

----------------------------------------------------------------------
George P. Richardson G.P.Richardson@Albany.edu
Rockefeller College of Public Affairs and Policy Phone: 518-442-3859
University at Albany - SUNY, Albany, NY 12222 Fax: 518-442-3398
-----------------------------------------------------------------------

[PADDY@cxa.ie]

Greetings All,

Does anyone know if SD has been successfully applied to the area of
credit rating?

SD teaches us that structure influences behaviour. I am interested
in understanding how the structure of an individuals personal
circumstances results in bahaviour patterns which lead to bad debts.

Any guidance or suggestions would be greatly appreciated.

Thanks,

Patrick Murphy, paddy@cxa.ie
Delta Performance Systems, http://www.dpsnet.com

[Patrick Murphy]

> George Richardson replied:
> That suggests that the problem statement needs a bit of revision. We
...
George,

Thank you for replying to my query. I take your point about re-defining the system
boundary. In fact this is my main problem at present. Maybe I should offer some more
information . . .

Financial institutions are, among other things, in the business of unsecured lending.
In this respect, a number of software credit scoring systems have been developed with
the purpose of helping to minimise bad debts. From what I understand, the software
comprises of a series of questions with different weightings and a black box which
generates a score - based on question responses. This score represents the likelihood
of the customer meeting all his loan repayments. The software ships with default
weightings and the challenge to the financial institution is to adjust these weighting
to better reflect their own target population. Unfortunately, this is not a science and
it is difficult to know how key variables such as age, number of years at present
employment, etc. relate to what has become known as credit-worthiness.

The credit scoring systems apply statistical techniques to short-term historical data
with a view to predicting short-term future behaviour. They are interested only in
what happens, not why it happens. Indeed, some claim that there is no causal
relationship as such.

My problem is, faced with the challenge of applying the system dynamics method to this
situation, where do I begin. In the first instance, I thought that this problem was
ideally suited for SD - a complex dynamic system coupled with a lack of understanding.
But what exactly should I be trying to model - the customer, the existing system, the
relationship between the two? Indeed, is this problem suitable for SD at all or should
I use another approach - such as CBR?

Again, any guidance will be much appreciated.

Regards,

Patrick.
Patrick Murphy
haven@iol.ie

["Robert M. Kane"]

Patrick Murphy asks whether modeling the internal motivations of an individual
is appropriate to SD..

Im not an SD insider but have made some observations concerning what insiders
are discussing and where I think they may mean to go.

I believe that SD the right place to model individual behavior based on
complex external economic forces. The problem of predicting individual
behavior is simply identifying the factors that operate on the individual and
then weighting them appropriately. Specifically, when faced with the options
of making a payment on a recreational vehicle or the exclusive alternative of
making a dental appointment, does the actor weight is health above
recreational opportunity? Does the actor believe that the lender will
actually repossess the vehicle? Does the dental problem affect the appearance
of the actor? Is the actor single and thus motivated to use the vehicle as a
social feature? Assuming that the vehicle attracts potential partners, does
the actor perceive that the dental problem counteracts the added
attractiveness of the vehicle? Of course the questions (and factor/weighting
pairs) continue without end.

The complexity of the underlying issues is beyond the discussions I have seen
here. SD aims to takes the weights and factors as described by subject-matter
experts and package them into a dynamic network. Practioners in this
discussion dont seem to be motivated to discover the underlying principles.

I thought that SD might support my need to understand the complexity of third-
world farmer participation in organized cooperative agriculture but have
pretty much given up the idea. I see now that I must first identify the
factors and weights (which probably is sociologically and psychologically
based). If I can accomplish this then SD might provide a structure to test
and predict impacts of varying external inputs.

My conclusion is that given the highly dynamic problem presented by individual
decision-making processes, a relatively static model is of little other than
academic value. An effective snap-shot of the dynamic process can provide
useful paradigms that are the basis of effective people-management activities.
This is a legitimate academic endeavor but probably isnt a direct path to a
practical solution.

I continue to follow the discussions and find it interesting to see the sorts
of problems presented.

Robert M. Kane, North American Coordinator
Institute for Sustainable Agriculture in the Tropics (SIAT)
Jaen, Peru
robert_kane@ds.Cubic.COM
http://www.ran.org
an
an_campaigns/amazonia/coffee.html

["Robert J. Glitz, LtCol, USAF" <]

Robert M. Kane wrote:
>
> I believe that SD the right place to model individual behavior based on
> complex external economic forces. The problem of predicting individual
> behavior is simply identifying the factors that operate on the individual and
> then weighting them appropriately.

I certainly hope youre right! Ive been pitching SD as (among other things) a way to
methodically relate "hard" variables (headcount, budget levels, force structure) with
"soft" variables like motivation, perception of quality, propensity to enlist, etc.

We are starting a fairly large project to investigate the structure and dynamics of
diversity recruitment and enlistment levels in the Air National Guard. To build a
robust model, we are going to have to add stocks that measure things like
personal satisfaction, sense of belonging, and prejudice. We hope to measure the
effect of low female and minority levels on our larger system by simulating the stocks
of political influence, public perceptions of fairness, etc.

Many of these stocks may not be measureable in standard units, but we intend to work
hard to build shared mental models of the interrelationships and incorporate them in a
Powersim MFS.

Robert Glitz, LtCol, USAF
rglitz@erols.com