ISTANBUL TECHNICAL UNIVERSITY « GRADUATE SCHOOL M.Sc. THESIS JUNE 2023 INVESTIGATION OF THE EFFECT OF CDS PREMIUMS ON HOUSING PRICES IN TÜRKİYE Aybala DEMİR Department of Real Estate Development Real Estate Development Programme Department of Real Estate Development Real Estate Development Programme JUNE 2023 ISTANBUL TECHNICAL UNIVERSITY « GRADUATE SCHOOL INVESTIGATION OF THE EFFECT OF CDS PREMIUMS ON HOUSING PRICES IN TÜRKİYE M.Sc. THESIS Aybala DEMİR (516191001) Thesis Advisor: Assoc. Prof. Dr. Kerem Yavuz ARSLANLI Gayrimenkul Geliştirme Anabilim Dalı Gayrimenkul Geliştirme Programı HAZİRAN 2023 ISTANBUL TEKNİK ÜNİVERSİTESİ « LİSANSÜSTÜ EĞİTİM ENSTİTÜSÜ TÜRKİYE'DE CDS PRİMLERİNİN KONUT FİYATLARINA ETKİSİNİN ARAŞTIRILMASI YÜKSEK LİSANS TEZİ Aybala DEMİR (516191001) Tez Danışmanı: Doç. Dr. Kerem Yavuz ARSLANLI v Thesis Advisor : Assoc. Prof. Dr. Kerem Yavuz ARSLANLI .............................. İstanbul Technical University Jury Members : Prof. Dr. Bülent GÜLOĞLU ............................. Istanbul Technical University Prof. Dr. Mehmet TOPÇU .............................. Konya Technical University Aybala DEMİR, a M.Sc. student of İTU Graduate School student ID 516191001, successfully defended the thesis/dissertation entitled “INVESTIGATION OF THE EFFECT OF CDS PREMIUMS ON HOUSING PRICES IN TÜRKİYE”, which she prepared after fulfilling the requirements specified in the associated legislations, before the jury whose signatures are below. Date of Submission : 16 May 2023 Date of Defense : 16 June 2023 vi vii FOREWORD I would like to extend my gratitude to my advisor, Assoc.Prof.Dr. Kerem Yavuz ARSLANLI for his unwavering support and invaluable contributions throughout the entirety of my thesis journey. His expertise and insightful suggestions have played a crucial role in shaping this thesis.I would also like to extend my heartfelt thanks to my dear friend, Onur ÜLKÜ, for his continuous encouragement and support. I hope that this thesis will contribute to the existing literature and for the future research in the field. June 2023 Aybala DEMİR (Real Estate Consultant/Appraiser) viii ix TABLE OF CONTENTS Page FOREWORD ............................................................................................................ vii TABLE OF CONTENTS .......................................................................................... ix ABBREVIATIONS ................................................................................................... xi LIST OF TABLES .................................................................................................. xiii LIST OF FIGURES ................................................................................................. xv SUMMARY ............................................................................................................ xvii ÖZET ….……………………………………………………………………...xix INTRODUCTION .................................................................................................. 1 Hypothesis .......................................................................................................... 3 Purpose of The Thesis ........................................................................................ 3 Literature Review ............................................................................................... 4 Credit Default Swaps ........................................................................................ 16 DATA AND METHODOLOGY ......................................................................... 19 Vector Autoregressive (VAR) Model .............................................................. 20 Unit Root Tests ................................................................................................. 24 The Variance Decomposition Analysis ............................................................ 27 Impulse-Response Analysis ............................................................................. 28 Granger Causality Analysis .............................................................................. 30 CONCLUSION ..................................................................................................... 33 REFERENCES ......................................................................................................... 39 CURRICULUM VITAE .......................................................................................... 43 x xi ABBREVIATIONS AIC : Akaike Information Criteria ADF : Augmented Dickey–Fuller BIST : Borsa Istanbul CDS : Credit Default Swaps CPI : Consumer Price Index CBRT : Central Bank of the Republic of Türkiye DJ : Downs Jones FTSE100 : The Financial Times Stock Exchange 100 Index FPE : Final Prediction Error GARCH : Generalized Autoregressive Conditional Heteroskedasticity GDP : Gross Domestic Product HPI : House Price Index HQ : Hannan-Quinn Information Criterion KPSS : Kwiatkowski-Phillips-Schmidt-Shin Unit Root LR : The likelihood ratio test statistic NASDAQ : Nasdaq Stock Exchange PP : Philips Perron Unit Root SIC : Schwarz Information Criteria S&P500 : The Standard and Poor's 500 TCMB : Türkiye Cumhuriyeti Merkez Bankası TL : Turkish Lira VAR : Vector Autoregression Model VECM : Vector Error Correction Model VIX : Chicago Board Options Exchange's Volatility Index xii xiii LIST OF TABLES Page Table 1.1 :Variables used in previous CDS determinants studies. .............................. 8 Table 1.2:Variables used in previous Housing prices determinants studies. .............. 9 Table 1.3: Papers examine the relationship between CDS premiums and housing prices. ...................................................................................................... 15 Table 2.1: Variables Definitons. ............................................................................... 19 Table 2.2: Table of Descriptive Statistics of non -stationary time series. ................. 22 Table 2.3: Table of ADF Unit Roots Test Results. ................................................... 25 Table 2.4: Lag Length Criteria .................................................................................. 26 Table 2.5: VAR Analysis. ......................................................................................... 26 Table 2.6: The variance decomposition results for CDS and HPE variables. ........... 27 Table 2.7: The Granger causality tests results. .......................................................... 31 xiv xv LIST OF FIGURES Page Figure 2.1: Graphs of original time series. ................................................................ 23 Figure 2.2: Graphs of time series with first differences. ........................................... 24 Figure 2.3: Impulse-response graphs of HPE variable. ............................................ 29 Figure 2.4: Impulse-response graphs of CDS variable. ............................................ 30 xvi xvii INVESTIGATION OF THE EFFECT OF CDS PREMIUMS ON HOUSING PRICES IN TÜRKİYE SUMMARY Studies in the literature show a decrease in lending by financial institutions and banks in countries due to increased financing costs associated with increases in CDS premiums. As a result, a noticeable decrease in housing prices is observed (Benbouzid, Mallick, & Pilbeam, 2018). In this context, this thesis examines how the Türkiye CDS premium affects housing prices and how an increase in CDS premiums affects housing prices. Additionally, this thesis will statistically analyze other macroeconomic variables that affect housing prices and the interrelationships among these variables. Studies on the relationship between CDS premiums and housing prices have shed light on how changes in CDS premiums can affect credit availability for homebuyers and overall housing demand, providing insights into potential mechanisms through which CDS premiums can impact housing prices. These studies emphasize the importance of considering other factors, such as economic conditions and other market factors, that can influence this relationship. Previous research on the determinants of Türkiye's CDS premiums has shown that volatility in premiums is more influenced by global than domestic variables. However, the high volatility in premiums is believed to stem from political and economic issues. This thesis aims to investigate the relationship between CDS premiums and Turkish housing prices, as well as other macroeconomic variables that influence this relationship. A vector autoregression (VAR) model has been used to statistically analyze variables such as the housing price index, effective exchange rate, BIST 100 (stock market index), inflation, interest rates, and Türkiye's CDS premiums. Econometric approaches such as VAR, impulse-response analysis, and Granger causality tests have been employed to determine the causal relationship among the variables. The analysis covers the period from September 2010 to December 2022, and the variables have been examined on a monthly basis. The data used in this thesis have been obtained from reliable sources such as the Central Bank, and Bloomberg. The selected timeframe of 2010-2022 has provided a comprehensive understanding of the relationship between Türkiye's CDS premiums, selected macroeconomic variables, and housing prices. The thesis hypothesises that increasing CDS premiums will lead to decreasing housing prices in Türkiye. Based on the findings of previous studies and the literature review, CDS premiums are known to be influenced by various variables such as interest rates, exchange rates, inflation, and the stock market. The results of the multivariate VAR model reveal that changes in Türkiye's 5-year CDS premiums have a positive impact on housing prices in the medium term. Furthermore, the model yields results indicating the role of other macroeconomic variables in the relationship between CDS premiums and housing prices. xviii In the VAR technique, the time series of the variables involved in the analysis must be stationary and does not have unit roots. Therefore, the analysis includes the variables by taking their logarithms. Subsequently, the Augmented Dickey-Fuller (ADF) unit root test has been applied to the series, and it has been observed that the series are not stationary. Taking first differences has been applied to make the logarithmic series stationary. However, when the housing price index variable is differenced once in the ADF test, it is still not stationary. Therefore, a Perron unit root test that takes into account structural breaks has been applied to this variable, and it has been observed that it is stationary after the first difference. After determining the optimum lag length based on information criteria in the VAR analysis, the second lag length with the minimum values of all criteria has been selected. After determining the optimum lag length, the VAR model has been constructed. After constructing the VAR model, the analysis proceeds to variance decomposition. The results of variance decomposition show that in the first period, the variance of the housing price index variable changes solely due to its own shocks and does not receive any contribution from other variables. Starting from the second period, it is observed that the variance of the housing price index variable is explained to varying degrees by other variables. In the second period, it is seen that 92.88% of the volatility in the housing price index is caused by its own shocks, 3.60% by the real effective exchange rate, and 2.97% by other variables such as inflation. According to the results of variance decomposition, the effects of the real effective exchange rate and inflation on the variance of the housing price index variable increase over ten periods. This indicates that over time, these variables become more important in explaining the movements of the housing price index variable. After ten periods, the most critical variables that cause changes in the housing price index are inflation (10.7%) and the real effective exchange rate (8.7%). Overall, the variance decomposition analysis demonstrates that the fluctuations in the housing price index variable are caused by both its own shocks and shocks from other variables, especially inflation and the real effective exchange rate. According to the variance decomposition results for the variable of Türkiye's 5-year CDS premiums, in the first period, 99.8% of the variance is solely due to its own shocks. According to the results of variance decomposition analysis, the impact of Türkiye's 5-year CDS premiums on the variance of the housing price index variable has been minimal over ten periods, explaining only 0.2% of the total variance. These results indicate that inflation (10.7%) is the most significant explanatory variable for housing prices and CDS premiums. The findings of this thesis contribute to our understanding of factors influencing housing prices, particularly in developing economies like Türkiye. The study emphasizes the importance of considering CDS premiums and macroeconomic variables when analyzing the housing market. Overall, this thesis contributes to our understanding of the driving forces behind increasing housing costs in developing economies and provides insights for policymakers in Türkiye regarding housing policy. xix TÜRKİYE'DE CDS PRİMLERİNİN KONUT FİYATLARINA ETKİSİNİN ARAŞTIRILMASI ÖZET Literatürde, CDS primlerindeki artışlarla birlikte artan finansman maliyetleri nedeniyle ülkelerdeki finansal kuruluşların ve bankaların daha az kredi verdiğini gösteren çalışmalar bulunmaktadır. Bunun sonucunda da konut fiyatlarında gözle görülür bir düşüş yaşandığı görülmektedir. (Benbouzid, Mallick ve Pilbeam, 2018) Bu bağlamda bu tez, Türkiye CDS priminin konut fiyatlarını nasıl etkilediğini ve CDS primindeki artışın konut fiyatlarını ne kadar etkileyeceğini inceleyecektir. Ayrıca bu tez, konut fiyatlarını etkileyen diğer makroekonomik değişkenleri ve bu değişkenlerin birbiri ile ilişkisini istatistiksel olarak inceleyecektir. CDS primleri ile konut fiyatları arasındaki ilişkiye yönelik yapılan çalışmalar CDS primlerindeki değişikliklerin konut alıcıları için kredi kullanılabilirliğini ve genel konut talebini nasıl etkileyebileceğine ışık tutmuştur ve bu ilişkiyi etkileyebilecek ekonomik koşullar ve diğer piyasa faktörleri gibi diğer faktörlerin dikkate alınmasının önemini vurgulamıştır. Türkiye'nin CDS primini etkileyen değişkenlerin belirlenmesine yönelik yapılan çalışmalarda, CDS primindeki oynaklığın yerel değişkenlerden çok küresel değişkenlerden etkilendiği görülmüştür. Ancak, primlerdeki yüksek oynaklığın siyasi ve ekonomik sorunlardan kaynaklandığı düşünülmektedir. Bu tezin amacı, CDS primleri ile Türkiye konut fiyatları arasındaki ilişkiyi ve bu ilişkiyi etkileyen diğer makroekonomik değişkenleri araştırmaktır. Vektör otoregresyon modeli , konut fiyat endeksi, efektif döviz kuru, BIST 100, enflasyon, faiz oranları ve Türkiye CDS primleri gibi değişkenleri istatistiksel olarak analiz etmek için kullanılmıştır. Değişkenler arasındaki nedensellik ilişkisini belirlemek için VAR (Vektör Otoregresyon), Etki-Tepki Analizi ve Granger Nedensellik Testleri gibi ekonometrik yaklaşımlar kullanılmıştır. Analiz, Eylül 2010- Aralık 2022 tarihleri arasını kapsamaktadır ve veriler aylık bazda incelenmiştir.Bu tezde kullanılan veriler Türkiye Cumhuriyeti Merkez Bankası ve Bloomberg gibi güvenilir kaynaklardan elde edilmiştir. 2010-2022 dönemi seçilen zaman aralığı, Türkiye'deki CDS primleri ve seçili makroekonomik değişkenler ile konut fiyatları arasındaki ilişkinin kapsamlı bir anlayışını sağlamıştır. Tezin hipotezi, CDS primlerindeki bir artışın Türkiye'deki konut fiyatlarında bir düşüşe yol açacağı yönündedir. Önceki çalışmaların bulguları ve literatür taramasının desteklediği şekilde CDS primlerinin, faiz oranları, döviz kuru, enflasyon, hisse senedi piyasası gibi çeşitli değişkenlerden etkilendiği bilinmektedir. Çok değişkenli VAR modeli sonuçları, Türkiye’nin 5 yıllık CDS primlerindeki değişikliklerin konut fiyatları üzerinde orta vadede az da olsa pozitif bir etkiye sahip olduğunu ortaya koymuştur. Ayrıca, model diğer makroekonomik değişkenlerin CDS primleri ile konut fiyatları arasındaki ilişkide rol oynadığına dair sonuçlar elde etmiştir. xx VAR tekniğinde, analizde yer alan değişkenlerin zaman serilerinin durağan olması ve birim köke sahip olmaması gerekmektedir. Bu nedenle, değişkenlerin logaritmaları alınarak analize dahil edilmiştir. Ardından, serilere ADF birim kök testi uygulanmış ve serilerin durağan olmadığı görülmüştür. İlk dereceden farklar alınarak logaritmik seriler durağan hale getirilmiştir. Konut Fiyat Endeksi değişkeni ADF testinde ilk farkı alındığında durağan olmamaktadır. Bu yüzden bu değişkene yapısal kırılmaları da dikkate alan Perron birim kök testi uygulanmıştır ve ilk farkında durağan olduğu görülmüştür. VAR analizi bilgi kriterlerine göre optimum gecikme uzunluğunu belirlemek için kriterlerin sonuçları incelenmiştir ve tüm bilgi kriterlerinin minimum olduğu 2. gecikme uzunluğu seçilmiştir. Optimum gecikme uzunluğu belirlendikten sonra VAR modeli oluşturulmuştur. VAR modeli oluşturulduktan sonra analizin varyans ayrıştırması aşamasına geçilmiştir. Varyans ayrıştırma sonuçları, Konut Fiyat Endeksi değişkenini varyansının ilk dönemde tamamen kendi şoklarıyla değiştiğini ve diğer değişkenlerden herhangi bir katkı almadığını göstermektedir. İkinci dönemle birlikte, Konut Fiyat Endeksi değişkeninin varyansının farklı derecelerde diğer değişkenler tarafından açıklandığı görülmektedir. İkinci dönemde, Konut Fiyat Endeksindeki oynaklığın %92,88'i kendi şoklarından, %3,60'ı reel efektif döviz kurundan ve %2,97'si enflasyon gibi diğer değişkenlerden kaynaklandığı görülmüştür. Varyans ayrıştırma sonuçlarına göre Reel Efektif Döviz Kuru ve Enflasyonun, Konut Fiyat Endeksi değişkeninin varyansı üzerindeki etkisinin on dönem boyunca arttığı görülmektedir. Bu durum da zaman içinde bu değişkenlerin Konut Fiyat Endeksi değişkeninin hareketlerini açıklamada daha önemli hale geldiğini göstermektedir. On dönemden sonra, Konut Fiyat Endeksindeki değişikliklere neden olan en önemli değişkenler Enflasyon (%10,7) ve Reel Efektif Döviz Kuru değişkeni (%8,7)'dir. Genel olarak, varyans ayrıştırma çalışması, konut fiyat endeksi değişkeninin dalgalanmalarının hem kendi şoklarından hem de özellikle enflasyon ve reel efektif döviz kuru gibi diğer değişkenlerin şoklarından kaynaklandığını göstermektedir. Türkiye'nin 5 yıllık CDS primleri değişkeni için varyans ayrıştırma analizinin sonuçlarına göre ise ilk dönemde varyansın %99,8'i sadece kendi şoklarından kaynaklanmaktadır. Varyans ayrıştırma analizine göre Türkiye'nin 5 yıllık CDS primlerinin Konut Fiyat Endeksi değişkeninin varyansı üzerindeki etkisi on dönem boyunca minimal olmuş ve toplam varyansın sadece %0,2'sini açıklamıştır. Bu sonuçlar, enflasyonun (%10,7) hem konut fiyatları hem de CDS primleri için en önemli açıklayıcı değişken olduğunu göstermektedir. Sonuç olarak, bu çalışma konut fiyatlarındaki değişikliklerin kendi dinamikleri ve enflasyon ile reel etkili döviz kuru gibi makroekonomik değişkenler gibi çeşitli faktörlerden etkilendiğini göstermiştir. Analiz ayrıca Türkiye'nin 5 yıllık CDS primlerindeki dalgalanmaların esas olarak CDS piyasası içindeki içsel şoklardan kaynaklandığını ve konut fiyatlarının bu primlerdeki değişikliklere önemli bir etki yapmadığını ortaya koymaktadır. Ayrıca, etki-tepki analizi aracılığıyla, CDS primlerindeki artışların konut fiyatları üzerinde orta vadede pozitif bir etkisi olduğu gözlemlenmiştir. Bu, CDS primlerindeki dalgalanmaların konut piyasasını gecikmeli bir yanıtla dolaylı olarak etkileyebileceğini göstermektedir. Bu bulgular, konut fiyatları, CDS primleri ve makroekonomik faktörler arasındaki etkileşimi analiz ederken dikkate almanın önemini vurgulamaktadır. Bu ilişkileri anlamak, konut ve finansal piyasalarla ilişkili potansiyel riskler ve zayıflıkları xxi değerlendirmek için politika yapıcılar ve piyasa katılımcılarına değerli içgörüler sağlayabilir.Genel olarak, bu tez, konut fiyatları ve CDS primleri arasındaki karmaşık dinamikleri aydınlatarak hem akademik araştırma hem de konut ve finans sektörlerindeki pratik karar süreçleri için değerli sonuçlar sağlamak suretiyle mevcut literatüre katkıda bulunmaktadır. xxii 1 INTRODUCTION Fluctuations in financial markets significantly affect economies, and one of the most important indicators of these fluctuations is a country's Credit Default Swap (CDS) premiums. CDS (Credit Default Swaps) premiums are used to calculate and assess the country's risk perceptions. They are calculated and evaluated by international investors' risk perceptions of the country. An increase in CDS premiums, following a worsening in the country's economic and financial indicators or political instability, increases the risk premium. (Kilci, 2017). High premiums can cause an increase in a country's borrowing costs, a decrease in foreign investment, and a slowdown in economic growth. Changes in CDS premiums have a considerable impact on the credit costs of a country's financial institutions and serve as an essential indication of the economy's performance. Numerous studies in the literature investigate the relationship between a country's credit risk premium, the risk premiums of the country's financial institutions, and consequently their financing costs. Risk premiums are significant for investors in developing countries such as Türkiye or countries with high uncertainty. Investors tend to avoid investing in these countries unless they are hedged. An increase in risk premiums dramatically raises the borrowing costs of these countries' banks and businesses, directly impacting their economy. As a result, monitoring CDS premiums by construction and housing industry investors is critical, particularly in countries like Türkiye, whose economy is strongly based on the construction sector. It is critical to understand the dynamics of changes in CDS premiums and the causes influencing these changes and their relationship with house prices. In recent years, Türkiye has been in the spotlight with its high CDS premiums. Türkiye's CDS premiums are high due to geopolitical risks, high inflation, low-interest rates, and high external debt levels. This situation causes an increase in Türkiye's borrowing costs and a decrease in foreign investment. Therefore, research on Türkiye's 2 CDS premiums has increased recently."Due to Türkiye's high connection with foreign exchange markets, CDS, which provides insurance against the default of Eurobonds issued by the Ministry of Treasury and Finance, can also be used by investors to take a position against TL." (Financial Stability Report, TCMB, 2020). Credit default swap (CDS) premiums measure the perceived risk of default on a particular debt obligation, such as a mortgage. When CDS premiums increase, it can indicate that lenders view the borrower as having a higher risk of default, making it more difficult for the borrower to obtain financing. This can lead to a decrease in housing demand, which can lead to a decrease in housing prices. Investigating the effect of CDS premiums on housing prices is an important topic because it can provide insights into the factors that influence the housing market and the potential consequences of changes in CDS premiums for the housing market. In Türkiye, as in other countries, the housing market can be influenced by various factors, including economic conditions, interest rates, and government policies. It is possible that an increase in CDS premiums could be one factor that contributes to a decline in housing prices. Still, it is crucial to consider the overall economic and political context in which the housing market operates. In Türkiye, as in other countries, the housing market can be influenced by various factors, including economic conditions, interest rates, and government policies. It is possible that an increase in CDS premiums could be one factor that contributes to a decline in housing prices. Still, it is crucial to consider the overall economic and political context in which the housing market operates. This thesis aims to examine the relationship between country CDS premiums, which are thought to be a good predictor of country default risk, housing prices, and macroeconomic indicators, as well as the amount to which they impact one another.The housing market is considered an essential indicator of economic growth. Therefore, examining the impact of Türkiye's CDS premiums on housing prices can provide an essential indicator of the country's economic situation. In line with this goal, monthly data between September 2010- December 2022 were used, using the VAR analysis method. In macroeconomic models, there is a general dynamic feedback mechanism between variables. It is not known with certainty whether the trend of any given time series in the system is independent of the trend of another time series in the system. When such symmetric interactions exist in multivariate systems involving 3 time series, VAR (Vector Autoregression) methods are used. VAR analysis is important because it allows researchers to investigate causal relationships between multiple variables and make predictions about the behavior of the system in response to various shocks or policy changes. (Kazdagli, 1996) This thesis will contribute to understanding the relationship between Türkiye's CDS premiums and housing prices. Additionally, the results of this thesis will serve as an essential source of information for Türkiye's economic policies and future strategies. Hypothesis This thesis hypothesises that increasing CDS premiums will lead to decreasing housing prices in Türkiye. This hypothesis suggests a negative relationship exists between CDS premiums and housing prices in Türkiye, and that an increase in CDS premiums will be associated with a decrease in housing prices. To test this hypothesis, a multivariate var model was fitted to the data, with CDS premiums as the independent variable and housing prices as the dependent variable. Then the model was used to estimate the relationship between CDS premiums and housing prices and test whether this relationship is statistically significant. In addition to examining the overall relationship between CDS premiums and housing prices, the role of other economic and financial variables in the relationship between these two variables also examined. It has used multivariate var analysis to examine the relationship between CDS premiums and housing prices controlling for variables such as interest rates, inflation, exchange rates, BIST100. Overall, the investigation of the effect of CDS premiums on housing prices in Türkiye is an important topic that could shed light on the factors that influence the housing market in Türkiye and on the potential consequences of changes in CDS premiums for the housing market. Purpose of The Thesis This thesis examines the relationship between Türkiye's CDS (Credit Default Swap) premiums and housing prices. Additionally, it aims to separately analyze the relationship between macroeconomic variables, CDS premiums, and housing prices. Monthly data on Türkiye's CDS premiums, housing price index, inflation, real effective exchange rate, BIST100 index, and policy interest rate from September 2010 4 to December 2022 will be used to investigate these linkages. Econometric techniques such as VAR (Vector Autoregression) analysis and Granger Causality analysis will be employed in this study, focusing on financial time series analysis. The first differences of non-stationary time series will be transformed into stationary series, and Granger Causality Test will be utilised to determine if the lagged values of one variable can explain another variable. Furthermore, this study will contribute to understanding the causes of housing prices in developing economies and provide insights for policymakers in Türkiye to make informed decisions regarding housing policies. The data used in this thesis are obtained from the Central Bank, and Bloomberg. During the data collection, Bloomberg provided data on Türkiye's CDS premiums, while the Central Bank supplied data on housing prices and other financial variables. The selected period for analysis covers September 2010 to December 2022. Literature Review In the literature section of the thesis, the studies regarding the relationship between CDS premiums and housing prices were examined under two main categories: studies investigating the variables influencing CDS premiums and studies focusing on macroeconomic variables affecting housing prices. Firstly, we will examine the "Studies Investigating the Variables Influencing CDS Premiums” section. In the studies, the variables that have significant relations with CDS are divided into global and local variables. Which is shown in the Table 1.1. The variables affecting CDS premiums are generally GDP, growth rate, interest rates, exchange rate, inflation, stock market, fear indices, export growth rate, risk-free interest rate, debt/GDP, debt/exports, unemployment rate, foreign debt, political stability, default history, stock market. Local variables: growth rates, cpi, real effective exchange rate, stocks, bond yields, stock market, risk appetite, interest rates, external debt balance. Global variables are SP500, VIX, S&P 500, NASDAQ indices, stocks and bond market. KOY, (2014) examined the CDS and Eurobond premiums of Germany, France, Italy, Spain, Portugal, Ireland, Türkiye and Greece between January 2009 and November 2012, including the initial period of the European Debt Crisis, and applied the Granger causality test. 5 The study found that the change in CDS premiums of France, Italy and Türkiye drove the change in Eurobond premiums, and that the CDS premiums and Euro-bond premiums of Ireland, Spain, Portugal and Greece were statistically correlated. Basarir and Keten (2016) analyzed the relationship between monthly stock market indices, exchange rates, and CDS premiums of 12 developing countries between 2010 and 2016. They examined the short-term relationship with the Granger causality test and the long-term relationship with the Panel Johansen cointegration test.They found a bidirectional causality between CDS premiums and stocks in the short run. CDS premiums were found to have a one-way causality to exchange rates. In the long term, however, no relationship was found between the CDS premiums of the countries examined and exchange rates or stock indices. Bektur and Malcıoğlu (2017) used two different causality tests to examine daily data from Türkiye's CDS premiums and BIST100 index from 2000 to 2017. First, using the "Hacker and Hatemi-J bootstrap causality test," they discovered a one-way relationship between CDS premium and BIST100 index. Later, using the Hatemi-J (2012) Asymmetric Causality Test, this relationship was divided into positive and negative shocks. The test results revealed a causal relationship between negative shocks from BIST100 and CDS premiums. Positive shocks in CDS premiums have been determined to be the Granger cause of positive shocks to the BIST100 index, however, no such relationship between the Stock Exchange and CDS premiums has been observed. Kılcı (2017) investigated the correlation between Türkiye's 5-year CDS premiums and economic variables such as the growth rate, unemployment rate, inflation, CPI, current account deficit/GDP, real effective exchange rate, capital adequacy ratio, non- performing loans/total loans and BIST30 throughout 2010-2016. Engle-Granger and Johansen Cointegration Tests were employed to analyze the data. The Engle-Granger test revealed that no long-term relationship existed between 5-year CDS premiums and unemployment rate, growth rate, CPI, and current account deficit/GDP ratio. However, financial variables such as non-performing loans/total loans, banking sector capital adequacy, BIST 30 values, the real effective exchange rate, and Türkiye's 5-year CDS premiums have been found to have a long-term relationship. 6 Danaci and Sit (2017) used unit root tests and the Toda-Yamamoto causality test to examine the relationship between Türkiye's 5-year CDS premiums and growth rate from 2009 to 2015. The Toda-Yamamoto causality analysis determined a one-way causality relationship between CDS premiums and growth rates. The results of Hacker and Hatemi-J (2006)'s Bootstrap-based Toda-Yamamoto causality analysis showed that there was a bidirectional causality relationship between CDS and growth rates. Aksoylu and Görmüş (2018) analyzed the causality relationship between the CDS premiums of 9 developing countries between 2005-2015 and the US 10-year government bond interest rate, the dollar exchange rate and the VIX index using the Granger causality test and the Hatemi J asymmetric causality tests. In Indonesia, Portugal, Mexico and Argentina, it was observed that there was asymmetric causality from exchange rate to CDS premium in at least one of the positive and negative shocks. The Granger causality analysis of the 10-year US government bond interest concluded that CDS premiums were not Granger causes. In the Hatemi-J asymmetric causality test, it was observed that there is a causal relationship between Argentina, Indonesia and Portugal's 10-year US government bond interest to CDS premiums in positive shocks. In negative shocks, a causal relationship was found from 10-year US government bond interest rates to CDS premiums of Türkiye, Indonesia, Philippines, Portugal and Poland. VIX fear index to the CDS premiums of Indonesia, Argentina, Philippines, Poland and Malaysia were found to have a causal relationship. Kunt and Taş (2008) The risk-free interest rate, reference asset return, and return volatility, which are thought to have an impact on CDS premiums and premium prices, were examined in the study conducted to estimate Türkiye's CDS premiums between 2000 and 2008. The analyses revealed a long-term relationship between the CDS premium and the reference asset's return, volatility, and risk-free interest rate. Akyol and Baltaci (2019) used the ARDL bounds test to determine the global and local variables influencing Türkiye's 5-year CDS premiums between 2005 and 2018. They used the BIST100 index returns, real interest rates, current account balance, portfolio investments, and inflation rates as local variables, and the MSCI Europe index, oil prices, FED interest rates, the VIX index, and US economic/monetary policy uncertainties as global variables. 7 According to the findings, BIST100 index returns, current account balances, and portfolio investments all negatively impact CDS premiums. In the long run, inflation, the real exchange rate, and real interest rates have been observed to positively affect CDS premiums. In their study using monthly data from 2005 to 2017, Ozpinar and Ozman (2018) investigated the effect of exchange rates and benchmark bond interest rates on CDS premiums. The Granger causality test determined one-way causality from the dollar rate to the CDS premium. Along with the Johansen cointegration test, it was determined that this relationship was a long-term positive one. No causality was detected between the benchmark interest and the CDS premium. Benbouzid a, Sushanta K. Mallick b, and Ricardo M. Sousa examined annual data from 58 banks in 15 countries from 2004 to 2011 to determine how country financial structures explain 5-year CDS spreads. They examined bank-level and country-level factors in explaining CDS spreads. In the study, it was seen that housing prices and CDS spreads were negatively related. At the bank level, it was seen that the improvement in asset quality, profitability and liquidity was associated with low credit risk. At the country level, it was seen a negative relationship between financial stability and CDS spread 8 Table 1.1 : Variables used in previous CDS determinants studies. Author Method Study Term Bond Stock Inflation Unemployment VIX US Bond Market GDP Growth Rate BIST30 Exchange Rate Interest Rate Norden and Weber (2009) VAR Analysis, Granger Causality (2000- 2002) x x Brandorf, C.-J. Holmberg (2010) Regression Analysis (2004- 2009) x x x Koy (2014) Unit Root Test and Granger Caus. (2009- 2012) x Kargı (2014) Granger Causality Johansen Cointegration test (2005- 2013) x x Başarır and Keten (2016) Granger Causality, Johansen Cointegration (2010- 2016) x Kılcı (2017) Unit-Root Tests. (2010- 2016). x x x x 9 Table 1.1 (continue) : Variables used in previous CDS determinants studies. Table 1.2:Variables used in previous Housing prices determinants studies. Author Method Study Term Bond Stock Inflation Unemployment VIX US Bond Market GDP Growth Rate BIST30 Exchange Rate Interest Rate Danacı and Şit (2017) ADF Unit Root Test, PP Unit Root Test (2009- 2015) x Longstaff vd. (2011) Regression Analysis (2000- 2010) x x Zhu, Haibin. (2006) Panel Data and VECM (1999- 2002) x x Author Method Study Term CPI Unemploy ment Rate Stock Market GDP Exchange Rate Interest Rate Income Mortgage Interest R. Construc tion Cost Economi c growth Martinez- Garcia and Grossman (2020) Recursive (Right- Tailed) Unit Root Test 23 Country (1975- 2015) x x x x Ucal and Gokkent (2010) VAR Model (1987- 2005) x x x x 10 Table 1.2 (continue): Variables used in previous Housing prices determinants studies. Author Method Study Term CPI Unemploy ment Rate Stock Market GDP Exchange Rate Interest Rate Inco me Mortgage Interest R. Construction Cost Economic growth Nneji, Brooks, and Ward's(2013) , Three- Regime Markov Switching Model (1960- 2011) x x x Apergis (2003) ECVAR x x x Rahal (2016) PVAR (2007- 2014) x x x Antonakakis and Floros (2016) Diebold and Yilmaz (1997- 2015) x x McQuinn and O'Reilly's (2008), Unit Root Tests, Johansen Cointegra tion (1980- 2005) x x Adams and Füss (2010) Panel Cointegra ion Test, Unit Root Tests (1975- 2007) x x x x Korkmaz (2019) ARDL test (2010- 2019) x Sertkaya (2016) VAR Model (2008- 2014) x x x 11 Literature have identified various relationships and correlations involving CDS premiums. Firstly, an inverse correlation has been discovered between CDS premiums and the BIST100 index, showing that when CDS premiums increase, the stock market tends to decline. Additionally, a strong negative relationship has been observed between CDS premiums and stocks, indicating that an increase in CDS premiums is typically associated with a decrease in stock prices. Moreover, there is a bidirectional causality relationship between CDS premiums and interest rates, which means that changes in CDS premiums can affect interest rates and vice versa. Furthermore, the banking sector's performance has been highlighted as a crucial factor driving fluctuations in CDS premiums. Specifically, the banking sector's performance has a notable effect on the fluctuations of CDS premiums. Additionally, a long-term relationship has been established between the real effective exchange rate and CDS premiums, showing that changes in the exchange rate can influence CDS premiums in the long run. In the long term, inflation, the real exchange rate, and real interest rates have positively impacted CDS premiums. In terms of economic growth, a bidirectional causal relationship between economic growth and CDS premiums has been identified. This shows that changes in CDS premiums can impact economic growth, and that economic growth can impact CDS premiums. Moreover, it has been observed that increases in ratios such as the Foreign Debt Service/Exports, Foreign Debt Service, External Debt/GDP, and Foreign Direct Investments/GDP tend to coincide with increases in CDS premiums, except during extraordinary periods. CDS premiums tend to decline during periods of high investor risk appetite and low systematic risk. Furthermore, studies have shown that risk-tolerant indicators such as the inflation rate and the VIX (Volatility Index) can influence a country's risk premium. The current account deficit negatively affects CDS premiums, risk-free interest rate, real exchange rate, risk appetite, and CDS premiums. On the other hand, the debt-to-GDP ratio has a positive association with the inflation rate and CDS premiums. When we examine the "Studies Focusing on Macroeconomic Variables Affecting Housing Prices" section in the literature, which is shown in the Table 1.2, Martinez- Garcia and Grossman (2020) investigated financial market spillovers and discovered periods of real housing price exuberance in 23 countries between 1975 and 2015. The 12 study shows that interest rate spreads, real stock market growth, real personal disposable income per capita growth, and inflation were among the most significant predictors of these episodes using a dynamic panel logit/probit methodology. The study shows that reasonable bubbles in housing prices and fundamental forces can cause modestly explosive dynamics. Ucal and Gokkent (2010) utilized a vector autoregression model to investigate the macroeconomic factors influencing Türkiye's housing markets. The study discovered that inflation plays a substantial role in accounting for housing price fluctuations. The study emphasizes the housing market's vulnerability to lengthy deviations from intrinsic value and asset price bubbles due to supply rigidities, poor information, and imperfect financial markets. Nneji, Brooks, and Ward's (2013), study shows that fundamental economic factors, particularly income growth and interest rates, are said to have a bigger influence on house price dynamics. Apergis (2003) investigates how macroeconomic variables such as inflation, employment, and mortgage rates affect Greek housing prices. According to the study, mortgage rate changes had the most significant percentage impact on housing prices, followed by inflation and employment. Rahal (2016) investigated how housing markets have responded to unusual monetary policy shocks following the 2008 financial crisis. The research used quarterly and monthly housing market datasets from eight OECD countries to estimate a variety of specifications in panel vector autoregressions. The results show a positive and sustained reaction in housing prices, which peaked one to two years after a policy shock. From 1997 to 2015, Antonakakis and Floros (2016) investigated the dynamic interdependence of the UK housing market, stock market, policy uncertainty, and macroeconomy. The study discovers that shocks from the housing market, stock market, and economic policy uncertainty have considerable spillover effects on inflation, economic growth, and monetary policy stance demonstrating the spread of the housing and financial crises to the real economy and policy responses. According to McQuinn and O'Reilly's (2008), income and interest rates are significant determinants of house prices in the long run. They found that a 1% increase in income led to a 0.83% increase in house prices, while a 1% increase in interest rates caused a 13 0.54% decrease in house prices. Adams and Füss (2010) , identify economic activity, long-term interest rates, and construction costs as significant determinants of house prices in 15 countries, with a 1% increase in economic activity leading to a 0.6% increase in house prices.This shows that economic growth is beneficial to the housing market. They also discovered that long-term borrowing rates and building costs have a considerable impact on housing prices. Duca, Muellbauer and Murphy (2010), highlight the role of financial innovation, misaligned incentives, global financial imbalances, and leverage in fueling credit expansions, housing demand, development, and pricing. The study shows how the housing slump has impacted other countries through risk premiums in global financial markets and current account imbalances. In a study by Gaspareniene, Remeikiene, and Skuka (2016) examining the impact of macroeconomic factors on Lithuanian house prices from 2008 to 2015, it was found that the availability of bank loans and interest rates had the most significant impact, accounting for 49.23% and 79.03% of the fluctuations, while inflation and GDP had a smaller impact, accounting for 39.35% and 0.58% of the fluctuations, respectively. The article emphasizes the major influence of macroeconomic factors on home prices, such as GDP, inflation rates, and bank loans, whereas other factors such as construction costs, wages, consumer purchasing power, and employment rates have a lesser impact. Korkmaz (2019) explored The relationship between housing prices and the inflation rate in Türkiye utilizing several econometric techniques such as the Phillips-Perron test, the ARDL boundary test, and the Granger causality test. The study examines the causation relationship between the housing price index (HPI) and the consumer price index (CPI) in 26 Turkish areas from 2010 to 2019. The significant finding reveals that inflation pressure in HPI affects some locations. Dilber and Sertkaya (2016) used three-month time series data from 2008 to 2014 to investigate the links between the House Price Index, Inflation Rate, Real Effective Exchange Rate, and Housing Loan Interest Rates. They discovered a unidirectional association between the house price index and the inflation rate and between the house price index and the housing loan interest rate in their research. They also discovered a bidirectional causative association between the home price index and the real effective 14 exchange rate and a unidirectional causal relationship between the real effective exchange rate and the housing loan interest rate. Demary's (2009) study found that real estate prices considerably impacted key macroeconomic variables, with housing demand shocks accounting for 12% to 20% of output changes and 10% to 20% of price swings. According to the study, rising housing prices enhance households' net worth, which leads to higher consumption expenditures and promotes aggregate demand, resulting in higher output and inflationary pressures. According to the study, monetary policy should take into account housing market changes in order to preserve overall economic stability. Gebeşoglu (2019) investigates the relationship between Turkish housing prices and other macroeconomic indicators. According to the study, Borsa Istanbul returns contribute to a drop in housing prices, and exchange rate volatility might cause macroeconomic imbalances that affect housing prices. In conclusion, the literature shows that macroeconomic variables including inflation, interest rates, stock markets, exchange rate, GDP, and bank loans have a significant impact on home prices. Two studies in the literature examine the relationship between CDS premiums and housing prices, which will be especially examined for this thesis. These two studies shown in Table 1.3. One of the two studies examining CDS and housing prices in the literature was made for the UK housing market and the other for the Turkish market. Benbouzid, Mallick and Pilbeam analyzed monthly 5-year CDS premiums in the UK banking sector between January 2004 and April 2011. As the explanatory variables of CDS premiums, the UK housing price index, the yield difference between the UK 30- year Treasury bills and 3-month Treasury bills, the UK TED spread and the UK FTSE100 index as the last variable was used. After performing the variables' stationarity and unit root tests, the VAR model was established by choosing the most appropriate lag length. As a result of the study, it was observed that a positive shock to housing prices significantly increased CDS premiums. The other result is that the most vital variable explaining CDS premiums is the housing price index. In addition, it has been observed that the housing price index is also affected by the yield difference and TED margin. 15 Table 1.3: Papers examine the relationship between CDS premiums and housing prices. Study/Authors Method Variables Conclusion The housing market and the credit default swap premium in the UK banking sector: A VAR approach, Res. Int. Business Finance. (2018). Benbouzid N, Mallick S, Pilbeam K VAR approach house prices, the yield spread, the UK TED spread and the FTSE 100 index. (2004-2011) The results of the study showed that a positive shock to the CDS premium significantly reduced housing demand and housing prices. Spillovers Between Turkish House Pricing, Stock Exchanges, Gold, CDS and Exchange Rate (2019, Master Thesis). Şentürk, E VAR approach BITS100, CDS, SP500, HPE house prices, exchange rates, gold prices, stock exchange rates, credit default swaps. (2003-2018) The results of the study showed that gold is the most effective variable for house price index in the long run in Türkiye when it is compared with other financial instruments. In a study conducted in Türkiye, Şentürk examined the relationship of house prices with gold prices, credit default swaps, exchange rates and stock market indices between January 2003 and September 2018. The VAR approach was used to examine the relationship between these variables. Since all series must be stationary for the VAR approach, stationarity and unit root tests were performed on the variables. It has been observed that the variables other than the housing price index are stationary. Perron unit root test was applied to the housing price index time series and it was seen that the series was stationary when structural breaks were taken into account. Then, impulse-response analysis and variance decomposition were performed. The study's results showed that gold is the most effective variable for the housing price index in Türkiye in the long run compared to other financial instruments. 16 Credit Default Swaps "Credit Default Swaps" briefly CDSs in its most basic terms, is the securitization of insurance contracts that protect the investor against the risk of non-payment of the loan. In case of non-payment of this loan, uncollectible debts are transferred to the CDS seller and the seller is obliged to pay this debt to the buyer. Therefore, it can also be defined as swapping risks without moving the investment asset. (Haibin 2006) In return for this insurance, a portion of the expected return on the investment is paid as a premium. In financial terminology, these premiums are paid in basis points. 100 basis points correspond to 1%. In return for the insurance, 100 basis points, or 1%, is paid to the CDS seller as a premium. The increase in CDS premiums results in higher borrowing costs for both countries and companies operating within those countries. When countries and corporations borrow, lending institutions incorporate the country's CDS premiums into the interest rates as a basis point. Consequently, this leads to borrowing at higher interest rates. CDS contracts can be created for the debt securities of both companies and countries. The greater the default risk of the borrower, the higher the CDS premiums. Therefore, CDS premiums of a country give an idea of the investors in the country and the ability of the country to pay its debts. CDSs protect against default, allowing investors to invest in countries or assets with low credit ratings. CDSs cover risks such as bankruptcy, credit downgrade, and default. While CDS maturities are between 1 and 10 years, the most frequently traded CDS is 5 years. (Schönbucher 2003, s.15-17). Credit default swaps (CDS) are financial instruments used to transfer the risk of default on a debt obligation from one party to another. They are typically used in the bond market and are designed to protect investors against the risk of default on a bond. A CDS contract specifies the terms under which the protection buyer will receive compensation in the event that the issuer of the bond defaults on its debt obligations. The protection buyer pays a periodic premium to the protection seller in exchange for this protection. If the issuer of the bond defaults, the protection seller is required to pay the protection buyer the face value of the bond, less any recovery amount that the protection buyer is able to obtain from the issuer. One of the critical features of CDS contracts is that they can be traded on the secondary market, which allows investors to buy and sell protection on a bond. This can be useful 17 for investors looking to hedge their exposure to a particular bond or for those looking to take on additional risk in search of higher returns. CDS contracts have been used extensively in the bond market and have played a role in a number of significant financial events, including the 2008 global financial crisis. During the crisis, the widespread use of CDS contracts was seen as one of the contributing factors to the collapse of Lehman Brothers, as the firm's failure led to a large number of CDS contracts being triggered, which put further strain on the financial system. (Dumontaux, N., & Pop, A. (2013). Overall, CDS contracts can be a valuable tool for managing risk in the bond market. Still, they also have the potential to contribute to instability in the financial system if they are used excessively or if the underlying bonds are not adequately evaluated. As such, it is crucial for investors and regulators to carefully consider the use of CDS contracts and to ensure that they are used in a responsible and transparent manner. 18 19 DATA AND METHODOLOGY This thesis, investigates the relationship between CDS premiums and Turkish housing prices, as well as other macroeconomic variables that influence this relationship. A vector autoregression (VAR) model has been used to statistically analyze variables such as the housing price index, effective exchange rate, BIST 100 (stock market index), inflation, interest rates, and Türkiye's CDS premiums. Econometric approaches such as VAR, impulse-response analysis, and Granger causality tests have been employed to determine the causal relationship among the variables. The literature study examined local variables that are thought to be related to Türkiye's 5-year CDS premium and housing prices. Local variables; House price index, Real effective exchange rate, BIST 100, inflation and Interest rates. Table 2.1 shows variable definisions. Interest Rates: Interest rates significantly impact housing markets because they are a significant cost component for housing loans. The interest rate variable is included in the analysis to assess the impact on Turkish house prices. According to the literature, Inflation has a significant impact on housing markets and CDS premiums. High inflation rates raise costs, which can cause home prices to rise. In order to assess the impact on Turkish house prices, the inflation rate variable is also included in the analysis. Table 2.1: Variables Definitons. Variables Definition Explanation HPE House Price Index Türkiye’s house price index CDS Credit Default Swaps 5 Years Türkiye’s credit default swaps INF Inflation Türkiye’s consumer price index INT Interest Rate Interest rate applied to one-week repo transactions BIST100 Borsa Istanbul 100 The index is used as the main index for Borsa Istanbul Equity Market USD USD Currency Sales USD-TRY Currency Sales 20 Exchange Rate: Studies in the literature show that the Turkish Lira's depreciation causes changes in Türkiye's CDS premiums. In this respect, exchange rate variable analysis was included in the analysis to understand better the relationship between the exchange rate and Türkiye's CDS premiums. BIST100: The literature shows that the stock market's performance is often linked to economic growth. It has been speculated that a growing economy can lead to an increase in housing demand and the stock market's strong performance can cause housing prices to rise. Since investors are risk aware, declines in the stock market may cause an increase in CDS premiums as well as a decrease in housing prices due to increased risk perception. As a result, the BIST100 variable was included in the analysis in order to understand the relationship between the BIST100 index and CDS and housing prices. The data for the thesis is from September 2010 to December 2022. These data were obtained from the Central Bank (CBRT) websites and Bloomberg. In macroeconomic models, there is a general dynamic feedback mechanism between variables. It is unknown whether the trend of any given time series in the system is independent of the trend of another time series. When such symmetric interactions exist in multivariate systems involving time series, VAR (Vector Autoregression) methods are used. VAR analysis is essential because it allows researchers to investigate causal relationships between multiple variables and make predictions about the system's behaviour in response to various shocks or policy changes. (Kazdagli,1996:42) Vector Autoregressive (VAR) Model The VAR model was developed by Sims in 1980, based on the assumption that each variable in an econometric model can affect other variables and be influenced by them (Sims, 1980: 1-49). The Vector Autoregressive (VAR) model is based on the idea that the values of multiple variables, which are interdependent over time, can be explained by the values of other variables in the econometrics. Each variable is used to describe its dynamic interactions with other variables, and it is a model frequently used to conduct causal tests written in an equation system. Combining these equations helps describe how each variable in the system behaves over time. 21 This model operates on the assumption that variables in a series of time series are related to each other. In the VAR model, multiple variables that are dependent on each other are analyzed in a single model. This model is expressed as a matrix describing how the variables are related to each other. This matrix allows us to understand how each variable depends on other variables' current and past values. In the VAR model, it can also enable us to understand the causal relationships of the variables in the time series with each other. In this study, the Granger causality test was applied to the variables. In the VAR model, Impulse-Response analyzes are also performed to measure the effect of a particular variable on other variables. The complexity of the interactions between the variables in financial time series makes it more challenging to decide which variables should be used as dependent and independent in econometric models. These uncertainties significantly affect the consistency of the results obtained. Therefore, some restrictions are imposed on structural models to overcome these complexities. (Adrian and Darnell, 1990: 114- 116). In VAR analysis, it is possible to analyse the relationships between economic variables without imposing any structural restrictions on the econometric model. Sims (1980) introduced VAR models by extending univariate autoregressive models. These models are widely chosen in time-series analysis in the literature because they offer dynamic interactions without placing any limitations on the structural model.In addition, including lagged dependent variables in VAR models provides strong future forecasts (Kumar et al., 1995: 365; Ozcan and Ari, 2011). The VAR model can investigate causal relationships between variables while determining the causality direction. As a result, by employing the VAR model in numerous financial time series, the impact of one variable on other variables and which variables influence others may be investigated. Using differentiated variables instead of level values in VAR models reduces the spurious regression problem. In order to use VAR model in structural analysis, Impulse-Response Analysis and Variance Decomposition analyses should be performed. When a one standard deviation shock is given to the variables in the model, the response of other variables is measured by Impulse-Response Functions. Enders (1995) states that Variance Decomposition is another technique used to analyse residuals, which measures the variance of forecast errors determined from model estimation. This technique is used to analyse the effects of statistical shocks on variables (Ozcan and Ari, 2011). 22 The method used in the study is the Vector Autoregressive (VAR) Model. The bivariate VAR Model can be expressed in the standard form as follows: 𝑦! = 𝑎" + ∑𝑏"𝑖 𝑦!#$ + ∑𝑏%𝑖𝑋!#$ +𝑣"𝑡 2.1 𝑥! = 𝑐" + ∑𝑑"𝑖 𝑦!#$ + ∑𝑑%𝑖𝑋!#$ +𝑣%𝑡 2.2 The formula above represents a bivariate Vector Autoregressive (VAR) model used to analyse the relationships between two economic variables. The variables 𝑦! and 𝑥! represent two distinct economic variables, where 𝑦! is expressed by a constant term 𝑎" and coefficients 𝑏" and 𝑏%, which establish the relationship between the variables 𝑦!#$ and 𝑋!#$ at past times. Additionally, 𝑣"𝑡 represents the unpredictable error term of the model. Similarly, the variable 𝑋! is expressed by the constant term 𝑐! and coefficients 𝑑" and 𝑑%, which determine the relationship between the variables 𝑦!#$ and 𝑋!#$ at past times. The term 𝑣%𝑡 represents the unpredictable error term of this variable. The analysis commenced with variance decomposition, impulse-response analysis, and applying the Granger causality test within the VAR model. The thesis aimed to determine the causal direction between the variables through the application of the Granger Causality Test. CDS data was collected from Bloomberg data terminal and other data from the CBRT data portal, and econometric analyses were conducted using the Econometric Views (Eviews, version 10) software package. In the thesis, first the logarithms of the variables used in the model were taken to ensure the variance balance and eliminate the extreme values. And logarithmic values were used in the analysis. Descriptive statistics for the variables and a correlation matrix were created. According to these results and the skewness and kurtosis coefficients of the variables, it is seen that in the Table 2.2 the series does not show a normal distribution. Table 2.2: Table of Descriptive Statistics of non -stationary time series. DESCRIPTIVE STATISTICS LNCDS5 LNBIST100 LNHPI LNINT LNINF LNUSD Mean 5.604388 6.867457 4.598528 2.224357 2.471440 1.320270 Median 5.531707 6.758180 4.540628 2.079442 2.274678 1.155811 Maximum 6.731293 8.614167 6.496322 3.178054 4.448633 2.928446 Minimum 4.784654 6.239625 3.848018 1.504077 1.383791 0.355872 Std.Dev. 0.456963 0.428981 0.604016 0.459577 0.654631 0.699636 23 Table 2.2 (continue): Table of Descriptive Statistics of non -stationary time series. DESCRIPTIVE STATISTICS LNCDS5 LNBIST100 LNHPI LNINT LNINF LNUSD Skewness 0.494823 1.654371 1.283493 0.635944 1.572234 0.636446 Kurtosis 2.442273 6.318812 4.573361 2.413728 5.251817 2.460097 Jarque-Bera 7.957838 135.4341 55.90013 12.09538 92.24324 11.78911 Probability 0.018706 0.000000 0.000000 0.002363 0.000000 0.002754 Sum 829.4495 1016.384 680.5822 329.2049 365.7731 195.4000 Sum Sq. Dev. 30.69588 27.05158 53.63082 31.04802 62.99559 71.95518 Observations 148 148 148 148 148 148 In order to determine whether the variables to be used in the model have a trend, their graphs were examined first. When the Figure2.1 graphs are examined, it is seen that the series has an increasing trend. When the Figure 2.2 graphs are examined it is seen that with the first differences, all variables except the HPI variable are stationary. Figure 2.1: Graphs of original time series. 24 Figure 2.2: Graphs of time series with first differences. Unit Root Tests Econometric models are built on stationary time series data. This can be explained as follows: Stationary series are characterized by the absence of systematic changes over time and the rapid disappearance of the effects of shocks they experience. These shocks can be transitory or permanent and can cause series to exhibit trend or seasonal fluctuations If a permanent shock induces a trend that does not converge to a specific value, then the series is considered non-stationary. Stationarity is important in expressing a real or spurious regression relationship in regression analysis. This stochastic trend, as it is called, can create a spurious regression relationship between series. Therefore, stationarity is important in econometric analysis, and the analysis of non-stationary series can lead to misleading results. (Yurdakul,2000;2) In this thesis, whether time series are stationary or not, and at what level they are stationary, have been determined using the Augmented Dickey Fuller (ADF) unit root test developed by Dickey and Fuller (1981). In addition to the ADF test, the PP unit root test which changing the error terms, better suited for detecting structural breaks, has also been utilized. This correction entails including MA (Moving Averages) corrections in addition to AR (Autoregressive) corrections in the Dickey Fuller and 25 Augmented Dickey Fuller models. As a result, the PP test can be written as an ARMA (Autoregressive Moving Average) process (Philips and Perron, 1988). Stabilization of variables is an important step to avoid spurious regression and obtain more accurate analyses results. Both stationary and trend stationary unit root analysis were performed to test the stationary of variables. It was concluded that the variables are not stationary. Since the variables are time series variables, it is common for them not to be stationary. Therefore, first differences were taken to make the variables stationary. According to the Table 2.3, ADF test result, it was observed that all variables were not stationary at logarithmic levels but stationary at the first difference of logarithmic values. Table 2.3: Table of ADF Unit Roots Test Results. Variables ADF Statistics (non-stationary) ADF Statistics (stationary) ADF Statistics (with structural break) LNCDS5 -3.12 -12.08 LNBIST100 1.48 -10.68 LNHPI 0.18 -2.73 -5.63 LNINT -1.96 -11.13 LNINF -2.46 -7.90 LNUSD -0.74 -7.34 * indicates that series are significant at 1% level of significance Because financial variables are in a relationship with one another across time, rather than accepting a single variable as a dependent variable, a VAR model was developed to look at how variables interact. Once the stationarity of the series was verified, the lag length was determined for the VAR analysis. Akaike Information Criteria (AIC), Schwarz Information Criteria (SC), Likelihood Ratio Test (LR), Final Prediction Error (FPE), and Hannan-Quinn Information Criteria (HQ) are checked to determine the optimal lag length. The optimal lag length is determined by selecting the lag with the lowest value among these criteria. Table 2.4 shows the SIC and AIC criteria used to determine the lag length of the VAR model. The 2. lag showing the lowest value in the AIC and SIC criteria was chosen as the most appropriate lag length for the VAR model which shows in the Table 2.5. 26 Table 2.4: Lag Length Criteria Lag LogL LR FPE AIC SC HQ 0 88.6899 NA 1.24E-08 -1.181284 -1.055214 -1.130053 1 1179.96 2073.42 3.51E-15 -16.25659 -15.3741 -15.89797 2 1316.82 248.297 8.33e-16* -17.69740* -16.05849* -17.03140* 3 1342.42 44.250 9.73E-16 -17.54882 -15.15348 -16.57543 4 1373.51 51.088 1.06E-15 -17.47878 -14.32702 -16.198 5 1402.36 44.912 1.20E-15 -17.37653 -13.46835 -15.78836 * indicates lag order selected by the criterion Table 2.5: VAR Analysis. DLNHPI DLNCDS5 DLNBIST100 DLNINF DLNINT DLNUSD DLNH PI (-1) 0.8574 0.8009 7.00E-01 0.35541 -0.662096 0.329359 -0.0370 -0.5657 -2.77E-01 -0.45858 -0.44101 -0.13539 [23.1469] [1.41563] [2.52591] [0.77503] [-1.50133] [2.43273] DLNC DS5 (-1) -0.0116 -0.1332 -7.25E-02 0.089244 0.113636 0.115504 -0.00744 -0.11361 -5.56E-02 -0.09209 -0.08856 -0.02719 [-1.55738] [-1.17275] [-1.30286] [0.96907] [1.28311] [4.24823] DLNBI ST100 (-1) 0.00779 -0.59465 -6.08E-02 0.24282 -0.064986 0.027383 (0.01507) (0.23016) -1.13E-01 -0.18656 -0.17941 -0.05508 [0.51684] [-2.58366] [-0.53919] [ 1.30157] [-0.36222] [ 0.49717] DLNIN F (-1) 0.01062 0.17310 -0.044251 0.211624 0.108043 -0.043719 -0.00705 -0.10775 -0.05276 -0.08734 -0.08399 -0.02578 [ 1.50580] [ 1.60652] [-0.83868] [ 2.42306] [ 1.28636] [-1.69554] DLNIN T (-1) -0.01201 -0.145432 -0.012307 -0.030621 0.028762 -0.018695 -0.00719 -0.10974 -0.05374 -0.08895 -0.08554 -0.02626 [-1.67146] [-1.32523] [-0.22902] [-0.34424] [ 0.33622] [-0.71189] DLNU SD (-1) 0.08878 -0.555681 0.246436 0.696202 0.303232 0.221334 -0.02232 -0.34096 -0.16696 -0.27637 -0.26578 -0.08159 [ 3.97692] [-1.62975] [1.47599] [2.51906] [ 1.14090] [ 2.71261] 0.00116 0.01144 -0.000326 -0.011827 0.006552 0.007319 C -0.00101 -0.01539 -0.00754 -0.01248 -0.012 -0.00368 [ 1.15131] [ 0.74330] [-0.04331] [-0.94805] [ 0.54617] [ 1.98732] 27 The Variance Decomposition Analysis Table 2.6: The variance decomposition results for CDS and HPE variables. Varia nce Deco mp. of DLN HPI SE DLNHPI DLNCDS5 DLNBIST1 00 DLNINF DLNINT DLNUS D 1 0.009475 100 0.000000 0.000000 0.000000 0.000000 0.000000 2 0.012908 92.87697 0.053523 0.7068 2.9668 0.799712 3.596185 3 0.015442 88.20061 0.707534 1.898893 3.9785 0.99082 6.223596 4 0.017299 86.21814 1.353074 2.043945 4.8573 1.23157 7.295915 5 0.018718 85.27771 1.646281 2.120179 5.788 1.37639 7.791354 6 0.019838 84.7083 1.796934 2.170917 6.7516 1.466764 8.105445 7 0.020742 84.30434 1.899863 2.212459 7.727 1.526262 8.330036 8 0.021481 84.00809 1.977071 2.242117 8.7086 1.569526 8.494498 9 0.022088 83.78634 2.035422 2.264074 9.6947 1.602071 8.617306 10 0.022592 83.61646 2.080085 2.280835 10.6841 1.627095 8.711386 Varia nce Deco mp. of DLN CDS5 SE DLNHPI DLNCDS5 DLNBIST1 00 DLNINF DLNINT DLNUS D 1 0.144723 0.154869 99.84513 0.000000 0.000000 0.000000 0.000000 2 0.150938 0.164819 91.79323 5.022744 2.7441 1.244614 1.030409 3 0.151596 0.185679 91.40846 5.206107 3.9214 1.249179 1.029105 4 0.151643 0.200065 91.35187 5.208489 4.9339 1.248414 1.057254 5 0.151679 0.220827 91.32125 5.210403 5.9335 1.248129 1.065876 6 0.151698 0.239405 91.30131 5.209132 6.9334 1.248953 1.067792 7 0.151712 0.255093 91.28438 5.208585 7.9335 1.249179 1.069243 8 0.151725 0.268184 91.26989 5.208173 8.9336 1.249322 1.07079 9 0.151735 0.279306 91.25752 5.207842 9.9337 1.249438 1.072177 10 0.151744 0.288761 91.24704 5.207536 10.9337 1.249548 1.073337 Table 2.6 shows the variance decomposition table displays the findings of the variance decomposition analysis conducted for DLNHPI, DLNCDS5, DLNBIST100, DLNINF, DLNINT, and DLNUSD. The variance decomposition of two variables, DLNHPI and DLNCDS5, is shown in this table, along with four other explanatory variables: DLNBIST100, DLNINF, DLNINT, and DLNUSD. The variance decomposition calculates the proportion of the total variance in the dependent variable (DLNHPI or DLNCDS5) that can be attributed to each explanatory variable, both individually and collectively. Table 2.6 shows that the variance of DLNHPI is entirely changed by its own shocks in the first period, with no contribution from other variables. From the second period onwards, we see that the variance of DLNHPI is explained to varying degrees by other 28 factors. In the second period, 92.88% of the volatility in DLNHPI is due to its own shocks, with minor contributions from DLNUSD (3.60%) and DLNINF (2.97%). The impacts of DLNUSD and DLNINF on the variance of DLNHPI rise over the ten periods, indicating that these factors become more relevant in explaining DLNHPI movements over time. After the ten periods, the most important variables leading to changes in HPI are DLNINF (10.7%) and DLNUSD (8.7%). By the end of the ten periods, we observe that DLNCDS5 (%2) has a minor impact on the changes in housing prices. Overall, the variance decomposition study shows that DLNHPI fluctuations are caused by both its own shocks and shocks in other variables, particularly DLNINF and DLNUSD. The table shows the results of the variance decomposition analysis for the DLNCDS5 variable, which shows that 99.8% of the variance was driven exclusively by its own shocks in the first period. Beginning with the second period, DLNCDS5 shocks account for 91.8% of the variation, with minor contributions from DLNBIST100 (5%) and DLNINF (2.74%). The effects of DLNINF on DLNCDS5 variance increased during the ten periods, whereas the effects of DLNBIST100 remained constant. After ten periods, the most relevant variables affecting changes in DLNCDS5 were found to be DLNINF (10.9%) and DLNBIST100 (5.2%). According to the variance decomposition analysis in Table 2.6, the HPI variable has only accounted for a 0.2% contribution to the changes in the CDS5 variable at the end of the ten periods. These results indicate that DLNINF (10.7%) is the most important explanatory variable for both DLNHPI and DLNCDS5. These findings are regarded as important for comprehending the dynamics of the housing market and making sound policy decisions. Impulse-Response Analysis Impact-response analysis is a method that is used to determine the effect of a random shock on other variables within a system. The impact-response function measures the response of a dependent variable to a randomly chosen shock, reflecting its effect on the current and future values of internal variables. Impact-response functions show the effects of shocks on variables and when they occur. 29 In the Impact-Response function graphs, the horizontal axis shows the duration of the response, and the vertical axis shows the size of the response. While the interrupted lines in the graphs indicate the confidence interval of ± 2 standard errors, the straight lines represent the response of the dependent variable over time to a shock of 1 standard error.( Kilian, L., & Lütkepohl, H. (2017). Figure 2.3: Impulse-response graphs of HPE variable. In Figure 2.3 and Figure 2.4, the responses of all variables to a unit shock applied to themselves and all other variables were measured. When examining the Impulse- Response Graphs, it is observed that the housing price index responds positively to its own shocks. These positive responses initially decrease until the second period and then stabilize. As can be understood from the graphs, the largest effect of a one- standard-deviation shock is on itself. In the case of the housing price index, it is observed that CDS premiums had no effect on the housing market until the second period, but had a positive impact from the second period to the third period and this effect remained stable in the following periods. When examining the reactions in the housing price index, it can be observed that a one-standard deviation shock in BIST100 variable results in a positive response in the housing price index up to the third period and increases, followed by a decrease until the fourth period and stabilization thereafter. A one-standard deviation shock in the inflation variable also elicits a positive response in the housing price index up to the second period, increasing and being followed by a decrease until the third period and stabilization thereafter. With respect to the interest rate variable, a negative response is observed up to the second period and stabilization thereafter. As for the effective exchange rate, it rapidly 30 increases in a positive direction up to the second period, followed by a slower increase until the fourth period and then a negative response thereafter. Figure 2.4: Impulse-response graphs of CDS variable. Granger Causality Analysis The Granger Causality Test is a statistical test that determines whether one time series can be used to predict another. The test was named after Nobel Laureate economist Clive Granger, who proposed the idea in the 1960s. The Granger Causality Test is widely used in econometrics, finance, and other fields that analyze time series data. The test is based on the assumption that if one time series causes another, then changes in the first time series should be useful in predicting changes in the second time series. The null hypothesis of the Granger Causality Test is that a time series' past values have no predictive power over another time series' future values. If the null hypothesis is rejected, then it can be concluded that there is evidence of Granger causality between the two-time series. (Granger, C. W. J. (1969)) Granger causality analysis checks whether the current value of one of the variables and the lagged values of other variables create causality. If the explanatory power of the model created to explain the value of the variable (Y) at time t increases when the lagged values of the other variable (X) are included; It is concluded that variable X is the Granger cause of variable Y. 𝑦! = 𝑎" +∑ 𝛽$& $'" 𝑋!#$ + ∑ 𝛿() ('" 𝑌!#(+𝑒*! (2.1) 31 𝑥! = 𝑎% + ∑ 𝜃$& $'" 𝑋!#$ +∑ 𝛾() ('" 𝑌!#(+𝑒+! (2.2) In the causality analysis, the significance of the H0 and H1 hypotheses is checked. Rejecting the H0 hypothesis means a Granger causality relationship exists between the variables. Table 2.7: The Granger causality tests results. Dependent variable: DLNHPI Excluded Chi-sq df Prob. DLNCDS5 2.425438 1 0.1194 DLNBIST100 0.267124 1 0.6053 DLNINF 2.267422 1 0.1321 DLNUSD 15.81587 1 0.0001 DLNINT 2.793784 1 0.0946 All 25.48061 5 0.0001 Dependent variable: DLNCDS5 Excluded Chi-sq df Prob. DLNHPI 2.004022 1 0.1569 DLNBIST100 6.675294 1 0.0098 DLNINF 2.580892 1 0.1082 DLNUSD 2.656091 1 0.1032 DLNINT 1.756226 1 0.1851 All 13.75318 5 0.0173 Dependent variable: DLNBIST100 Excluded Chi-sq df Prob. DLNHPI 6.380205 1 0.0115 DLNCDS5 1.697445 1 0.1926 DLNINF 0.703380 1 0.4017 DLNUSD 2.178548 1 0.1399 DLNINT 0.052451 1 0.8189 All 9.729071 5 0.0833 Dependent variable: DLNINF Excluded Chi-sq df Prob. DLNHPI 0.600666 1 0.4383 DLNCDS5 0.939090 1 0.3325 DLNBIST100 1.694089 1 0.1931 DLNUSD 6.345662 1 0.0118 DLNINT 0.118500 1 0.7307 All 14.66282 5 0.0119 32 Table 2.7(continue): The Granger causality tests results. Dependent variable: DLNUSD Excluded Chi-sq df Prob. DLNHPI 5.918154 1 0.0150 DLNCDS5 18.04748 1 0.0000 DLNBIST100 0.247181 1 0.6191 DLNINF 2.874854 1 0.0900 DLNINT 0.506783 1 0.4765 All 40.02624 5 0.0000 Dependent variable: DLNINT Excluded Chi-sq df Prob. DLNHPI 2.254005 1 0.1333 DLNCDS5 1.646359 1 0.1995 DLNBIST100 0.131204 1 0.7172 DLNINF 1.654733 1 0.1983 DLNUSD 1.301653 1 0.2539 All 12.44962 5 0.0291 Table 2.7 shows the Granger causality tests for the five-variable VAR (vector autoregression) model: 146 observations were investigated between September 2010 and December 2022. By testing whether previous values of one variable may predict the current values of another, the VAR model illustrates the direction and strength of causal links. For each dependent variable, the table shows the excluded variable, the chi-squared statistic, the degrees of freedom (df), and the probability (Prob.) of the null hypothesis that the excluded variable does not Granger-cause the dependent variable. The Granger causality analysis table shows that the independent variable DLNUSD is significant at the 1% level, indicating that changes in the US Dollar exchange rate can be used to predict changes in the Turkish House Price Index (HPI), and thus the USD variable is the Granger cause of the HPI variable. Furthermore, it is clear that the DLNBIST100 independent variable is the Granger cause of CDS5. The DLNBIST100 variable has a significant effect on the credit default swap margin (CDS5), indicating that fluctuations in the BIST 100 index can predict changes in Türkiye's CDS. In terms of DLNINF, the results show that DLNUSD has a strong Granger-causal impact at a 1% level of significance, showing that past DLNUSD values can help anticipate current DLNINF values. 33 CONCLUSION The number of studies on CDS premiums and housing prices has been increasing in the literature, particularly after the 2008 financial crisis. Numerous studies focus on the determinants of countries' CDS premiums and their relationship with financial variables. Similarly, many studies investigate the macroeconomic variables that affect housing prices. However, a limited amount of research specifically examines the relationship between CDS premiums and housing prices. Stability and confidence in countries play a crucial role in increasing investment volume and consequently revitalizing housing demand (KPMG, 2018:10). Therefore, CDS premiums are considered a reliable indicator for predicting changes in housing prices. High CDS premiums have a direct impact on a country's borrowing costs, leading to an increase in borrowing interest rates and the tightening of lending conditions by lending institutions. When CDS premiums are high, it indicates a higher perception of risk associated with lending to that particular country. To mitigate their risk exposure, lending institutions respond by charging higher interest rates on loans or imposing stricter lending requirements. This helps them compensate for the perceived higher default risk and potential losses. As a result, the government and private companies face higher borrowing costs, making it more expensive to access financing. This, in turn, can lead to liquidity issues for local banks, a decrease in credit supply, and overall difficulties in obtaining necessary funds. Consequently, the increased borrowing costs and reduced credit availability have a contracting effect on economic activity. Hence, the relationship between CDS premiums and borrowing costs plays a critical role in shaping a country's financial stability and economic growth. Therefore, governments and policymakers need to focus on maintaining economic and financial stability in order to keep CDS premiums low and investor confidence high. Several studies in the academic literature indicate that the escalation of CDS premiums leads to elevated financing costs for financial institutions and banks, consequently resulting in a decline in lending. This, in turn, has been observed to have a significant impact on housing prices, leading to a noticeable decrease (Benbouzid, Mallick, & Pilbeam, 2018). Numerous studies in the literature have indicated a negative impact of inflation on 34 housing prices, apart from CDS. Furthermore, several research works have examined the influence of exchange rates on housing prices, with most studies demonstrating a negative effect of exchange rates on housing prices. Studies exploring the relationship between policy interest rates and housing prices have revealed that policy interest rates are a significant factor affecting housing prices. In this thesis, the impact of these variables on Türkiye's CDS premiums was also be investigated. Studies on the relationship between CDS premiums and housing prices have shed light on how changes in CDS premiums can affect credit availability for homebuyers and overall housing demand, providing insights into potential mechanisms through which CDS premiums can impact housing prices.. Previous research on the determinants of Türkiye's CDS premiums has shown that global variables more influence volatility in premiums than domestic variables. However, the high volatility in premiums is believed to stem from political and economic issues. This thesis aims to investigate the relationship between CDS premiums and housing prices in Türkiye, as well as the other macroeconomic variables influencing this relationship. The vector autoregression (VAR) model has been employed to statistically analyze variables such as the housing price index, effective exchange rate, BIST 100, inflation, interest rates, and Türkiye's CDS premiums. Econometric approaches such as VAR (Vector Autoregression), Impulse-Response Analysis, and Granger Causality Tests have been employed to determine the causal relationship between variables. The analysis covers the period from September 2010 to December 2022, with monthly data on the variables under investigation. The data utilized in this thesis has been obtained from reliable sources such as the Central Bank and Bloomberg. The selected time range from 2010 to 2022 has provided a comprehensive understanding of the relationship between CDS premiums, selected macroeconomic variables, and housing prices in Türkiye. Moreover, the study also examined the role of other economic and financial variables, such as interest rates, inflation, real effective exchange rates, and BIST100, in the relationship between CDS premiums and housing prices, using multivariate VAR analysis. In the VAR approach, the time series of the variables included in the analysis should be stationary and not have a unit root. Therefore, the logarithms of the variables were first taken and included in the analysis as logarithms. Then, the ADF unit root test was 35 applied to the series, and it was seen that the series were not stationary. The logarithmic series are made stationary by taking the first-order differences. The criteria' results were examined to select the optimum lag length according to the VAR analysis information criteria. The optimum lag length was determined as the second lag where all information criteria are minimum. After the optimum lag length was determined, the VAR model was created. After the VAR model was created, variance decomposition was done. The variance decomposition results show that the variance of house prices is entirely changed by its own shocks in the first period without any contribution from other variables. Starting from the second period, we see that house prices are explained to varying degrees by other factors. In the second period, 92.88% of the volatility in housing prices was caused by their own shocks, while 3.60% was caused by the real effective exchange rate and 2.97% inflation. The effects of the real effective exchange rate and inflation increase over ten periods, indicate that these factors have become more relevant in explaining house price movements over time. It has been determined that the most critical variables that cause changes in housing prices after ten periods are still inflation (10.7%) and real effective exchange rate 8.7%. The Granger causality analysis also shows that the real effective exchange rate is significant at the 1 % level, indicating that changes in the real effective exchange rate can be used to predict variations in the Türkiye House Price Index (HPI). Furthermore, the BIST100 variable serves as the Granger cause for the 5-year CDS premiums. The significant impact of the BIST100 on the credit default swap margin (CDS5) indicates that fluctuations in the BIST100 index can be utilized to forecast variations in Türkiye's 5-year CDS premiums. In general, the variance decomposition study shows that the fluctuations in housing prices are caused by their own shocks and shocks in other variables, especially inflation and real effective exchange rate. According to the variance decomposition results of Türkiye's 5-year CDS premiums, it shows that 99.8% of the variance in the first period is due only to its own shocks. On the other hand, housing prices contributed only 0.2% to the changes in Türkiye's 5-year CDS premiums at the end of the decade. Starting from the second period, shocks from Türkiye's 5-year CDS premiums account for 91.8% of the change, with minor contributions from BIST100 (5%) and inflation (2.74%). effects remained constant. 36 After ten periods, the variables that most affected the changes in 5-year CDS premiums were inflation with 10.9% and BIST100 with 5.2%. The results show that inflation (10.7%) is the most important explanatory variable for both house prices and CDS premiums. When the impact response analysis of the housing price index is examined, it is seen that CDS premiums did not have any effect on housing prices until the second period, but had a positive effect from the second to the third period, and this effect remained constant in the following periods. The lack of initial response of housing prices to CDS shocks indicates that the short-term effects of these shocks are limited. The analysis results reveal that there is no immediate reaction of housing prices to the shocks occurring in CDS, and no significant change is observed in housing prices. However, the increase in housing prices from the second period to the third period indicates that housing prices become more responsive to CDS shocks over time. During this period, it is observed that when a CDS shock occurs, housing prices are positively affected and increase.When the reactions in the housing price index are analyzed, it is seen that a shock of one standard deviation in the BIST100 variable causes a positive reaction in the housing price index until the third period and then a decrease until the fourth period. A shock of one standard deviation in the inflation variable also reacts positively to the housing price index until the second period, increases until the third period, and stabilizes. In the interest rate variable, a negative reaction is observed until the second period, and then stability. The effective exchange rate, on the other hand, increases rapidly in the positive direction until the second period, shows a slower increase until the fourth period, and then reacts negatively. In conclusion, this study has demonstrated that the changes in housing prices are influenced by various factors, including their own dynamics and macroeconomic variables such as inflation and real effective exchange rate. The analysis also reveals that the fluctuations in Türkiye's 5-year CDS premiums are primarily driven by internal shocks within the CDS market, and housing prices do not exert a significant impact on the changes in these premiums. Furthermore, through the impulse-response analysis, it has been observed that increases in CDS premiums have a positive effect on housing prices in the medium term. This shows that fluctuations in CDS premiums can indirectly influence the housing market, albeit with a delayed response. 37 These findings highlight the importance of considering the interplay between housing prices, CDS premiums, and macroeconomic factors when analyzing the dynamics of the housing market. Understanding these relationships can provide valuable insights for policymakers and market participants in assessing the potential risks and vulnerabilities associated with housing and financial markets. Overall, this thesis contributes to the existing literature by shedding light on the complex dynamics between housing prices and CDS premiums, providing valuable implications for both academic research and practical decision-making processes in the housing and financial sectors. 38 39 REFERENCES Akkaya, M. (2017). Türk Tahvillerinin CDS Primlerini Etkileyen İçsel Faktörlerin Analizi. Maliye Finans Yazıları,130-145. Aksoylu, E., and Görmüş, Ş. (2018). Gelişmekte Olan Ülkelerde Ülke Riski Göstergesi Olarak Kredi Temerrüt Swapları: Asimetrik Nedensellik Yöntemi. Ekonomik ve Sosyal Araştırmalar Dergisi, 14(1), 15-33. Akyol, H., and Baltacı, N. (2019). CDS Primlerinin Makroekonomik Belirleyicilerinin İncelenmesi: ARDL Sınır Testi Yaklaşımı. Global Journal of Economics and Business Studies, 8(16), 33-49. Angelopoulos, F. (2017). The Relationship between CDS spreads and Macroeconomic factors of the Countries of The Eurozone. Unpublished master's thesis, Tilburg University, School of Economics and Management . Antonakakis, N., and Floros, C. (2016). Dynamic interdependencies among the housing market, stock market, policy uncertainty and the macroeconomy in the United Kingdom. International Review of Financial Analysis, 4. Badurlar, İ. (2021). Makroekonomik Göstergeler ve Ülke Risk Primi İlişkisinin İncelenmesi: Türkiye Örneği. Yaşar Üniversitesi E-Dergisi, 16(61), 310-329. Başarır, Ç., and Keten, M. (2016). Gelişmekte Olan Ülkelerin Cds Primleri İle Hisse Senetleri Ve Döviz Kurları Arasındaki Kointegrasyon İlişkisi: A Coıntegratıon Analysis Between Cds Premiums, Stock Indexes And Exchange Rates In Emerging Countries. Mehmet Akif Ersoy Üniversitesi Sosyal Bilimler Enstitüsü Dergisi, 8(15), 369-380. Bektur, Ç., and Malcıoğlu, G. (2017). Kredi Temerrüt Takaslari İle Bist 100 Endeksi Arasindaki İlişki: Asimetrik Nedensellik Analizi. Bolu Abant İzzet Baysal Üniversitesi Sosyal Bilimler Enstitüsü Dergisi, 17(3), 73-83. Benbouzid, N., Mallick, S., and Pilbeam, K. (2018). The housing market and the credit default swap premium in the UK banking sector: A VAR approach. Research in International Business and Finance, 44, 1-15. Chan, K. C., Fung, H. G., and Zhang, G. (2009). On the relationship between Asian credit default swap and equity markets. Journal of Asian Business Studies, 4(1), 2-11. Danacı, M. C., Şit, M., and Şit, A. (2017). Kredi Temerrüt Swaplarının (CDS’lerin) Büyüme Oranıyla İlişkilendirilmesi: Türkiye Örneği. Aksaray Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi, 9(2), 67-78. 40 Demary, M. (2009). The Link Between Output, Inflation, Monetary Policy and Housing Price Dynamics. Munich Personal Repec Archive MPRA Paper No: 15978, University Library of Munich, Germany. Değirmenci, N., and Pabuçcu, H. (2016). Borsa İstanbul Ve Risk Primi Arasındaki Etkileşim: Var ve NARX Model. Akademik Sosyal Araştırmalar Dergisi, 4, 248-261. Dickey, D. A., and Fuller, W. A. (1979). Distribution of the Estimators for Autoregressive Time Series with a Unit Root. Journal of the American Statistical Association, 74, 427-431. Dilber, İ., & Sertkaya, Y. (2016). 2008 Finansal Krizi Sonrası Türkiye’de Konut Fiyatlarının Belirleyicilerine Yönelik Analiz. Anemon Muş Alparslan Üniversitesi Sosyal Bilimler Dergisi, 4(1), 11-29. Dumontaux, N., and Pop, A. (2013). Understanding the market reaction to shockwaves: Evidence from the failure of Lehman Brothers. Journal of Financial Stability, 9(3), 269-286. Engle, R. F., and Granger, C. W. (1987). Cointegration and Error Correction: Representation. Econometrica, 55(2), 251-276. Eren, M., and Başar, S. (2016). Makroekonomik Faktörler Ve Kredi Temerrüt Takaslarının Bıst-100 Endeksi Üzerindeki Etkisi: Ardl Yaklaşımı. Atatürk Üniversitesi İktisadi ve İdari Bilimler Dergisi, 30(3), 576-589. Gebeşoğlu, P. F. (2019). Housing Price Index Dynamics in Türkiye. Yaşar Üniversitesi E-Dergisi, Jourrnal of Yasar / Special Issue on Applied Economics and Finance, 100-107. Granger, C. (1969). Investigating Causal Relations by Econometric Models and Cross-Spectral Methods. Econometrica, 37(3), 424-438. Hancı, G. (2014). Kredi temerrüt takasları ve BIST-100 arasındaki ilişkinin incelenmesi. Maliye ve Finans Yazıları, 102, 9-22. John V. D., Muellbauer J., Anthony Murphy.(2010) Housing markets and the financial crisis of 2007–2009: Lessons for the future, Journal of Financial Stability, Volume 6, Issue 4, 2010, 203-217. Kargı B. (2014). Credit Default Swap: The Analysis of Time Series For The Interaction With The Interest Rates and The Growth in Turkish Economy”. Montenegrin Journal of Economics, 10(1), 59-66. Karadaş, H. A., and Salihoglu, E. (2020). Seçili Makroekonomik Değişkenlerin Konut Fiyatlarına Etkisi: Türkiye Örneği. Ekonomik ve Sosyal Araştırmalar Dergisi, 16(1), 63-80. Kazdagli, H. (1996). Türkiye Cumhuriyeti Merkez Bankası'nın Kuruluş Tarihçesi ve 1934-1938 Dönemindeki Para Politikasının VAR Yöntemiyle Analizi. Hacettepe Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi, 14(2), 35-52. Kilian, L., and Lütkepohl, H. (2017). Structural Vector Autoregressive Analysis (Themes in Modern Econometrics). Cambridge: Cambridge University Press. 41 Korkmaz, Ö. (