ANALYSIS OF TURKEY'S DOMESTIC DEBT STOCK INCREMENT WITH RR AND PCR

There are many variables that affect domestic debt. But whenvariables are taken into model together, it is usually seen that they have strongrelationship which points out to multicollinearity problem. In a multiple linearregression model, in order to decrease the standard error of related coefficientsbetween dependent and independent variables and in order to have more sensitive predictions, multicollinearity problem have to be solved. Some techniquesovercome multicollinearity problem by using data gathering and variable elimination methods. Also in some methods biased predicting methods are usedsince they correct multicollinearity problem without eliminating variables butthose methods give biased results.One of the biased methods is Ridge Regression (RR) which has been widelyused in case of multicollinearity.Another method is Principal ComponentRegression (PCR), which constitutes regression model by collecting relatedvariables in a single variable where all of the formed variables are uncorrelatedand orthogonal to each other. In this study, an application to domestic debtincrement by using RR and PCR methods has been conducted.In analysis, a multilinear regression model with 8 variables constructedbetween the years 1985-2010 intuitively thought to be affecting domestic debtincrement. However Least Square (LS) method, RR Method and PCR methodshowed that some of the variables were insignificant.Thus, after the firstanalysis, number of independent variables were limited to four and the finalanalyses were conducted with four independent variables.Analysis showed that, while RR and PCR coefficients matched with thetheoretical expectations, LS coefficients gave inconsistent results. RR resultsindicated that the public sector borrowing requirement and the internal debtservice were the only two variables that affects domestic debt increment. Onthe other hand PCR showed that beside those two variables the exchangerate and the government budget deficit were also effective on domestic debtincrement. As a result RR and PCR applied in case of multicollinearity and inboth methods significant variables met the theoretical expectations. HoweverRR results are limited since there were only two significant variables whereasPCR results gave relatively better results in which four significant variableswere found

Keywords:

-,

PDF

___

Alpar, R., C¸ ok de˘gi¸skenli istatistiksel y¨ontemler, Nobel,Ankara,2003.
Ayyıldız, H. and K¨ur¸sat ¨O., A study on sales forecasting for pharmaceutical industry in Turkey. Journal of Economics and Administrative Sciences, University of Eski¸sehir Os- mangazi (2011),6(1),71-102.
B¨uy¨ukuysal, M.C¸ ., Ridge regresyon analizi ve bir uygulama, Uluda˘g University, The Institute of Health Sciences,(2010), Master Thesis.
Cank¨uyer, E. and S¨onmez, H., The study and an application of the problem of multi- collinearlty in regression analysis, Anatolian University, Journal of Science Faculty (1996), no.2, 173-187. State planning organization: economic and social Indicators 1950-2010, http://www.dpt.gov.tr/PortalDesign/PortalControls/WebIcerikGosterim.aspx?Enc=83D5A6 FF03C7B4FC5A73E5CFAD2D9676
El-Deren, M and Rashwan, N. I., Solving multicollinearity problem using ridge regression models, Int. J. Contemp. Math. Sciences, (2011), Vol. 6, no. 12, 585 600.
Gujarati, D. N. and Porter C. D., Basic econometrics, 5th Edition, McGraw-Hill/Irwin, New York, 2008.
Hoerl, A. E. and Kennard, R. W., Ridge regression: applications to non-orthogonal problems, Technometrics, (1970), Vol. 12, No. 1, 69-82.
Ipek, O., A study on ridge regression., V. National Econometrics and Statistics Symposium, C¸ ukurova University, (2001).
Karakaya, E., The comparison of the efficiency and bias of ridge and liu estimators, Anadolu Univesity, Statistics Program, (2011), Master Thesis.
Ka¸sko, Y.,The observed type I error and power of logistic regression model under multi- collinearity, Ankara University, Department of Animal Science, (2007), Master Thesis.
Montgomery, D.C. and Peck, E.A., Introduction to linear regression analysis, John Wiley and Sons, No: 641, New York,1991.
Sonuvar, E., Ridge regression analysis of Turkey’s export, Yildiz Technical University, Grad- uate School of Natural and Applied Science, (2007), Master Thesis,.
S¸en, H. and Sa˘gba¸s, I., B¨ut¸ce a¸cıkları: teori ve T¨urkiye uygulaması, Se¸ckin Yayınevi, 2004.
Vinod, H. D. (1978), A survey of ridge regression and related techniques for improvements over ordinary least squares, The Review of Economics and Statistics, (1978), Vol. 60, no. 1, 121
Yavuz, A., T¨urkiyede i¸c bor¸c sto˘gundaki de˘gi¸simin analizine y¨onelik bir regresyon analizi ¸cal¸sması, S¨uleyman Demirel University The Journal of Faculty of Economics and Adminis- trative Sciences, (2003), Vol. 8, Issue: 1, 339-356.
Yildiz Technical University, Department of Statistics, Istanbul TURKEY E-mail address: idemir@gmail.com