스트레스 테스트와 Monte Carlo 시뮬레이션을 통한 국내금융기관의 주택담보대출 신용위험분석

박연우(Yun-Woo Park),방두완(Doo-Won Bang) 한국주택학회 2011 주택연구 Vol.19 No.4

본고에서는 스트레스 테스트와 Monte Carlo 시뮬레이션 방법론을 사용하여 국내 금융기관의 주택담보대출의 예상손실을 추정한다. 스트레스 테스트는 이례적인 거시경제 충격으로 인한 금융기관의 예상손실을 측정하는 방법으로 본고에서는 거시경제의 충격으로 발생하는 스트레스 요인 중 주택가격 폭락과 차입자의 상환능력급락을 고려한다. 1998년 외환위기 당시 대표적인 부동산 시장지표인 국민은행 아파트가격지수 기준으로 서울, 강남, 전국 아파트 가격은 각각 17.8%, 17.6%, 14.9% 하락하였다. 이를 기초로 이례적인 경기침체상황이 발생한 경우 주택가격이 20%까지 하락하고 차입자의 신용등급이 3등급까지 하락할 때 발생하는 주택담보대출 손실규모를 산출한다. 본 논문의 특징은 지금까지 국내에 보고된 스트레스 테스트 논문과는 달리 거시경제에 이례적 충격이 발생했을 때 주택담보대출의 부도율과 부도대출손실률(loss given default; LGD)의 변화를 미시적 자료를 사용하여 실증적으로 추정하여 손실규모를 추정하였다는 것이다. 구체적으로 주택가격하락이 담보가치하락을 통해 부도율과 LGD에 미치는 영향을 추정하였다. 상환능력하락도 없고 주택가격도 변화하지 않을 때 주택담보대출 포트폴리오의 예상손실률은 0.27%; 상환능력하락 없이 주택가격만 10%, 20% 하락할 때 예상손실률은 0.34%, 0.44%; 주택가격이 20% 하락하고 상환능력하락이 1등급, 2등급, 3등급 하락할때 예상손실률은 1.50%, 2.32%, 3.49%로 각각 추정되었다. 그러나 주택가격이 30%, 40%, 50% 하락하고 신용등급 3등급 하락하는 스트레스 상황(미국의 경우와 같은 극심한 스트레스 상황과 유사) 하에서 예상손실률은 2.66%~3.41%로 증가하는 것으로 추정 되었다. Monte Carlo 시뮬레이션을 통한 분석결과 주택가격 20%하락과 신용등급 1등급, 2등급, 3등급 하락시 99.9% VaR(최대 손실예상치)는 각각 1.518%, 2.261%, 3.307%로, 99.99% VaR는 각각 1.678%, 2.421%, 3.466%로 추정되었다. 국내주택담보대출은 담보부 대출로 신용대출에 비해 안정성이 높은데다 국내금융기관들의 낮은 담보인정비율로 인해 주택가격하락 스트레스에는 높은 안정성을 보이나 가계의 상환능력하락 스트레스에는 취약함을 확인하였다. In this paper, we estimate the expected loss of mortgage loans held by Korean financial institutions using stress test as well as Monte Carlo simulation. Stress test attempts to measure the expected loss incurred by the financial institutions caused by an extraordinary level of shocks in the macro-economy. As macro-shocks we consider a dramatic fall in the house price and the impairment in the borrower's debt repayment ability. The Kookmin house price index for Seoul, Kangnam, and the whole country, fell by 17.8%, 17.6%, 14.9%, respectively, following the 1998 foreign exchange crisis in Korea. On the basis of this, we estimate the mortgage loan loss caused by the fall in the house price of 20% as well as up to 3 notch drop in the borrower's credit rating. The most salient aspect of this paper is the fact that unlike the other research papers which use the stress test methodology we estimate the default rate, LGD (loss given default) and the loan loss regression equations using micro data empirically. Specifically, we estimate the effect of the house price change, which reduces the value of the loan collateral, on the default rate and the LGD. We also model the impact of the reduction in the borrower ability to repay the loans on the default rate empirically. In the base case, where there is no macro-stress, the expected loan loss is 0.27%. When the house price falls by 10% and 20% and the borrower's credit rating does not fall, the expected loan losses are 0.34% and 0.44%, respectively. We find that when the house price falls by 20% and the borrower's credit rating falls by 1 notch, 2 notch, and 3 notch the expected loan losses are 1.50%, 2.32%, 3.49%, respectively. Furthermore, under more extreme stress conditions where the house price falls by 30%, 40% and 50% and the credit ratings of prime borrowers, middle credit risk borrowers as well as high credit risk borrowers drop by three notches (similar to the stress in the US), the expected mortgage loan loss rate rises to 2.66%-3.41%. On the other hand, the results of Monte Carlo simulation shows that when the house price falls by 20% and the borrower's credit rating falls by 1 notch, 2 notch, and 3 notch the expected loan loss is such that the 99.9% VaRs (value-at-risk) are 1.518%, 2.261%, 3.307%, respectively and the 99.99% VaRs (value-at-risk) are 1.678%, 2.421%, 3.466%, respectively. The Korean mortgages show a high stability in the face of the house price drop stress since the lending institutions apply a very low loan-to-value ratio in addition to the mortgage loans being collateralized with residential properties. On the other hand, they are sensitivie to the stress caused by the reduction in the borrowers repayment ability.

고정금리 주택담보대출 조건부 조기상환율의 결정요인 분석

박연우(Yun-Woo Park),방두완(Doo-Won Bang) 한국주택학회 2011 주택연구 Vol.19 No.3

본 연구에서는 주택담보대출 풀의 월별 조기상환 시계열을 사용하여 고정금리 장기주택담보대출의 조건부 조기상환율(CPR: conditional prepayment rate)에 영향을 미치는 요인을 분석하였다. 금리변동, 주택가격변화, 주택담보대출의 경과기간, 이사계절성을 설명변수로 설정하여 조건부 조기상환율 모형을 추정하였다. 한국주택금융공사가 2004년부터 2006년까지 유동화한 주택담보대출의 풀별 데이터를 사용하여 분석한 결과 주택담보대출 계약이자율과 현재 주택담보대출시장의 이자율과의 차이, 대출경과기간, 주택가격 상승률 그리고 주택거래량은 조기상환율에 모두 양(+)의 영향을 미치는 것으로 나타났다. 이러한 현상은 주택가격 상승기뿐만 아니라 주택가격 하락기에서도 확인되었고 금융위기 이전과 이후에서 모두 확인되었다. 주택담보대출금리가 1%p 하락하면 CPR은 연 3%p-4%p 증가하고 주택가격이 월 1%p(연 12%p) 증가하면 CPR은 연 1%p-2%p 증가한다. 주택가격 상승기와 하락기, 금융위기 이전과 이후의 소기간별로 추정된 조기상환 회귀식의 모수가 동일하지 않음으로 모수의 기간별 불안정성이 존재함을 확인하였다. 또한 주택가격효과, 즉 주택가격이 CPR에 양의 영향을 미치는 현상은 주택가격이 증가(감소)하면 거래량이 증가(감소)하고 거래량이 증가(감소)하면 조기상환율이 증가(감소)하기 때문인 것을 확인하였다. 풀 고정효과, 시간(연 및 월)고정효과, 자기상관성, 이분산성, 내생성, 시차효과을 통제하여도 추정된 모수는 안정적으로 나타났다. 개별대출 분석 방법(로짓모형이나 위험모형 등)에 비해 풀 시계열 회귀분석은 조건부 조기상환율을 직접 추정할 수 있다는 장점이 있어 주택담보대출 유동화증권(MBS)의 설계 및 가격결정에 유용하다. We analyse the determinants of the conditional prepayment rates (CPR) of fixed rate mortgage loans using pool level time-series regression approach. Using the prepayment rates time series observations of pools of mortgages originated and securitized for the 2004-2006 period, we estimate the CPR regression equation. We find that the CPR is positively related to the spread between the mortgage loan interest rate and the current market mortgage loan interest rate, the house price increase, the loan age and the transaction volume. CPR rises 3%p-4%p for 1%p fall in interest rate, rises about 1%p for 1%p-2%p rise in house price index. The model coefficients are relatively stable even when we control for pool fixed effects, year fixed effects, month fixed effects, time lag, autocorrelation and heteroscedasticity. We find some evidence of change in CPR model parameters caused by external macro shocks. We also find that the house price effect on the CPR is in part accounted for by the transaction volume effect. This study contributes to the development of the MBS markets in Korea by deriving CPR models, which predict the CPR directly so that they can be readily used to design and price the MBS.

평가기반 아파트가격지수에서의 비대칭 평활화 현상에 관한 연구

박연우(Yun-Woo Park),방두완(Doo-Won Bang) 한국주택학회 2011 주택연구 Vol.19 No.2

본 연구에서는 대표적인 평가기반 아파트가격지수인 국민은행 아파트가격지수의 평활화 현상(smoothing)을 국토해양부가 공표하고 있는 실거래가격으로 만든 반복매매지수(repeat sales index)를 사용하여 조사하였다. 실증분석결과를 요약하면 평가기반 아파트가격지수는 상승시장보다는 하락시장에서 평활화가 큰 것으로 나타났다. 상승시장에서는 평가기반 아파트가격지수는 실거래 아파트가격지수를 상대적으로 빨리 추적하고 하락시장에서는 상대적으로 천천히 추적하는 것으로 나타났다. 또한 평가기반 아파트가격지수의 변동성도 실거래 아파트가격지수의 변동성에 비해 상승시장보다는 하락시장에서 더욱 작은 것으로 분석되었다. 새로운 거래가격이 평가에 반영되는 정도는 하락시장에서는 20%, 상승시장에서는 45% 정도인 것으로 나타났다. 평가기반 아파트가격지수의 측정오차(measurement errors)모형의 분석결과 하락기에 확대되는 평가기반 아파트가격지수의 평활화 현상은 지속되는 경향이 강하다는 것을 확인하였다. 평가기반 아파트가격지수의 평활화현상이 아파트가격 상승기보다 하락기에 더 큰 것은 적은 거래량의 영향도 있지만 하락기 고유의 영향도 있음을 확인하였다. 본 연구의 결과를 전체적으로 요약한다면 평가기반 아파트가격지수는 상승시장보다는 하락시장에서 평활화 효과가 크게 나타나는 것으로 분석되었다. We investigate the smoothing phenomenon in the Kookmin Bank apartment price index, the most widely used appraisal based apartment price index in Korea using a repeat sales index based on the transaction data published by the Ministry of Land, Transport and Maritime Affairs. In summary, we find that there is a greater degree of smoothing in the appraisal based apartment price index during the up market than the down market, that is, the appraisal based apartment price index tracks the sales price based index relatively slowly during the down market while it tracks the sales price based index relatively fast during the up market. The volatility in the sales price based index is far greater than the volatility in the appraisal based price index during the down market, but not during the up market. Only about one quarter of the new sale prices are priced in the appraisal during the down market while the new sale prices account for almost half of the appraisal based apartment price index during the up market. We also show that during the down market the appraisal based apartment price index reduces the gap from the sales price based index in subsequent periods only limitedly so that the smoothing during the down market tends to be persistent. We find that the transaction volume, which is a measure of information in sale prices, has a negative influence on the gap between the appraisal based apartment price index and the sales price based index. However, even when we control for the volume effect, the down market effect is still present suggesting that the asymmetrical smoothing between the down market and the up market is caused not only by the information effect of volume but also by other factors present in the down market. Our findings are generally consistent with the asymmetrical smoothing hypothesis where the smoothing in the appraisal based apartment price index is greater during the down market than during the up market.

조기상환과 콜한도가 CMO(다계층 MBS)의 듀레이션에 미치는 영향분석: 부분이체(Partial Path-through) 구조를 중심으로

박연우 ( Yun Woo Park ),방두완 ( Doo Won Bang ) 한국파생상품학회(구 한국선물학회) 2012 선물연구 Vol.20 No.4

주택담보대출은 조기상환위험에 노출되어 있고 이런 주택담보대출을 유동화한 주택담보대출 유동화증권(MBS)도 같은 조기상환에 노출되어 MBS 투자자들은 조기상환으로 인한 만기의 축소를 예측하여 재투자위험을 관리하는 것이 필요하다. 본 논문은 기초자산의 조기상환 회귀함수를 추정하고 이를 사용하여 한국주택금융공사(KHFC)가 발행하는 CMO(다계층 MBS)의 듀레이션을 분석하였다 특히 트랜치별로 콜 금지기간과 콜한도를 지정하여 콜 금지가간 이후에도 발행기관이 조기상환위험을 부분적으로 투자자에게 이전하는 구조를 분석하였다 CMO 듀레이션에 대한 실증적 연구가 없어 실제로 CMO의 듀레이션이 어느 정도인지 CMO듀레이션의 결정요소는 무엇인지 잘 알려지지 않았다 . 따라서 본고는 CMO의 듀레이션이 어느 정도인지 CMO 듀레이션의 결정요소가 무엇인지를 보여주고 있다 본고는 실제로 듀레이션이 법정만기보다 현저하게 짧다는 것을 확인하여 듀레이션이 보다 장기적인 MBS구조 도입이 모색되어야함을 보여주고 있다. 금리확률과정하에서 CMO 듀레이션을 추정한 결과 조기상환속도가 증가함에 따라 CMO 듀레이션이 감소하고 원금대비 콜한도가 감소함에 따라 듀레이션이 증가하는 것을 확인하였다 본고는 발행기관이 조기상환위험의 일부를 인수할 때 듀레이션과 요구수익률에 미치는 영향을 구체적으로 추정하여 조기상환구조의 설계에 중효한 근거를 제공한다. Pesidential mortgage loans as well as the MBS(mortgage-backed security), which securitizes these loans. are exposed to prepayment risk We examine the effect of prepayment process on the duration of the CMO (multi-tranche MBS).ln particular. we examine the effect of partial pass-through where there is a call limit expressed as a percentage of initial tranche balance Due to the absence of empirical research on the CMO duration. neither the catual CMO duration nor the determinants of the CMO duration have been reported. Our study reports the actual CMO duration and the determinants of the CMO duration. By showing that the CMO duration is much shorter than the nominal time-to-maturity we point to the need to search for longer duration MBS structures We find that in both the deterministic and stochastic interest rate environments duration is reduced as prepayment speed rises and duration rises as call limit decreases.We make contribution to the literature by shedding light on the effect of prepayment and call limit on the duration of multi-tranche MBS. ln particular, this research characterizes the impact of the partial pass-through structuring approach on the CMO duration as well as CMO pricing Finally,it assists CMO investors in better assessing and managing reinvestment risks of pass-through products.

주택가격과 은행대출의 상관관계에 관한 연구

박연우 ( Yun Woo Park ),방두완 ( Doo Won Bang ) 한국금융연구원 2012 금융연구 Vol.26 No.1

Most households fund house purchases using mortgage loans from banks or other financial institutions since residential properties tend to be costly. This leads to an interaction between bank loans, more specifically, residential mortgage loans and house prices. While there have been numerous studies which examine the relationship between bank lending and house prices, there is no consensus on the direction of causality between the two variables. In spite of the close relationship between the housing market and the banking soundness there has been little study on the causality between bank lending and house prices in Korea. This paper aims to shed light on the relationship between bank lending and house prices and derive policy implications on housing finance. The study periods are 1986Q1~2010Q4, which we break into the real estate cycle 1 (January 1986~December 1998) and the real estate cycle 2 (January 1999~December 2010), the latter of which we break up in turn into up market in cycle 2 (January 1999~December 2006) and down market in cycle 2 (January 2007~December 2010) while the markets examined are Korea as a whole as well as Seoul and three regional markets (Daegu, Gwangju, and Busan). We examine the long run relationship using the co-integration tests and the causality using the dynamic response models, namely, VECM. We measure house prices using the Kookmim Bank apartment price index while we measure bank credit using the total bank lending as well as residential mortgage loans. Since the house price dynamics and the characteristics of banking tend to be different between Seoul and regional markets, we divide the Korean residential market into Seoul and three regional markets. In addition, we also consider other factors that can influence house prices such as household income, money supply, interest rate, the number of new housing units permitted and housing construction costs. The empirical findings are as follows. Cointegration tests show that long run relationship exists between house prices and bank lending in Korea. As for the causality between house prices and bank lending based on dynamic response models we find that for the study periods bank lending causes house prices and vice versa in all markets. However, housing demand side factors such as GDP, M2 and interest rates and housing supply side factors such as housing permits and housing construction costs do not have a consistent effect on either house prices or bank lending across various markets. Furthermore, we find that if imbalance occurs between house prices and bank lending it is reduced over time so as to maintain the long run equilibrium between house prices and bank lending. Unlike the hypothesis that the causality relationship between bank lending and house prices for Seoul would be different from that for regional markets, we find positive relationship between bank lending and house prices consistently for Seoul and regional markets. This result suggests that as the Korean residential mortgage markets mature there is a gradual convergence of the housing financing markets between Seoul and the regional markets. Our findings suggest that, since bank lending causes house prices, sudden rise (fall) causes sudden rise (fall) in bank lending leading to banking instability. They also suggest that, since house prices cause bank lending, ability-to-repay tests such as DTI serve as an effective house price stabilizer.

건설업체의 재무건전성모형 추정과 건설업의 스트레스 테스트

박연우 ( Yun Woo Park ),방두완 ( Doo Won Bang ) 한국부동산분석학회 2011 不動産學硏究 Vol.17 No.3

We examine the influence of the macro variables on the financial soundness of the construction industry. We estimate the financial soundness model with three macro variables, namely, the change in house price, interest rates and the change in GDP as explanatory variables using the panel data of the listed as well as the unlisted construction firms in Korea for the 2000-2009 period. We find that the house price and the GDP have a positive influence, while interest rates have a negative influence, on the financial soundness of the construction firms. Next, we estimate the impact of two macro shocks in the form of a house price drop and an interest rate increase on the probability of financial distress in the construction industry. We find that unlisted firms are more vulnerable to the housing market and interest rate stresses than listed firms showing that loans to the smaller firms without access to the capital markets have a greater credit risk suggesting that the financial institutions which have lent to the unlisted firms face a greater credit loss.

ISO 표준 기반 정보보호관리체계 보안성 검토 온라인 운영 관리 시스템 개발

홍선수(Seon Soo Hong),박연우(Yeon Woo Park),윤수빈(Su Bin Yun) 한국통신학회 2020 한국통신학회 학술대회논문집 Vol.2020 No.2

[논문]콩나물 중 Cinnamanilide 잔류성 분석

김택제(Taek-Jae Kim),어연우(Yun-Woo Eo),이재성(Jae-Seong Rhee),박경배(Kyong-Pae Park) 한국분석과학회 1991 분석과학 Vol.4 No.3

한국 모기지시장의 채무불이행 및 조기상환 분석

방두완 ( Doo Won Bang ),박세운 ( Sae Woon Park ),박연우 ( Yun Woo Park ) 한국금융연구원 2010 금융연구 Vol.24 No.4

The purpose of this study is to investigate the default and the prepayment behaviors in the Korean mortgage markets. In particular, we examine the factors that play major roles in determining default and prepayment, investigate the seasoning effect in the default and the prepayment, and explore whether there is any difference between Seoul (hot housing market) and Pusan (cold housing market). We use 145,782 mortgage loans, which KHFC (Korea Housing Finance Corporation) securitized from January 2004 to December 2007. Of these there are 21,069 mortgage loans originated in Seoul and 12,503 mortgage loans in Pusan. The KHFC mortgages in the sample have the maximum maturity of 20 years and the maximum LTV of 70%. The average LTV of the overall sample is 60.3%; the average LTV is 57.8% in Seoul while the average LTV is 60.9% in Pusan. The initial loan amount is KRW 74 million in the overall sample, KRW 111 million in Seoul and KRW 66 million in Pusan. The prepayment penalties of KHFC mortgages are 2% if prepayment occurs within 1 year, 1.5% within 3 years and 1% within 5 years. We use 90 days arrears as the proxy for loan default consistent with the Basel II. Contrary to our expectation that the default rate in the regional market as such Pusan would be higher than that in Seoul, which experienced a sharp price run-up during the study period, we find that the default rate in Seoul is higher and the difference is statistically significant. During the study period of 2004~2007 the cumulative default rate is 0.78% in the overall markets, 1.40% Seoul and 1,10% in Pusan. We find that the default is higher, lower the credit rating, lower the past house price increase, higher the mark-to-market loan-to-value ratio (MLTV), and higher the DTI. The self-employed borrowers show a lower default rate. These findings are consistent with economic theories and findings reported in other markets. We find little difference in the factors that influence the mortgage default between Seoul and Pusan. The prepayment rate rises as the mortgage rate drops below the contract mortgage rate. It rises right after the 12 months and 36 months of loan age to coincide with the reduction in the prepayment penalty. The prepayment rate is higher, higher the MLTV and higher the mortgage balance. We find little difference in the factors that influence the mortgage prepayment between Seoul and Pusan. During the study period, the prepayment rate is 20.06% in the overall markets, 20.55% in Seoul and 17.73% in Pusan. From the CDR(conditional default rate) and the CPR(conditional prepayment rate) we estimate using the Kaplan-Meier product limit method, we find that the seasoning effect exists in the Korean mortgage markets as in the US mortgage markets. Therefore, using the US benchmark default model (SDA) and the US benchmark prepayment model (PSA) the estimated CDR is approximately 50~100% SDA in the overall markets and in Seoul and 50~150% SDA in Pusan while the estimated CPR is approximately 150~250% PSA in the overall markets and Seoul and 100~200% PSA. The estimated CDR and CPR statistical models, which express the CDR and CPR as a function of the age of the mortgage loan respectively, provide a valuable benchmark for practitioners. Financial institutions carrying a significant amount of mortgage loan assets would find the estimated default model useful for the credit risk management while the estimated prepayment model provides useful information for the duration management of the mortgage loan assets and the MBS that are based on these assets.

전력계통 구조물의 상태진단용 자가발전 무선 센서 노드 개발 및 평가

김창일 ( Chang Il Kim ),정영훈 ( Young-hun Jeong ),윤지선 ( Ji Sun Yun ),홍연우 ( Youn Woo Hong ),장용호 ( Yong-ho Jang ),최범진 ( Beom-jin Choi ),박신서 ( Shin-seo Park ),손천명 ( Chun Myung Son ),서덕기 ( Duck Ki Seo ),백종후 ( 한국센서학회 2016 센서학회지 Vol.25 No.5

A self-powered piezoelectric energy harvester was developed for the application in wireless sensor node. The energy harvester was evaluated with power generation characteristics for the wireless sensor node for structural diagnosis of the electric power system. The self-powered wireless sensor node was set to measure temperature, vibration frequency of the electric power system. A piezoelectric harvester composed of 7 uni-morph cantilevers (functionalized as 6 generators and 1 vibration sensor) was connected to be an array and revealed to produce significantly high output power of approximately 10 mW at 120 Hz under 3.4 g((1 g = 9.8 m/sec²). The wireless sensor node could work as the electric power generated by the developed piezoelectric harvester.