You are on page 1of 14

CUYdU!

1TQ`dYfU>Ugc")(

?VfVcdTYVZ_f_XV_
Z_UVcDTYcZWeV_cVZYVUVdD73

Im folgenden finden Sie eine Liste


der seit dem letzten Newsletter fer-
tiggestellten Schriftenreihen. Falls
Sie Interesse an einer der im folgen-
den angeführten Ausgaben der Re-
port- oder Working Paper Series
haben, so wenden Sie sich bitte an
unsere Homepage oder notfalls an
Frau Mag. Doris Prato (313 36/45
87, doris.prato@wu-wien.ac.at):

CVa`ce?`"%
May 1998
Neal M. Stoughton, Josef Zechner
IPO-Mechanisms, Monitoring and Ownership Strucutre
This paper was accepted for publication in: Journal of Financial Economics
Abstract
This paper analyzes the effect of different IPO mechanisms on the structure of share
ownership and explores the role of underpricing and rationing in determining investors’
shareholdings. We focus on the agency problem that results when large institutions are
the only investors capable of monitoring the firm whereas small shareholders free-ride
on these activities. The major conclusion of this paper is that some well-known aspects
of IPOs may be explained as rational responses by the issuer to the existence of regula-
tory constraints in public capital markets. We find that there is a two-stage offering
mechanism in which an investment banker acting in the interests of the issuer optimally
rations the allotment of shares to small investors in order to capture the benefits associ-
ated with better monitoring by institutions. Importantly, in our model, the existence of
underpricing (and oversubscription) is an indication that the issuer has received a higher
ex ante price than would have been obtained through a competitive Walrasian-type of-
fering process.
1TQ`dYfU>Ugc")( CUYdU"

CVa`ce?`"&
May 1998
Christian Schittenkopf, Georg Dorffner, Engelbert J. Dockner
Volatility Prdiction with Mixture Density Networks
This paper was accepted for publication in: Proceedings of the International Conference
on Artificial Neural Networks, September 2-4 1998, Skövde, Schweden.
Abstract
Despite the lack of a precise definition of volatility in finance, the estimation of volatil-
ity and its prediction is an important problem. In this paper we compare the performance
of standard volatility models and the performance of a class of neural models, i.e. mix-
ture density networks (MDNs). First experimental results indicate the importance of
long-term memory of the models as well as the benefit of using non-gaussian probability
densities for practical applications.

CVa`ce?`"'
May 1998
Thomas Reutterer
Combined Market Structure and Segmentation Analysis Based on Brand Choice
Data: Overcoming the Limitations of Conventional Techniques with Topologically
Ordered Feature Maps
This paper was accepted for publication in: Proceedings of the 1998 AMA Exchange
Colloquium July 23-25, Vienna.
Abstract
The objective and related issues of simultaneously performing competitive market
structure (CMS) and market segmentation analysis is well-documented in the marketing
literature. In this paper, an artificial neural network based approach of Kohonen (1982)
for the formation of topological ordered feature maps in introduced into the context of
brand choice data based combined CMS/segmentation analysis. Selected aspects of the
methodological basis are discussed and in a demonstration study using household-level
brand choice probabilities derived from diary household panel data the failure of con-
ventional methodology is opposed to the promising properties of the suggested neuro-
computing approach.
CUYdU# 1TQ`dYfU>Ugc")(

CVa`ce?`"(
June 1998
Friedrich Leisch, Andreas Weingessel, Evgenia Dimitriadou
Competitive Learning for Binary Valued Data
This paper was accepted for publication in: Proceedings of ICANN ’98, International
Conference on Artificial Neural Networks, Sköde, Sweden, September 2-4, 1998.
Springer Verlag.
Abstract
We propose a new approach for using online competitive learning on binary data. The
usual Euclidean distance is replaced by binary distance measures, which take possible
asymmetries of binary data into account and therefore provide a ``different point of
view'' for looking at the data. The method is demonstrated on two artificial examples
and applied on tourist marketing research data.

CVa`ce?`")
June 1998
Peter Tino
Spatial Representation of Symbolic Sequences through Iterative Function Systems
This paper is under review for publication in: IEEE Trans. on Systems, Man, and Cy-
bernetics (B)
Abstract
Jeffrey (1990) proposed a graphic representation of DNA sequences using Barnsley's
iterative function systems. In spite of further developments in this direction (Oliver et.
al, 1993; Roman-Roldan et. al, 1994, Li, 1997), the proposed graphic representation of
DNA sequences has been lacking a rigorous connection between its spatial scaling char-
acteristics and the statistical characteristics of the DNA sequences themselves. We 1)
generalize Jeffrey's graphic representation to accommodate (possibly infinite) sequences
over an arbitrary finite number of symbols, 2) establish a direct correspondence between
the statistical characterization of symbolic sequences, 3) show that for general symbolic
dynamical systems, the multifractal spectra in the sequence space coincide with the
spectra on spatial sequence representations.
1TQ`dYfU>Ugc")( CUYdU$

CVa`ce?`"*
June 1998
Peter Tino, Georg Dorffner
Constructing finite-context sources from fractal representations of symbolic se-
quences
This paper under review publication in: Machine Learning
Abstract
We propose a novel approach to constructing predictive models on long complex sym-
bolic sequences. The models are constructed by first transforming the training sequence
n-block structure into a spatial structure of points in a unit hypercube. The transforma-
tion between the symbolic and Euclidean spaces embodies a natural smoothness as-
sumption (n-blocks with long common suffices are likely to produce similar continua-
tions) in that the longer is the common suffix shared by any two n-blocks, the closer lie
their point representations. Finding a set of prediction contexts is then formulated as a
resource allocation problem solved by vector quantizing the spatial representation of the
training sequence n-block structure. Our predictive models are similar in spirit to vari-
able memory length Markov models (VLMMs). We compare the proposed models with
both the classical and variable memory length Markov models on two chaotic symbolic
sequences with different levels of subsequence distribution structure. Our models have
equal or better modeling performance, yet, their construction is more intuitive (unlike in
VLMMs, we have a clear idea about the size of the model under construction) and easier
to automize (construction of our models can be done in a completely self-organized
manner, which is shown to be problematic in the case of VLMMs).

CVa`ce?`#!
June 1998
Peter Tino, Georg Dorffner, Igor Farkas
Building stochastic prediction machines from fractal representations of symbolic
sequences
This paper is under review for publication in: NIPS’98
Abstract
We propose a novel approach to constructing predictive models on symbolic sequences.
The models are constructed by first transforming the n-block structure of the training
sequence into a spatial structure of points in a unit hypercube, such that the longer is the
common suffix shared by any two n-blocks, the closer lie their point representations.
Such a transformation embodies a natural smoothness assumption -- n-blocks with long
common suffices are likely to produce similar continuations, whereas n-blocks with
different suffices may lead to a range of different future scenarios. Finding a set of pre-
diction contexts is then formulated as a resource allocation problem solved by vector
CUYdU% 1TQ`dYfU>Ugc")(

quantizing the spatial n-block representation. Our predictive models are similar in spirit
to variable memory length Markov models (VLMMs). We compare the proposed mod-
els with both the classical and variable memory length Markov models on two chaotic
symbolic sequences with different levels of subsequence distribution structure. Our
models have a superior performance, yet, their construction is more intuitive (unlike in
VLMMs, we have a clear idea about the size of the model under construction) and easier
to automize (construction of our models can be done in a completely self-organized
manner, which is shown to be problematic in the case of VLMMs).

CVa`ce?`#"
June 1998
Peter Tino, Georg Dorffner
Recurrent neural networks with Iterated Function Systems dynamics
This paper was accepted for publication in: International ICSC/IFAC Symposium on
Neural Computation (NC’98), Technical University in Vienna, Austria, September 23-
25, 1998.
Abstract
We suggest a recurrent neural network (RNN) model with a recurrent part correspond-
ing to iterative function systems (IFS) introduced by Barnsley (1988) as a fractal image
compression mechanism. The key idea is that 1) in our model we avoid learning the
RNN state part by having non-trainable connections between the context and recurrent
layers (this makes the training process less problematic and faster), 2) the RNN state
part codes the information processing states in the symbolic input stream in a well-
organized and intuitively appealing way. We show that there is a direct correspondence
between the entropy spectra characterizing the input stream and the spectra of general-
ized dimensions of activations inside the RNN state space. We test both the new RNN
model with IFS dynamics and its conventional counterpart with trainable recurrent part
on two chaotic symbolic sequences. In our experiments, RNNs with IFS dynamics out-
perform the conventional RNNs with respect to information theoretic measures com-
puted on the training and model generated sequences.
1TQ`dYfU>Ugc")( CUYdU&

H`c\Z_XARaVc?`(
April 1998
Sara Dolnicar, Friedrich Leisch, Andreas Weingessel, Christian Buchta,
Evgenia Dimitriadou
A Comparison of Several Cluster Algorithms on Artificial Binary Data Scenarios
from Travel Market Segmentation
Introduction
Social scientists confronted with the problem of segmenting individuals into plausible
subgroups usually encounter two main problems: First: there is very little indication
about the correct choice of the number of clusters to search for. Second: different cluster
algorithms and even multiple replications of the same algorithm result in different solu-
tions due to random initializations and stochastic learning methods. In the worst case
numerous solutions are found which all seem plausible as far as interpretation is con-
cerned. The consequence is, that in the end clusters are postulated that are in fact “cho-
sen” by the researcher, as he or she makes decisions on the number of clusters and the
solution chosen as the “final” one. In this paper we concentrate on the power and stabil-
ity of several popular clustering algorithms under the condition that the correct number
of clusters is known. Artificial data sets modeled to mimic typical situations from tour-
ism marketing are constructed. The structure of these data sets is described in several
scenarios, and artificial binary data are generated accordingly. These data, ranging from
very simple to more complex, real-data-like structures, enable us to systematically ana-
lyze the “behavior” of the cluster methods. Section 3 gives an overview of all cluster
methods under investigation. Section 4 describe our experimental results, comparing
first all scenarios and then all cluster methods. To accomplish this task, several evalua-
tion criteria for cluster methods are proposed. Finally: Sections 5 and 6 draw some
conclusions and give an outlook on future research.

H`c\Z_XARaVc?`)
May 1998
Robert Heinkel, Alan Kraus, Josef Zechner
The Effect of Green Investment of Corporate Behavior
Abstract
This paper examines the effect of exclusionary “ethical investing” on corporate behav-
ior. We assume a simple world with a finite number of firms that possess one of two
production technologies, “clean” or “polluting”. We also assume two types of risk
averse investors, “neutral” and “green”. Neutral investors are willing to consider in-
vestments in clean or polluting firms relative to clean firms. When this price differential
just covers the cost of a firm switching from the polluting technology to the clean tech-
nology, share-price-maximizing firms will change their technologies, e.e., corporate
behavior is affected by ethical investing. However, our numerical examples, constructed
CUYdU' 1TQ`dYfU>Ugc")(

with empirically reasonable parameter values, indicate that a large proportion of green
investors is required to overcome reasonable switching costs. In addition, the average
cost of a capital across all firms in the economy is shown to be nonmonotonically con-
cave in the proportion of green investors. That is, the economy’s cost of capital is lowest
if there are either very few green investors, of very many. Our casual empirical observa-
tions lead us to conclude there are very few green investors.

H`c\Z_XARaVc?`*
May 1998
Neal M. Stoughton, Kit Pong Wong, Josef Zechner
IPOs and Product Quality
Abstract
In this paper we study a product market motive for going public. We model a firm that
produces a good with unknown quality. When a firm makes an IPO, consumers use in-
formation incorporated into the stock market price to make an inference about the pro-
duct quality. Firms that produce higher quality products are more likely to proceed with
an IPO. The model is extended to allow for product market competition with a rival
firm. We study the announcement effects on the rival’s stock price when it is publicly
traded and the likelihood of a competitor going public when the rival is privately held.
Our results predict the presence of “hot issue” markets depending on the distribution of
the market size uncertainty.

H`c\Z_XARaVc?`"!
May 1998
Vojislav Maksimovic, Alex Stomper, Josef Zechner
Capital Structure, Information Acquisition and Investment Decisions in Industry
Equilibrium
Abstract
We analyze how financial structure affects the incentives of firms to acquire costly in-
formation about demand or cost conditions in the industry. In industry equilibrium, the
value of information obtained by building prototypes or postponing investment deci-
sions is determined by the rival firms’ use of such information. Although the number of
firms which acquire costly information adjusts until they have the same value as unin-
formed firms, their cash flows generally differ in their riskiness. This implies that a
firm’s financial structure is a major determinant of its information acquisition strategy.
As a consequence there is a relationship between the distribution of financial structures
in an industry and the equilibrium in the market for information. We characterize the
distribution of financial structures that supports the industry equilibrium. We find that if
the new investment must be financed internally, then leverage is negatively related to
1TQ`dYfU>Ugc")( CUYdU(

information acquisition expenditure whereas if the new investment is financed exter-


nally, the relationship is the converse. We show how the existence of a tax advantage of
debt affects the firms’ incentives to acquire information and the number of informed
firms.

H`c\Z_XARaVc?`""
May 1998
Christian Schittenkopf, Georg Dorffner, Engelbert J. Dockner
Identifying Stochastic Processes with Mixture Density Networks
Abstract
In this paper we investigate the use of mixture density networks (MDNs) for identifying
complex stochastic processes. Regular multilayer perceptrons (MLPs), widely used in
time series processing, assume a gaussian conditional noise distribution with constant
variance, which is unrealistic in many applications, such as financial time series (which
are known to be heteroskedastic). MDNs extend this concept to the modeling of time-
varying probability density functions (pdfs) describing the noise as a mixture of gaus-
sians, the parameters of which depend on the input. We apply this method to identifying
the process underlying daily ATX (Austrian stock exchange index) data. The results
indicate that MDNs modeling a non-gaussian conditional pdf tend to be significantly
better than traditional linear methods of estimating variance (ARCH) and also better
than merely assuming a conditional gaussian distribution.

H`c\Z_XARaVc?`"#
March 1998
Thomas Reutterer
Panel-Data Based Competitive Market Structure and Segmentation Analysis Using
Self-Organizing Feature Maps.
Abstract
In this paper the ‚Self-Organizing (Feature) Map‘ (SOM) methodology as originally
proposed by Kohonen (1982) is employed in the context of Competitive Market Struc-
ture (CMS) and segmentation analysis using household-specific brands preferences de-
rived from diary panel data as input patterns for SOM training. The adaptive SOM algo-
rithm results in a representation of competitive structures among rival brands at the
segment-level, i.e. for submarkets with (internally) more homogeneous brand choice
features. This property of SOM-based CMS/segmentation analysis allows the two inter-
dependent tasks to be performed simultaneously, as it is frequently claimed in the mar-
keting literature.

H`c\Z_XARaVc?`"$
CUYdU) 1TQ`dYfU>Ugc")(

June 1998
Friedrich Leisch, Andreas Weingessel, Kurt Hornik
On the Generation of Correlated Artificial Binary Data
Abstract
The generation of random variates from multivariate binary distributions has not gained
as much interest in the literature as, e.g., multivariate normal or Poisson distributions.
Binary variables are important in many types of applications. Our main interest is in the
segmentation of marketing data, where data come from customer questionnaires with
‘‘yes/no’’ questions. Artificial data provide a valuable tool for the analysis of segmenta-
tion tools, because data with known structure can be constructed to mimic situations
from the real world (Dolnicar et al. 1998). Questionnaire data can be highly correlated,
when several questions covering the same field are likely to be answered similarly by a
subject. In this paper we present a computationally fast method to simulate multivariate
binary distributions with a given correlation structure. The implementation of the algo-
rithm in R, an implementation of the S statistical language, is described in the appendix.

D732\eZgZezeV_"**)&BfRceR]Z^vSVcS]ZT\

Folgende Aufstellung bietet einen Überblick über Aktivitäten des SFB in den Vormo-
naten:

April
01 16.00 Ini 2 – Statistical H.H. Bock Wahrscheinlichkeitsmodelle in
Modelling (Uni Aachen) der Cluster Analyse
02 16.15 Ini 2 – Statistical Monika Dörfler Strukturtheorie für rekurrente
Modelling neuronale Netze
23 09.00- Ini 3 – Market C. Buchta Adaptive Modelling Workshop:
12.00 Segmentation S. Dolnicar Zwischenbericht über Quanti-
and Product K. Grabler sierungsergebnisse für autologi-
Positioning A. Weingessel stisch generierte Binärdaten
13.00- F. Leisch Bericht zum Ausbaustand des
15.00 M. Natter Simulationsmodells ‚Biermarkt‘
23 17.00 Ini 2 – Statistical Seminar Strasser u. Schacher-
Modelling mayer
24 13.00- Ini 3 – Market H. Mitlöhner Verteiltes Rechnen mit PVM-
Open Segmentation Parallel Virtual Machine; Pro-
End and Product grammierung adaptiver Agenten
Positioning an Finanzmarktbeispielen
1TQ`dYfU>Ugc")( CUYdU!

29 14.00 J. van Schuppen Systemtheory and systemidenti-


(CWI Amsterdam) fication of compartmental
systems

May
07 16.15 Ini 2 – Statistical H.-G. Feichtinger Ökonometrisches Seminar:
Modelling (UNI Wien) Gabor Analysis: Methoden der
lokalen Fourier Transformation
07 17.00 Ini 2 – Statistical Seminar Strasser u. Schacher-
Modelling mayer
08 14.00- Adaptive Friday Duffie Darell Research Initiative 6
15.30

13 16.00 Ini 2 – Statistical Jörg Rahnenführer Simple competitive learning:


Modelling (Uni Düsseldorf) der Gewissenalgorythmus

June
17 16.00 Ini 2 – Statistical Usula Gather/Uni Robuste Dimensionsreduktion
Modelling Dortmund in Regressionsmodellen
19 13.30 Adaptive Friday Research Initiative 6

19 18.00 Mitarbeiterver-
sammlung
24 16.00 Ini 2 – Statistical Iris Pigeot Grafische Modelle
Modelling
26 16.00 Ini 3 – Market Adaptive Modelling Workshop
Segmentation
and Product Po-
sitioning
Ini 2 – Statistical
Modelling

D73>ZeRcSVZeVcfaUReV

Als neue SFB-Mitarbeiter begrüßen wir


CUYdU!! 1TQ`dYfU>Ugc")(

Mag. Thomas Dangl.


Er hat seine Tätigkeit am 01. 04 1998 aufgenommen und stellt seitdem seine Dienste
dem SFB ganztags als wissenschaftlicher Mitarbeiter der Initiative 6 zur Verfügung.
Ebenso freuen wir uns
Univ.-Doz. Gerhard Sorger
seit 08. 05. 1998 als stellvertretenden Projektleiter für Herrn Prof. Otrubas Initiative 6
‚Artificial Financial Markets‘ beim SFB willkommen heißen zu dürfen!
Aus dem SFB ausgeschieden sind
Mag. Thomas Hudetz (per 01.04. 1998, Initiative 2)
Mag. Reinhard Zimmermann (mit 31.05. 1998, Initiative 6)

D73<]Rfdfc+
$ERXVKVZeWc:_eVcUZdkZa]Z_RcZeze

Die SFB-Klausur vom 01. 10. 98 – 03. 10. 98 im Kurort Reichenau an der Rax bietet
Ihnen die einmalige Gelegenheit, bereits bestehende Arbeitsbeziehungen zu intensivie-
ren oder neue Verbindungen mit unbekannten Kollegen zu knüpfen. Wagen Sie einen
weiteren Schritt in unerforschte Tiefen des SFB, indem Sie mit Forschungsmitarbeitern
über vergangene Projekte diskutieren sowie Ideen über neue interdisziplinäre Ziele und
Kooperationsmöglichkeiten vorantreiben. Als Resümee wird am 02. 10. 98 eine Über-
sichtspräsentation stattfinden, bei der FWF-Präsident Arnold Schmidt zu Gast sein wird.

Beherbergt werden Sie in Einzelzimmern im Hotel Marienhof, umgeben von einem


prachtvollen Park, inmitten des Rax-Gebirges. Alle Zimmer sind mit Bad oder Dusche,
WC, Telefon, Radio, Minibar und Fernseher ausgestattet. Teilweise mit Balkon und
Blick auf die Rax. Das Hotel bietet freundliche Tagungsräume für Seminare und indivi-
duelle Betreuung der Teilnehmer sowie ein reichhaltiges Gastronomieangebot. Das sich
im Hotel befindliche, renommierte Kurzentrum mit Hallenbad wird wegen seiner exklu-
siven Therapiemöglichkeiten bei Stressymptomen empfohlen.

Sollten Sie sich bis heute noch nicht via e-mail angemeldet haben, tun Sie dies bitte jetzt
(über unsere termin-homepage oder mail an doris.prato@wu-wien.ac.at), da der Anmel-
deschluß zur Teilnahme an der Klausur gleichzeitig mit dem Erscheinen dieses News-
letters festgesetzt wurde. Die Teilnahmekosten - ausgenommen sind Extras, die Sie sich
gönnen möchten - übernimmt selbstverständlich der SFB.

6Z_]RUf_Xkf^D73H`c\dY`aUVc
>Rc\eW`cdTYVc"!±"#@\e`SVc"**)
1TQ`dYfU>Ugc")( CUYdU!"

Diese Kooperationsveranstaltung unseres SFB-AM mit dem Berliner SFB (im großen
Sitzungssaal der WU Wien) dient dem Informationsaustausch über Forschungsergebnis-
se und -vorhaben, der Suche nach gemeinsamen Forschungsinteressen und möglichen
gemeinsamen Projekten. Mitarbeiter ALLER INITIATIVEN des SFB sind herzlich ein-
geladen, daran teilzunehmen. Um ausreichende Mengen an Kaffee und ähnlich banaler
Ressourcen zu garantieren, bitten wir allerdings um Anmeldung (e-mail).

Programmübersicht
(Die Vortragstitel sind derzeit noch nicht vollständig bekannt)

Samstag, 10. 10.


13.00 – 13.30 Begrüßung
13.30 – 15.00 Vorträge des SFB Berlin
15.00 – 15.30 Kaffeepause
15.30 – 17.00 Vorträge des SFB Berlin

Sonntag, 11. 10.


9.00 – 10.30 Harald Hruschka:
Dynamische Markenwahlmodelle, Modellierung der Mar-
kenwahl mittels Multilayer-Perzeptronen
10.30 – 11.00 Kaffeepause
11.00 – 12.30 Harald Hruschka:
Nichtlineare Verallgemeinerungen des MNL-Modells
12.30 – 13.30 Mittagspause
13.30 – 15.00 SFB-AM Initiative 3:
Kombinierte Positionierungs / Segmentierungs-Analysen
15.00 – 15.30 Kaffeepause
15.30 – 17.00 SFB-AM Initiative 3:
Vektorquantisierung mit künstlichen Binärdaten

Montag, 12. 10.


9.00 – 10.30 Plenarsitzung:
Gemeinsamkeiten – Unterschiede – Kooperationsmöglich-
keiten
10.30 – 11.00 Kaffeepause
11.00 – 12.30 Diskussion
12.30 Ende des Workshops
CUYdU!# 1TQ`dYfU>Ugc")(

Von hier aus geht’s zu allen SFB-Informationen:

:^acVddf^
Herausgeber und Medieninhaber:
Spezialforschungsbereich ‘Selbstlernende Modelle in den Sozial-
und Wirtschaftswissenschaften’
Wirtschaftsuniversität Wien
Augasse 2-6, A-1090 Wien
Der Spezialforschungsbereich ‘Selbstlernende Modelle in den Sozial-
und Wirtschaftswisschenschaften’ wird aus Mitteln des Fonds zur
Förderung der wissenschaftlichen Forschung finanziert.
Verlags- und Herstellungsort: Wien
Für den Inhalt verantwortlich: Josef Mazanec (Sprecher des SFB)
Redaktion & Produktion: Sara Dolnicar
Grundlegende Richtung:
SFB-internes Informationsmedium
2URaeZgV?Vhd
]VeeVc
Jahrgang 2 – Nr. 1 - Juni 1998 http://www.wu-wien.ac.at/am/news.html

Inhalt

Neuerscheinungen in der Schriftenreihe des SFB .............. 1


SFB Aktivitäten 1998 (5. Quartal) im Überblick ................... 9
SFB-Mitarbeiterupdate ...................................................... 10
SFB-Klausur: 3 Tage Zeit für Interdisziplinarität................ 11
Einladung zum SFB-Workshop der Marktforscher ........... 11
Impressum......................................................................... 13

You might also like