Validation of the coherent market hypothesis using neural networks and JSE securities exchange data

Myburgh, Gustav (2001-12)

Thesis (MBA)--Stellenbosch University, 2001.

Thesis

ENGLISH ABSTRACT: Much research effort has been spent over the past few decades in the field of capital market analysis and modelling. This research was mostly based on static linear models or derivatives thereof such as the Efficient Market Hypothesis, the Capital Asset Pricing Model and the Arbitrage Pricing Theory. This study project takes an interesting look at a contemporary capital market hypothesis, which is fundamentally based on a non-linear statistical model. The Coherent Market Hypothesis (CMH) was first formulated by Tonis Vaga in 1990. It is based on a theory of social imitation, taking factors such as the underlying fundamental situation and the level of crowd behaviour into account. It also includes the phenomenon of “random walk” as a special case. The CMH departs from the premise of rational investors and normally distributed share returns. In turn, it offers a series of “market states” ranging from trendless (random walk), through unstable transition into coherent bull or bear phases and ultimately into periods of chaotic fluctuation (panics and crashes). The CMH is mathematically formulated and therefore it offers many opportunities for experimentation. This study project is an investigation of the validity and application of the CMH using real JSE data. Artificial Neural Networks were applied as computational aids. The main objective was to demonstrate the CMH’s usefulness as a forecasting tool in both a quantitative as well as qualitative capacity. The results of the quantitative analysis were not as significant or valuable as initially expected. However, the usefulness of the CMH was demonstrated in a more qualitative sense. It is shown that the CMH offers a rich theoretical framework for interpretation, understanding and recognising of market dynamics.

AFRIKAANSE OPSOMMING: Gedurende die afgelope paar dekades is aansienlike hoeveelhede navorsing gedoen in die veld van kapitaalmark analise en modellering. Hierdie navorsing was hoofsaaklik gebaseer op statiese, lineêre modelle of afgeleides daarvan, naamlik die Efficient Market Hypothesis, die Capital Asset Pricing Model en die Arbitrage Pricing Theory. Hierdie werkstuk kyk vanuit ‘n interessante oogpunt na ‘n meer hedendaagse kapitaalmark hipotese wat fundamenteel gebaseer is op ‘n nie-lineêre statistiese model. Die Coherent Market Hypothesis (CMH) is oorspronklik geformuleer deur Tonis Vaga in 1990. Dit is gebaseer op ‘n teorie van sosiale nabootsing en dit neem faktore in ag soos die onderliggende fundamentele situasie asook die vlak van groepgedrag. Die verskynsel van “random walk” word ook ingesluit as ‘n spesiale geval. Die CMH wyk af van die aanname dat beleggers rasioneel optree asook van die aanname dat aandeel opbrengste normaal verspreid is. In teendeel, die CMH omvat ‘n reeks marktoestande wat wissel van die tendenslose (random walk) deur onstabiele oorgang na koherente bul- of beerfases en uiteindelik in tydperke van chaotiese skommelings (markineenstortings). Die CMH is wiskundig geformuleer en daarom bied dit vele geleenthede ten opsigte van eksperimentering. Hierdie werkstuk is ‘n ondersoek na die geldigheid en toepassing van die CMH met die gebruik van JSE aandeledata. Kunsmatige Neurale Netwerke is gebruik as berekeningshulpmiddels. Die hoofoogmerk was om die bruikbaarheid van die CMH as voorspellingshulpmiddel te demonstreer in beide ‘n kwantitatiewe sowel as kwalitatiewe opsig. Die resultate van die kwantitatiewe analise was nie so beduidend as aanvanklik verwag nie. Die bruikbaarheid van die CMH was wel gedemonstreer in ‘n meer kwalitatiewe opsig. Dit is ook aangetoon dat die CMH ‘n omvangryke teoretiese raamwerk bied vir die interpretasie, begrip en uitkenning van markdinamika.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/52601
This item appears in the following collections: