RISK ASSESSMENT AND MEASUREMENT METHOD AND SYSTEM FIELD OF INVENTION
The present invention relates to the financial and banking sectors of business. In one form, the present invention is a residential real estate (RRE) capital value risk identification and estimation tool. In other form(s), the present invention is applicable to any financial instrument such as, property, shares, etc. against which a financial facility, for example a loan, might be granted and which financial instrument, fluctuates/grows in its value or market price. The present invention, in one form, has been developed for use in the financial sector, particularly in the field of banking, but also has application in mortgage securitisation, mortgage insurance and other areas where the identification and measurement of the probability of capital loss and/or negative equity for residential real estate is important. BACKGROUND OF INVENTION The traditional residential real estate (RRE) risk assessment tool for lending and mortgage insurance is the Loan to Value ratio (LVR). The LVR is defined as the mortgage amount at loan commencement divided by the security value at loan commencement. For a number of years, any loan that exceeds a Loan to Value ratio (LVR) of 80% is considered to be "risky" and generally requires mortgage insurance. Loans written at less than a 80% loan to value ratio are considered to be "safe". Lenders and Mortgage insurers alter policy from time to time, depending upon local geographic conditions and may consider not lending or insuring at any price. To date, these risk adjustments tend to be intuitive or semi-empirical. Banking and insurance are considered mature, highly competitive industries with little (considered) chance for differentiation in the market. Banks constantly strive to develop and offer new products and reduce costs.
Banks are in the process of reviewing the role of the traditional Valuer / Appraiser in the lending process in order to reduce cost and speed up process. Statistically derived, automated valuations are becoming more common in the USA and Canada. Australia is starting to follow this path. In essence, the current trend in relation to residential real estate risk assessment is to continue to use a loan to value ratio, but in a manner which is quicker and cheaper and which
works on the basis of a "portfolio" of risks. The portfolio of risks is believed to enable exposure to relatively risky loans to be offset with exposure to relatively less risky loans.
The concept of negative equity is considered important in residential financial areas including lending, insurance securitisation, and investment. Negative equity is the term used when the value of a property falls below the outstanding loan amount of the property. It has been clearly demonstrated that negative equity often becomes a trigger for loan delinquency, default and mortgage insurance claims. Standard and Poor's notes that "the major" causes of default in Australia are generally considered to be:
• Loan affordability
• Loss of income
• Personal crisis OR any of the above accompanied by declining property values.
According to Robert Schiller (1996,-Journal of Real Estate ResearchNol.7, Issue 2, p248), one of the world's most well respected finance researchers and practitioners:
"there is no shortage of evidence on the importance of house prices and equity in the default decision".
Quercia and Stegman (1992, Journal of Housing Research, Vol.3, Issue 2, p375) reviewed 29 empirical studies done over a 30-year period and concluded:
"Consistently, home equity, or the related measure of loan to value ratio, has been found to influence the default decision. There is consensus in most recent default studies that the correct measure of a borrower's net equity is the contemporaneous market value of the property less the contemporaneous value of the loan "
Kau, Keenan and Kim (1994, Journal of Urban Economics, Vol.35, Issue 3, p287) reached the same conclusion: "There exists a significant literature examining the causes of default. In conformity with this paper's approach, considerable empirical evidence exists showing that it is the house versus the mortgage value, rather than such personal characteristics as the homeowner's liquidity position, that explains default."
In a study using a discrete proportional hazard model, micro-level mortgage data from Freddie Mac, and weighted repeat sales price indices, Quigley, Van Order and Deng (1993, the Competing Risks for Mortgage termination by Defaults and Prepayments in the Residential Housing market, paper presented at National Bureau of Economic Research Summer Institute, Cambridge, MA, USA, at p24) stated:
"The results show that the probability of negative equity ratio is the main time varying covariate influencing mortgage holders default decisions"
In a recent investigation into mortgage defaults in Australia, Berry et. al. (1999, Falling Out of Home Ownership, Mortgage Arrears and defaults in Australia, University of Queensland Press, ISBN: 1 875997 32 6 at p.37) summarized findings:
"mortgage default is viewed from an option taking perspective in which the level of owner's equity, captured by the LVR, is the key causal factor. There is increasing agreement that the correct measure of equity is the difference between current market value of the dwelling and the current market value of the mortgage, rather than the level of equity implicit in the initial LVR." SUMMARY OF INVENTION
It is an object of the present invention to provide a method, system and computer program product which enables capital risk identification and an associated estimation tool therefore.
The present invention provides a method, system and computer program product as outlined in the specification and as defined in the attached claims.
Furthermore, the invention as outlined in the specification and as defined in the attached claims provides a method of providing a financial facility for a transaction involving a financial instrument.
The present invention has come about by realising that the loan to value ratio is a gross generalisation in respect to residential real estate risk. In coming to realise the present invention, research suggests that residential real estate sub-markets, measured at a very fine level of disaggregation, have very different risk profiles. An example of a sub-market is a real estate market specified in terms of geographic location (zip/postcode), type (house / non house) and price (subdivisions /categories). Equally, it should be understood that sub-markets
may be determined for any other financial instrument, such as shares for example, where it is suggested that such sub-markets, measured at a very fine level of disaggregation, have very different risk profiles
In most cases, the 80% loan to value ratio marker is considered a gross overstatement or understatement of the real sub-market risk. In addition, the measure of risk for a given sub-market is dynamic, and will tend to change depending upon underlying factors such as economic cycles, price levels and/or other economic and non-economic events.
In addition, it has been realised that the real residential real estate or other properties risk occurs not only at the time of arranging the financial facility but is significant for a period of years. That period varies on financial conditions and the terms of the mortgage, but 5 years is considered a reasonable period for domestic mortgages. During this period, the loan principal has usually been reduced by a relatively small amount, whereas property markets and economies may have changed significantly. It is during this 3 to 5 year period from the commencement of the loan that the residential real estate security risk exists. The present invention is directed, in one aspect to the identification and measurement of the probability of "negative equity" or the outstanding loan amount exceeding the value of the financial instrument at any time during this period. It is considered that this is the real measure of risk, not the initial loan to value ratio(LVR).
The present invention has a number of advantages, which include: That it will allow users in the finance industry to improve their risk assessment in the RRE market. In turn, this will allow them to better direct their business and minimise losses. Another benefit of the present invention is to allow increased returns on a risk adjusted basis. That it will enable estimation of the probability of negative equity (value less than outstanding loan balance) at a user set confidence level for a period 3 to 5 years in the future. This calculation can be undertaken for any financial instrument, for example, residential property type or value in any location, where sufficient historical data is available. An example output may be as follows: Example 1 - Ballarat decile 1 (lowest 10%) houses:
1. To be 95% confident of no negative equity in this residential real estate sub-market, lend only up to 89% (approx) of the current value.
2. There is a 23.3% chance that this residential real estate sub-market will fall below its current value within the next 5 years. 3. There is a 76.7% chance that this residential real estate sub-market will not fall below it's current value within the next 5 years. Example 2 - Toorak decile 1 (lowest 10%) houses:
1. To be 95% confident of no negative equity in this residential real estate sub-market, lend only up to 47.5% (approx) of the current value. 2. There is a 75.7% chance that this residential real estate sub-market will fall below its current value within the next 5 years.
3. There is a 24.3% chance that this residential real estate sub-market will not fall below it's current value within the next 5 years
In determining the present invention, the Applicant's research indicates that based on historical data, residential property sub markets specified in terms of geographic location (zip / postcode), type (house / non house) and price (subdivisions /categories) have:
• Substantially different risks of negative equity,
• Substantially different financial performance characteristics, • Different trend growth rates and,
• Different probabilities of market value rise or fall.
Because the present invention is forward looking i.e., looking at what might happen at any time in the next 5 years, and expressed in terms of probabilities, the results can be used to help any business involved in residential finance to: • Identify and measure detailed subdivisions of risk
• Manage and mitigate that risk,
• Set explicit risk policies and strategies,
• Direct business to where the greatest risk adjusted rates of return can be gained, • Provide for capital adequacy on a more accurate basis,
• Provide for VAR (Value at Risk) input,
• Move away from intuitive or semi empirical measures of risk.
Equally, the present invention may be applicable to many other financial instruments (property, shares, etc ...) against which a financial facility such as a loan might be granted and which instrument fluctuates/changes in value or market price. Most commonly, this would be property, but the principles could be applied generally. The type of investment against which the loan is granted would need to be able to be grouped into sufficiently homogenous subsets for analysis to be meaningful. For example, loans secured against business operations would need to be assessed against the on-going value of a business. Analysis of loans secured against a share portfolio of, for example, the All Ordinaries, may also be contemplated in accordance with the present invention. DESCRIPTION OF PREFERRED EMBODIMENT
A preferred embodiment of the present invention, which relates to residential real estate, will now be described. It is to be borne in mind that the present invention also has similar application to many other financial instruments or properties: 1. DATA :
An embodiment of the present invention, directed at residential real estate uses residential real estate historical sales data that is sourced from either the
Valuers General office's or from private data resellers. In the USA, data is more freely available than in Australia and thus it should be easier to obtain the necessary data.
In the first instance, historical sales data from as many years as possible is required. 10 years would probably be a minimum. In Victoria, we are using 16 years of data.
Because the data is from different sources, it contains a variety of fields, formats, and is of varying quality. The data needs to be standardised in its format, fields, and the like before they can be agglomerated into one data file. The required data fields are price, date of transaction, property type (house or flat/unit) and postcode. Data for land sales are not analysed.
In Victoria, there are 1.6 million sales recorded between 1985 and 2001. In all, the model would use about 6 million records of base data and each record would have 4 fields.
The database associated with the present invention would be updated periodically and probabilities re-calculated. Each year, another 500,000 records
could be added. If updated on a quarterly basis, 125,000 records would be added.
Historical output from the present model would also be collected so that, in future, the dynamics of the RRE markets could be displayed and analysed. 2. DATA CLEANING AND CHECKING:
The data would be cleaned and checked using a purpose built, rules based algorithm. We wish to identify and remove mathematically intractable, inconsistent, illogical and substandard observations, outliers, incomplete sets etc.
A preferred set of rules may reflect that the data is reviewed to ensure that it consists of values that could represent a single house or flat at a given point in time, and that it may provide a reasonable basis for analysis.
2A. Initially the data is reviewed to ensure that the number of sales per annum provided for a dwelling type and suburb is greater than 20. The basis for this is that the property prices are divided into deciles. With less than 20 observations, the value of a decile can be too heavily influenced by a single data point/sale. In our research, this test omitted the most data in comparison to the other criterion below. Out of 58,180 observations, 329 were omitted, as there was too few observations to maintain statistical accuracy or relevancy.
"Points" may be just sales data values, by year, by Flat/House, and / or by suburb. They may be the individual data records.
2B. The values must be between the set ranges. For our purposes we have set the low log price to 9 ($8,103) and the high to 17 ($24,154,953)
In the same data set as in 1., 100 prices were deemed to be too small, and
62 too large. 2C. The price must be within a ratio of ± 2 from the nearest price. Note that the logged prices are estimated to 1 decimal place. The above criterion means that the value can be no more than 7 log-points from the last observation.
For example, the last logged price observed for Boronia houses was 13.4.
The next observation is 14.3. There are (14.3-13.4)/0.1 = 9 logged price points between the two observations. This equates to a ratio of 2.46 which is outside the range. The price difference in dollar terms is $963,343.
2D. There must be at least 3 contiguous points when the data are grouped in bands of 0.1 by log-price. This ensures that the points are not too scattered.
Data is tested for sufficient residual numbers to determine if calculations will be statistically significant or the whole set should be identified and removed.
It has been determined that if the data is too sparse, we cannot offer a service in respect of that section of the market based on that sample having sufficient statistical accuracy.
3. DATA SPLITTING:
Data are split into their respective sub-markets. A residential real estate property sub-market is a transaction:
• in a given postcode • either a house or a unit/flat and
• within a yearly dynamic price quantile, (at this stage a decile) described by the total range of a given years sales data. This process involves sorting the data in ascending order, and then counting, for example, one-tenth of the sorted data to determine or get the first decile point. A useful and advantageous approach at present is to take the mid-point to represent the value for the range. For example, the value of the first 20th-quantile is taken to represent the value for the first decile.
When this process is complete, there are mathematically tractable data sets for some 150,000 property sub-markets across Australia. The historical performance characteristics of these property markets are now capable of being analysed, described statistically and boundary conditions set for the future based on their historical performance.
Values are obtained at the mid-points of the deciles, and these are taken as being proxies for the whole decile, and ratio values are obtained across years to get year-on-year rates of change in value of that decile. The "boundary condition parameters" are the mean, standard deviation and serial correlation values derived for such a data set, comprising the yields obtained for one decile in one house/flat set for one suburb.
This database will be unique in Australia and could form the basis of other useful work.
4. RATE OF CHANGE:
Decile mid points are calculated for each year for each sub-market and rates of change between each year's midpoints are calculated and stored. For
each sub-market we now have one datum point per year - 15 rates of change for 16 years. The rates of change are converted to the "force of interest" equivalents, by taking the natural logarithm of the 1 + the rate of change expressed as a decimal. For example, a rate of change of 10% is expressed as ln(1.1) = 0.09531.
These data point series form the basis of subsequent calculations : mean, standard deviation, and serial correlation values.
5. DESCRIPTIVE CALCULATIONS:
The 150,000 RRE sub-markets with 15 data points each have mean, standard deviation and serial correlations with 1 , 2 and 3 year lags calculated.
6. CALCULATION OF BOUNDARY CONDITIONS AND PROBABILITIES: We are using a 5 parameter algorithm at this stage, but in the future, and depending upon data available, more or fewer parameters may be used. The key outputs are expressed as : if a loan is granted at x% of valuation, what is the likelihood that the price will fall below the loan amount over the next 5 years.
Alternatively, if the lender wants there to be only a 5% chance that there will be negative equity over the next 5 years, how much should be lent as a % of sale price.
For example, with one set of parameters, if the lender is prepared to accept a 20% chance of negative equity at any time in the first 5 years, 80.6% of valuation could be lent. For a more conservative 5% chance, only 68.1% would be made available to the purchaser.
The process proceeds by assuming no capital repayments in the first 5 years, and using the Normal distribution assumption in conjunction with the parameter values calculated to derive the probabilities.
The outputs from this multi parameter algorithm are stored in a database with pre-defined fields.
The following description outlines one process of determining Loan to
Value Ratios (LVR), in accordance with the present invention, by assessment of the likelihood of property values falling to any given / predetermined level.
Obviously, other formulae or techniques could be used. Pairs of { } indicate the start and end of a given sub-process.
Determine or get the relevant yields from year to year, for a given sub- market criteria, such as, for each suburb, for houses/flats separately, and for each decile
where r
t,i is Yield in year t and the ith quantile;
Pt,i is the Consideration in year t and the ith quantile.
and calculate mean, delta, by
where ry is Yield in year t and the ith quantile; n is the number of years with 'calculated' yields;
Mean: Sum of all values divided by the number of observations Determine or get the relevant standard deviation and serial correlation parameter;
Standard deviation:
where s is the standard deviation of delta; n is the number of years with 'calculated' yields' X
j,i is the natural log of 1 plus the yield in year j and the ith quantile
Serial correlation parameters: these were approximated using a similar formula to below
-ΣLwfo - **--* -*) .
Construct the Cholesky matrix C, such that C'.C produces the matrix formed by s'.R.s where
s is the column matrix of standard deviations, and R is the square matrix of serial correlations. For example, for serial correlations po , pi and p2 construct as
R = 11 Ppoo Ppii p2
Po 1 Po Pi
Pi Po 1 Po
P2 Pi Po 1 arge number of simulations (e.g. 10000) { Set an arbitrary initial property value Construct column matrix X of the most recent yields in reverse order. For example, if the most recent log yields are y-i, y2, y3, construct as y3, y2, yi.O For each of five years {
Solve for z such that z=C" (X-μ) where μ is a column matrix of the mean returns. Derive z4 as a Normal(0,1) variate.
(We do this using the Box-Muller formulae, but any method would suffice)
Calculate simulated returns as μ + C*z
Retain the z values for all but the most recent year, and re-position the remainder so that they relate to one year earlier;
From the current year yield returned, calculate the revised estimate of a property value as last value times Exp(Current Year's return value).
} Take the lowest value of the property over the five year period
Divide the lowest value by the original property price at time 0 to obtain a price ratio
The % chance that a residential real estate sub-market will fall below its current value within the next 5 years is estimated accordingly. }
7. DATABASE:
Database Front End:
At this stage, the output data stored in the database can be interrogated, picked, re-arranged and re-expressed in a number of ways to answer many different types of questions.
An example given previously was: Example 1 - Ballarat decile 1 (lowest 10%) houses:
A. To be 95% confident of no negative equity in this RRE sub-market, lend up to 89% (approx) of the current value. B. There is a 23.3% chance that this RRE sub-market will fall below its current value within the next 5 years.
C. There is a 76.7% chance that this RRE sub-market will not fall below it's current value within the next 5 years.
Example 2 - Toorak decile 1 (lowest 10%) houses: A. To be 95% confident of no negative equity in this RRE sub-market, lend up to 47.5% (approx) of the current value.
B. There is a 75.7% chance that this RRE sub-market will fall below its current value within the next 5 years.
C. There is a 24.3% chance that this RRE sub-market will not fall below it's current value within the next 5 years.
Problems & Solutions:
The following is a summary of the current problems or limitations that various business types within the financial sector experience and how the present invention(LR) helps to solve these problems.
Similar industry relevant issues are applicable to mortgage insurance, securatisation and ratings processes. As the present invention may be embodied in several forms without departing from the spirit of the essential characteristics of the invention, it should be understood that the above described embodiments are not to limit the present invention unless otherwise specified, but rather should be construed broadly within the spirit and scope of the invention as defined in the appended claims. Various modifications and equivalent arrangements are intended to be included within the spirit and scope of the invention and appended claims.