A Business Intelligence System for Risk Management in the Real Estate Industry

Transcription

1 A Busness Intellgence System for Rsk Management n the Real Estate Industry Raul Valverde John Molson School of Busness Concorda Unversty 1455 de Masonneuve Blvd Montreal, Qc, Canada ABSTRACT The purpose of ths work s to buld a Busness Intellgence System (BIS) for the real estate ndustry. The BIS be useful for rsk managers nterested n real estate nvestment. The paper dentfes the man rsk management related decsons that real estate professonals made on a daly bass. Rsk management decson models are selected from an extensve lterature revew. These models are ncorporated nto an nformaton system. A set of reports that wll support rsk management real estate decsons are desgned and mplemented n a prototype bult to demonstrate the potental of ths technology n the real estate ndustry. The database of the nformaton system s populated wth a sample of data collected from the real estate market of the state of Calforna USA and results were generated to llustrate how reports can be used to support real estate rsk management decsons. Keywords Fnancal Informaton Systems, Data mnng, Real Estate, Busness Intellgence, Fnance, Portfolo Management, Rsk Management. 1. INTRODUCTION A Busness Intellgence System (BIS) can be descrbed as an nteractve, computer-based system desgned to help decsonmakers to solve poorly structured problems. Usng a combnaton of models, analytcal technques, and nformaton retreval, such systems help develop and evaluate approprate alternatves. BIS should focus on strategc decsons, not operatonal ones. More specfcally, they should contrbute to reduce the rsk faced by managers when they need to make decsons regardng future optons. It s n ths lght that we have chosen to develop a BIS to assst real estate admnstrators n fnancal rsk management decsons. In the last two decades the rsk management fnance theory has been appled to the real estate decson makng process but wth not much effort of the busness or academa to develop a rsk management BIS that ntegrates ths knowledge, and transmts the theory to real-world practce. Property managers face every day crtcal rsk management decsons as determnng the prce for sell or rent of a property, choce of fnancng, nvestment analyss, real estate portfolo management, real estate valuaton. In these cases a BIS can be very valuable n order to mnmze the rsk of potental losses due to wrong decsons. The objectve of ths research project s to desgn and develop a rsk management BIS usng exstng decson models to assst managers n real estate decsons. Followng, the decson process wll be analyzed and the models to support the decsons wll be dentfed. Fnally, wth the use of nformaton technology the BIS wll be desgned and mplemented n order to satsfy the needs of the real estate ndustry. 2. LITERATURE REVIEW In the real estate ndustry are two dstnct but nterrelated markets: the market for tenant space and the market for nvestment captal. The decson makng n each market s dfferent snce the space market s more concerned wth the use decson whereas the nvestment decson s made n the captal market. In ths research, the focus would be the nvestment captal market. In the captal market, decsons are made based on nvestment analyss, n whch the prmary objectve s the maxmzaton of the captal return gven a certan rsk, or gven a return, to mnmze the rsk. In order to acheve ths objectve, nvestors use decson support models. The decson support models for captal market have emphass on modern portfolo theory. Portfolo Theory assumes that unsystematc rsk s reduced by ncludng real estate n a portfolo, and perhaps by dversfyng across economc areas and property types. The model assumes that are no constrants on the amount of any gven asset class. In practce these models are hard to use because real estate n partcular lacks prce data on frequent transactons that are comparable to prce data for publcly traded stocks and bonds. Investors tradtonally have thought of equty real estate as an neffcent market n whch the key to success s n the skll wth whch an ndvdual nvestment s selected and negotated. The general approach seems to be buyng propertes when they become avalable f they look lke "good deals", wth lttle regard for the equally mportant ssue of how acquston fts wth other holdngs n the portfolo and what effect, f any, t wll have on the overall rsk and return objectves of the portfolo. Investors need to assess how the real estate segment fts nto ther entre portfolo. Ths means settng rsk and return objectves for the equty real estate portfolo as a whole that are compatble wth the goals for the nvestor's entre portfolo. Devsng a strategy for achevng these objectves and evaluatng the extent to whch ndvdual transactons conform to the strategy and are lkely to further portfolo objectves Real estate dversfcaton s the key n real estate nvestment. In dversfcaton the combnaton of property type and economc regons affect the rsk and return characterstcs of a portfolo. Spreadng assets geographcally has great affect n rsk f the dversfcaton s done across economc ndependent regons. 14

2 In the Captal Market the most mportant decson that needs to be supported s how much to put nto dfferent categores of assets and the overall rsk level of the portfolo. In order to support ths decson, real estate portfolo decson models need to be mplemented. These models should acheve hgher-than average levels of return, and the nvestor must construct a portfolo nvolvng greater-than-average rsk. It also should be possble and useful to measure the rsk and return and to develop, n an approxmate manner, a portfolo strategy that balances the trade-off between these two performance crtera The total rsk on any nvestment can be decomposed nto systematc and unsystematc component. Portfolo decson models focuses on the unsystematc rsk that wll largely dsappear as an nfluence on the return of a well-dversfed portfolo. The rsk from changes n economc condtons throughout the country s systematc and wll nfluence any portfolo, no matter how large and well dversfed, because t nfluences each of the parts. The captal market may have a lack of actve management of real estate assets, along wth changes n the envronment surroundng corporate owned real estate that may result n sgnfcant value that s undetected by managers and nvestors alke. The potental hdden value n real estate s a functon of: changes n captal market condtons, changes n frm prospects, n utlzaton of real estate, taxes, changes n accountng, and n factors affectng agency costs of the frm [1]. Changes n a frm's operatng prospects are due ether to changes n return prospects or changes n rsk exposure havng a smlar prospect. The expected return on real estate may justfy ownershp (as opposed to leasng) of real estate n perods when the frm prospects are poor, but ths relatonshp can reverse when frm prospects mprove. The magntude of the value ncrement created by changes n the use of a frm's real estate obvously depends on how much value the new use adds to the property. The potental to ncrease frm value suggests that frms should consstently revew the performance and the value of ther real estate. A corporaton must understand how ts real estate holdngs are affectng ts total market value, n order to determne how to utlze ths asset. Ths requres a valuaton model of real estate wthn the corporate settng. Also, there s a need of decdng whch fnancng alternatve s the most approprate. On decdng whch alternatve s best we should have n mnd the objectves for corporate real estate management suggested by Mles et al. [1] whch are: Cash Generaton Takeover preventon More effectve utlzaton of the tax laws Mnmzng agency cost The use of real estate fnancng as a market sgnal Playng the local real estate market by usng the corporate advantage generated by the corporaton's longtme horzon. Mantanng flexblty gven the frm's current and expected space needs, flexblty s an mportant consderaton whle pursung the objectves above Snce the most common choce of fnancng s the mortgage [2], t s mportant to analyze the most mportant components n the mortgage choce. A basc component of mortgage choce s the loan-to-value decson, the choce of mortgage nstrument and the decson to prepay or default. Prepayment may be caused by a household's decson to move, by refnancng to obtan a lower rate, or to obtan more money. Follan [2] argues that any model of mortgage choce should be consstent wth several stylzed facts about mortgage choce. In the case of Loan-to-Value Decson (LTV) the followng facts must be taken n to consderaton. LTV for the populaton as a whole s qute low, around 30%; ndeed, over one thrd have no mortgage at all. LTV among recent home buyers are much hgher LTVs declne rapdly wth length of stay n the house LTVs among the elderly are specally low Lendng rules lmt mortgage payment to ncome ratos Follan [2] fnds that the only relevant varables n the LTV models are length of stay, age and ncome. He fnds that there s a postve relatonshp between ncome and LTV whereas the LTV decreases as age and length of stay ncrease. Concernng the facts that affect the Mortgage Instrument Choce [2] he proposes that any model that supports ths decson must take n to account that: The fxed-rate-mortgage works well durng nonnflatonary tmes and t remans extremely popular. The surge n adjusted-rate-mortgage was demand drven, beng strongly related to movements n the level of nterest rates Demand n ARM's ncreases modestly wth ncome and length of stay. Short tmers prefer the use of ARMs The demand for other nstruments besde ARM's and FRM's s low In Refnancng and Default payment decson models the followng facts must be consdered: Refnancng does not take place as rapdly as pure opton prcng models would suggest. Default seems to depend upon LTV Foreclosure s relatvely uncommon An mportant aspect of rsk management s dversfcaton. Dversfcaton s akn to not puttng all your eggs n one basket. Lots of academc research shows that a dversfed portfolo reduces rsk for a gven rate of return Wllams [3] was the frst to suggested that the greater the relatve balance of return from operatng and reverson, the more dversfed the portfolo, and thus the better the portfolo performance. Later Kwame Addae-Dapaah [4] found that the assocaton between the cash flow concentraton level and the portfolo performance ndex, and between the dversfcaton ndex and the portfolo performance ndex was stronger than depcted by Wllams[3]. Norman, Srmans and Benjamn [4] suggested that real estate s mportant for portfolo dversfcaton, although there s no consensus on the optmal amount; studes lookng at longer tme perod hold more percentage of portfolo n real estate. There were further studes about the role of the real estate n a 15

3 portfolo nvestment together wth other nvestment such as bonds and stocks. The studes also nvestgated how economcal factors affect the allocaton between dfferent type assets [6]. There are many studes nvestgatng questons lke how we are gong to dversfy? ; what are the dmensons of dversfcaton?. Ths opens topc of cluster analyss, whch parttons observed cases nto relatvely homogenous groups to produce an operaton [7]. Abraham, Goetzmann and Watcher [8] have been usng ths n real estate market. The object of such analyss s to sort observatons nto groups, called clusters, so that the degree of smlarty s hgh among members of the same group and low among members of dfferent groups. Ths nvolves two consderatons: (1) whch varables to use n the classfcaton; and (2) how to calculate smlarty between the observatons. For varables (dmensons) to use n the classfcaton, the most popular ones are geographc and property dversfcaton. For property type categores, they dffer among research, nsttutonal nvestment and government sources, and these descrptons are generally references offce, warehouse, manufacturng, retal hotel and multfamly property types [8]. Whle property type should not be the only dversfcaton strategy used, ths study fnds that property type allocatons may enhance nvestor returns over real estate market and/or economc cycles [9]. For geographc clusterng, The studes was carred across many countres such USA [10,11,12], UK [13], and Chna [14]. Many fndngs support the noton that homogenous geographc groupng for optmzed dversfcaton must recognze the relatve mportance of the underlyng economc factors for dfferent property markets. Although there are clear sgns of geographc nfluence, the clusters are not at all synonymous wth pure geographc regons of contguous states [15]. Veser evaluated "Wthn Real Estate" dversfcaton strateges [16] by addng controls to the expermental desgn to determne why one strategy s superor to another. The best strategy used sxteen dmensons (four property types n four geographc regons). In our study, our decson staton are usng smlar strategc dmenson. After dscussng dmensons of dversfcaton and how to cluster, another mportant theme s optmal dversfcaton model. Tradtonally everythng s based on Markowtz Portfolo Theory (MPT) [17]. Markowtz consdered each nvestment s return as beng represented by a probablty dstrbuton of expected returns over a perod of tme. The rskness of each asset s based on the varablty of these expected returns. Although the portfolo s mean return s a weghted average of the ndvdual asset s mean return, the portfolo s rskness cannot be calculated the same way due to the correlatons between each asset s returns[18]. Multple and less than perfectly correlated assets, allow mnmzaton of unsystematc rsk and choce of exposure to systematc rsk for any gven level of return. Portfolo dversfcaton reduces unsystematc real estate nvestment rsk [19]. Effcent market theores suggest that nvestors are not rewarded for bearng unsystematc rsk and therefore should seek to mnmze ts presence. There are dfferent ssues on MPT. Gold [19] beleves that asset allocaton n ntervals rather than pont estmates n tradtonal MPT make t easer to manage a real estate portfolo. Ths accounts for the fuzzness of the real estate effcent fronter. Because of ths fuzzness, t would be possble that an effcent portfolo developed by meanvarance analyss may not be any more effcent than a navely dversfed portfolo. Cheng and Lang [20] nvestgated that n practcal terms, effcent portfolos may not be superor to navely dversfed portfolos n a statstcal sense usng an effcency test developed by Gbbons, Ross, and Shanken [21]. Tradton MPT doesn t take nto account of the tme varant behavor of varables of the real estate market, ths leads to development of stuated decson supportng system n our study. Another problem MPT ncludes fracton of asset but real estate asset s not dvsble. We need to modfy MPT to accept dscrete nput behavor, ths leads to development of the Fndlay model [22]. 3. METHODS AND REALIZATION The objectve of our system s to support rsk management decsons n the real estate ndustry. The prmary users of the system are nvestors; they wll use the system to support ther nvestment decsons as the knd of property to nvest n order to maxmze ther return. They wll use the system to montor the performance of the corporate buldng and to support how to fnance the property and the amount of prepayment. The proposed system wll address the decsons below: How much to put nto dfferent categores of assets Select the optmum overall rsk level of the portfolo What categores of assets to nclude n the portfolo When to sell an asset n the portfolo Choce of fnancng Amount of prepayment Value of property The mathematcal models to be used n our DDS are: Fndaly model [22]: To estmate the assets to nclude n the nvestment portfolo, ths model was selected due to the fact that assumes that the assets ncluded are ndvsble, whch s the case n real lfe. Fndlay [22] has proposed a portfolo model for the real estate that works fne for portfolo szes lmted to ten assets, because most unsystematc rsk reducton occurs over the frst ten assets. The objectve s to mnmze the overall portfolo rsk. Z ( e) Where Z(e) = The objectve functon that calculates the varance of a portfolo of ndvsble assets. X = Ether 0 or 1, representng the decson to ether nvest or not nvest n the asset, C Cj mn M M 1 j1 ( X M j1 = The cost of the th asset, = The cost of the jth asset, C C j X j J C j j X J ) (1) 16

4 j j = The standard devaton of the th asset = The standard devaton of the jth asset = The correlaton coeffcent between assets I and j, and CjXj = The total portfolo outlay M= Number of assets The Tucker model [23] s used to decde whch fnancng choce to use. The model was selected for ts powerful capablty of predctng whch method (fxed rate or adjustable rate) has the better probablty of better performance. The model suggests than n an effcent market, borrowers wll select the least expensve mortgage. Snce the mortgage most common choces are the adjustable-rate (ARM) and fxed-rate mortgage (FRM), t concentrates n these two choces. Tucker [23] based hs model on the hypothess that the probablty of borrowers selectng ARMs over FRMs ncreases as the nterest-rate margn between the two types of mortgages ncreases, and decreases wth ncreases n one-year constant maturty Treasury bll rates, and ncreases wth FRM rate ncreases. Logt analyss was used to determne f fnancal varables are sgnfcant n determnng borrower selecton between fxed-rate and adjustable-rate mortgage. The probablty of a borrower selectng s specfed as equaton 2: 1 P (1 Exp Z= T - BILL INTDIFF CPI (2) Where: P T-bll = CPI = Probablty of selectng ARM over FRM, One-year constant maturty-treasury bll rate = Percent change n consumer prce ndex from one year pror, INTDIFF= (FRM nterest rate) (ARM nterest rate). The Mles et al. performance model [1]: To model the performance of a corporate property n terms of cost of captal, the model was selected gven ts capablty of determnng f the property s not beng underutlzed. Mles et al. [1] suggests that frms should consstently revew the performance and the value of ther real estate. A corporaton must understand how to utlze ths asset, and ths requres a valuaton model of real estate wthn the corporate settng. The model assumes that the cash flow after taxes comng from the frm s constant and that the frm has an nfnte number of years. Equaton 3 shows the value of the frm ncludng the real estate property. Where: VR n1 0 ( Z) Da( T ) ( Da RS) n (1 Kw ) (1 Kw ) Da = S - ( VC * US + FC ) n1 Db ( (1 Kw ) (3) Vc = Value of the frm at today; Da T = Cash flow of the company before taxes ncludng real estate; = Taxaton rate; Kw = Cost of captal; n RS = Number of years that real estate; = Real estate estmated sell value after n years; Db = Cash flow of the company wthout real estate; S = Gross ncome before taxes of the frm; VC = Varable cost; US = Unts sold; FC = Fxed cost. The Hartzall model [24]: To value a property n terms of nvestment. The model was selected gven the avalablty of the requred data n the publc nformaton system. Hartzell [24] begns wth the premse that the rate of ncrease n real estate ncome s a functon of the nflaton rate modfed by lease structure, real supply and demand condtons, and the degree of product enhancement or deteroraton that occurs over tme. In ths form, real estate can be vewed as a bond whose prncpal s nflaton-ndexed and whose coupons range from zero to full ndexaton. Thus, the prce of real estate can be reduced to the followng equaton. (Current Property Value) = (Present Value of Net Rents Over Next T Years + Present Value of Expected Market Prce n T years) T NR P0 ( ) t (1 K0 ) Where: P0 = Present value of future cash flows generated by the property; T = Term of lease; NR = Net rental ncome on lease (fxed over T years) NR0= Current level of market rents; g0 = Current expected growth rate n property (4) t1 NR0 1 E g0 u1 1 g0 u2...1 g0 ut value, whch reflects the expected Economy wde nflaton rate; (1 k0 ut = Unexpected growth rate n rents n year T that reflects unexpected nflaton, local supply T ) 17 XMT

5 120 Internatonal Journal of Computer Applcatons ( ) and demand mbalances, as well as obsolescence and enhancements, whch are nterrelated wth local market condtons; MT = Prce-to-rent multple n year T; k0 = Current requred rate of return; and E[..]= Expected property value n T years The Goldman prepayment mortgage model [25] s also ncluded gven ts powerful capablty of predcton of the prepayment rate for a mortgage. The model captures four mportant economc effects: 1. The refnancng ncentve; 2. Seasonng or age of the mortgage; 3. The amount of the year (seasonalty); and 4. Premum burnout. The Refnancng Incentve: Ths measures the refnancng ncentve as the weghted average of recent values of C/R, the mortgage coupon rate dvded by the mortgage-refnancng rate. For values of C/R below one, the homeowner s prepayment opton s out-of-the money, and the refnancng ncentve s relatvely small. Conversely, when C/R exceeds one, the coupon s above the refnancng rate, and the ncentve to refnance ncreases dramatcally. Seasonng: It s well known that mortgage prepayment rates rse from very low levels at ssue to much hgher levels as the mortgage age. The model captures the nteracton between seasonng and coupon by makng the seasonng effect a functon of the mortgage s current C/R. Fgure 1 shows the relatve seasonng effects for a dscount pool wth C/R = 0.8, a par pool wth C/R = 1.0 and C/R = 1.2. %SEASONED 100% Age n Months Fgure 1Seasonalty effect Coupon = 0.8 Coupon = 1.0 Coupon = 1.2 Goldman [25] captures the effect n a seres of curves, but he doesn t say how to model t mathematcally. The model seems to have certan lnearty and although t s dffcult to determne a unversal model, a lnear mathematcal model can approxmate the effect that may be used for decson support purposes. The lnear model that captures ths effect s approxmated usng the equaton 5. Ths equaton assumes that most of the tme the term C/R s more than 1, snce a C/R of less than one s not attractve to analyze due to the fact that the prepayment n ths case would be zero. S = Where S= Seasonng factor; L = Length of mortgage; C/R = The mortgage coupon rate dvded by the A = mortgage refnancng rate; Age n months of the mortgage. The basc archtecture of the proposed BIS s shown n fgure 2: Fgure 2 The BIS Archtecture The report generator s used for customzed presentaton of knowledge n terms of templates (comprsed of forms). It can help decson-makers quckly spot patterns, trend, and proportons that exst n data The reports that our BIS wll generate are: (L/A) C/R (5) Fnancng choce report: It wll suggest the best fnancng choce (ARM or AFM), and the probablty assocated wth t. Performance Report: It wll show how the corporate buldng s used n terms of proftablty and wll suggest actons to be taken n order to mprove proftablty of the corporaton related to the buldng. Investment report: It wll suggest the assets to be bought n order to acheve a mnmum rsk, and the expected level of return of nvestment for ths choce. Investment property valuaton: It wll show the nvestment potental value of a property. Amount of prepayment for a mortgage: It wll suggest the percentage of prepayment for the mortgage. Fgure 3, shows the state transton dagram of our graphcal user nterface desgn. 18

6 User selects Performance Perfomance Report Report Securty Auyhorzaton Corporate Real Estate Performance System User selects cancel or close Logn successful Real Estate DDS System Investment Database Management System User selects cancel or close User selects portfolo management User Selects Fnancng User System selects cancel or Close User selects close or cancel Fnancng DDS System Portfolo Management System User selects Fnancng Report User selects cancel or close User selects User selects Portfolo report cancel or close Portfolo Report Fnanng Report Fgure 3 State transton dagram for the graphcal user nterface 4. RESULTS AND DISCUSSIONS A prototype of the DDSS was mplemented n Vsual Basc.Net. The system s man menu s shown n fgure 4. The man menu s organzed n four major functonal areas: Investment decson support utltes, fnancng decson support tools and performance evaluaton support tools. The nvestment decson support utltes are composed of the nvestment database and the portfolo management system. The nvestment database keeps all the data needed to calculate the real estate nvestment portfolo. The database (Fgure 5) has three felds: Locaton, rate and type. The locaton feld keeps the geographcal place of the property (NW, NE, SW, SE, Center), the rate keeps the captalzaton rate on a year for the property and the type keeps the type of property (ndustral, mxed-used, retal, offce). Fgure 4 System's man menu The database has the capablty of updatng, deletng and addng new data. In order to test the system, the database was populated wth 300 records for dfferent property types and places for the estate of Calforna U.S for the perod of

7 ths case the suggested prepayment s 8.25% of the left mortgage captal. Fgure 5 Investment database The portfolo management system receves as nput several nvestment opportuntes (no maxmum lmt) and the user has to enter the locaton, type and prce for each property. The system wll calculate the Fndlay [3] rsk for all possble nvestment portfolos and t wll report the portfolo wth the lowest rsk. Fgure 6 shows an example wth thrteen propertes as nput, where the suggested portfolo was composed of three assets wth a Fndlay rsk of 0.16 out of the maxmum possble rsk of A new set of 4 propertes were are added and the DSS selected a new asset that gves a fnal result of four assets n the portfolo wth a rsk of out of the maxmum rsk of 12.23, whch supports the Fndlay theory that the more assets you have n the portfolo the lower the nvestment portfolo rsk. Fgure 7 Prepayment module The fnancng module suggests what knd of fnancng (ARM or FRM) shall the user chose n order to get a better rate. The module also gves the probablty of the accuracy of the decson. The nput varables are the regon CPI, the ARM rate, the FM rate and one the year Treasury bll rate. In the example of fgure 8 the CPI s 11%, the ARM rate s 6%, the FRM rate s %8 and the treasury bll rate s %4, for ths example the module suggests the ARM mortgage wth a success probablty of Fgure 6 Portfolo management system wth four three assets output The fnancng decson support system s composed of the fnancng and prepayment systems. The prepayment system helps to decde the amount of prepayment n percentage of the mortgage when the user negotates the mortgage at a lower rate than the orgnal rate. The module needs the orgnal mortgage coupon rate, the mortgage-refnancng rate, the length of the mortgage, the age of the mortgage at the tme of the negotaton and the month when the payment s done. Fgure 7 shows an example where the mortgage orgnal rate s 12%, the refnancng rate 11%, the length of the mortgage s 240, the age s 60 and the month of payment s March, n Fgure 8 The fnancng module The performance decson support tools are composed of the nvestment valuaton and corporate performance study modules. The frst evaluates a property n terms of nvestment, gven a property value, yearly rent, yearly operatng costs, property number of years gvng cash flow, resale value after the gven number of years and the desred rate of return. The module evaluates the nvestment and reports the net present value of the property, whch can be used for decson purposes. Fgure 9 shows an example of ths module, where the property value s $105,000, yearly rent $680, yearly expenses of $3000, number of years of 25, resale value of $40,000 and desred rate of return of %10, gvng a result of net present value of - $122,561. In ths case the system suggests that the nvestor should not buy the property. 20

8 Fgure 9 Investment valuaton module The corporate performance study module calculates the dfference of the net present value of the frm wth and wthout the real estate. In fgure 10 the module calculates the net present value of a dfference of a lawyer frm wth an annual ncome of $1,500,0000, annual fx cost of $300,0000, tax rate of %40, buldng value of $700,0000, buldng lfe of 25 years, resale value of $270,000, average cost of captal of %6, estmated annual rent of $60,000, estmated annual operatng cost of $20,000. The frm rents an offce of hs buldng to an accountant for $5,000 a year. For ths example the module calculates the dfference at -$283,183, for whch t suggests to sell the property and rent a equvalent buldng snce the sell wll add net present value to the frm. Fgure 10 The corporate performance study module 5. CONCLUSIONS Two dfferent knds of markets for the real estate ndustry were dentfed: the space and captal markets. For the captal market, the decson process was analyzed, and the requrements to support each decson were specfed. In order to support the decsons, several decson models were analyzed and those that satsfed the requrements the best were selected as part of the BIS. A BIS was desgned accordng to the specfcatons and t was mplemented usng the Vsual Basc programmng language and the Mcrosoft Access database. The system s database was populated wth data from Calforna U.S.A and the system was tested wth dfferent examples and satsfactory reports were obtaned from these smulatons. Durng the project the most mportant research n real estate decson support was covered and several technologes were combned as artfcal ntellgence, data mnng and database management systems for the purpose of supportng the decsons made n the real estate ndustry. Several lmtatons were encountered as the avalablty of relable data and the lttle research n the topc. On the other hand the Fndlay portfolo management theory gave good results, the same for fnancng and performance modules although there was not enough tme to measure performance gven tme constrants. Fnally there s a need for more research n the topc, snce there s great opportunty to mplement BIS n the ndustry gven the great benefts that were presented n ths research. 6. REFERENCES [1] Mles, M., J. Prngle and B. Webb, Modelng the Corporate Real Estate Decson, Journal of. Real Estate Research, 4(3), pp 47-66, [2] Follan, J. R., "Mortgage Choce" Amercan Real Estate and Urban Economc Assocatons Journal, 18(2), , [3] Wllams, J. E., Real Estate Portfolo Dversfcaton by Sources of Return, In A.L. Schwartz Jr. and S.D. Kappln (Eds.), Alternatve Ideas n Real Estate Investment, Boston, MA: Kluwer Academc Publshers, [4] Kwame Addae-Dapaah*, Ser Gek Wee**, M. Shahd Ebrahm***, Real Estate Portfolo Dversfcaton by Sources of Return, Volume 8, Number 1, Journal of Real Estate Portfolo Management 2002 [5] Norman, E. J., G. S. Srmans and J. D. Benjamn, The Hstorcal Envronment of Real Estate Returns, Journal of Real Estate Portfolo Management, 1995, 1:1, [6] Benjamn J.D., Srmans G.D and Zetz, E N., Returns and Rsk on Real Estate and Other Investments: More Evdence Volume 7, Number 3, the Journal of Real Estate Portfolo Management [7] Glberto M., Hamelnk F., Hoesl M., and MacGregor B., Optmal Dversfcaton wthn Mxed-Asset Portfolos usng a Condtonal Heteroskedastcty Approach: Evdence from the U.S. and the U.K. Journal of Real Estate Portfolo Management, 1999, 5:1, [8] Abraham, J., Goetzmann W. and Wachter S., "Homogeneous Groupngs of Metropoltan Housng Markets," Journal of Housng Economcs 3(1994) [9] Anderson Arthur & Company, Managng the Future, Chcago: Insttute of Real Estate Management, Fall [10] Mueller, G. R., & Laposa, S. P. (1995). Property type dversfcaton n real estate portfolos. The Journal of Real Estate Portfolo Management, 1(1), [11] Cheng P. and Black R.T, Geographc Dversfcaton and Economc Fundamentals n Apartment Markets: A Demand Perspectve :2, [12] Theron R. Nelson, and Susan L. Nelson, Regonal Models for Portfolo Dversfcaton, Volume 9, Number 21

9 1, 2003 of the Journal of Real Estate Portfolo Management. [13] Wllams J.E., Real Estate Portfolo Dversfcaton and Performance of the Twenty Largest MSAs, Volume 2, Number 1, 1996 of the Journal of Real Estate Portfolo Managemen [14] Weng S., Huang Y., Potental for Portfolo Dversfcaton Across Chna's Real Estate Markets, AsRES 2004 Conference [15] Png Cheng and Roy T. Black, Geographc Dversfcaton and Economc Fundamentals n Apartment Markets: A Demand Perspectve :2, [16] Veser T., Evaluatng "Wthn Real Estate" Dversfcaton Strateges, Volume: 6, Issue Number: 1, 2000 of Journal of Real Estate Portfolo Management [17] Markowtz, Harry M. (1952). Portfolo selecton, Journal of Fnance, 7 (1), [18] Tucker W. Francs, Real Estate Investment Thess A Real Estate Portfolo Optmzer Utlzng Spreadsheet Modelng wth Markowtz Mean-Varance Optmzaton And Monte Carlo Smulaton, Johns Hopkns Unversty, 2000 [19] Gold R, Why the Effcent Fronter for Real Estate s "Fuzzy", 1:1, 59-66, 1995 of Journal of Real Estate Portfolo Management [20] Png Cheng, Youguo Lang, Optmal Dversfcaton: Is It Really Worthwhle?, 6:1, 7-16, 2000 of Journal of Real Estate Portfolo Management [21] Gbbons, Mchael R., Stephen A. Ross, and Jay Shanken, 1989, A test of the effcency of a gven portfolo, Econometrca 57, [22] Fndlay, M. C., Hamlton, C. W., Messner, S. D. and Yormark, J. S., Optmal. Real Estate Portfolos, AREUEA Journal, 7 (3), , [23] Tucker, M. Adjustable-Rate and Fxed-Rate Mortgage Choce: A Logt Analyss. Journal of Real Estate Research, 4(2), pp , [24] Hartzell, D. J., Shulman, D. G., Langeteg, T. C. and Lebowtz, M. L. A look at real estate duraton, Journal of Portfolo Management 15(1), 16-24, [25] Rchard R. and Roll R. 1989, Modelng prepayments on fxed-rate mortgage-backed securtes, Journal of Portfolo Management, 15(3), pp