ࡱ> z`  (bjbj 7*H, &   .6 $ZZZn8N nH_v"(~>^^^^^^^$`hTc^9Zggg^Z Z _gZ pZ Zg^j Z PI2u;_0H_c cdjcZ 'lnp [^^;dH_ggggnnn*nnnnnnZ Z Z Z Z Z  Predicting loss given default (LGD) for residential mortgage loans: a two-stage model and empirical evidence for UK bank data Abstract With the implementation of the Basel II regulatory framework, it became increasingly important for financial institutions to develop accurate loss models. This work investigates the Loss Given Default (LGD) of mortgage loans using a large set of recovery data of residential mortgage defaults from a major UK bank. A Probability of Repossession Model and a Haircut Model are developed and then combined into an expected loss percentage. We find the Probability of Repossession Model should comprise of more than just the commonly used loan-to-value ratio, and that estimation of LGD benefits from the Haircut Model which predicts the discount the sale price of a repossessed property may undergo. Performance-wise, this two-stage LGD model is shown to do better than a single-stage LGD model (which directly models LGD from loan and collateral characteristics), as it achieves a better R-square value, and it more accurately matches the distribution of observed LGD. Keywords: Regression, Finance, Credit risk modelling, Mortgage loans, Loss Given Default (LGD), Basel II 1. Introduction With the introduction of the Basel II Accord, financial institutions are now required to hold a minimum amount of capital for their estimated exposure to credit risk, market risk and operational risk. According to Pillar 1 of the new Basel II capital framework, the minimum capital required by financial institutions to account for their exposure to credit risk can be calculated using two approaches, either the Standardized Approach or the Internal Ratings Based (IRB) Approach. The IRB approach is further split into two and can be implemented using either the Foundation IRB Approach or the Advanced IRB Approach. Under the Advanced IRB Approach, financial institutions are required to develop their own models for the estimation of three credit risk components, viz. Probability of Default (PD), Exposure at Default (EAD) and Loss Given Default (LGD), and this for each section of their credit risk portfolios. The portfolios of a financial institution can be broadly divided into either the retail sector, consisting of consumer loans like credit cards, personal loans or residential mortgage loans, or the wholesale sector, which would include corporate exposures such as commercial and industrial loans. The work here pertains to residential mortgage loans. In the United Kingdom, as in the US, the local Basel II regulation specifies that a mortgage loan exposure is in default if the debtor has missed payments for 180 consecutive days (The Financial Services Authority (FSA)  ADDIN EN.CITE Financial Services Authority200971717146Financial Services Authority,Prudential Sourcebook for Banks, Building Societies and Investment Firms2009http://fsahandbook.info/FSA/html/handbook/(2009), BIPRU 4.3.56 and 4.6.20; Federal Register  ADDIN EN.CITE Federal Register200732323246Federal Register,Risk-Based Capital Standards: Advanced Capital Adequacy Framwork - Basel II; Final Rule2007http://www.federalreserve.gov/generalinfo/basel2/FinalRule_BaselII/(2007)). When a loan goes into default, financial institutions could contact the debtor for a re-evaluation of the loan whereby the debtor would have to pay a slightly higher interest rate on the remaining loan but have lower and more manageable monthly repayment amounts; or banks could decide to sell the loan to a separate company which works specifically towards collection of repayments from defaulted loans; or, because every mortgage loan has a physical security (also known as collateral), i.e. a house or flat, the property could be repossessed (i.e. enter foreclosure) and sold by the bank to cover losses. In this case there are two possible outcomes: either the sale of the property is able to cover the value of the loan outstanding and associated repossession costs with any excess being returned to the customer, resulting in a zero loss rate; alternatively, the sale proceeds are less than the outstanding balance and costs and there is a loss. Note that the distribution of LGD in the event of repossession is thus capped at one end. The aim of LGD modelling in the context of residential mortgage lending is to accurately estimate this loss as a proportion of the outstanding loan, if the loan were to go into default. In this paper, we will empirically investigate a two-stage approach to estimate mortgage LGD on a set of recovery data of residential mortgages from a major UK bank. The rest of this paper is structured as follows. Section 2 consists of a literature review and discusses some current mortgage LGD models in use in the UK, followed by Section 3 which lists our research objectives. In Section 4, we describe the data available, as well as the pre-processing applied to it. In Sections 5 and 6, we detail the Probability of Repossession Model and the Haircut Model respectively. Section 7 explains how the component models are combined to form the LGD Model. In Section 8, we look at some possible further extensions of this work and conclude. 2. Literature review Much of the work on prediction of LGD, and to some extent PD, proposed in the literature pertains to the corporate sector (see Schuermann  ADDIN EN.CITE Schuermann20046666665Schuermann, TilShimko, DavidWhat Do We Know About Loss Given DefaultCredit Risk: Models and Management2ndnew basel accordcredit risk2004Risk Bookshttp://www.newyorkfed.org/research/economists/schuermann/Schuermann_LGD-What_do_we_know_1.5.pdfLGD models' explanatory variables include seniority of loan on capital structure, industry of firm, collateral commiteed towards loan, and some variable representing the state of the economy.(2004), Gupton & Stein  ADDIN EN.CITE Gupton200240404027Gupton, Greg M.Stein, Roger M.LOSSCALC: Model for Predicting Loss Given Default (LGD)2002Moody's KMV Research Publicationhttp://www.moodyskmv.com/research/files/wp/losscalc_methodology.pdfPredictive variables of recovery are type of collateral, type of debt and seniority, type of firm, industry and macroeconomic conditions. Developed a robust and predictive LGD model which can be applied to a number of defaulted instruments. This model was consistently able to outperform LGD models that rely on historical averages.(2002), Jarrow  ADDIN EN.CITE Jarrow200147474717Jarrow, RobertDefault Parameter Estimation Using Market PricesFinancial Analysts JournalFinancial Analysts Journal75-925752001http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.139.4754Similar to Frye (2000a, 2000b) Introduced a recovery model incorporating debt and equity.(2001), Truck et al.  ADDIN EN.CITE Truck200575757527Truck, StefanHarpaintner, StefanRachev, Svetlozar T.A Note on Forecasting Aggregate Recovery Rates with Macroeconomic Variablesbusiness cyclerecovery ratemultiple regressionCBOE volatility index2005Unpublished paper as part of thesis.https://statistik.ets.kit.edu/download/doc_secure1/tr_a_note_on_forecasting.pdf(2005), Altman et al.  ADDIN EN.CITE Altman200544417Altman, EdwardI.Brady, BrooksResti, AndreaSironi, AndreaThe Link between Default and Recovery Rates: Theory, Empirical Evidence, and ImplicationsThe Journal of BusinessThe Journal of Business2203-22287862005http://www.journals.uchicago.edu/doi/abs/10.1086/497044doi:10.1086/497044Corporate sector PD and LGD prediction. Benchmarking study. No single macroeconomic variable is adequate, but they do benefit statistically from the inclusion of a variable which represents the economy.(2005)), which can be partly explained by the greater availability of (public) data and because the financial health or status of the debtor companies can be directly inferred from share and bond prices traded on the market. However, this is not the case in the retail sector, which partly explains why the LGD models are not as developed as those pertaining to corporate loans. 2.1. Risk models for residential mortgage lending Despite the lack of publicly available data, particularly on individual loans, there are still a number of interesting studies on credit risk models for mortgage lending that use in-house data from lenders. However, the majority of these have in the past focused on the prediction of default risk, as comprehensively detailed by Quercia & Stegman  ADDIN EN.CITE Quercia199263636317Quercia, Roberto G.Stegman, Michael A.Residential Mortgage Default: A Review of the LiteratureJournal of Housing ResearchJournal of Housing Research341-379321992http://www.knowledgeplex.org/kp/text_document_summary/scholarly_article/relfiles/jhr_0302_quercia.pdfReview of mortgage lending models, mostly on Probability of Default models.(1992). One of the earliest papers on mortgage default risk is by von Furstenberg  ADDIN EN.CITE von Furstenberg196976767617von Furstenberg, George M. Default Risk on FHA-Insured Home Mortgages as a Function of the Terms of Financing: A Quantitative AnalysisThe Journal of FinanceThe Journal of Finance459-4772431969Blackwell Publishing for the American Finance Association00221082http://www.jstor.org/stable/2325346Paper on mortgage default risk. Characteristics of mortgage loan, like LTV, term of loan, age and income of debtor, can be used to predict whether default will occur. Work on mortgage risk models mostly focused on PD.(1969) where it was found that characteristics of a mortgage loan can be used to predict whether default will occur. These include the loan-to-value ratio (i.e. the ratio of loan amount over the value of the property) at origination, term of mortgage, and age and income of the debtor. Following that, Campbell & Dietrich  ADDIN EN.CITE Campbell198322222217Campbell, Tim S.Dietrich, J. KimballThe Determinants of Default on Insured Conventional Residential Mortgage LoansThe Journal of FinanceThe Journal of Finance1569-15813851983Blackwell Publishing for the American Finance Association00221082http://www.jstor.org/stable/2327587Investigated the impack of macroeconomic variables on mortgage loan default risk. Find that LTV is significant factor. Economy does affect probability of default, especially local unemployment rates.(1983) further expanded on the analysis by investigating the impact of macroeconomic variables on mortgage default risk. They found that loan-to-value ratio is indeed a significant factor, and that the economy, especially local unemployment rates, does affect default rates. This is confirmed more recently by Calem & LaCour-Little  ADDIN EN.CITE Calem200423232317Calem, Paul S.LaCour-Little, MichaelRisk-based capital requirements for mortgage loansJournal of Banking & FinanceJournal of Banking & Finance647-672283CapitalMortgageCredit risk20040378-4266http://www.sciencedirect.com/science/article/B6VCY-4851CNV-4/2/1a0d1f76c0fac243f5516457cededb37Doi: 10.1016/s0378-4266(03)00039-6Single-stage LGD model for mortgage loans. LTV is a significant factor for probability of default. Estimated recovery by employing spline regression to accommodate the non-linear relationships observed between both loan-to-value ratios (LTV and DLTV) to recovery rate. Achieved R-square of 0.25.(2004), who looked at estimating both default probability and recovery (where recovery rate = 1 LGD) on defaulted loans from the Office of Federal Housing Enterprise Oversight (OFHEO). Of interest was how they estimated recovery by employing spline regression to accommodate the non-linear relationships that were observed between both loan-to-value ratios (LTV at loan start and LTV at default) and recovery, which achieved an R-square of 0.25. Similarly to Calem & LaCour-Little  ADDIN EN.CITE Calem200423232317Calem, Paul S.LaCour-Little, MichaelRisk-based capital requirements for mortgage loansJournal of Banking & FinanceJournal of Banking & Finance647-672283CapitalMortgageCredit risk20040378-4266http://www.sciencedirect.com/science/article/B6VCY-4851CNV-4/2/1a0d1f76c0fac243f5516457cededb37Doi: 10.1016/s0378-4266(03)00039-6Single-stage LGD model for mortgage loans. LTV is a significant factor for probability of default. Estimated recovery by employing spline regression to accommodate the non-linear relationships observed between both loan-to-value ratios (LTV and DLTV) to recovery rate. Achieved R-square of 0.25.(2004), Qi & Yang  ADDIN EN.CITE Qi200962626217Qi, MinYang, XiaolongLoss given default of high loan-to-value residential mortgagesJournal of Banking & FinanceJournal of Banking & Finance788-799335Loss given defaultResidential mortgageDefaultRecoveryDownturnBasel II20090378-4266http://www.sciencedirect.com/science/article/B6VCY-4TGXPD0-2/2/5661d2c5f5175e96830254e52c1b3c31DOI: 10.1016/j.jbankfin.2008.09.010Single-stage LGD model for mortgage loans, in particular for loans with high LTV ratios that have gone into default. Data from private mortgage insurance companies. Achieved high values of R-square (0.6), probably because they have near-perfect expert-based information at time of default, which is not usually available to lenders.(2009) also modelled loss directly using characteristics of defaulted loans, using data from private mortgage insurance companies, in particular on accounts with high loan-to-value ratios that have gone into default. In their analysis, they were able to achieve high values of R-square (around 0.6) which could be attributed to their being able to re-value properties at time of default (expert-based information that would not normally be available to lenders on all loans; hence one would not be able to use it in the context of Basel II which requires the estimation of LGD models that are to be applied to all loans, not just defaulted loans). 2.2. Single vs. two-stage LGD models Whereas the former models estimate LGD directly and will thus be referred to as single-stage" models, the idea of using a so-called two-stage" model is to incorporate two component models, the Probability of Repossession Model and the Haircut Model, into the LGD modelling. Initially, the Probability of Repossession Model is used to predict the likelihood of a defaulted mortgage account undergoing repossession. It is sometimes thought that the probability of repossession is mainly dependent on one variable, viz. loan-to-value, hence some probability of repossession models currently in use only consist of this single variable. This is then followed by a second model which estimates the amount of discount the sale price of the repossessed property would undergo. The Haircut Model predicts the difference between the forced sale price and the market valuation of the repossessed property. These two models are then combined to get an estimate for loss, given that a mortgage loan would go into default. An example study involving the two-stage model is that of Somers & Whittaker  ADDIN EN.CITE Somers200767676717Somers, MarkWhittaker, JoeQuantile regression for modelling distributions of profit and lossEuropean Journal of Operational ResearchEuropean Journal of Operational Research1477-14871833RegressionBasel IICredit scoringHaircut distributionKernel regressionLoss given defaultProfit assessmentRevenue modelling20070377-2217http://www.sciencedirect.com/science/article/B6VCT-4MM1P37-2/2/cb4912ee43a7cce9cc516bc6aa7a8ac2DOI: 10.1016/j.ejor.2006.08.063Acknowledged the two-stage approach for modelling of mortgage loans LGD, but not detailed. Focused on the estimation of haircut, using quantile regression, because it was observed to have a non-normal distribution.(2007), who, although they did not detail the development of their Probability of Repossession Model, acknowledged the methodology for the estimation of mortgage loan LGD. In their paper, they focus on the consistent discount (haircut) in sale price observed in the case of repossessed properties and because they observe a non-normal distribution of haircut, they propose the use of quantile regression in the estimation of predicted haircut. Another paper that investigates the variability that the value of collateral undergoes is by Jokivuolle & Peura  ADDIN EN.CITE Jokivuolle200348484817Jokivuolle, EsaPeura, SamuIncorporating Collateral Value Uncertainty in Loss Given Default Estimates and Loan-to-value RatiosEuropean Financial ManagementEuropean Financial Management299-31493credit riskcollateralloss given defaultloan-to-valueBasel II2003Blackwell Publishing Ltd1468-036Xhttp://dx.doi.org/10.1111/1468-036X.0022210.1111/1468-036x.00222Work on corporate loans with collaterals. Find that variability in value of collateral affects recovery rates, thus highlighting the correlation between the value of the collateral and recovery amount. Connection between variability of LGD and uncertainty in collateral value suggest that LGD is correlated to likelihood of default and some other macroeconomic variables.(2003). Although their work was on default and recovery on corporate loans, they highlight the correlation between the value of the collateral and recovery. In summary, despite the increased importance of LGD models in consumer lending and the need to estimate residential mortgage loan default losses at the individual loan level, still relatively few papers have been published in this area apart from the ones mentioned above. 3. Research objectives From the literature review, we observe that the few papers which looked at mortgage loss did so either by directly modelling LGD (single-stage" models) using economic variables and characteristics of loans that were in default or did not look at both components of a two-stage model, i.e. haircut as well as repossession. This might be due to their analysis being carried out on a sample of loans which had undergone default and subsequent repossession, and thus removed the need to differentiate between accounts that would undergo repossession from those that would not. We note also that there was little consideration for possible correlation between explanatory variables. Hence, the two main objectives of this paper are as follows. Firstly, we intend to evaluate the added value of a Probability of Repossession Model with more than just one variable (loan-to-value ratio). Secondly, using real-life UK bank data, we would also like to empirically validate the approach of using two component models, the Probability of Repossession Model and the Haircut Model, to create a model that produces estimates for LGD. We develop the two component models before combining them by weighting conditional loss estimates against their estimated outcome probabilities. 4. Data The dataset used in this study is supplied by a major UK Bank, with observations coming from all parts of the UK, including Scotland, Wales and Northern Ireland. There are more than 140,000 observations and 93 variables in the original dataset, all of which are on defaulted mortgage loans, with each account being identified by a unique account number. About 35 percent of the accounts in the dataset undergo repossession, and time between default and repossession varies from a couple of months to several years. After pre-processing (see later in Section 5), we retain about 120,000 observations, with accounts that start between the years 1983 and 2001 (note that loans predating 1983 were removed because of the unavailability of house price index data for these older loans) and default between the years of 1988 and 2001, with at least a two year outcome window (for repossession to happen, if any). Note that this sample does not encompass observations from the recent economic downturn. Under the Basel II framework, financial institutions are required to forecast default over a 12-month horizon and resulting losses at a given time (referred to here as observation time"). As such, LGD models developed should not contain information only available at time of default. However, due to limitations in the dataset, in which information on the state of the account in the months leading up to default (e.g. outstanding balance at observation time) are unavailable, we use approximate default time instead of observation time. When applying this model at a given time point, a forward-looking adjustment could then be applied to convert the current value of that variable, for example, outstanding balance, to an estimate at time of default. Default-time variables for which no reasonable projection is available are removed. 4.1. Multiple defaults Some accounts have repeated observations, which mean that some customers were oscillating between keeping up with their normal repayments and going into default. Hereby, each default is recorded as a separate observation of the characteristics of the loan at that time. Because the UK Basel II regulations state that the financial institution should return an exposure to non-default status in the case of recovery, and record another default should the same exposure subsequently go into default again (The Financial Services Authority (FSA)  ADDIN EN.CITE Financial Services Authority200971717146Financial Services Authority,Prudential Sourcebook for Banks, Building Societies and Investment Firms2009http://fsahandbook.info/FSA/html/handbook/(2009), BIPRU 4.3.71), we include all instances of default in our analysis, and record each default that is not the final instance of default as having zero LGD (in the absence of further cost information). We note that other approaches to deal with repeated defaults could be considered depending on the local regulatory guidelines. 4.2. Time on Book Time on Book is calculated to be the time between the start date of the loan and the approximate date of default. The variable time on book exhibits an obvious increasing trend over time (cf. Figure 1) which might be partly due to the composition of the dataset. In the dataset, we have defaults between years 1988 and 2001, which just about coincides with the start of the economic downturn in the UK of the early nineties. We observe that the mean time on book for observations that default during the economic downturn is significantly lower than the mean time on book for observations that default in normal economic times.  Figure 1: Mean time on book over time with reference to year of default 4.3. Valuation of security at default and haircut At the time of the loan application, information about the market value of the property is obtained. As reassessing its value would be a costly exercise, no new market value assessment tends to be undertaken thereafter and a valuation of the property at various points of the loan can be obtained by updating the initial property value using the publicly available Halifax House Price Index (all houses, all buyers, non-seasonally adjusted, quarterly, regional). The valuation of security at default is calculated according to Equation 1:  EMBED Equation.3  (1) Using this valuation of security at default, other variables are then updated. One is valuation of property as a proportion of average property value in the region, which gives an indication of the quality of the property relative to other properties in the same area; another is LTV at default (DLTV) which is the ratio of the outstanding loan at default to the valuation of the security at default; and yet another is haircut, which we define as the ratio of forced sale price to valuation of property at default quarter (only for observations with valid forced sale price). For example, a property estimated to have a market value of 1,000,000 but repossessed and sold at 700,000 would have a haircut of  EMBED Equation.3 . 4.4. Training and test set splits To obtain unbiased performance estimates of model performance, we set aside an independent test dataset. We develop each component model on a training set before applying the models onto a separate test set that was not involved in the development of the model itself, to gauge the performance of the model and to ensure there is no over-fitting. To do so, we split the cleaned dataset into two-third and one-third sub-samples, keeping the proportion of repossession the same in both sets (i.e. stratified by repossession cases). These are then used as the respective training and test set for the Probability of Repossession Model. However, since a haircut can only be calculated in the event of repossession and sale, all non-repossessions will subsequently be removed from the training and test sample for the second Haircut Model component. 4.5. Loss given default When a loan goes into default and the property is subsequently repossessed by the bank and sold, legal, administrative and holding costs are incurred. As this process might take a couple of years to complete, revenues and costs have to be discounted to present value in the calculation of Loss Given Default (LGD), and should include any compounded interest incurred on the outstanding balance of the loan. However, in our analysis, we simplify the definition of LGD to exclude both the extra costs incurred and the interest lost, because we are not provided with information about the legal and administrative costs associated with each loan default and repossession. Hence, LGD is defined to be the final (nominal) loss from the defaulted loan as a proportion of the outstanding loan balance at (year end of) default, and where loss is defined to be the difference between outstanding loan at default and forced sale amount, if the property was sold at a price that is lower than the outstanding loan at default (i.e. outstanding loan at default > forced sale amount). If the property was able to fetch an amount greater than or equal to the outstanding loan at default, then loss is defined to be zero. If the property was not repossessed, or repossessed but not sold, loss is also assumed to be zero, in the absence of any additional information. With loss defined to be zero, LGD is of course also 0. 5. The Probability of Repossession Model Our first model component will provide us with an estimate for the probability of repossession given that a loan goes into default. 5.1. Modelling methodology We first identify a set of variables that are eligible for inclusion in the Repossession Model. Variables that cannot be used are removed, including those which contain information that is only known at time of default and for which no reasonably precise estimate can be produced based on their value at observation time (e.g. arrears at default), or those that have too many missing values, are related to housing or insurance schemes that are no longer relevant, or where the computation is simply not known. We also then check the correlation coefficient between pairs of remaining variables, and find that none are greater than %0.6%. Using these, a logistic regression is then fitted onto the repossession training set and a backward selection method based on the Wald test is used to keep only the most significant variables (p-value of at most 0.01). We then check that the signs of each parameter estimate behave logically, and that parameter estimates of groups within categorical variables do not contradict with intuition. 5.2. Model variations Using the methodology above, we obtain a Probability of Repossession Model R1, with four significant variables, loan-to-value (LTV) ratio at time of loan application (start of loan), a binary indicator for whether this account has had a previous default, time on book in years and type of security, i.e. detached, semi-detached, terraced, flat or others. In a second model, we replace LTV at loan application and time on book with LTV at default (DLTV), referred to as Probability of Repossession Model R2. Including all three variables (LTV, DLTV and time on books) in a single model would cause counter-intuitive parameter estimate signs. Another simpler repossession model fitted on the same data, against which we will compare our models, is Model R0. The latter model only has a single explanatory variable, DLTV, which is often the main driver in models used by the retail banking industry. 5.3. Performance measures Performance measures applied here are accuracy rate, sensitivity, specificity, and the Area Under the ROC Curve (AUC). In order to assess the accuracy rate (i.e. total number of correctly predicted observations as a proportion of total number of observations), sensitivity (i.e. number of observations correctly predicted to be events in this context: repossessions as a proportion of total number of actual events) and specificity (i.e. number of observations correctly predicted to be non-events in this context: non-repossessions as a proportion of total number of actual non-events) of each logistic regression model, we have to define a cut-off value for which only observations with a probability higher than the cut-off are predicted to undergo repossession. How the cut-off is defined affects the performance measures above, as it affects how many observations shall be predicted to be repossessions or non-repossessions. For our dataset, we choose the cut-off value such that the sample proportions of actual and predicted repossessions are equal. However, we note that the exact value selected here is unimportant in the estimation of LGD itself as the method later used to estimate LGD does not require selecting a cut-off. The Receiver Operating Characteristic (ROC) curve is a 2-dimensional plot of sensitivity and 1 specificity values for all possible cut-off values. It passes through points (0,0), i.e. all observations are classified as non-events, and (1,1), i.e. all observations are classified as events. A straight line through (0,0) and (1,1) represents a model that randomly classifies observations as either events or non-events. Thus, the more the ROC curve approaches point (0,1), the better the model is in terms of discerning observations into either category. As the ROC curve is independent of the cut-off threshold, the area under the curve (AUC) gives an unbiased assessment of the effectiveness of the model in terms of classifying observations. We also use the DeLong, DeLong and Clarke-Pearson test (DeLong et al.  ADDIN EN.CITE DeLong198831313117DeLong, Elizabeth R.DeLong, David M.Clarke-Pearson, Daniel L.Comparing the Areas under Two or More Correlated Receiver Operating Characteristic Curves: A Nonparametric ApproachBiometricsBiometrics837-845443jackknifingmann-whitney testreceiver operating characteristic (ROC) curvestructural componentsu-statistics1988International Biometric Society0006341Xhttp://www.jstor.org/stable/2531595(1988)) to assess whether there are any significant differences between the AUC of different models. 5.4. Model results Applying the DeLong, DeLong and Clarke-Pearson test, we find that the AUC values for model R2 is significantly better than that for R0 (cf. Table 1), whereas R1 performs worse. Hence, model R2 is selected for further inclusion in our two-stage model. Table 2 gives the direction of parameter estimates used in the Probability of Repossession Model R2, together with a possible explanation. The parameter estimate values and p-values of all repossession model variations can be found in Appendix, Tables A.7, A.8 and A.9. Table 1: Repossession model performance statistics ModelAUCCut-offSpecificitySensitivityAccuracyR1, Test Set (LTV, time on books, Security, Previous default)0.7270.43557.44975.68869.186R2, Test Set (DLTV, Security, Previous def)0.7430.43259.39876.20370.213R0, Test Set (DLTV)0.7370.43658.62676.00869.812DeLong et al p-value, R1 vs. R0<0.001DeLong et al p-value, R2 vs. R0<0.001 Table 2: Parameter estimate signs for Probability of Repossession Model R2 VariableRelation to probability of repossession (given default)ExplanationDLTV (LTV at default)+If a large proportion of loan is tied up in security, likelihood of repossession increasesPrevious default+Probability of repossession increases if account has been in default beforeSecurity-Lower-range property types such as flats are more likely to be repossessed in the case of default 6. The Haircut Model The Haircut Model is only applicable to observations that have undergone the repossession and forced sale process, where haircut is defined to be the ratio of forced sale price to valuation of security at default. Therefore, securities that were not repossessed, or repossessed but not sold do not have a haircut value, and are thus excluded from the development of the Haircut Model. An OLS model is also developed to explicitly model haircut standard deviation, as a function of time on books, as suggested by Lucas  ADDIN EN.CITE Lucas200652525247Lucas, AlanBasel II Problem SolvingConference on Basel II & Credit Risk Modelling in Consumer Lending 2006Southampton, UKhttp://www3.imperial.ac.uk/portal/pls/portallive/docs/1/7287866.PDFAn OLS model to explicitly model haircut standard deviation, as a function of time on books (binned). Combine Probability of Repossession Model and Haircut Model by taking into account the probabilities of other values of haircut occurring, and the different levels of loss associated with these different levels(2006).  Figure 2: Distribution of haircut (solid curve references the normal distribution) The distribution of haircut is shown in Figure 2 with the solid curve referencing the normal distribution. Statistics from the Kolmogorov-Smirnov and Anderson-Darling Tests (Peng  ADDIN EN.CITE Peng200457575747Peng, GuangbinTesting Normality of Data Using SASPharmaSUG2004San Diego, Californiahttp://www.lexjansen.com/pharmasug/2004/posters/po04.pdf(2004)) suggest non-normality with p-values of <0.01 and <0.005 respectively, but for the purposes of the prediction of LGD, we approximate haircut by a normal distribution. 6.1. Modelling methodology The top and bottom 0.05 percent of observations (26 cases) for haircut are truncated before we establish the set of eligible variables to be considered in the development of an OLS linear regression model for the Haircut Model. We also check the relationship between variables and haircut. In particular, the valuation of security at default to average property valuation in the region ratio displays high non-linearity (cf. Figure 3) and is binned into 6 groups for model development. Backward stepwise regression is used to remove insignificant variables and individual parameter estimate signs are checked for intuitiveness. We also check for intuition within categorical variables, and examine the Variance Inflation Factors (VIF).  Figure 3: Relationship between haircut and (ranked) valuation of security at default to average property valuation in the region ratio 6.2. Model variations Using the methodology above, we obtain a Haircut Model H1, with seven significant variables, loan-to-value (LTV) ratio at time of loan application (start of loan), a binary indicator for whether this account has had a previous default, time on book in years, ratio of valuation of property to average in that region (binned), type of security, i.e. detached, semi-detached, terraced, flat or other, age group of property and region. In a second model, we replace LTV at loan application and time on book with LTV at default (DLTV), referred to as Haircut Model H2; note that, as previously, including all three variables (LTV, DLTV and time on books) in a single model would cause counter-intuitive parameter estimate signs. Comparative performance measures for the two models are reported in the following section. 6.3. Performance measures The performance measures considered here are the R-square value, Mean Squared Error (MSE) and Mean Absolute Error (MAE). To create a graphical representation of the results, we also present a binned scatterplot of predicted haircut value bands against actual haircut values, where predicted haircut values are put into ascending order and binned into equal-frequency value bands; the mean actual haircut value is then compared against the mean predicted haircut value in each haircut band. 6.4. Model results First, we note that all parameters for all models have low VIF values, the only ones above 2 belonging to geographical indicators. In the Haircut Model, the combination of LTV and time on books seems to be able to capture the information carried in DLTV because, as it is observed from Table 3, Model H1 gives the better performance. This could be because LTV gives an indication of the (initial) quality of the customer whereas values of DLTV could be due to changes in house prices since the purchase of the property. Based on this, Model H1 is selected as the Haircut Model to be used in the LGD estimation. Table 3: Haircut model performance statistics ModelMSEMAE EMBED Equation.3 H1, Test Set0.0390.1470.143H2, Test Set0.0390.1480.131 Table 4: Parameter estimate signs of Haircut Model H1 VariableRelation to haircutExplanationLTV +Refer to Figure 4 and explanation in Section 6.4Ratio of valuation of security at default to average property valuation in that region, binned+/-Medium-end properties (relative to the region the property is in) have higher haircut than lower-end properties, but higher-end properties tend to have lowest haircut (cf. Figure 3 in Section 6.1)Previous default+Haircut is higher for accounts that have previously defaultedTOB (Time on book in years)+Older loans imply greater uncertainty and error in estimation of value of security at default, so higher haircut is possibleSecurity+Haircut tends to be higher for higher-end property types (e.g. detached)Age group of property (oldest to newest)+Haircut tends to be higher for newer propertiesRegionN/AHaircut differs across regions Table 4 details parameter estimate signs. From it, we see that a greater LTV at start implies a higher haircut (i.e. a higher forced sale price). This would mean that the larger the loan a debtor took at time of application in relation to property value, the higher the forced sale price of the security would be in the event of a default and repossession. At first, it might seem as though this parameter estimate sign might be confused due to the number of variables in the Haircut Model, or due to some hidden correlation between variables. In order to rule out this possibility, we look at the relationship between LTV at start and haircut. From Figure 4, we observe that there indeed appears to be a positive relationship between haircut and LTV. Part of the explanation for this might be found in policy decisions taken by the bank. For loans with high loan-to-value ratios, due to the large amount (relative to the loan) the bank has committed towards the property, when the account does go into default and subsequent repossession, the bank may be reluctant to let the repossessed property go unless it is able to fetch a price close to the current property valuation. Another possible reason could be that borrowers with a low LTV are likely to sell early and only end up in repossession when they know the house is in a bad state and unlikely to make anything near its indexed valuation.  Figure 4: Relationship between haircut and (ranked) LTV at time of loan application To further validate the model, we also include in Figure 5 a scatter plot of mean (grouped) predicted and actual haircut. From it, we observe that our model produces unbiased estimates of haircut. Parameter estimates of all models can be found in Appendix, Tables A.10 and A.11.  Figure 5: Prediction performance of haircut test set 6.5. Haircut standard deviation modelling To be able to produce an expected value for LGD (see later, Section 7.1), we will not only require a point estimate for haircut but also a model component for haircut variability. Further inspection reveals that the standard deviation of haircut increases with longer time on books (cf. Figure 6), which can be expected because the valuation of a property is usually updated using publicly available house price indices (instead of commissioning a new valuation process), and the longer an account has been on the books, the greater the uncertainty and error in the estimation of current valuation of property, which will affect the error in the prediction of haircut as well. As suggested by Lucas (2006), to model this relationship, a simple OLS model was fitted that estimates the standard deviation for different time on books bins. Time on books is binned into equal-length intervals of 6 months, and standard deviation of haircut is calculated for each group based on the mean haircut in that group. This model will later on be used to calculate the expected values for LGD (cf. Section 7.1). Performance statistics for this auxiliary model are detailed in Table 5; parameter estimates can be found in Appendix, Table A.12.  Figure 6: Mean haircut standard deviation by time on book bins Table 5: Haircut Standard Deviation Model performance statistics ModelMSEMAE EMBED Equation.3 Training Set0.00010.00460.9315Test Set0.00020.01050.8304 7. Loss given default model After having estimated the Probability of Repossession Model, the Haircut Model and the Haircut Standard Deviation Model, we now combine these models to get an estimate for Loss Given Default. Here we illustrate two ways of combining the component models, report their respective LGD predictions, and advocate use of the more conservative approach producing an expected value for LGD that takes into account haircut variability. We also compare these results against the single-stage model predictions and performance statistics. 7.1. Modelling methodology A first approach referred to in our paper as the haircut point estimate" approach would be to keep the probabilities derived under the Probability of Repossession Model and apply the Haircut Model onto all observations. The latter would give all observations a predicted haircut value in the event of repossession. Using this predicted value of haircut, predicted sale price and predicted loss (outstanding balance at default less sale proceeds), if any, can be calculated. We then find predicted LGD by multiplying the probability of repossession with this predicted loss if repossession happens. Although this method does produce some estimate for LGD, regardless of whether the observation is predicted to enter repossession or not, it uses only a single value of haircut (although it is the most probable value). However, if the true haircut happens to be lower than predicted, sale proceeds would be overestimated, which would mean that a loss could still be incurred (provided that haircut falls below DLTV). This is an illustration of how misleading LGD predictions could be produced if the component models were not combined appropriately. Hence, to produce a true expected value for LGD, one should also take into account the distribution to the haircut estimate and the associated effect on loss in its left tail. Hence, the second and more conservative approach, suggested e.g. by Lucas (2006) and referred to here as the expected shortfall" approach, also takes into account the probabilities of other values of haircut occurring, and the different levels of loss associated with these different levels. To do so, we first apply the Probability of Repossession Model to get an estimate of probability of repossession given that an account goes into default. We then apply the Haircut Model onto the same dataset to get an estimate for haircut,  EMBED Equation.3  , for each observation j, regardless of whether the security is likely to be repossessed. A minimum value of zero is set for predicted haircut, as there is no meaning to a negative haircut. The Haircut Standard Deviation Model is then applied onto each observation j to get a predicted haircut standard deviation,  EMBED Equation.3 , depending on its value of time on books (see Section 6.5). From these predicted values, we approximate the distribution of each predicted haircut by a normal distribution,  EMBED Equation.3 . For simplicity, the subscript j, which represents individual observations, will be dropped from here on. As long as the haircut (sale amount as a ratio of valuation of property at default) is greater than DLTV (outstanding balance at default as a ratio of valuation of property at default), and ignoring any additional administrative and repossession-associated costs, the proceeds from the sale would be able to cover the outstanding balance on the loan, i.e. there would be no shortfall. Hence, the expected shortfall expressed as a proportion of the indexed valuation of property is:  EMBED Equation.3  (2) where p(.) denotes the probability density function of the distribution for h. To convert the latter into a standard normal distribution, we let:  EMBED Equation.3 ;  EMBED Equation.3  Hence, Expected Shortfall can easily be derived as follows:  EMBED Equation.3  (3) where  EMBED Equation.3  and  EMBED Equation.3  denote the cumulative distribution function and probability density function of the standard normal distribution, respectively. Expected loss given default is then obtained from the probability of non-repossession and the expected shortfall calculated for the repossession scenario (cf. Equation 4, below). The probability of an account undergoing repossession given that it has gone into default is multiplied against the expected LGD the account would incur in the event of repossession. We also multiply the probability of an account not going into repossession against the expected LGD for non-repossessions (denoted by c). We can use the average observed LGD for actual non-repossessions as the expected conditional LGD for non-repossessions.  EMBED Equation.3  (4) where c is the loss associated with non-repossessions (assumed to be 0 in the absence of additional information). Finally, we obtain predicted LGD by taking the ratio  EMBED Equation.3  to (estimated) outstanding balance at default. 7.2. Alternative single-stage model To be able to compare this two-stage model, we also developed a simple single-stage model using the same data. A backward stepwise selection on the same set of eligible variables used earlier in the two-stage model building was applied, and resulting model parameter estimates are added in Appendix, Table A.13. However, it is noted that whatever the results of the single-stage model, because it directly predicts LGD based on loan and collateral characteristics, it does not provide the same insight into the two different drivers (i.e. repossession risk and sale price haircut) of mortgage loss, and as such does not provide as rich a framework for stress testing. The performance measures of this single-stage model are then compared against those of the preferred two-stage model developed in the previous section (i.e. using the expected shortfall approach), as well as the two-stage model that would result from the so-called haircut point estimate" approach. 7.3. Model performance Using the same performance measures as those used for the Haircut Model, we compare the MSE, MAE and R-square values of our two-stage and the single-stage models (cf. Table 6). It is observed that both two-stage model variations achieve a substantially better R-square of just under 0.27 (compared to 0.233 for the single-stage model) on the LGD Test set, which is competitive to other LGD models currently used in the industry. Table 6: Performance measures of two-stage and single-stage LGD Models Method, DatasetMSEMAE EMBED Equation.3 Single stage, Test set0.0260.1210.233Two-stage (haircut point estimate), Test set0.0250.1080.268Two-stage (expected shortfall), Test set0.0250.1010.266 The distributions of predicted LGD and actual LGD for all LGD models are shown in Figure 7. In the original empirical distribution of LGD (see top section of Figure 7), there is a large peak near 0 (where losses were zero either because there was no repossession, or because the sale of the house was able to cover the remaining loan amount). Firstly, we observe that the single-stage model (shown in the bottom section of Figure 7) is unable to produce the peak near 0. Moreover, note that the two-stage model using the haircut point estimate seemingly is the model that most closely reproduces the empirical distribution of LGD, as it is able to bring out the peak near 0. Although the R-square values achieved by the two two-stage LGD model variations are very close (see Table 6), their LGD distributions are quite different. The haircut point estimate approach is shown to underestimate the average loss (cf. mean predicted LGD from haircut point estimate method being lower than mean observed LGD). Unlike the former approach, the expected shortfall method takes a more conservative approach in its estimation of LGD, which takes into account the haircut distribution and its effect on expected loss based on probabilities of different haircut values occurring. This will make a difference especially for observations that would be predicted to have low or zero LGD under the haircut point estimate method because these very accounts are now assigned at least some expected loss amount, hence moving observations out of the peak and into the low LGD bins.  Figure 7: Distribution of observed LGD (Empirical), predicted LGD from two-stage haircut point estimate model (HC pt. est), two-stage expected shortfall model (E.shortfall), single-stage model (single stage) (from top to bottom) To further verify to what extent these various models are able to produce unbiased estimates at an LGD loan pool level, we create a graphical representation of the results. We look at a binned scatterplot of predicted LGD value bands against actual LGD values, where predicted LGD values are put into ascending order and binned into equal-frequency value bands. For each method we used in the calculation of LGD, we plot the mean actual LGD value against the mean predicted LGD value (for that LGD band) onto a single graph, included in Figure 8. Observe that both of the two-stage models are able to consistently estimate LGD fairly closely, whereas the single-stage model either overestimates (in the lower-left hand region of the graph which represents observations that have low LGD) or underestimates LGD (in the upper-right hand region of the graph which represents observations with high LGD). Furthermore, the expected shortfall approach is shown to produce the more reliable estimates in the lower-LGD regions, outperforming the haircut point estimate approach in the lower-left part of the graph, where the haircut point estimate approach indeed underestimates risk (i.e. the estimates fall below the diagonal).  Figure 8: Scatterplot of predicted and actual LGD in LGD bands Finally, in order to check robustness of our two-stage LGD model, we have also experimented with re-estimating the two component models this time including only the first instance of default for customers with multiple defaults (i.e. not all instances of default included for observations with multiple defaults). Detailed results are not reported here, but for both component models and the LGD model itself, we obtained the same parameter estimate signs, and parameter estimates were similar in size. 8. Conclusions and further research In this paper, we developed and validated a number of models to estimate the LGD of mortgage loans using a large set of recovery data of residential mortgage defaults from a major UK bank. The objectives of this paper were two-fold. Firstly, we aimed to evaluate the added value of a Probability of Repossession Model with more than just LTV at default as its explanatory variable. We have developed a Probability of Repossession Model with three variables, and showed that it is significantly better than a model with only the commonly used DLTV. Secondly, we wanted to validate the approach of using two component models, a Probability of Repossession Model and a Haircut Model, which consists of the Haircut Model itself and the Haircut Standard Deviation Model, to create a model that produces estimates for LGD. Here, two methods are explained, both of which will produce a value of predicted LGD for every default observation because the Haircut Model, which gives a predicted sale amount and predicted shortfall, shall be applied to all observations regardless of its probability of repossession. However, we then show how the first method, which uses only the haircut point estimate, would end up underestimating LGD predictions. The second and preferred method (expected shortfall) derives expected loss from an estimated normal haircut distribution having the predicted haircut from the Haircut Model as the mean, and with the standard deviation obtained from the Haircut Standard Deviation Model. For comparison purposes, we also developed a single-stage model. This model produced a lower R-square value, and was also unable to fully emulate the actual distribution of LGD. Having shown that the proposed two-stage modelling approach works well on real-life data, in our further research, we intend to explore the inclusion of macroeconomic variables in either or both the Probability of Repossession Model and the Haircut Model. These macroeconomic variables might include the unemployment rate, the inflation rate, the interest rate or some indication of the amount of borrowing in each economic year. Finally, we also consider the use of alternative methods, for example, survival analysis to better predict and estimate the time periods between each milestone (repossession and sale) of a defaulted loan account. 9. Acknowledgements We thank the bank who has kindly provided the dataset that enabled this work to be carried out and Professor Lyn Thomas for his guidance throughout this work. We also thank the editor and reviewers who have contributed invaluably with their insightful feedback and recommendations. Any mistakes are solely ours. 10. References  ADDIN EN.REFLIST Altman, EdwardI., Brady, B., Resti, A., Sironi, A., 2005. The Link between Default and Recovery Rates: Theory, Empirical Evidence, and Implications. The Journal of Business 78, 2203-2228. Calem, P.S., LaCour-Little, M., 2004. Risk-based capital requirements for mortgage loans. Journal of Banking & Finance 28, 647-672. Campbell, T.S., Dietrich, J.K., 1983. The Determinants of Default on Insured Conventional Residential Mortgage Loans. The Journal of Finance 38, 1569-1581. DeLong, E.R., DeLong, D.M., Clarke-Pearson, D.L., 1988. Comparing the Areas under Two or More Correlated Receiver Operating Characteristic Curves: A Nonparametric Approach. Biometrics 44, 837-845. Federal Register, 2007. Risk-Based Capital Standards: Advanced Capital Adequacy Framwork - Basel II; Final Rule. Fernandez, G.C.J., 2007. Effects of Multicollinearity in All Possible Mixed Model Selection, PharamaSUG Conference (Statistics & Pharmacokinetics), Denver, Colorado. Financial Services Authority, 2009. Prudential Sourcebook for Banks, Building Societies and Investment Firms. Gupton, G.M., Stein, R.M., 2002. LOSSCALC: Model for Predicting Loss Given Default (LGD). Jarrow, R., 2001. Default Parameter Estimation Using Market Prices. Financial Analysts Journal 57, 75-92. Jokivuolle, E., Peura, S., 2003. Incorporating Collateral Value Uncertainty in Loss Given Default Estimates and Loan-to-value Ratios. European Financial Management 9, 299-314. Lucas, A., 2006. Basel II Problem Solving, Conference on Basel II & Credit Risk Modelling in Consumer Lending Southampton, UK. Peng, G., 2004. Testing Normality of Data Using SAS, PharmaSUG, San Diego, California. Qi, M., Yang, X., 2009. Loss given default of high loan-to-value residential mortgages. Journal of Banking & Finance 33, 788-799. Quercia, R.G., Stegman, M.A., 1992. Residential Mortgage Default: A Review of the Literature. Journal of Housing Research 3, 341-379. Schuermann, T., 2004. What Do We Know About Loss Given Default, in: Shimko, D. (Ed.), Credit Risk: Models and Management, 2nd ed. Risk Books. Somers, M., Whittaker, J., 2007. Quantile regression for modelling distributions of profit and loss. European Journal of Operational Research 183, 1477-1487. Truck, S., Harpaintner, S., Rachev, S.T., 2005. A Note on Forecasting Aggregate Recovery Rates with Macroeconomic Variables. von Furstenberg, G.M., 1969. Default Risk on FHA-Insured Home Mortgages as a Function of the Terms of Financing: A Quantitative Analysis. The Journal of Finance 24, 459-477.  Appendix Table A7: Parameter estimates for Probability of Repossession Model R0 VariableVariable explanationEstimateStdErrWaldChiSqProbChiSqIntercept--3.0690.02812235.289<0.01DLTVLoan to value at default2.8210.0299449.349<0.01 Table A8: Parameter estimates for Probability of Repossession Model R1 VariableVariable ExplanationEstimateStdErrWaldChiSqProbChiSqIntercept--1.1380.040795.605<0.01LTVLoan to value at loan application2.1010.0402809.703<0.01TOBTime on books (in years)-0.1880.0032899.616<0.01Previous defaultIndicator for previous default0.1020.0348.869<0.01security0 (base)Flat or other----security1Detached-0.6250.031413.989<0.01security2Semi-detached-0.6700.024787.436<0.01security3Terraced-0.4210.021395.497<0.01 Table A9: Parameter estimates for Probability of Repossession Model R2 VariableVariable ExplanationEstimateStdErrWaldChiSqProbChiSqIntercept--2.5700.0345769.803<0.01DLTVLoan to value at default2.6790.0298295.648<0.01Previous defaultIndicator for previous default-0.4710.032211.064<0.01security0 (base)Flat or other----security1Detached-0.4610.031219.425<0.01security2Semi-detached-0.5460.024503.458<0.01security3Terraced-0.3430.022253.470<0.01 Table A10: Parameter estimates for Haircut Model H1 VariableVariable ExplanationEstimateStdErrProbTVIFIntercept-0.5080.009<0.010.000LTVLoan to value at loan application0.2430.007<0.011.136TOBTime on book (in years)0.0050.001<0.011.251VVAratio1 (base)Value of property / region average <= 0.9----VVAratio20.9 < Value of property / region average <= 1.2-0.0050.0040.2481.134VVAratio3 1.2 < Value of property / region average <= 1.5-0.0590.006<0.011.149VVAratio4 1.5 < Value of property / region average <= 1.8-0.0920.008<0.011.127VVAratio5 1.8 < Value of property / region average <= 2.4-0.0900.009<0.011.161VVAratio6 Value of property / region average > 2.4-0.1380.009<0.011.226Previous defaultIndicator for previous default0.0420.006<0.011.168Propage1Very old property (before 1919)-0.0850.003<0.011.273Propage2Old property (1919-1945)-0.0320.004<0.011.194Propage3 (base)Built after 1945----security0 (base)Flat or other----security1Detached0.1650.006<0.011.875security2Semi-detached0.1290.004<0.011.764security3Terraced0.0940.003<0.011.739region1North-0.1120.010<0.011.753region2Yorkshire & Humberside-0.0950.008<0.012.898region3North West -0.0990.008<0.013.163region4East Midlands -0.1000.008<0.012.489region5West Midlands -0.0650.008<0.012.449region6East Anglia -0.0670.009<0.011.968region7Wales -0.1150.009<0.012.140region8South West-0.0470.008<0.013.272region9South East-0.0620.007<0.016.348region10Greater London-0.0100.0070.1665.214region11Northern Ireland -0.0340.0140.0171.256region12 (base)Scotland or others / missing ---- Table A11: Parameter estimates for Haircut Model H2 VariableVariable ExplanationEstimateStdErrProbTVIFIntercept-0.5910.008<0.010.000DLTVLoan to value at default0.1620.005<0.011.175VVAratio1 (base)Value of property / region average <= 0.9----VVAratio20.9 < Value of property / region average <= 1.2-0.0110.004<0.011.126VVAratio3 1.2 < Value of property / region average <= 1.5-0.0690.006<0.011.141VVAratio4 1.5 < Value of property / region average <= 1.8-0.1080.008<0.011.116VVAratio5 1.8 < Value of property / region average <= 2.4-0.1080.009<0.011.149VVAratio6 Value of property / region average > 2.4-0.1580.009<0.011.209Previous defaultIndicator for previous default0.0640.005<0.011.010Propage1Very old property (before 1919)-0.0790.003<0.011.261Propage2Old property (1919-1945)-0.0300.004<0.011.193Propage3 (base)Built after 1945----security0 (base)Flat or other----security1Detached0.1620.006<0.011.874security2Semi-detached0.1260.004<0.011.761security3Terraced0.0920.003<0.011.736region1North-0.1090.010<0.011.752region2Yorkshire & Humberside-0.0940.008<0.012.897region3North West -0.0980.008<0.013.159region4East Midlands -0.1120.008<0.012.497region5West Midlands -0.0760.008<0.012.454region6East Anglia -0.1020.009<0.012.007region7Wales -0.1250.009<0.012.141region8South West-0.0800.008<0.013.325region9South East-0.0950.007<0.016.489region10Greater London-0.0420.007<0.015.323region11Northern Ireland -0.0300.0140.0401.256region12 (base)Scotland or Others / Missing ---- Table A12: Parameter estimates for Haircut Standard Deviation Model VariableVariable ExplanationEstimateStdErrProbTIntercept-0.181<0.001<0.01TOB binsTime on book (in years)0.010<0.001<0.01 Table A13: Parameter estimates for single-stage LGD model VariableVariable ExplanationEstimateStdErrProbTVIFIntercept--0.0930.005<0.010.000DLTVLoan to value at default0.2300.002<0.011.263secondappSecond applicant present-0.0030.0010.0121.105VVAratio1 Value of property / region average <= 0.9-0.0490.004<0.018.976VVAratio20.9 < Value of property / region average <= 1.2-0.0500.004<0.015.416VVAratio3 1.2 < Value of property / region average <= 1.5-0.0350.004<0.013.093VVAratio4 1.5 < Value of property / region average <= 1.8-0.0180.005<0.012.148VVAratio5 1.8 < Value of property / region average <= 2.4-0.0180.005<0.012.037VVAratio6 (base)Value of property / region average > 2.4----Previous defaultIndicator for previous default-0.0320.002<0.011.018Propage1Built before 19190.0230.002<0.011.653Propage2 (base)Built between 1919 and 1945----Propage3Built after 1945-0.0100.001<0.011.536Propage4Age unknown-0.1330.014<0.011.017security0 Flat or other0.0650.002<0.011.370security1Detached-0.0200.002<0.011.628security2Semi-detached-0.0130.002<0.011.370security3 (base)Terraced----region0Others or Missing0.0540.013<0.011.054region1North0.0410.004<0.011.758region2Yorkshire & Humberside0.0410.003<0.013.011region3North West 0.0470.003<0.012.940region4East Midlands 0.0520.004<0.012.233region5West Midlands 0.0370.004<0.012.364region6East Anglia 0.0470.004<0.011.782region7Wales 0.0470.004<0.012.047region8South West0.0380.003<0.012.936region9South East0.0500.003<0.015.244region10Greater London0.0300.003<0.014.265region11Northern Ireland 0.0280.006<0.011.333region12 (base)Scotland----  Date of default was estimated by the bank using the arrears status and amount of cumulated arrears at the end of each year for each account because we are not explicitly given default date in the original dataset.  Due to a data confidentiality agreement with the data provider, the scale for the y-axis has been omitted in some of the reported figures.  Available at: http://www.lloydsbankinggroup.com/media1/research/halifax_hpi.asp  A more common definition of haircut is the complement,  EMBED Equation.3 . However, in this paper, we will use the term haircut to refer to the ratio, rather than its complement, to facilitate the interpretation of parameter signs and further notation. We note that which of these two definitions is used does not affect the actual modelling but will make a difference for the interpretation of coefficient signs.  If variables within the model are highly correlated with each other, it would be reflected in a high value of VIF. Any value above 10 would imply severe collinearity amongst variables while values less than 2 would mean that variables are almost independent ( ADDIN EN.CITE Fernandez200733333347Fernandez, George C J.Effects of Multicollinearity in All Possible Mixed Model SelectionPharamaSUG Conference (Statistics & Pharmacokinetics)2007Denver, Coloradohttp://www.crda.unr.edu/crda/publications/pharmasuggf.pdfFernandez, G.C.J., 2007. Effects of Multicollinearity in All Possible Mixed Model Selection, PharamaSUG Conference (Statistics & Pharmacokinetics), Denver, Colorado.).  Since regional differences may not persist over time, alternatively, one can choose to omit the geographic dummy variables from the model. Our robustness tests indicate that the model fit is slightly lower without these; parameter estimate signs and estimates of the other variables remained stable.  Alternatively, because standard deviation of haircut is different for different groups of observations, the weighted least squares regression method was considered to adjust for heteroscedascity in the OLS model developed in Section 6.4. Two different weights were experimented with - the error term variance of each observation (from running an OLS model for haircut) and time on books. Both models produced similar parameter estimates to the selected haircut model, which suggests that the OLS model was able to produce robust parameter estimates even though the homoscedasticity assumption was violated. Also, because both models did not explicitly model and produce standard deviation of haircut, which is required in the calculation of expected LGD, a separate OLS model for standard deviation is necessary.     PAGE  PAGE 33 CD}~ S T ^ ɸraOa>- h{:hfCJOJQJ^JaJ h{:hsCJOJQJ^JaJ#h{:h'6CJOJQJ^JaJ h{:h'CJOJQJ^JaJ h{:hHCJOJQJ^JaJ h{:hyZCJOJQJ^JaJ#h{:h5CJOJQJ^JaJ#h{:hf5CJOJQJ^JaJ h{:hC?CJOJQJ^JaJ#h{:h/5CJOJQJ^JaJ#h{:h5CJOJQJ^JaJ#h{:hY(5CJOJQJ^JaJ~S T ^   77/808QQ````dhgd{:H5AA ( RSYZ*+12[\m/1ehi!"IJɻɥɻɥޔޔޔޔޔރޔޔrrޔޔޔޔ h{:h<CJOJQJ^JaJ h{:hHCJOJQJ^JaJ h{:h-CJOJQJ^JaJ+h{:hHCJOJQJ^JaJmHnHuh~5CJOJQJ^JaJ)jh{:hHCJOJQJU^JaJ h{:hyZCJOJQJ^JaJ h{:hf[CJOJQJ^JaJ,LM  kl56JK   "-%ͻ͘rdUh~5CJOJQJ^JaJ*h~5CJOJQJ^JaJ)jh{:hQ\CJOJQJU^JaJ h{:hQ\CJOJQJ^JaJ h{:hteCJOJQJ^JaJ#h{:hd#5CJOJQJ^JaJ#h{:hf5CJOJQJ^JaJ h{:hf[CJOJQJ^JaJ h{:hyZCJOJQJ^JaJ h{:h-CJOJQJ^JaJ!-%.%4%5%E%F%G%()))))),~------0b1c1i1j1z1{13~66667788òꤕꤕꤕꤕÄr`#h{:ho6CJOJQJ^JaJ#h{:hW6CJOJQJ^JaJ h{:h CJOJQJ^JaJh~5CJOJQJ^JaJ*h~5CJOJQJ^JaJ h{:hQ\CJOJQJ^JaJ h{:hteCJOJQJ^JaJ+h{:hQ\CJOJQJ^JaJmHnHu)jh{:hQ\CJOJQJU^JaJ$888 888&8.8/808999=======ABBBɷ۷۷oaRo h{:h CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:h<CJOJQJ^JaJ+h{:h CJOJQJ^JaJmHnHuh~5CJOJQJ^JaJ*h~5CJOJQJ^JaJ)jh{:h CJOJQJU^JaJ h{:hteCJOJQJ^JaJ h{:hf[CJOJQJ^JaJ#h{:hD6CJOJQJ^JaJ#h{:hV*6CJOJQJ^JaJ6t7t=t>ttttuuvvxxzz{{{m|o|~~67õÓÓp^MÓ< h{:h]p/CJOJQJ^JaJ h{:h2CJOJQJ^JaJ#h{:hd#5CJOJQJ^JaJ h{:hvCJOJQJ^JaJ#h{:hf5CJOJQJ^JaJ h{:hf[CJOJQJ^JaJ h{:h{:CJOJQJ^JaJh{:CJOJQJ^JaJ h{:hteCJOJQJ^JaJ+h{:h CJOJQJ^JaJmHnHu)jh{:h CJOJQJU^JaJ`ttuuvvxxzz{{~~67NOpqrsJdhgd{:7:;<EFNOsЃу12pq &'BCɷɦoaRo#h{:hW6CJOJQJ^JaJ#h{:h*5CJOJQJ^JaJ#h{:h.`5CJOJQJ^JaJ h{:h*CJOJQJ^JaJ h{:hr /CJOJQJ^JaJ h{:h@CJOJQJ^JaJ h{:hUtCJOJQJ^JaJ h{:hf[CJOJQJ^JaJ#h{:h56CJOJQJ^JaJ#h{:h&9'6CJOJQJ^JaJ#h{:h:\*6CJOJQJ^JaJ`bcdeoz{ɷɦweSA#h{:hy6CJOJQJ^JaJ#h{:h'6CJOJQJ^JaJ#h{:hW6CJOJQJ^JaJ h{:hvCJOJQJ^JaJh{:hUtCJ^JaJ h{:hUtCJOJQJ^JaJ h{:hf[CJOJQJ^JaJ#h{:h:\*6CJOJQJ^JaJ#h{:h6CJOJQJ^JaJ#h{:ho6CJOJQJ^JaJ#h{:h.`6CJOJQJ^JaJ&'()*ʹtbP>#h{:h6CJOJQJ^JaJ#h{:h'6CJOJQJ^JaJ#h{:hW6CJOJQJ^JaJ h{:h8CJOJQJ^JaJ h{:hk[CJOJQJ^JaJ h{:h5g"CJOJQJ^JaJ#h:hUt6CJOJQJ^JaJ h{:hUtCJOJQJ^JaJ h{:hf[CJOJQJ^JaJ#h{:hP 6CJOJQJ^JaJ#h{:ha?6CJOJQJ^JaJ&'AB !޲ dh$Ifgd{:dhgd{:*+,9@AB !WX"#ɷraL>Lh~5CJOJQJ^JaJ)jh{:hk[CJOJQJU^JaJ h{:h.`CJOJQJ^JaJ h{:h3CJOJQJ^JaJ h{:h CJOJQJ^JaJ h{:hf[CJOJQJ^JaJ#h{:h*6CJOJQJ^JaJ#h{:h76CJOJQJ^JaJ#h{:h 6CJOJQJ^JaJ#h{:h.`6CJOJQJ^JaJ#h{:ho6CJOJQJ^JaJ#)*ò|jXF4"#h{:h|6CJOJQJ^JaJ#h{:hA6CJOJQJ^JaJ#h{:hz%6CJOJQJ^JaJ#h{:h.`6CJOJQJ^JaJ#h{:ho6CJOJQJ^JaJ#h{:hL6CJOJQJ^JaJ#h{:h'6CJOJQJ^JaJ#h{:hW6CJOJQJ^JaJ h{:hvCJOJQJ^JaJ h{:h CJOJQJ^JaJ)jh{:hk[CJOJQJU^JaJ+h{:hk[CJOJQJ^JaJmHnHu "$<>]_%2GJNrtu³ó̳˹˹˹˹˹˨tcRc h{:hmCJOJQJ^JaJ h{:h8&CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:h CJOJQJ^JaJ#h:hL:.6CJOJQJ^JaJ h{:hL:.CJOJQJ^JaJ#h:hf6CJOJQJ^JaJ h{:hfCJOJQJ^JaJ h{:hf[CJOJQJ^JaJ#h{:h76CJOJQJ^JaJPV\cjA55555 dh$Ifgd{:kd$$Iflֈ >1!' &&&N&`&!644 la[p<jqr5kd$$Ifl,ֈ >1!' &&&N&`&!644 la[p< dh$Ifgd{:Գڳ5kd$$Ifl,ֈ >1!' &&&N&`&!644 la[p< dh$Ifgd{:̳ӳ 68:<>?ANOUVXbckluvͻͩͻޘ޻އvevTvTvC h{:hDFCJOJQJ^JaJ h{:h+ CJOJQJ^JaJ h{:h@CJOJQJ^JaJ h{:h(CJOJQJ^JaJ h{:hr /CJOJQJ^JaJ h{:h-x"CJOJQJ^JaJ#h:h6CJOJQJ^JaJ#h:hL:.6CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:hL:.CJOJQJ^JaJ h{:h&VCJOJQJ^JaJ5kd$$Ifl,ֈ >1!' &&&N&`&!644 la[p< dh$Ifgd{: !"B5kd$$Ifl,ֈ >1!' &&&N&`&!644 la[p< dh$Ifgd{:BIJKLMN5kd$$Ifl,ֈ >1!' &&&N&`&!644 la[p< dh$Ifgd{:NO۴[jkd!$$IfTlF J(  t6    44 laT dh$Ifgd{:dhgd{: vehl̵͵ӵ*+ＪޙޙwffUD h{:hd#CJOJQJ^JaJ h{:h`CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:h8&CJOJQJ^JaJ h{:hRLCJOJQJ^JaJ h{:h}CJOJQJ^JaJ#h:hRL6CJOJQJ^JaJ h{:hI#CJOJQJ^JaJ h{:h+ CJOJQJ^JaJ h{:hDFCJOJQJ^JaJ h{:h(CJOJQJ^JaJ[\mo dh$Ifgd{:jkd!$$IfTlF J(  t6    44 laTŵǵ) dh$Ifgd{:jkd"$$IfTlF J( t6    44 laT)*+@A÷ķ 9:;<Xdhgd{:jkd"$$IfTlF J(  t6    44 laT+?@A߶÷ķIJ ʹʹo^I8 h{:hvCJOJQJ^JaJ)j#h{:hvCJOJQJU^JaJ h{:h`CJOJQJ^JaJ+h{:hk[CJOJQJ^JaJmHnHuh~5CJOJQJ^JaJ)jh{:hk[CJOJQJU^JaJ h{:hoCJOJQJ^JaJ h{:h=aCJOJQJ^JaJ h{:hf[CJOJQJ^JaJ#h{:hd#5CJOJQJ^JaJ#h{:h.`5CJOJQJ^JaJ9:<=?@ABLX;<>ò}k}YGY0,jh}hf[CJH*OJQJU^JaJ#h{:hb6CJOJQJ^JaJ#h{:h6CJOJQJ^JaJ#h{:ho6CJOJQJ^JaJ#h{:hf[6CJOJQJ^JaJ#h{:hW6CJOJQJ^JaJ h{:hf[CJOJQJ^JaJ h{:hyCJOJQJ^JaJ h{:h=aCJOJQJ^JaJ)jh{:hk[CJOJQJU^JaJ+h{:hk[CJOJQJ^JaJmHnHuXY>A/0/0 dh$Ifgd{:dhgd{:>?@_aɷq`qO>- h{:hezCJOJQJ^JaJ h{:hfCJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:h`CJOJQJ^JaJ#h:h=a6CJOJQJ^JaJ h{:h=aCJOJQJ^JaJ h{:hoCJOJQJ^JaJ#h{:hez6CJOJQJ^JaJ#h{:ho6CJOJQJ^JaJ h{:hvCJOJQJ^JaJ)j5h{:hoCJOJQJU^JaJ h{:hOCJOJQJ^JaJ/0!'(./0r`N<#h{:h46CJOJQJ^JaJ#h{:h6CJOJQJ^JaJ#h{:h8&6CJOJQJ^JaJ#h{:hz%6CJOJQJ^JaJ#h{:hW6CJOJQJ^JaJ h{:hu~CJOJQJ^JaJ h{:h=aCJOJQJ^JaJ h{:hoCJOJQJ^JaJ#h{:ho6CJOJQJ^JaJ#h{:hez6CJOJQJ^JaJ#h{:h%r6CJOJQJ^JaJ^`PR ,-34޻~iWWF5 h{:hsCJOJQJ^JaJ h{:h%rCJOJQJ^JaJ#h:hr /6CJOJQJ^JaJ)j|[h{:h`CJOJQJU^JaJ,jP h{:h`CJOJQJU^JaJ h{:h`CJOJQJ^JaJ)jh{:h`CJOJQJU^JaJ h{:hr /CJOJQJ^JaJ#h:h=a6CJOJQJ^JaJ h{:h=aCJOJQJ^JaJ h{:hk[CJOJQJ^JaJ maaaa dh$Ifgd{:kd]$$IfTlU\i  %%%%M644 lap(T %+maaaa dh$Ifgd{:kdQ^$$IfTl,\i  %%%%M644 lap(T+,-clmeeYYY dh$Ifgd{:dhgd{:kd_$$IfTl,\i  %%%%M644 lap(T46@AIJRSZ[_`bcx%')޼ޙ͙wwwfwwwwU h{:h&CJOJQJ^JaJ h{:h5CJOJQJ^JaJ h{:hmCJOJQJ^JaJ h{:hRCJOJQJ^JaJ h{:hhCJOJQJ^JaJ#h:h8&6CJOJQJ^JaJ h{:h8&CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:hB@CCJOJQJ^JaJ h{:h%rCJOJQJ^JaJ! dh$Ifgd{:jkd_$$IfTlF;    t6    44 laT%) dh$Ifgd{:jkd``$$IfTlF;    t6    44 laT@ dh$Ifgd{:jkd`$$IfTlF;   t6    44 laT@A]_ dh$Ifgd{:jkd`a$$IfTlF;   t6    44 laTAF0GZ9:;n]H])j h{:hvCJOJQJU^JaJ h{:hvCJOJQJ^JaJ)jch{:hvCJOJQJU^JaJ h{:h -CJOJQJ^JaJ h{:hfCJOJQJ^JaJ,jh}hENCJH*OJQJU^JaJ h{:hsCJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:hmCJOJQJ^JaJ h{:hhCJOJQJ^JaJ1 dh$Ifgd{:jkda$$IfTlF;   t6    44 laT12[] dh$Ifgd{:jkdRb$$IfTlF;   t6    44 laT dh$Ifgd{:jkdb$$IfTlF;   t6    44 laT9:< dhgd{:jkdDc$$IfTlF;    t6    44 laT RS#$%tnYH7 h{:h`CJOJQJ^JaJ h{:hifCJOJQJ^JaJ)jh{:h>rCJOJQJU^JaJ,jh}hfCJH*OJQJU^JaJ h{:hfCJOJQJ^JaJ h{:hvCJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:hoCJOJQJ^JaJ#h{:ho6CJOJQJ^JaJ#h{:hez6CJOJQJ^JaJ#h{:h36CJOJQJ^JaJ#$%flpt dh$Ifgd{:dhgd{: tu­꜋ygVE3#h{:hW6CJOJQJ^JaJ h{:hQOCJOJQJ^JaJ h{:h{K1CJOJQJ^JaJ#h{:h95CJOJQJ^JaJ#h{:ho5CJOJQJ^JaJ h{:hoCJOJQJ^JaJ h{:hCJOJQJ^JaJ)jh{:hLCJOJQJU^JaJ,jwP h{:hLCJOJQJU^JaJ h{:h`CJOJQJ^JaJ)jh{:h`CJOJQJU^JaJ maaaa dh$Ifgd{:kd$$IfTl\ ^ %%]%]%%644 lap(Tmaaaa dh$Ifgd{:kdc$$IfTl,\ ^ %%]%]%%644 lap(TRmeeeeeeeeeedhgd{:kd-$$IfTl,\ ^ %%]%]%%644 lap(T ij}~۸p[D22#h}h{K16CJOJQJ^JaJ-jhh}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ)jhh}CJOJQJU^JaJ h{:hCJOJQJ^JaJ h{:h{K1CJOJQJ^JaJ h{:hoCJOJQJ^JaJ#h{:h/6CJOJQJ^JaJ#h{:h6CJOJQJ^JaJ#h{:ho6CJOJQJ^JaJîq`O>,#h}h{K16CJOJQJ^JaJ h{:h2CJOJQJ^JaJ h{:hjCJOJQJ^JaJ h{:hbCJOJQJ^JaJ-j/h -h}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH  h{:h{K1CJOJQJ^JaJ)jhh}CJOJQJU^JaJ-jhh}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ [\,VWvfg,-Ydhgd}dhgd{: XYZ[ɴڌ{jYHYH6YH#hn2h6CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:h&CJOJQJ^JaJ h{:hutCJOJQJ^JaJ h{:h{:CJOJQJ^JaJ h{:hl\CJOJQJ^JaJ-jʵhENh}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ)jhh}CJOJQJU^JaJ h{:h2CJOJQJ^JaJ[\`fw͸{fO>> h{:hr3CJOJQJ^JaJ-jh}h}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH -jh}h}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ)jhh}CJOJQJU^JaJ h{:h2CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:h CJOJQJ^JaJ#$%&'(+,-23F޸{jYH7︧ h{:h.mBCJOJQJ^JaJ h{:hl\CJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:h&CJOJQJ^JaJ-jhENh}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ)jhh}CJOJQJU^JaJ h{:hr3CJOJQJ^JaJ h{:h2CJOJQJ^JaJ h{:h CJOJQJ^JaJFGHINObcde|Ӿp_p_pN_pN= h{:h`CJOJQJ^JaJ h{:h.mBCJOJQJ^JaJ h{:hr3CJOJQJ^JaJ h{:h CJOJQJ^JaJ-jhh}CJEHOJQJU^JaJ)jDP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ)jhh}CJOJQJU^JaJ-jhh}CJEHOJQJU^JaJ)jCP h}CJPJUVaJnH tH VWXklmnoqstuv޻iXG6G h{:hl\CJOJQJ^JaJ h{:hzlCJOJQJ^JaJ h{:h{:CJOJQJ^JaJ-jmh}h}CJEHOJQJU^JaJ)j DP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ)jhh}CJOJQJU^JaJ h{:hc+uCJOJQJ^JaJ#h}h`6CJOJQJ^JaJ h{:h`CJOJQJ^JaJ h{:hENCJOJQJ^JaJvw{|}~ 23̻ﻙވsbM)jDP h}CJPJUVaJnH tH  hh}CJOJQJ^JaJ)jhh}CJOJQJU^JaJ h{:hl\CJOJQJ^JaJ h{:hKCJOJQJ^JaJ h{:hiCJOJQJ^JaJ h{:h:CJOJQJ^JaJ#h}h:6CJOJQJ^JaJ h{:hc+uCJOJQJ^JaJ h{:h&CJOJQJ^JaJ345efghlw±|jXjFj4#h~5hz%6CJOJQJ^JaJ#h~5h:6CJOJQJ^JaJ#h~5hrq6CJOJQJ^JaJ#h~5h:a>6CJOJQJ^JaJ#h~5h{:6CJOJQJ^JaJ#h~5hW6CJOJQJ^JaJ h{:hOICJOJQJ^JaJ h{:h CJOJQJ^JaJ h{:hKCJOJQJ^JaJ)jhh}CJOJQJU^JaJ-jhhh}CJEHOJQJU^JaJ YZ[_fqr!ͻt_N7,jP h{:hCJOJQJU^JaJ h{:hCJOJQJ^JaJ)jh{:hCJOJQJU^JaJ h{:h -CJOJQJ^JaJ#h~5h86CJOJQJ^JaJ#h~5hrq6CJOJQJ^JaJ#h~5h{:6CJOJQJ^JaJ#h~5hW6CJOJQJ^JaJ h{:h.jCJOJQJ^JaJ h{:h`CJOJQJ^JaJ h{:h{:CJOJQJ^JaJ YZqr !hx| dh$Ifgd{:dhgd{: @bcdIOPsij|kVkH6#h~5h~55CJOJQJ^JaJh~5CJOJQJ^JaJ)jh{:hvCJOJQJU^JaJ h{:hvCJOJQJ^JaJ h{:hLCJOJQJ^JaJ)jlh{:h -CJOJQJU^JaJ h{:h:CJOJQJ^JaJ h{:hvf CJOJQJ^JaJ h{:h -CJOJQJ^JaJ)jh{:hCJOJQJU^JaJ)jh{:hCJOJQJU^JaJrffff dh$Ifgd{:kd$$IfTl\0j % %{%{{644 lapTrffff dh$Ifgd{:kdv$$IfTl,\0j % %{%{{644 lapT,28>rffff dh$Ifgd{:kd"$$IfTl,\0j % %{%{{644 lapT>?@abdIJVWrjjjjjjjjjjjdhgd{:kd$$IfTl,\0j % %{%{{644 lapT WOPtu` a   R$8hgd~5dhgd{:stu " ` a _˺˩˺˘ucRA0 h{:h@RiCJOJQJ^JaJ h{:hKCJOJQJ^JaJ h{:hVCJOJQJ^JaJ#h~5hre5CJOJQJ^JaJ#h~5h~55CJOJQJ^JaJ h{:hreCJOJQJ^JaJ h{:hM iCJOJQJ^JaJ h{:hvCJOJQJ^JaJ h{:hBNCJOJQJ^JaJ h{:h`CJOJQJ^JaJ h{:h~5CJOJQJ^JaJ#h~5hV5CJOJQJ^JaJͼu`O`A.`%h~5CJOJQJ^JaJmHnHuh~5CJOJQJ^JaJ h{:hCJOJQJ^JaJ)jh{:hCJOJQJU^JaJ h{:hM iCJOJQJ^JaJ#h~5h[Q5CJOJQJ^JaJ#h~5h~55CJOJQJ^JaJ#h~5hW5CJOJQJ^JaJ h{:hreCJOJQJ^JaJ h{:h@RiCJOJQJ^JaJ h{:hKCJOJQJ^JaJ h{:h@LCJOJQJ^JaJ+37@U^eoy dh$Ifgd{:dhgd{: 0^`0gd~5gd~5hgd~5"./46IJμxxfUD3 h{:hKCJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:hYCJOJQJ^JaJ#h:h 6CJOJQJ^JaJ h{:hDFCJOJQJ^JaJ h{:hLCJOJQJ^JaJ h{:h CJOJQJ^JaJ h{:h?44CJOJQJ^JaJ#h~5h-5CJOJQJ^JaJ#h~5h?445CJOJQJ^JaJh~55CJOJQJ^JaJ h{:h]CJOJQJ^JaJyz=1111 dh$Ifgd{:kdu$$IfTlֈ v %%{%%#%N%1644 lap<T1kd$$IfTlֈ v %%{%%#%N%1644 lap<T dh$Ifgd{:1kd$$IfTlֈ v %%{%%#%N%1644 lap<T dh$Ifgd{:&/DMT^h dh$Ifgd{:dhgd{:#% ot4BCOPUWch    ( ޻ުުވ޻wުވf h{:h-CJOJQJ^JaJ h{:h)7CJOJQJ^JaJ h{:h3CJOJQJ^JaJ h{:htCJOJQJ^JaJ h{:hKCJOJQJ^JaJ#h:h 6CJOJQJ^JaJ h{:hDFCJOJQJ^JaJ h{:h CJOJQJ^JaJ h{:hLCJOJQJ^JaJ&hisu|=1111 dh$Ifgd{:kd$$IfTlֈp y "&i&.""" 644 la[p<T1kd $$IfTlֈp y "'i&.""" 644 la[p<T dh$Ifgd{:1kd $$IfTlֈp y "'i&.""" 644 la[p<T dh$Ifgd{:  dh$Ifgd{:  =CI=1111 dh$Ifgd{:kd $$IfTlֈp y "'i&.""" 644 la[p<TIOUVgu1kd $$IfTlֈp y "'i&.""" 644 la[p<T dh$Ifgd{:uwy{}~1kd $$IfTlֈp y "'i&.""" 644 la[p<T dh$Ifgd{:~ dh$Ifgd{:=1111 dh$Ifgd{:kd$$IfTlֈp y "'i&.""" 644 la[p<T1kd$$IfTlֈp y "'i&.""" 644 la[p<T dh$Ifgd{: 1kd$$IfTlֈp y "'i&.""" 644 la[p<T dh$Ifgd{:Xav dh$Ifgd{:dhgd{:=1111 dh$Ifgd{:kd$$IfTlֈ 9L &&_&.&&&S 644 la[p<T1kd$$IfTlֈ 9L &'_&.&&&S 644 la[p<T dh$Ifgd{:1kd$$IfTlֈ 9L &'_&.&&&S 644 la[p<T dh$Ifgd{:.5;CI dh$Ifgd{:IJ[ikm=1111 dh$Ifgd{:kd$$IfTlֈ 9L &'_&.&&&S 644 la[p<Tmoqr|1kd$$IfTlֈ 9L &'_&.&&&S 644 la[p<T dh$Ifgd{:1kd$$IfTlֈ 9L &'_&.&&&S 644 la[p<T dh$Ifgd{: dh$Ifgd{:=1111 dh$Ifgd{:kd$$IfTlֈ 9L &'_&.&&&S 644 la[p<T   1)dhgd{:kd$$IfTlֈ 9L &'_&.&&&S 644 la[p<T dh$Ifgd{:( ) 0 1 6 8 9 k l n p q r       ! !N!O!!!!!L"M""""""#̻wffffffwU h{:htCJOJQJ^JaJ h{:h[ CJOJQJ^JaJ h{:hKCJOJQJ^JaJ h{:h:CJOJQJ^JaJ h{:hDCJOJQJ^JaJ h{:h6CJOJQJ^JaJ h{:h]CJOJQJ^JaJ#h:h 6CJOJQJ^JaJ h{:h CJOJQJ^JaJ h{:hCJOJQJ^JaJ! 9 B W ` g m q  dh$Ifgd{:dhgd{:q r | ~    B66666 dh$Ifgd{:kd$$Iflֈ3PX &3&a &8&&&X 44 lalp<       6kd$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:       6kd$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:  ! ! !!F!6kd$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:F!H!J!L!N!O!Y!6kd$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:Y!!!!!!!6kd$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:!!!!!!! dh$Ifgd{:!!"3":"@"F"B66666 dh$Ifgd{:kdv$$Iflֈ3PX &3'a &8&&&X 44 lalp<F"L"M"X""""6kd\ $$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:"""""""6kdB!$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:"""""# #6kd("$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{: #&#,#2#8#9#B#6kd#$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:#8#9#:#@#A#{#|#}########$ $ $5$6$f$g$$$$$$$!%"%R%S%%%%%%%& &5&6&g&h&&&&&&&&&޼މ h{:h+-uCJOJQJ^JaJ h{:htiCJOJQJ^JaJ h{:h3CJOJQJ^JaJ h{:h><CJOJQJ^JaJ h{:hKCJOJQJ^JaJ h{:h:CJOJQJ^JaJ h{:h6CJOJQJ^JaJ4B#b#i#o#u#{#|#6kd#$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:|####### dh$Ifgd{:#######B66666 dh$Ifgd{:kd$$$Iflֈ3PX &3'a &8&&&X 44 lalp<####$$$6kd%$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:$$ $ $$$#$6kd&$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:#$)$/$5$6$@$N$6kd'$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:N$T$Z$`$f$g$q$6kdr($$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:q$z$$$$$$6kdX)$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:$$$$$$$ dh$Ifgd{:$$$$$$$B66666 dh$Ifgd{:kd>*$$Iflֈ3PX &3'a &8&&&X 44 lalp<$$$$%%%6kd$+$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:%%!%"%*%9%@%6kd ,$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:@%F%L%R%S%[%j%6kd,$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:j%q%w%}%%%%6kd-$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:%%%%%%%6kd.$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:%%%%%%% dh$Ifgd{:%%%%%%&B66666 dh$Ifgd{:kd/$$Iflֈ3PX &3'a &8&&&X 44 lalp<&& &&&#&)&6kd0$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:)&/&5&6&?&N&U&6kdn1$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:U&[&a&g&h&q&&6kdT2$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:&&&&&&&6kd:3$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:&&&&&&&6kd 4$$Iflֈ3PX &3'a &8&&&X 44 lalp< dh$Ifgd{:&& ''''0'7'='A' dh$Ifgd{:dhgd{:&&&''';'<'='@'g'k'')(.()))))))****K*u**ͻ͙͈ͪͪwf͙͙͙͙U h{:h><CJOJQJ^JaJ h{:htCJOJQJ^JaJ h{:h*$CJOJQJ^JaJ h{:h[ CJOJQJ^JaJ h{:hKCJOJQJ^JaJ h{:hDCJOJQJ^JaJ#h:h66CJOJQJ^JaJ h{:h6CJOJQJ^JaJ h{:hLCJOJQJ^JaJ h{:hCJOJQJ^JaJA'B'L'N'T'Z'=1111 dh$Ifgd{:kd5$$IfTlֈ@pk!&'] '0'<''<!644 la[p<TZ'`'f'g'l''1kd+6$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:''''''1kdB7$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:''''''' dh$Ifgd{:'''(#()(=1111 dh$Ifgd{:kdK8$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T)(/(5(6(A(q(1kdT9$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:q(x(~((((1kd]:$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:((((((( dh$Ifgd{:((()")()=1111 dh$Ifgd{:kdf;$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T().)4)5)@)i)1kdo<$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:i)p)v)|)))1kdx=$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:))))))) dh$Ifgd{:)))))*=1111 dh$Ifgd{:kd>$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T*****1*1kd?$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:1*8*>*D*J*K*1kd@$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:K*[*l*n*p*r*t* dh$Ifgd{:t*u*****=1111 dh$Ifgd{:kdA$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T**h-n-o-------------. .(...6.=.>........../ͼ͚͉͉͉ͫͫͫͫxxxxgxVx h{:hKCJOJQJ^JaJ h{:h\CJOJQJ^JaJ h{:h-CJOJQJ^JaJ h{:htCJOJQJ^JaJ h{:hDCJOJQJ^JaJ h{:hCJOJQJ^JaJ h{:hyCJOJQJ^JaJ h{:hS nCJOJQJ^JaJ h{:h6CJOJQJ^JaJ h{:h3CJOJQJ^JaJ!******1kdB$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:******1kdC$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:******* dh$Ifgd{:**+ +++=1111 dh$Ifgd{:kdD$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T++%+&+.+4+1kdE$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:4+;+A+G+M+N+1kdF$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:N+V+m+t+z+++ dh$Ifgd{:++++++=1111 dh$Ifgd{:kdG$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T++++++1kdH$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:++++++1kdI$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:+++, ,,, dh$Ifgd{:,,,,,3,9,=1111 dh$Ifgd{:kdJ$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T9,?,E,F,N,U,1kdK$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:U,\,b,h,n,o,1kdL$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:o,w,,,,,, dh$Ifgd{:,,,,,,=1111 dh$Ifgd{:kdN$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T,,,,,,1kdO$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:,,,,,,1kdP$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:,---#-)-/- dh$Ifgd{:/-0-@-^-`-b-=1111 dh$Ifgd{:kd#Q$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<Tb-d-f-g-h-1)dhgd{:kd,R$$IfTlֈ@pk!&'] &0&<&&<!644 la[p<T dh$Ifgd{:h--------AkdCS$$IfTlrb & %h%$$$644 lap2T dh$Ifgd{:dhgd{:-----.. .".Ikd5T$$IfTlrb & %h%$$$644 lap2T dh$Ifgd{:".(./.5.6.7.q.z.IAAdhgd{:kd'U$$IfTlrb & %h%$$$644 lap2T dh$Ifgd{:z...... dh$Ifgd{:......>2222 dh$Ifgd{:kdV$$IfTlֈu D!t$'*'A' ''0'#'!44 la p<T......2kd-W$$IfTlֈu D!t$'*'A' &&0&#&!44 la p<T dh$Ifgd{:...///2kdAX$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:///)/D/M//0)12191s1{1111112?2I2R2W222H5I5 6!6666޼޼ͫޚޚxމaPaPaP hr /h{:CJOJQJ^JaJ-jhr /h{:0JCJOJQJU^JaJ h{:h3CJOJQJ^JaJ h{:hHnCJOJQJ^JaJ h{:h><CJOJQJ^JaJ h{:htCJOJQJ^JaJ h{:h[ CJOJQJ^JaJ h{:hKCJOJQJ^JaJ h{:h-CJOJQJ^JaJ h{:hCCJOJQJ^JaJ//*/1/7/=/C/ dh$Ifgd{:C/D/O/y///>2222 dh$Ifgd{:kdAY$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T//////2kdAZ$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{://////2kdA[$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{://"0)0/050;0 dh$Ifgd{:;0<0G0w0~00>2222 dh$Ifgd{:kdA\$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T0000002kdA]$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:0000002kdA^$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:00 1"1$1&1(1 dh$Ifgd{:(1)1:1Y1`1f1>2222 dh$Ifgd{:kdA_$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<Tf1l1r1s1|112kdA`$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:1111112kdAa$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:1111111 dh$Ifgd{:111112>2222 dh$Ifgd{:kdAb$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T2 2222%22kdAc$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:%2,22282>2?22kdAd$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:?2J2X2^2d2j2p2 dh$Ifgd{:p2q2{2222>2222 dh$Ifgd{:kdAe$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T2222222kdAf$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:2222222kdAg$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:2222222 dh$Ifgd{:222 333>2222 dh$Ifgd{:kdAh$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T33%3&3.3432kdAi$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:43:3@3F3L3M32kdAj$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:M3U3l3r3x3~33 dh$Ifgd{:333333>2222 dh$Ifgd{:kdAk$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T3333332kdAl$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:3333332kdAm$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:33334 44 dh$Ifgd{:444'4-434>2222 dh$Ifgd{:kdAn$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T3494?4@4H4O42kdAo$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:O4U4[4a4g4h42kdAp$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:h4p4{44444 dh$Ifgd{:444444>2222 dh$Ifgd{:kdAq$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T4444442kdAr$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:4444442kdAs$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:44 5555$5 dh$Ifgd{:$5%555>5@5B5>2222 dh$Ifgd{:kdAt$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<TB5D5F5G5H52*dhgd{:kdAu$$IfTlֈu D!t$'*&A& &&0&#&!44 la p<T dh$Ifgd{:H5 6668q=>AAAAAAAAAAAAAAh]hgdB7M &`#$gdS$a$gdk[ $7$8$H$a$gdf[$a$gd< $7$8$H$a$gd@ $7$8$H$a$gdf6777'7778797L7M7N7O7777888889״ף{{hQ@ hh{:CJOJQJ^JaJ-jhh{:0JCJOJQJU^JaJ$heh{:OJPJQJ^JnH tH  heh{:CJOJQJ^JaJ-jOvh(h}CJEHOJQJU^JaJ!jvCP h}PJUVnH tH )jh(h{:CJOJQJU^JaJh{:CJOJQJ^JaJ h(h{:CJOJQJ^JaJ-jh(h{:0JCJOJQJU^JaJ99<<m=n=q=r=>>>AAAAAAAAAAAAAAAA굞|t|t|t|t|jdjd`I-jhh{:0JCJOJQJU^JaJh{: h{:0Jjh{:0JUjhiLUhiLhf>[h{:0JCJaJ hf>[h{:CJOJQJ^JaJ-jhf>[h{:0JCJOJQJU^JaJ hh{:CJOJQJ^JaJ+hh{:CJOJQJ^JaJmHnHuh~5CJOJQJ^JaJ)jhh{:CJOJQJU^JaJAAAAAAAAB( ( ( ( h{:h-CJOJQJ^JaJhmVYUhiLh{:)hn20JCJOJQJ^JaJmHnHu-jhh{:0JCJOJQJU^JaJ$hh{:0JCJOJQJ^JaJ AAAAAAAB(((((((( ( ( (dhgd{:h]hgdB7M 5 01h:pB7M. A!"#$% Dd<3t  C PA8Mean TOB over default yearsbOBftJ:\24if+D n#BftJ:\24ifPNG  IHDRtWgI PLTEgdIDATxnpjF:Rģ=>)K+ bRUd48\S'ħ^lܐ$Izo~sZܸUwvYmO\Q BYojޞ ԗ5l e{qY ðm6ВKLiP:|xȪ{2 *٪K9ǐu[RevJ0Q-QEnKSՕ*f6jv*yuJfTmM2!SI:|n8s"3UUj ]nZ% @ݣoLYtzuzA @Ԫ-ԣbDn˴c%,SP7Ԍ>3w:U_;ԅ ;Tij5!P-MVmU(E@R2j[sC&dkyzJT4>pȩlzbjwծNd*ϕ> u#3Ս:&ief,̆깬dy[2ښ~&-@mѶPO[s6I5ͬjs:h2g鮴ޒ!@JsJ~U} YIZ%Ng:$_P_-y 3@- y__ HLkK&o0eׇyS0S;gii `qtڪuhcך.}[Ccz$3 \m6mC@S68 ,:?޹7:u'q3 }wcSg&$2s-1] Hs-`s-О Ԍj~sN?Nj`~ez~nWs&A&}q5p+h>VsȦ'&ZYvn@2 4 7@ .P?de8d[;Ģ/fz*^!ܪ!!ѯkP%澏3{C!GvPߙ/u*=P2=C@R6=sm3j@;$q @V'R3gLӗ~sd7 '@2~o CP/fb000000HMbQ{ pH````-W0000Qp' Dxt9qS?` 0zx'I`#iWX KqNh%q/yL'x.KqN(8q$K}`'W@A`%q^(t8@A`$ K]K Kc~ .0 V 郟T<Mw% xa? i8@C$0Jx _ҿO4JF@=h^XwbyNmX*Px=#4 @$  Ny PMF Sh @W#@ N^%qESh~ @O.Wo# :4Jz  +@Oa%q720.:@Cz uh1 ĝ{Q4J?6)g9@ R Ӂ2X?;4E pر߿XN ֯M iUt_]!I9A8I `w"6I,KVN`1/6@.hT_' 0TWB:sK ?*JG_3Bә O$T.wqCpK wvb'`9qeJy `%dl @!i$mUu(.\V 06}@Id3,,%`Wh"U5VMDgD#<O8 Q@nv:AD0E#5`')\c"p*\XXqeĕԴ`pE}d`Y@Eul`[XN[Xl?VEEh"2 /; [### 1 14<'2lV 2N`ZgX ;pFEH S "@k} %~#q3 pb @>hp||sw;AV`B@Co@vG8` 6+$:l#;, # "< #`7 =v`#`/ >v ?v;%q"`@ A!vZ"`W{=QD("`@ G#vwDIPe xP&&&*O4ǿ+pBiGAh?/4 <y!8>+.곀: "Ld~Eu7; x:B"x G4|06"Pǖn/_>9(il11\7=;y罱q&S4syu` L%]52|@1@ @` wh+w Td``` 0000V*w Td``` p000a$VϷ}StpݛV7`љ 4+X70h Y+Qr*v!nXYw".6{Xuoj?gؚ2f4-fJO=-VIj}Ѫiۮ,&vrZrEvinY-$-[VfkHN^qOnͺf1^U [ `f@nv' nt;Ec5NEԘŲ:66ի P^ oX]wYwYp]`[9Z\Lr:]ҫ\U^Xn)֋wI[ȇ3]j"`9pd @lèe q*odx}67bf^ <1000000008o}.?n/N,30H 8H-V`nBGP9 iu $C`n @r  U@D 0:4U'X%ȉ(L/ڂo7IENDB`.Dd b  c $A? ?3"`?2x/#b:8|OT[ `!L/#b:  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root EntryK Fy䲓 Data xWordDocumentJ7*ObjectPoolMpA;y䲓_1357136736FpA;pA;Ole CompObjfObjInfo  #$'*+,-0369<?@ABCDEFGJMPSVYZ[\]^_`adgjmnoqrstuvwyz{|} FMicrosoft Equation 3.0 DS Equation Equation.39qU|E Valuation of security default = HPI def yr, def qtr, Equation Native _1357136750 FpA;pA;Ole  CompObj fregion HPI start yr, start qtr, region Valuation of security start FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfo Equation Native _1357136834 'F`?`mBOle Uj0wT 700,0001,000,000=0.7 FMicrosoft Equation 3.0 DS Equation Equation.39q8|O`/qI(+x}So@~w5iAmR R*p.!$*:va``@ E aTaht{/(zr8c:b<2e]q%Y>cTG k f أ1% Bz'"r_?0oMMϟ<1xBO3_}X~ ,5Xf'Dڹ $Ziuo;vm$yJLwER7]^6LzwDv?9&MZqCO[7qu\W09BOIJy*a-_{ْ}9s &/"of?iՍʧ/PucoO_#Ka l[-vsy*اb>\M葳vvl;j]qnsۖݟ$\7ɓM~]^8?1Dԁvu@f _\{Dd Xb  c $A? ?3"`?2i#ϯzDҟK `!i#ϯzDҟK† @ 7gxuR1KPI lI$EKJC %[LnꬃkAqqDpT޽}}0/XuR\%cEǽH9LDJ yF7\,XӪ'M'pC:{ etVEoX~Z:?, D\!G_xTiq{4=@$YGc/z!k:]w&!ƨn{[ nOԾ)uRH?oRt56!ֶ' | 'Ǧ ojo]7۟o U_M@Z!;A颺MVݺN]hkp}ڟAZݲ&= [@8/oC=Ժ.J]mŐI$.]Lev,X@!@~^P> h~ɽ<6s>Od Nes2.hseʷWtM:_ӶWmnkr0m{m)yؾ?kooy辽uog#``s0@€8&?l~_H~i 398jr p`ؼL_7d8p? `x `s0hQ^',NH>d; `N\ \~ oy|31 ` * s kN8^0d`XX?9 dx`%f' p@X=@V п X[```M k0|9Y5=`o``M5,oĖ 0% [*[_7_kdoUj,Xh&mi.}%ot|6fX.pR6?} .qc@cwF|, `. X! W3Jgrz6g%m#p35X 7N0rravcO'0ˆN`.p @L#@#Aqk":`"5SW0*rKv{AP>'ܜ`'I s00g@\O. SD0 -kɅ@ M. `iq{jhĭr+?ܬ=<p=`EVjp`5nnhnxq|9\)d 0p@F8^'cv`X8;p@GC~f(`$v1 1£4p+@ Xqp; y raFD 0@0 Tq`H0`Tqq˧X>`7?߷}+I`8`, cO x*q0$ `t*xSA ΄vc $d>>[],X8vy@v" @iDDfBp+ IX$ `&Lh&  kf*@]y 'H@܄KD`0!M0%`DtLоm<`H 0-,p~@?#, LJlJt`D] XwRwAw0$``H@ @H@ܝc0iX`0 @Й&tIuG@_`HK̃X `0wD  L,;H:sqw`yaw01|l{HW'@{}l5uwlf@g`i_g0x~@ Yf3XL6w@g n&yGl&N3@ :\ n>#`fs`s0mo5$n0(x! k.hܼ=q2ghqvi~@@^x-`hD/0>pu| cPn{^[๎y]Khb=`~A@; <- и‰, zi(@.P?/ zYF/ e[}w.m-fEtqz1E B18 E*=;nA&uH4OQ.ǙF.9ls( ! +vTmT\=7aNP>=GnP쇃 `|#@=`['E_SL9 PWڢOⲥ9^H11ȹ@M h׽JYxݬR6IP*!*}"؇',dө8٨F,aL}"!6.*+2' WjPvאEYkQaٮ5u(}D1.AWhקCBpۇծ38OTLmw \uS׾i#6&DײW>;@ @ڇC#j6IF~!_J BLvA~AP^fcqaQ(DsRUf߭]{vۯ~kE|2h.@آõ}̮7u}.Dنم8#z/~pvC> 1h2r4 Ip0X효~ φg:MQ5g Ԗڇ1=0D}&n`چ(;[ ; g A5 QoaLU Q< B,ہI4c' +TJR^ELBչ2 W2Pu`@,.\kQV[6juIENDB`g&Dd<3  C pAXRelationship between Haircut and ranked VVAb%t{Hi'[%Y5 nS%t{Hi'PNG  IHDRtWgI PLTEgd$IDATxˮQZSOjKTO>S-4%J<\ ëVf؟[B"w @ @g&@ OAzh34 A_W;eo;!Zm%B@Y[EF.l4ύkhS"Vy2{s|0y,4a@/" =P KB\ 09M@SzI?+p* @/Anyr  @<1m*M̍}@f4,%BI { пw΋aҡ[C_O@ȷ a~o"" a6xgߗsKÊ@Ʀ;/05&@thk EI TAI2Io7>/*T* Ю-I@r% @CE ¼jNF @AgCor$jcavxq;ۮCKtZK0Ҵ]4/AKȾZ $o@'^M_` "a~S@ /Atƾn/paAsEoVDozAPt݅K pK9C62y74C/3 {sKY Q{&r)yq]lhn?n{qvXhHGp;)GbҡKp 9? F'dA?|C oi-.%+RmVtsHV)4W;Ư]sݕ]kWq_c{$&r⿌ y'mٸ]Õ2oe6+ ,~J( \M2Orb%HꆴU ~<ܿ1 (_,!_{6#LG_ Eʕr O\]A H@G y@ x$:UD@K@P jpoP4(ڵUo @7 }":@PFЖۇy*u9*SB_q`Y'ضdN(ZpƑh%ؖ)Q+Sn$jj*^ V0Pcs5$+7w@aP_:6JLg%k S*Y%V+A0sm@#nLWnKwqGg> 7R @ptQԎfO0RlNl)W7@\oxx)xL[u!5b.N xJ^y8XBmBjiߑ`ˁE3Vn2! oϭ]aj"7DŽ/U:6nMz6T nؒ_q5*M\,Bȑ@ƝM]Ŋ/՗i4g%O:YZC f '̺vhPKT7TDĥW̼X:Uo 0DT 󽄈BF+k} *x͟)l^ZB3`c@_IĪt֎9y ` x+϶z;`@jYs' 3%]F@9,sg ㍳ :͘& <@\Jj:˴L&Z5uԎ6*@Q- B++>y-@01TuAϺ40,k A`Z@Trt{@VoN& hc"rPHDeҚʛKHTeZWUrSRm5HxAh5dm!P@-YB!@IfP4вtub9^%&Z|?DQ=67Zy<Z3EsMe&_ʄ`2 `nV*pG^zvY<"lY))rZ.r+om…n>yEý@1 60J3Hi2LNǢ_h3譨]dmtܿ@e` kІtbʤћl%_h<'t,#Zh {ћ2u 5IKuGx欯؋\5%@T7ϛgpfT%z&A`sH9(ks+Ţ0˭sbJC|.Sǡ^z!r`*U;{60- 1\7 ?I @{vOiT;2X9wOr>+|@;,}jQ=6/o6|X+FB,C*BGVZKQ,@ VjzkЈUYop쏛Xcr^YX,o^v `B{]r^Y\01,b$D)/7BZ-#@]P"%7D-# @+D}`h{Y @h43@WMHz mBW{ )@W{NղRS+@t4jJ0|RTf e- V7 Z`P0;lfhBoѨ7[+gMN:_T nu54LZ Z~f `2 N@FR]1kKjPe;*`^jz*{VMq7 7C}lå\(r7#4~O6&iD9`Zex}CltSh*)z@&1J;G(mϝPnK\lf^:!.U: xA9>6Gqb2驓 tdž[8Լ q ֩b"g]@퉵b 0,{9( [9u7LhW(|XB೒v\7!FNc&4˙&j Y wf,kk/x@9-TC baz֥!KZ Y@щ)tܜbn[ Ŭb;*ˆOsXc%9jrycv=T~5XK+[bOKUiCs 3g>~gGٻ,6  mϵ=e_ $OPy} 62R ԋP [u_k jkZۏ}a[Ɏ-|.džuZήU2{nTA@_'P8(>e9enMk7߲/Vv0?Bչ>D^j68KТ6iKG<i@L*RqqX Jە.<Z}C9WJn 80hj3 Pqvڎ|jJH8awvR(~vm/ *#.bYOwvH&&Tj5,+|+/U֝ aB$.{ @0Z5NcRZ5_*zWɖC z%NdG) ['5pmzVRE !Dpʍbfk8 ++<bE :WIq`m~ji$8ɞMjW(b>&漝jh*#n3 ;f}5z)ϴ̢׾4X9oG/{P@pa$ )c/|_5cQ[ /|&,J'05]R2%:i56~Zv0 8_ @~fzi?m/YȺ3 `j1K'`?_,{{+ k'w;Kfa/.i鎵:r TKPVv |-$ Mq {d">^T+_Ӈ$)?Gh63"{lS<@n33gGN*? 9$?:G?0.jƴ@X fxS1 Xg;bDlP[rwbO\¥T8hyys57;>Z~kwx(Xl\<ncM&\'7> @tnP.(n ˽@I Vf@d)!6aI|?yu):f3X}.qF1aP3lmYp{.Ǿ 2NO*ӫzVΨPk;̶8&z!9[sۃ}#U*_$^=͸mpN:E4dJ/ɍ9LByA)A@y0W,@*a| +u~0- ][Didz+k:X 5O:q @Qvq@ EU?Jxrc5t4-}F[ݹz( q_u,݂M?cS;}PON5Vx#pBS@ AQ(F}"Ӳė_KT+BVj &jbڟG(`iן;O`8AS^ 7Ϲn<&@#\jН& gpl0:C#P0e@KS]9e{FNn=5Aho t5||8x>h-bɣWhe% <j;M;`>00Z59] (./*~8R!;7\m5u>D@:)$A9dyF+f 醞 I&qG_ D -(U9]]fV5| s oN{"4\c(Ж 8 b5)AYoUEq`Ÿ Pv$/@_;1PG#0Z5@@~U(3"!U ܇p쮭h>c[T.$tkU>q\S(k!*铗%8k@.VlKH9eC#-c<+f3|sf@:@ @R㜹|aTn-㯅G)W41sJ ODI||-DtgO'HA'1z%S>_t9N}9 =84N>EEZ*'w(} (>s0P>?<l*h Ԋʁ)"^D`$@t4qЪ%@ 8'H Dq u\ /#p"-BZNJjdC xjG9r-B @\s"r `sfvجi@s:-Blk%{; @GD` -a@q@wc \tٲ0 fJO@l @B @ Zb@] @E(6j*vv@I % @8I @ /!@ ZhB @"@ hE @@@ !@D -B Z@ -!@ ZhB @"@ hE @@@ !@ XI  @@@ !@D -B \^ K9hba@E% KuBG"@ @De:MddM]*&cp%~_~&~ \ M# =:螩{jb]v9L; V@ @ .zlǓagKL\~t;Wd[x˃n .@8 7hRr!(݅twa u vc?F_F~~S'P:}u,L52\܍O'"˃%JK)'v h\>Ƀ~$T`xp<ܟG@XqJ$, q׉w 1%TBCyH'' @c112BYL%bpus~ yTR2TbSTƫK:@ %/>AS d ozxx3M  G V1G [0St< ~ޘL)#G&x;H˯R=S-2tJ\"}~ɷHt%@O+@})$?AGk<{9/ T %gb} e7Py@ D@f]txrL[xU_&@ H:_hKgPM@ P-,@|R $J@fOQk -@5AQ15AU Я J]4xuOWkj'YÚ^ϫ\] bPgkQgR_ ~}/.X%Dk q"/GzFW@"rX%R_OЕL`g[|v@yLtWuˠ `Wyv{l+{~婼tK>WrL*Z[^ {&b_ͿWmy^y@nl!>Ś8_Hfb_e@(X-=lˌtO|x=e~NF_2 }c >XC? `pGnsbR8@4L0Wh=zh/ٹ7 P1mb8 qTq#‹DBExEUcI QQ<@F ^ R8**aC(%51`Uq B>XjE:@7c6l)PDj=`}$p]wĝq%o^K DDyS.t)_털)6YBѫVćLfr"dJs-- 'v{)_{Wu83U].S_ ثlsSaBBFoh^&h*ٕ.W0[a5n0ҼQJajFxabF$8# ueQ(Byu *o>u @Wio^DU6l=@Wipi zo*LD;q@Koj2WTD=0Z`H'{=3DPTw*Ö= @wM `0+ <d8lԖm.tLyf:4 ArtZyhDjN}Zm_6fLibo4@]rD;NXX3S&[ ש?|dC|%: }`")@|65 T@vljY}~aҫ<,!(t$5=@Vg$/}a+Wt,Lb@WKlk" J"Nn_>FHYW$Rxr@[iId٫F.\PGk{D[><E…| /¥ɻc1%a77n o(eyjHcYl{AIw]Z;ko(ǟ).˔ L9[׈wN@ۻ\vh{WOOиs nl=Fi$;rQFj#`nR)߫Vya;ό|pԚ / QX6qYk,Jf7hVs_2`ě7 ;˜ (کKLy@3Y@T 1QLpM)П,pujl՞Nz@xHTq <?#} (-5@H_!vΝb{y 1)yw=@N~O[珛|3`b0uubY4LU.1PV"@w>qEN27 d@DrT[dJP+6vWԁ9j ^~;|wOl=1"f@2`xe1(" }E-Zjmӫcoԯp~4W8@̌N0FW`af5 DN_٥ +XМ«t43rv|#: m9>yH@tvX[ .՘xp Fk蹿RcTނ]V|9 0.E `F9-3fLkR0!́ƅX(ㄼ|9 $SQ*@~l٭v%:!L %=8 =(I Ջ4`z9@4C]SH-Ts#Y&v6 @xA)0q28NvR V¼lfbYexleaYg|lj Rpe?CK#ª@P:ROhG%lXSSԇWnNuXg2-H l!Yx @&FE]_c|uq/Q\ ţ 5Zp0u`5hq@6Lw*8LsJjtg"S߮ @-ғ_/:\إ |lvTZS߱L@`..0,pt; {x]}u$ZX0$XLNcVٛK{cƻi4J;OpNgzN\ JTq~z038P/PsPgۨx&b[-L0 6[+̮P8_бiK90=b/*uX"퍗FX6?. G/yo:2^@N5q@SP .QĀbfNO 􄘻$ 30`ҕsR6؂W%븸N'os -p"csŠ0pei%;9tbcdn%9ɂ!÷K0;B0zx `t& @2s? Ћ{=`h>@%Gw\x Z<bz s ^\ &bs.d;:@a7Bj;G/:D"g,'XH_ S "ة3Pk5ɦ'S#30 @) `@Ot ]skk~$Rj<@JEp$Svt{@PP1 c%[$*aNN<@ptY a+ b`N-Obp d<]1 2 D{0a.,@ )8acdFHI2VJK|hH" @1-! @h;0oDj!BR'I cb?L4G PP?@iJG1bR%8h?:_-a5@08@L`N I<bн`=@@{:@A6RS *P d ](x (^ <@EseͶ^yk~>ٶmvwcvmsd`OPl_ Ή"OvN1`})P~R sJ6,`=8}\@uVK0ҏi @i۟{=4@_g^/g$(ˋKGIOx!U''7sJG?_r[(oy-d(-6VgcZSG\ 륏U#@0~tyjO ȃq$x ~rM$4 ?0 Fb{^I` 8膩]~x@U~z5A jzO^_.&]`5m׆%}`t2\c T:IENDB`Dd<3  C nAVPrediction performance of Haircut Test setbǫk14 ¬,{ZO nzǫk14 ¬,{ZPNG  IHDRtWgI PLTECxIDATxMʚp 0>MH,ÐUB>ErNB\Ս^1U.w1y$Kd;v]WؖJ<mCya"ߴ,acx x=iQ kbmH 4,*XW>WcZclyFS7%3@XO` i@9*BX̦~-C$׮ `[+@ἀb@94J/c;OQ4C2_(Kb_]@.MS㞎4c\A^ -N $ P,̼NKYI4cqHay zA_zF2s#qf_HYF1OMI%!+9/Jf5{Mb| seBxaT=6mJhJKě7?n)p+/"0+ѩ]|A(Li3=: fl L! .KN4,V d(;(2nWPuPr1j>8T\B/,8>% 2{iy^F)!4JB] /]>xO Q;(L$6Wp ltlj X&W*LI 5%T -K _%r@fVrH^GQL=` -??Y mUe=@HLtqW] DVZPJ`j,k%TbPt՚  ^ u zUCh pܴcR2O)b [^|.ᑛ|n d9םVU[n .ya@}{5XA}f\P{t2 3Jc}~"s}7yDG<mY2 4[e*W[(W ^u<>y< 5'$iqߚM論hƅZ-{'r&HOPsm\^(Q Vdu" U$fYhS (wAd&x^alQ 84B>A)h h h 0s`6 cdz@ǟݾ 8zC4ySc+nL&׮gΧӾY`zPN5p;*u8@\XE|60ཕ~%[)`E_N V }ǪX)U0F8z;)$@ ,n F^]p%Tka: z,]Xz׎P6W s`0tӁ3 j`=@7ü7^Pe:$`Ycෟ"R@-W`1‰,,*>0Q PqNzN\L;g~;];:hbgynʀ(8k*肊ÙƳq@yk[U(kK  󙼿lqxlic:_1|P֦$呚 ZJkc`~V-=?(O%=﮽og;75Qi_m '$+Q:/+@1 GTPтڼ/{-84E3׼o _rϕ{e_8 [ Y@IÅ$CR/єg=N!fE z_F"{-[>෨u8k-f>_-Zs ƀ kV;4df]'~Msz DZ(KPx/k>#mׯ/&#}/t0xԵ!TRoR ~vmJ@%!/3t)Վ_I_-P|g'59ѵ^|ۜY[`1PrĀJ XX/RծxV.^M׻vmXmɕN3C]=S|Pkۘ } RbPv[YPeo>s>Wtn;˞Λ3dϼhayZ};

3yrOT_u9"NE%4=ë7+gfÁTY?\N{lQ?Wʶ{=r Ϻ:O{JU\P xTځ;$H=e* ǴO_,0{O/MWzAS>z:T'= ;y>+,eoHEd!pypL0r[sN~y y:E.Mσ؜!lT;Lw4_mP'0i归^-~U9}jO7}lP>n7ʷ<{ ;vyf7,ۿEP Y⾁^qk Unf)tge?cͬwW TRˮ(Ry3~ʵm+}̎Jdw||7}uc껧+׿4n vJo'%uߕNzPW|/וNGqyOoVϖGKT;pҤ 7\Obܱ5XחwK7NC{2ոA~}wS`1П\.P+p7y_t|b) wu?sLgwq k)$ vÆ,b!@X] WT]~k6$[/96{jwqm8lVͣVyt3<1`og_;[8`G-KRo])Dҷ2^/ꇿO }O~/S\N!7)@{^[Ovp'eZcu2|b@# v;O@|`Ux<%>r}5)` [w!Sג,~k}-P/qe\j3Ɛ|c_)o໿rz=T$?~.Z`qh \W*5S8sg8h/V Vkvƀx}.V2_-3i `i pg]/kz%80 dc mL;=4ؽXUIްup2Z'07mub T ifry!XXr˳@ 0/=5W[ǀUznX\{P)\EoJ:va[,?!`as`w?U/rD|;@@-PWYk97Ttpkp^\~|C@qS1Šu t``9uPS=kq=wc}*7*tnqfU[Χ \!<BnwD xHp z 8qQ8\9pQ)p pz ],p1zYwdƏ( IB\1pAN1.5IC\1N1.5.5IC\1N1.5.5IC2( dJݰ#MZ}oC, @uht%qU߳¶:w2%U6WCompObjfObjInfoEquation Native :_1357136837F`mBPDUK 2H  j FMicrosoft Equation 3.0 DS Equation Equation.39qU  j FMicrosoft Equation 3.0 DS EqOle CompObjfObjInfoEquation Native 6_1357136832EFPDPDOle CompObjfObjInfo!uation Equation.39qUl(& h j ~N2H  j , j  2 () FMicrosoft Equation 3.0 DS Equation Equation.39qEquation Native "_1357136856"FPDPDOle %CompObj &fObjInfo!(Equation Native )'_1357136868$FPDPDOle .U `d Eshortfall percent| repossession()=ph()DLTV"h()dh ""DLTV +" FMicrosoft Equation 3.0 DS EqCompObj#%/fObjInfo&1Equation Native 2m_13571368731)FPDPDuation Equation.39qUQ`4 z=h"2H ~N0,1() FMicrosoft Equation 3.0 DS Equation Equation.39qOle 4CompObj(*5fObjInfo+7Equation Native 8VU:X4 D=DLTV"2H  FMicrosoft Equation 3.0 DS Equation Equation.39qU14 Eshortfall percent| r_1357136888.FPDPDOle :CompObj-/;fObjInfo0=Equation Native >M_1357136895,;3FPDPDOle HCompObj24Ifepossession()=pz()D"z()dz ""D +" =Dpz()dz ""D +" ()"pz()zdz ""D +" ()=DCDF Z D()""PDF Z D()() FMicrosoft Equation 3.0 DS Equation Equation.39qU6p@4 CDF Z D() FMicrosoft Equation 3.0 DS EqObjInfo5KEquation Native LR_13571368988FPDPDOle NCompObj79OfObjInfo:QEquation Native RR_13571369086@=FPDPDuation Equation.39qU6Z4 PDF Z D() FMicrosoft Equation 3.0 DS Equation Equation.39qFG      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxy|}~Ole TCompObj<>UfObjInfo?WEquation Native XXU<Z4 Eloss|default()=Eshortfall percent|repossession()indexed valuationPRepossession|default()[]+c1"PRepossession|default()()[] FMicrosoft Equation 3.0 DS Equation Equation.39qUL54 Eloss|default()_1357136923BFPDPDOle bCompObjACcfObjInfoDeEquation Native fh_1357136758GFPDP(GOle hCompObjFHif:)[7#۶"{lEUym[ RP[uRҾAжc+D[hOAmN[?1uRҾAV4V$^w@:M <@ܺjcn{lHl?G I*iOAVIj) h/R T)>J,lo PM@KG=Fn 8U]Y|a<@;x'(1@iL2Z. - pj~4=zPd8 Y i8+l N{"f 3zԯ+{B5YWzjhOߏw@{oeÓߗIENDB`Dd<3   C pA XMean haircut standard deviation by TOB binsb ֹ~dF3ۚ nֹ~dF3PNG  IHDRtWgI PLTEgd~IDATxn0a+W#QfOP]y)dT GXכFƆ27VsU%3V)6|&sUmn&2a.-d%ŠST|6l6TF7rE.y eZ \8rx[XFFyC+Z{UzOP2@V&U>gÿ؉6C6JgI r ^I/waɽ@xWjڶk4.N&=Bb-gU;@:v%R֓Bv Dж"Jd]Ycsδji# dRu )P $3 zVaɎaƐdrf[m K n YºRבLCٹM@w&gVCx$RTbBREX4gY >T  T˪ g@JO.BitR$JdU@Ihr;!TVƼK?!LoLVt$fQP: e.olpE]Mp M*O@*..M 4*&բ |8-gqB فau+@(ZPX'Z E-PU~pL@jj*9[ Z@ΝU("@Vn;:Bۺ Nll|8 iu YS5(Y LKM2%(k)ª`Fi k.YKl[1Ė`ɲ\t ͼeZ 3 k!#-0] n8MaUUh. KkP{-ҵ@'+D+2Ptg%ZBEN{Flc0ɉ*ªj)%%` `}{5 B&p7 }Prv+kd;N/h^dvVKOvXgK^+?~r04⮲x_ !lj=h) jip%Hi ) j+pJU PZH 3V;7D) jR@1=EP @T @ (>"@K,IZ*`HM%4eHe]) kد,p02uҔbk ?Ga{>{S8n?NU &j= _#ؿu'`@b_hGSPjJS8ɗ/4w rnZʙv>lt> ք3h W$>\!rY`x X `#J)"?8/+Tty:|< ]JKk|E)k-@ _w)?[A 2` bI_ұ/@ ߇k߇Ю`NqLQnjjP1@ b=`>)JN*P __!(@ҮjZKܛN)i( T [crXsW?`Q?q`@J0*>MIdI0IA %\$7 0I:ST=@x ป+‡\^_|srO1C S0O7W pUWLxUk O O nRz1Q;^@'r(^)Zj=ڳ ]e|{\{) RL{zu9 _i$` =E @OQڳ =E)n>`-'excp%DuNyLv12Ir[H5hT3Om7h4@1@ (sVϼ?}9Y6~=$0ʀ}!<-#o]KShP=*hx> b~@`  ;ts"`] 玒_^@sUe@:[Ia {` -٠џk#: ,y=+/Xw!n t3RV;`jjy5 qy7̪ `!/~`W:>qy XJxcdd``> @c112BYL%bpu 1Dl\ _ˣ-xXt9@{*!5Ljf/~12ps0*1L䂺 C3n``ÄI)$5d.P"CD|bb#3XB8H$$If!vh5Y 5q5p5#vY #vq#vp#v:V l6,55]5%9/ p(T$$If!vh5Y 5q5p5#vY #vq#vp#v:V l,6,55]5%9/ p(T$$If!vh5Y 5q5p5#vY #vq#vp#v:V l,6,55]5%9/ p(TDd Tb   c $A ? ?3"`? 2h_jC\gO;.D; `!<_jC\gO;. XJ@C xcdd`` @c112BYL%bpu`Tba.;s'\F=3vTrAC r`ca"5#RpeqIj.C\E.:OaQDd @|b   c $A ? ?3"`? 2d🯅&B~(@Y `!8🯅&B~(` 0xcdd``> @c112BYL%bpu1;Zf`3ًv\\M\>h`` #RpeqIj.= @ ]` bg!t?2m]-Dd Hb  c $A? ?3"`? 2w=?&jnS `!K=?&jn@%9(+xcdd`` @c112BYL%bpu@1sM20a~, *6x|i^. ( 88n7r<?̯YفԢԼx-eJrIqF~QIZbNt%:Tt220)|Q?BK\r \vZ* 0y{-#RpeqIj.;: `g!` XQճDd b  c $A? ?3"`?2?-/Oz.(-[l&; `!?-/Oz.(-[l& ` &Pxcdd``>$d@9`,&FF(`T)WRcgbR vĒʂT/&*de- ް`@dbabMa`ĵb]!Xk7o-`2rjT2ن6nndnM 2UIMse=5ln0~0WcGuP4F6]hnw ~P{^1vp &?. ;q#2[]1l.!f[M\^ `p`dbR ,.IeC D4<b@3X?uoDd Tb  c $A? ?3"`?2%IDh\ `!%IDh\ڠ `P[xcdd``` 2 ĜL0##0KQ* WÔ%d3H1)fYˁ"vĒʂT/&*de- b Y8ArȅL,L I9 CtM 1Dl\ _O]ˈK+oŕXcu yL`~dG%IwbԞitTrB cp bJ}t8l}qo v 0y{)I)$5E.Xݗ`VĂCDd lb  c $A? ?3"`?2[%,Zs7] `![%,Zs7 + 0DxڕKLQ\t"RZ)1&.!X1&!!mXDWlJ$1r#1,ܸpcbQ >@0Q13)ToI&߽9\2iCeGD 2P,ZIuچG1b Ygh ?SQle ͯQ cdS -8sr?k!~\u+C[=ݵذ[>CToez;4 awHy4PGD6+36Ɉ_;5+$t_ytGp?z~a\,sXHĊg(;8qX!7Qϳ+ Ӯ7~2`𼟨TRJ*}IVgn瑙z:+cP~ `!vG>:+cP  XJDxcdd``cd``baV d,FYzP1n:&B@?b ] UXRY7˿nA 27)?D.B ?'дm,@@@ڈq>3HF%!bLp? 2Bjn? Q !`wz`2" ý !ƽP00Y.n\x5`h, 0y{I)$5adE.X`4׈YDd Tb  c $A? ?3"`?2X)dW-2HЩX `!wX)dW-2HЩ XJExcdd``cd``baV d,FYzP1n:&B@?b . ㆪaM,,He``7LY \p=o!XkygH! P.P56ebiը$df0c12H*Єj.HhBHbF~n"؝Lv'\F Wu^s}\t>`2" ý O{9Ub`\RkX``S#RpeqIj.!dE.X`8I4Dd b  c $A? ?3"`?2E!88[C}a1(`c(@:ƘN,1$CY-mK2wK)>>m^y'aǕyKNw-~~䳵~m4 }*wJ l;D _ty~?Yi%Ww%~۴_ WӇ̘.;7V(b Q +JqoyqT'doru7KxBwLgF~;sR)gs|ވG3`/`kaM,#Y 6kA(gڢQDd Tb  c $A? ?3"`?2o NkGw `!oo NkG" FXJ=xcdd``dd``baV d,FYzP1n:&.! KA?H1\ qC0&dT20c&,ebabMa` :-PHq22ݫQIȼ\q? {d&܆ s*$f&4#nwBeW+;XSцdpM*'җ_\ SR M.p Kv0o8+KRsAaPdk[> 1hDd T,b  c $A? ?3"`?2[|٣ =z47 `!/|٣ =z4 XJxcdd``> @c112BYL%bpu 1Dl\ _˅ֱUsi#/ `UBk3Œ?&* g?m˭.nz}91ͯ'#s2GB5^6APw: [wox5 xVmI;}[E0}Cqk#74%J#Pk"Ecmbܒhy-/KQUkťXfS xٌ5Z.XiXCy(m/W֙b@r h4hT`,-M…x#b}kZ:CglƸ@ל5';w ^`괨цY $北o0zZXCy(m/W֙n*-3-Wb]a^hi*֥֦iXz[$SހM5t)L&@f+fZc'+I\nsQ(]s $P @j;+-,d/½pOq% `;&%)\`B(.5bi].-|q(dה93{`r=y w( #֧l3o\0F} sBv%QKCҍQ@hPq. {p0 PFRib#(Sin< 9htyz إQ ='8X@@86& /88KS:bV$#t87A$ /Y\GiTj]^?D6;1,#+'& RW 8  )vdoO-*]& tgRIC?=Zh~Kh4/35 o4e^spzn8,~?p,&>+zw G]!zi]O: ѹz&D-{mӗ bBCn+qP&@~[Yc b-=7(i/@aTo%@ 2gtU W#w&ʐ⃡!ptFH +3{B 0^b'*`fnB}.﹏ !.Z705DAw!EQ.xs@!}i~bH|#_GteHv?塪3?{E 0^PkN6 5ȧRѠb.zpQV<"\p:JSaʃ 2A<hu 6p # P -3/[. Z$B/|e-E{ /xxU:A0Gӏ>byOƟc|ߣ,g :?vg_S p]tty"T\5^zYLEЭoqETbO{]?Įb{ctms+Njv٦/zd{{C܅x$FlTr027[TS 1/Dcb:@Aͯ"PprM(( /_i8ø hmy >1ՙOv%PӍD(]P @-rATsu<<@w"`/3K! *d%F(e*R - >P]cOVRaA|f o8 9Z< @9D0d^IH/Sg*xr{> `8RaF1G6OS kn_, S ~Q3aTkL=I`zۜb8'&w~*4S)ƫX*^ k}oﱖO[\Y xnp t P@0jӕ!pR |ehHgUJ.J_8[䃡+6 @Gt683^~^B<@] Jco},+ ;1 0њ+Cuz@b`z EFF ֏?L_i8c2d ݹq,%2@tN.^(E{!U;P kޔg5W@; X % =xmV .0 |4@0x-3 7jn鍓 P P/Hyg|+/R )V-@ ]l/[QPP zHP[,յ(?~'ecgrS Pueg l<hxA YEnp \¥'ﯹ7\0l  p]tty>Gj8+8*7O@%bQ1cRYL|ҭ1U!zx-T c@0`4w @O龜!PcM_ 00@?ߚuJ޾ﻚcx@jn(Ѕc@2pӗ[~=cK;z})y8u p8<>+6r>OV|ewxd |y򵓊btpP@_K; EO }9D/i ƒ,xADbLk^"aoޏ~9&l2ALSX7Z }ٚW[] R"\3ENݣ'\w >ٕ=zӞ Pk4v@ F.x2[LP}Pq0D2C8aP' 0E{&. H^qx^(=zL0XQoP4@=8Kc@;]Ѡ[z]yz X_l+h1~ Tw}@uaC P7 -n0 ` /8#ɜNjʉ gdDT]T.mqV eӿ=~Vg?.D7 <(N~V>Z̧ f-ߖ6 MZYcW%B3w鼊Y>7+g)kbT^1:gemMF[Yێy5dmyqZi{|㴊sS(Wz|5pZ;_1,fڄSʹ8w .GS;jv~&sHU\prӟ  k>Xf8#. gyGƭ\0 s ?|Z@k&Эa hZ ۓ*|w*$v ŵ ƵP'=8-e1Ŏ2c-\ٮX\n0k(W1`I n6rDfp&=Qg0PNG  IHDRtWgI PLTECxIDATx=șڭ3 /qN{]X88hsgП]ld\ nkNjyHsG1!\VH;7:KnakA}v&"b\wV;|Jʄ I;S\㤤]4oV׹m-&g_-[5fV l"d>7dn[Q?+¼q>KYYLA>=Yή5Fms+I?",z/{E\Ut>T]@z (Uh0p?`|^`I`j j.tPR\!wTJ *Xl[ vȷ[MZ-`t<dCϭoݵh63 E]+2ALs=If&EALЎXFLp63}PpsC+[@p}t][;S:&n >tT%{ud nn+ bZ7lRۡ4"Us2@  O d Fκ)-[X:uGp#l *o]>uGճG K@:7wxL-bvfU}K?/~ANPAON eR־Fgn^1KK:H R[?v.Wn~p'*]=K1o=v!^eWQ(9ԏFͼu p;*Fv_U-B+` ab󟥭ܦ04[ R*ASmm{ׯ!B W HiA4%R%R$U(@( P68I@88hPhhPmOh9>6%Dt P2DMR"ħtFF88h4@ v$ O x=R*=R/(@LĈJ? H @8$R$RZ4=@bZ( P64I@B88d +Ptth_`"^~Pt@"6͔@ . lP4VF gPцPp%J `'I@88;; e-Hc (1~߱"JO<Ի%@1`+(觇T|%*nT?fGGmh =C% sL?yp-b(hn]\@Kp" RA&~l LJ&78&QG%@\ #8OzJnc< la38qP4%@ 4xf01@Ip `B|{|hQ4;@u@sy,:9m-.QWu%@JDK4iK@+  o,u@ >V38-0XF\`82P$lV"'w>Yevx;xKꀀbr6 {_X4*!@dW͠h;E``tt οK@/TG@ 4ca db41yh88@ ppa_޻݁0ϘK7Ĵ4@HXH,!2! `X* li3T/@vI, Pw Lf$ MGpXHHK)A}'Zx\m0o PM7>P$p. I0OA@W򭀻ЌE(4*$13@׀X`iw8@ `lEl ixfP'GQ z!&Ey'! X/<^\@u:ߓGݯGT7p@=H.l6;GD:4E|3:LYF9ФI@/ah6ZA,%26 pQ0b3$}}ll, h"`N?R@4% X4{E<Ir_ Fc]S=:{d#4|7%~%Mo0:i!R&PF @Q& `ILTH`5`>/48W^%u0~t2_)l2\A ! r<DMC<L`GIo@v @G *>>.O LÓI,ݓ@$ Aϓ:S'yh@5.h<8N5S 8:R;h_!&-k/o~̦p8&:u@ STtgDFMd(tDN3X2<px?*?Hf؈%G ԁHfwD`MiJ@ri\N@TK_;rNT߹1D8\ |En4> pP<j v[gϛopJ>jˏ!s5#& F(88inh^ kW7&R:& DBȽJK.ٔiu9&H|=mg#C@i PxJ>Pi@(1|):ÂH{W z$@.P&v&H3X谠덀 r@ %t;@O%h[fP&C9Yr=2h  K\`#rn6@ X 3NW^П<W $O~EE\ WA\Ap@+. ྞ/ad/@8(bS ZA߄GVi0?.Uf`=i`=iI_D+^[iW V p0 4ip"f&4:4@%p%z G%'Wi@wW7 ia-~cԿ"wm'VsPĥ9(b>/$T ͧ+L `I V*<*W 3/4U 8WhA?=|Й>w @a, };W 2,4@ p0@#4@;$@8] Xh Vi`-_a&_#M+֘C(aVPKH?P r @`=c0ixVSG`E]vZt.~F se}|DE|@(9UB)4G͠b+i w'l_cҽ萢|Ӽ@ pg( HȤ x* 4[O@I>Dxe󍱇Uw?4?8`͸"w?6_ؿ7`j>/؛6c` o}p>/6Z%`#%?uҀL<<!~|3=g4*9<I:B VwM&EW~-<{?2@ d)d 2@ d 2@ d 2@`I~SEcpbsT`A6i-ys6%0vn7e`H+0\2V(`PvP:C&ֆM&}2@ d 2@ d 2@ d3}0ٝ3^itsi7GZV/vE/ސޕo]o&lOD@IyT{u:lj6vVՕA>[(VS84ي3@ ^`]@W?,\ϖSa@ȵ}'M}ݜoyypӛ8I60;h.ۥ5+yVrP/>T-df%ʧ(ٝ ~ζ:kzz7eҭYoE5 J!% :`@Uv/nC"]>_ ytYgM~`iKYWn,R0F~@\7e 2@ d 2@ d 2@ d 2@@)&/lIENDB`$$If!vh5d5 5555V#vd#v #v#v#v#vV:V l6,55{55#5N519/ p<T$$If!vh5d5 5555V#vd#v #v#v#v#vV:V l6,55{55#5N519/ p<T$$If!vh5d5 5555V#vd#v #v#v#v#vV:V l6,55{55#5N519/ p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 655i5.59/ a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59/ a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59a[p<T$$If[!vh55i5.5r5J5#v#vi#v.#vr#vJ#v:V l 6,55i5.59/ a[p<T$$If[!vh55_5.555S#v#v_#v.#v#v#vS:V l 655_5.555S9/ a[p<T$$If[!vh55_5.555S#v#v_#v.#v#v#vS:V l 6,55_5.555S9/ a[p<T$$If[!vh55_5.555S#v#v_#v.#v#v#vS:V l 6,55_5.555S9a[p<T$$If[!vh55_5.555S#v#v_#v.#v#v#vS:V l 6,55_5.555S9a[p<T$$If[!vh55_5.5 FMicrosoft Equation 3.0 DS Equation Equation.39qUx 1"sale pricevaluation of security at defaultObjInfoIkEquation Native l1Table{TdSummaryInformation(Lp55S#v#v_#v.#v#v#vS:V l 6,55_5.555S9a[p<T$$If[!vh55_5.555S#v#v_#v.#v#v#vS:V l 6,55_5.555S9a[p<T$$If[!vh55_5.555S#v#v_#v.#v#v#vS:V l 6,55_5.555S9a[p<T$$If[!vh55_5.555S#v#v_#v.#v#v#vS:V l 6,55_5.555S9/ a[p<T$$Ifl!vh535a 58555#v3#va #v8#v:V lX 535a 5859/ alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859/ alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859alp<$$Ifl!vh535a 58555#v3#va #v8#v:V lX ,535a 5859/ alp<$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9/ a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9/ a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9a[p<T$$If[!vh55] 505<55<#v#v] #v0#v<#v#v<:V l!6,55] 505<55<9/ a[p<T$$If!vh55 555#v#v #v#v#v:V l6,5h55559/ p2T$$If!vh55 555#v#v #v#v#v:V l6,5h55559/ p2T$$If!vh55 555#v#v #v#v#v:V l6,5h55559/ p2T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!,5A5 5505#59/ a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!,5A5 5505#59/ a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T$$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59a p<T $$If !vh5A5 5505#5#vA#v #v#v0#v##v:V l!5A5 5505#59/ a p<TDd `'#b  c $A? ?3"`?2ۈN )2v `!ۈN )2|@#xuQ1KPwim T PAZY[HMiRKn'G8]]\`|8i}{ZV 3a m唛ՙXRQ9X`|,эq E5Ƅ[<.OL*bIE -Ht]9p(/# ;`kWoE8 1Iғ v>_qE V,9$fr f#[J0˟uw<ս!9^_m G1ߺv˛gy`D1D HQ u}`wӯҕ鞸ڥ2,(Qm!ף t:0 +B0-B|No*D@D NormalCJ_HaJmH nHsH tHZ@Z _ Heading 1$<@&5CJ KH OJQJ\^JaJ \@\ _ Heading 2$<@& 56CJOJQJ\]^JaJV@V _ Heading 3$<@&5CJOJQJ\^JaJDA@D Default Paragraph FontRi@R  Table Normal4 l4a (k(No List >@>  Footnote TextCJaJ@&@@ Footnote ReferenceH*jj :k Table Grid7:V06U@!6 :k Hyperlink >*B*phB'1B rComment ReferenceCJaJ<B< r Comment TextCJaJ@jAB@ rComment Subject5\HbH r Balloon TextCJOJQJ^JaJ4 @r4 B7MFooter  9r .)@. B7M Page NumberZYZ MA Document Map-D M CJOJQJ^JaJJOJ * Chapter Title dh@& 5;\ROR *Chapter Sub-Title$dh@&a$>*XOX J\2Figure Caption$dh@&a$6CJOJQJVOV ?44 Table Caption$dh@&a$6CJOJQJ66 _TOC 2 ^:CJ<@< _TOC 1 xx 5;CJ\88 _TOC 3 ^ 6CJ]66 _TOC 4 !^CJaJ66 _TOC 5 "^CJaJ66 _TOC 6 #^CJaJ66 _TOC 7 $^CJaJ66 _TOC 8 %^CJaJ66 _TOC 9 &^CJaJT#T _Table of Figures' ^` :CJ~O~ y+Normal + Justified,Line spacing: 1.5 lines($dha$4@4 Header ) 9r Pf;R 9ea)[  9              9*~ST^ ///000IIXXXXllmmnnpprrssvv6z7zNzOzpqrsJK}~)*Ɏʎّؑ]^_z{&'AB !ީPV\cjqrԪڪ !"BIJKLMNO۫[\moŬǬ)*+@AîĮ 9:;<XY>AȻ޻߻/0/0 %+,-cl%)@A]_12[]9:< #$%flptR [\,VWvfg,-YZqr !hx|,28>?@abdIJVWOPtu`aR   $  8    +37@U^eoyz&/DMT^hisu|  =CIOUVguwy{}~ Xav.5;CIJ[ikmoqr|9BW`gmqr|~  FHJLNOY3:@FLMX &,289Bbiou{|  #)/56@NTZ`fgqz!"*9@FLRS[jqw} #)/56?NU[aghq '07=ABLNTZ`fgl#)/56Aqx~ " ( . 4 5 @ i p v |             !!!!!1!8!>!D!J!K![!l!n!p!r!t!u!!!!!!!!!!!!!!!!!!!!!!" """"%"&"."4";"A"G"M"N"V"m"t"z""""""""""""""""""""# #####,#3#9#?#E#F#N#U#\#b#h#n#o#w#####################$$$#$)$/$0$@$^$`$b$d$f$g$h$$$$$$$$$$$$%% %"%(%/%5%6%7%q%z%%%%%%%%%%%%%%%%%%&&&&*&1&7&=&C&D&O&y&&&&&&&&&&&&&&"')'/'5';'<'G'w'~''''''''''''' ("($(&((()(:(Y(`(f(l(r(s(|((((((((((((((((() ))))%),)2)8)>)?)J)X)^)d)j)p)q){)))))))))))))))))))))) ****%*&*.*4*:*@*F*L*M*U*l*r*x*~********************+ ++++'+-+3+9+?+@+H+O+U+[+a+g+h+p+{+++++++++++++++++++++ ,,,,$,%,5,>,@,B,D,F,G,H, ---/q4588888888888888888 9 90000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00000000000000000000000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0000000000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00000000000000000000000000000000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00000000000000000000000000000000000000000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0@0@0@0@0@0@0@000 @0y00@0y00@0y00@0y00y00/y00y00@0@0y006y00t!y00]y008rsJK}~qr!"MN[\)* +,@A12R \,VWvf>?yzhi  UV}~IJqrqr  NOLM89{|  56fg!"RS 56ghABfg564 5     !!J!K!t!u!!!!!!!%"&"M"N"""""""##E#F#n#o#######/$0$f$g$$$%%5%6%%%%%&&C&D&&&&&;'<'''''(()(r(s((((())>)?)p)q)))))))%*&*L*M*******++?+@+g+h+++++++$,%,F,G,H, ---/q458 9{00,@y00+y00(y00'y00%y00y00y00y00y00y00@0y00@0y00{00{00{00hEhy0 0 {00Ehy0 0 {00Ehy0 0 {00Fhy0 0 {00HFhy0 0 {00Fhy0 0 {00by0 0 {00 by0 0 {0!0"by0 0 {0#0$,cy0 0 {0%0&dcy0 0 {0'0(cy0 0 {0)0*cy0 0 {0+0,0Ey0 0 {0-0.hEy0 0 {0/00Ey0 0 {0102Ey0 0 {0304Fy0 0 {0506HFy0 0 {0708Fy0 0 {090:Fy0 0 {0;0<Fy0 0 {0=0>x&y0 0 {0?0@&y0 0 y0 0 HTv{0 0{0 00 0{00Wv{00{000 0 0{00 {00{000{000 0 0{0000000 0{0[0\)0 {0]0^)0 {0_0`0*0 {0a0bh*0 {0c0d*0 {0e0f*0 {0g0h+0 {0i0jH+0 {0k0l+0 {0m0n/0 {0o0p(00 {0q0r`00 {0s0t00 {0u0v00 {0w0x10 {0y0z@10 {0{0|x10 {0}0~10 {0010 {00 20 {00X20 {0020 {0020 {00h:i0 y0+0,0 {00:i0 {00;i0 {00H;i0 {00;i0 {00;i0 {00;i0 {00(i0 {00Ui0 {00@Ui0 {00xUi0 {00Ui0 {00Ui0 {00 Vi0 {00XVi0 {00Vi0 {00Vi0 {00Wi0 {008Wi0 {00pWi0 {00Wi0 {00Wi0 {00Xi0 {00PXi0 {00Xi0 {00Xi0 {00Xi0 {000Yi0 {00hYi0 {00Yi0 {00(0 {00`0 {000 {000 {000 {00@0 {00x0 {000 {000 {00 0 {00X0 {000 {000 {000 {0080 {00p0 {000 {000 {00 0 {00P 0 {01 0 {01 0 {01 0 {010!0 {0 1 h!0 {0 1 !0 {0 1!0 {01 #0 {01X#0 {01#0 {01#0 {01$0 {018$0 {01p$0 {01$0 {01 $0 {0!1"%0 {0#1$P%0 {0%1&%0 {0'1(%0 {0)1*%0 {0+1,0&0 {0-1.h&0 {0/10&0 {0112&0 {0314'0 {0516H'0 {0718'0 {091:'0 {0;1<'0 {0=1>((0 {0?1@`(0 {0A1B(0 {0C1D(0 {0E1F)0 00 {00y00y00{00{00{00{00 00  %%%( -%8BW`6t7uf`*#̳v+>4t[Fv3s( #&*/69A (7If`JjBN[)X +@1Y>Wyh Iu~Im q    F!Y!!!F""" #B#|###$#$N$q$$$$%@%j%%%%&)&U&&&&A'Z'''')(q(((()i)))*1*K*t*****+4+N+++++,9,U,o,,,,,/-b-h--".z..../C////;0000(1f11112%2?2p22222343M33333434O4h44444$5B5H5A (      !"#$%&'()*+,-./012345689:;<=>?@ABCDEFGHJKLMNOPQRSTUVWXYZ[\]^_`abcdeghijklmnopqrstuvwxyz{|}~ ( R Y *1-4F!!!~%%%b)i)z)~..1555::'<@AKB'H.HJOOPKVRVD]cce6l=lp|&W")Iti}#%2FHNbdWkm24 9QQQQQQQQQQQQQQQQ::QQQ::::::::::::::Q !(! !g% :Q  b$4m0 u  }b$ۢS5<&#b$(PxY>L%b$E%IUjZHb$|Ȩ)Yuu.[ 'b$4˺h.f,b$)tRjmG:8Jb$XN@.`jb$pB6b$DJoJ@ 0(  B S  ?H0(   92 _Toc214451589 _Toc228185197 _Toc228185198 _Toc214451592 _Toc228185199 _Toc214451593 _Toc228185200 _Toc214451594 _Toc228185201 _Toc228185206 _Toc228185240 _Toc228185241 _Toc214451599 _Toc228185208 _Toc214451604 _Toc228185213 _Toc214451605 _Toc228185214 _Toc214451606 _Toc228185215 _Toc214451607 _Toc228185216 _Toc214451610 _Toc228185219 _Toc214451612 _Toc228185221 _Toc214451613 _Toc228185222 _Toc228185273 _Toc228185250 _Toc214451615 _Toc228185224 _Toc228185257 _Toc214451618 _Toc228185227 _Toc228185254 _Toc228185255 _Toc228185258 _Toc228185259 _Toc214451620 _Toc228185229 _Toc214451621 _Toc228185230 _Toc214451622 _Toc228185231 _Toc228185280 _Toc228185281 _Toc228185263 _Toc214451625 _Toc228185234//XXXmmsss JJ))'+@9H<EggZZ!)m 9  !"#$%&'()*+,-./01&0&0XmmrrMzMzIJpp((yy@@ݩ@VVǻ'.Cpp'gH 93܆T3;3;2;3;1p2;q<T1;2;T2;1;p q:T3;l\EDD pl pD䔍#pDpT=ave p4 CL9<<ov,St~:lD FFqsssss*[GGqq""""##N#N###$$@$@$****++H+H+++++5,5, 9      !"#$%&'()*+,-./012HHssssss7hMM""""*#*#S#S###$$H$H$****%+%+M+M+++ , ,=,=, 9  !"#$%&'()*+,-./0129)*urn:schemas-microsoft-com:office:smarttagsState8!*urn:schemas-microsoft-com:office:smarttagsCityV3*urn:schemas-microsoft-com:office:smarttagsplacehttp://www.5iantlavalamp.com/B2*urn:schemas-microsoft-com:office:smarttagscountry-region w3222232323)3332323!32323)3332323!32323)3332323!32327=z11115B:B=BCBI$IIIJJOOeeeeee!')/IO"( !,/4:@      H Y   8 >     +/`i 9A^denoxMST]^g`fgl067<$$$$%%%%&&H,E0Q0W6g67788888888888888 9DLҟԟacgj ,1v{' x  : ?   =     7     kn89|236H,6.8.3p488888888888888 93333333333333333333333333333333333IID^ccce6l7l=lq| &ssPPQQX"#)Ji&2eWn5G,H,H, - ----/03m4p45888888888888888 9H,88888888888888 9|Zz}4(~ /w@Nv7V̥U=p l(?&TU Pd)t|&6)2Ol=3Tdg!9R^}DrB9kF'"JNVc-0VYp^`.^`.^`.^`. ^`OJQJo( ^`OJQJo( ^`OJQJo( ^`OJQJo(hh^h`. hh^h`OJQJo(h ^`o(hH.^`OJQJ^JOh+'0 $ D P \ ht|0LOSS GIVEN DEFAULT MODELLING FOR MORTGAGE LOANSml303Normalmleow8Microsoft Office Word@ @ct@bX@T,wDocumentSummaryInformation8x`CompObj~q՜.+,00 hp  University of Southampton + 0LOSS GIVEN DEFAULT MODELLING FOR MORTGAGE LOANS Title  FMicrosoft Office Word Document MSWordDocWord.Document.89q      !"#$%&'()*+,-./01o(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH ^`OJQJo(n ^`OJQJo(n pp^p`OJQJo(n @ @ ^@ `OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n PP^P`OJQJo(n ^`OJQJo("  ^`OJQJo("  pp^p`OJQJo("  @ @ ^@ `OJQJo("  ^`OJQJo("  ^`OJQJo("  ^`OJQJo("  ^`OJQJo("  PP^P`OJQJo("  ^`OJQJo(n ^`OJQJo(q pp^p`OJQJo(n @ @ ^@ `OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n PP^P`OJQJo(nh^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hH^`OJPJQJ^Jo(-^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH ^`OJQJo(n ^`OJQJo(n pp^p`OJQJo(n @ @ ^@ `OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n PP^P`OJQJo(nh^`OJQJo(hHh ^`hH.h pLp^p`LhH.h @ @ ^@ `hH.h ^`hH.h L^`LhH.h ^`hH.h ^`hH.h PLP^P`LhH.^`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hHh ^`hH.h ^`hH.h pLp^p`LhH.h @ @ ^@ `hH.h ^`hH.h L^`LhH.h ^`hH.h ^`hH.h PLP^P`LhH. ^`OJQJo(n ^`OJQJo(n pp^p`OJQJo(n @ @ ^@ `OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n ^`OJQJo(n PP^P`OJQJo(nh^`OJQJo(hHhpp^p`OJQJ^Jo(hHoh@ @ ^@ `OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHhPP^P`OJQJ^Jo(hHoh  ^ `OJQJo(hH&d)"JNdg!9-0VY^}D9kF l~}|6)2l=3         bĜxq*ښpr /"I^SfO`8<f4`DӺt΂4tҔiv_s.t+@把N9X ފ         D z        <┅@ Z^X[Д:VɊXښ                           $2yj\'X.)FtJ,j9\-         r9wG.-~yHzpS4* lz[~; 3 >a%\ gk~*IG  lz[$ j6)"aeK]xDh|6E~F%\ @ CUY>a&z>a.,8TD-,8Ta2Q#Cc%\  >a&\F8'gk~*[+Pa,)"ai-)-/`b0G2)"a{2!4lz[E6>ac6T7wG9tY:H+:=@OY=$=W?lz[*%? 5?0@.A=kbB,8TcB CJD!,E(^EHs'fH6J7UkM5M,8TJN'LQlz[JR)"aGT,8TFUk6U,8TNV)"aTdV[]W)"a@JYaEZ4[lz[vS]>6_}Y``lz[)"aHzp>aGT?mb)"akTc9Sd>a8Fe>aifhajkAl>a7l $o1[cp%\ p utc8{Tt vlz[$+v>a*vm-zNzc8{^{lz[{s9|)"arO~>a} FQY /D0A.HENx y'<e"x|}M& K +efC.jm~4gEVR]R;PX % 6A [ e Z ! % 5 >>  N T IV >[ v^ c  )E F q   X$ + . A vf q} ]_lmu [Y) *$c)O?jTT']O| 88Q!9l\/`}$)|B_kf)yh$*k/Kefqst,n.?;v XC)`oo&N*@iqm&68Dkm/r1s)#+{}}ET(3H%X`!{IKQZTv%3&bOzlt{n@]sv{z} + - P !!i+"7"^"5g"-x"-#I#m#.$C$pF$EP$P$%r+%?%,k%q%z%~%i&|&O%&c&u&'&9'E (Y(Hj(9)qf)Ig)F*S*V*:\*}*g++4+U+?m+m+{+,e,s,/z, ---)-Q-UT-s-.7.L:.XR.Zr.x/r //]p/0o 0i%070P0W0l0~04?1{K1cc1Hr1u1122202J\2l2n2\v233D3Og3r3~3%4?44<4D4v4Q5535k5~5i6cZ6g6~67 7)7E,7W78r*8I68A8C8<L<T<yU<l^<@f<s<&=a=:a>?*?=F?@@.@|P@n@mAQAAMAElA }AB\B!B9B;ABJB[BlB.mBxmBCCB@CcQCy DD^(D OD_DlDEE[6EAEmMEDXEjEF6F NFf G+HBHl^HwHCzH>IHIRInIuI>J)K)KPKTKYK5]KhKlKC}K LLCL'L@LBLFLRL|LfMR MMB7MoBM4JM NZ N!>NBNENcNiN{9OQOwO4:P&WPhcP(DQUWQR .R&7R]RjR{SSxSS6SCSFS~SZ#T(T1T4T KTfTmTIuU;VVHVPVW3W6WTWoWsrWWGXscXpX2Y6YGYmVY`YtYuYuYHgZ*[;[f>[f[k[\6\=\SB\-S\V\j\\8]M]J^7^J>^X^__C4_w_4`v``g``U4a=a_aaaKoa>ua]bVbjjbVmbc{EcX]c4tcucs{cd|deee"e?ereff!f)6fifqf gMg0gC1gMKgYXgZgh9(hmAh8Bh!Yhoha|hhPiM iiQ"i,i3i>i@RiSi?jOjg_jjq4k5k6kRkjkxkl/l~rlB}lm@mPm_mamjmnmS npna n"n'nHncncnsunEn:?GL /D[ad JRoU-kJnwS2=dEH[Q6jbOheIx 1N`9hhhu~?K`YD)# HN@hi1mq~.*4A>H??OryRW]41TP\}%u z&kMQ} !'%7ap |7]qhy4XdutL .12Otia w="gm1%<G\`%hj"<GTTEWh r1:}e3M*U&-t *#8&,+x0yT&],;BV%}S"(OITH"Q"/fez:OWrSu$9=DG\L?Nj ,00=^Cvwq"!^QMm 2hRcd B2c0Z>IVJa8uzza4 KI?w$DX!-->@viLbb-M^@|O4ZjzDFaYB./%0>"AOS8_v '.Ijj:kv94Zr1WD.Qc '{(6>ARRWhm5'JM`~CXvd 'EE#KUdi@yi)4<ak=q L8&IR~\U&15+L%lp s?sGuh(2.ObZd#)BNR^kKWhE_LyM:VWJe/$6IV\l(u2MmCB#_hiiu'T^8",iWZ^et!4?7CwF/R&V_.6C?+I96Q^6w8gnA $C$6IPfz <F:;:KJ>TaW/XR[_[`Hhh2* N" ?kd*C g 15ndDg/+<Ib?SZqUeGfgJL'rtsF4RV2kt'v oRV\dxEz(.zR[j^r!uCHQ)<Yr`~bk]ORn||G%8]n3tBut5E 6''*-O =a?A]~dte1-4Q>rWo,]...Pam ;%V0;LFL[i -z1NOfQQyZd+D*r.0y4i5v6S5V.`hh8{:Qb'"5JMP\)t)x+Le$EQ\ ,29Ly./7V \{t5( C2_GG8F aRc}cMXX}z ANh MS*`DEIvR[t}(iP"Jl|/m9<NEzJ!8R}~mީPV\cjqrԪڪ !"BIJKLMNO۫[\moŬǬ)* %+,-cl%)@A]_12[]9: %flpt!hx|,28>?7@U^eoyz&/DMT^hisu|  =CIOUVguwy{}~ Xav.5;CIJ[ikmoqr|9BW`gmqr|~  FHJLNOY3:@FLMX &,289Bbiou{|  #)/56@NTZ`fgqz!"*9@FLRS[jqw} #)/56?NU[aghq '07=ABLNTZ`fgl#)/56Aqx~ " ( . 4 5 @ i p v |             !!!!!1!8!>!D!J!K![!l!n!p!r!t!u!!!!!!!!!!!!!!!!!!!!!!" """"%"&"."4";"A"G"M"N"V"m"t"z""""""""""""""""""""# #####,#3#9#?#E#F#N#U#\#b#h#n#o#w#####################$$$#$)$/$0$@$^$`$b$d$f$g$h$$$$$$$$$$$$%% %"%(%/%5%6%7%q%z%%%%%%%%%%%%%%%%%%&&&&*&1&7&=&C&D&O&y&&&&&&&&&&&&&&"')'/'5';'<'G'w'~''''''''''''' ("($(&((()(:(Y(`(f(l(r(s(|((((((((((((((((() ))))%),)2)8)>)?)J)X)^)d)j)p)q){)))))))))))))))))))))) ****%*&*.*4*:*@*F*L*M*U*l*r*x*~********************+ ++++'+-+3+9+?+@+H+O+U+[+a+g+h+p+{+++++++++++++++++++++ ,,,,$,%,5,>,@,B,D,F,G, 9m<33330333333333333<333333EN.InstantFormatu EN.Layoutv EN.Librariesxw<ENInstantFormat><Enabled>1</Enabled><ScanUnformatted>1</ScanUnformatted><ScanChanges>1</ScanChanges></ENInstantFormat>T<ENLayout><Style>J Banking Finance</Style><LeftDelim>{</LeftDelim><RightDelim>}</RightDelim><FontName>Tahoma</FontName><FontSize>11</FontSize><ReflistTitle></ReflistTitle><StartingRefnum>1</StartingRefnum><FirstLineIndent>0</FirstLineIndent><HangingIndent>720</HangingIndent><LineSpacing>0</LineSpacing><SpaceAfter>2</SpaceAfter></ENLayout>N<ENLibraries><Libraries><item>Southampton.enl</item></Libraries></ENLibraries>@G,G,QG,G,49 9@<@P@UnknownGz Times New Roman5Symbol3& z Arial5& zaTahoma?5 z Courier New;SimSun[SO;Wingdings"1hr&c&L,w .% 74d+68 :QXR 2/LOSS GIVEN DEFAULT MODELLING FOR MORTGAGE LOANSml303mleowd