Testiranje strukturnega preloma v modelih dinamičnih pogojnih korelacij

Žunko, Matjaž

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Testiranje strukturnega preloma v modelih dinamičnih pogojnih korelacij
Authors:	ID Žunko, Matjaž (Author) ID Jagrič, Timotej (Mentor) More about this mentor...
Files:	DOK_Zunko_Matjaz_2015.pdf (116,18 MB) MD5: C62E22B6CA48A0A2F63F251F02DBB6CC
Language:	Slovenian
Work type:	Doctoral dissertation
Typology:	2.08 - Doctoral Dissertation
Organization:	EPF - Faculty of Business and Economics
Abstract:	Pogosto uporabljena modela za modeliranje časovno variabilnih korelacij sta modela dinamičnih pogojnih korelacij, model DCC in model ADCC. Splošna predpostavka v analizi časovnih vrst je konstantnost parametrov modela. Če je v modelu prisoten strukturni prelom, lahko neupoštevanje spremembe parametrov privede do neustreznega modela in posledično vodi v napačno nadaljnjo uporabo modela. V doktorski disertaciji predstavimo novo metodologijo testiranja strukturnega preloma v modelih dinamičnih pogojnih korelacij. Specifikacijo modelov za vsak dinamični parameter razširimo z diferenčnimi parametri, kombiniranimi s slamnatimi spremenljivkami. Tako definirani diferenčni parametri nam neposredno pokažejo spremembe parametrov pred in po potencialnem strukturnem prelomu. Na njihovi osnovi imamo dosti več možnosti izvedbe statističnih testov kot v obstoječih metodologijah. Ker diferenčni parametri nastopajo v specifikaciji kot samostojni parametri, lahko pridobimo tudi standardne napake njihovih ocen ter kovariance z ocenami ostalih parametrov. Te lahko nato uporabimo v preizkušanju raznovrstnih domnev na podlagi t-testa in Waldovega testa. Vzorčne lastnosti nove metodologije temeljito preverimo z Monte Carlo simulacijskimi eksperimenti. Rezultati kažejo dobre vzorčne lastnosti. Testiranje konstantnosti posameznih parametrov s t-testi se je pokazalo kot boljše od uporabe testov z razmerjem verjetij, saj nanj manj vpliva izpolnjevanje predpostavke porazdelitve podatkov. Metodologijo apliciramo na treh raziskovalnih vprašanjih koreliranosti svetovnih delniških trgov ter delniških trgov držav Centralne in Vzhodne Evrope. Rezultati kažejo, da je v korelacijski strukturi na zahodnoevropskih delniških trgih ob uvedbi evra prisoten strukturni prelom, tako v dolgoročnih povprečjih kot v dinamičnem delu modela ADCC. Po pridružitvi držav Centralne in Vzhodne Evrope Evropski uniji leta 2004 so se povečale dinamične korelacije med njihovimi delniškimi trgi in dinamične korelacije s svetovnimi delniškimi trgi. Po pridružitvi so se povečala dolgoročna povprečja, sprememba v dinamičnih parametrih pa ni statistično značilna. Po izbruhu svetovne finančne krize so proučevani trgi postali še bolj korelirani, ponovno je prisoten strukturni prelom v dolgoročnih povprečjih, dinamične korelacije pa so postale manj stabilne, s strukturnim prelomom tudi v dinamičnem delu modela ADCC.
Keywords:	multivariatni GARCH, dinamične pogojne korelacije, asimetrija, strukturni prelom, slamnata spremenljivka, Monte Carlo simulacije
Place of publishing:	[Maribor
Publisher:	M. Žunko]
Year of publishing:	2015
PID:	20.500.12556/DKUM-47149
UDC:	330.4
COBISS.SI-ID:	11983132
NUK URN:	URN:SI:UM:DK:19XHXGVJ
Publication date in DKUM:	08.05.2015
Views:	2196
Downloads:	328
Metadata:
Categories:	EPF
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Secondary language

Language:	English
Title:	Testing for a structural break in dynamic conditional correlation models
Abstract:	Commonly used models for modeling time-varying correlations are dynamic conditional correlation models, the DCC model and the ADCC model. General assumption of applied time series analysis is constancy of the parameters of the model. If a structural break is present, ignoring a change in parameters can lead to unsuitable model and consequently leads to erroneous continued use of the model. In this dissertation we present a new methodology for testing a structural break in the dynamic conditional correlations models. We extend specification of the models for each dynamic parameter with differential parameters, combined with dummy variables. These differential parameters indicate us directly by how much the parameters in period after the structural break differs from the parameters before a structural break. On the basis of these parameters we have much more possibilities for execution of statistical tests, as in the existing methodologies. Since the differential parameters appear in the specification as an independent parameters, we can also obtain standard errors of the estimates and their covariances with estimates of other parameters. These can then be used in testing of a variety of assumptions on the basis of the t-test and Wald's test. We thoroughly check sample properties of the new methodology with Monte Carlo simulation experiments. The results show good sample properties. Testing the constancy of the individual parameters with t-tests proved to be better than with likelihood ratio tests, since the fulfillment of the distribution assumption has less impact on it. The methodology is applied to three research questions about correlations of global equity markets and equity markets of Central and Eastern Europe. The results show that there is a presence of a structural break in the correlation structure of the western stock markets at the introduction of the currency euro, so in the long-term averages, as in the dynamic part of the ADCC model. After the accession of some Central and Eastern European countries to the European Union in 2004, the dynamic correlations between their equity markets and dynamic correlation with global equity markets have increased. After this event, the long-term averages have increased, but a change in dynamic parameters was not statistically significant. Following the outbreak of the global financial crisis, the studied markets have become even more correlated, with again present a structural break in the long-term averages, while dynamic correlations have become less stable, with a structural break in the dynamic part of the ADCC model.
Keywords:	multivariate GARCH, dynamic conditional correlation, asymmetry, structural break, dummy variable, Monte Carlo simulation

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back