全球風(fēng)險(xiǎn)管理專業(yè)人士協(xié)會(huì)(GARP)致力于為風(fēng)險(xiǎn)管理?xiàng)l線上的各級(jí)人員，包括各大金融機(jī)構(gòu)的風(fēng)險(xiǎn)從業(yè)者和監(jiān)管機(jī)構(gòu)人員提供風(fēng)險(xiǎn)教育和最新行業(yè)資訊。

　　過(guò)去十年中，美國(guó)房地產(chǎn)市場(chǎng)緩慢復(fù)蘇。一年前，標(biāo)準(zhǔn)普爾Case-Shiller美國(guó)房?jī)r(jià)指數(shù)超過(guò)危機(jī)前的峰值水平，這是一個(gè)重要的里程碑。然而，整體上蓬勃發(fā)展的市場(chǎng)掩蓋了地方房產(chǎn)市場(chǎng)狀況的參差不齊。此外，抵押貸款評(píng)估的技術(shù)進(jìn)步帶來(lái)的風(fēng)險(xiǎn)疑云尚存。

　　最近，總部設(shè)在美國(guó)加利福尼亞州爾灣的房地產(chǎn)數(shù)據(jù)分析公司CoreLogic宣布，對(duì)美國(guó)前50大都市住房市場(chǎng)的分析中，發(fā)現(xiàn)有一半城市的房?jī)r(jià)被高估了。與此同時(shí)，在抵押貸款危機(jī)之后的幾年里，許多城市都在努力尋找立足點(diǎn)。異常低的庫(kù)存水平被認(rèn)為是導(dǎo)致丹佛、邁阿密和拉斯維加斯等城市房?jī)r(jià)高于長(zhǎng)期房?jī)r(jià)上漲趨勢(shì)10%以上的主要原因。

　　可負(fù)擔(dān)性問(wèn)題和經(jīng)濟(jì)放緩最終可能導(dǎo)致房?jī)r(jià)下跌。在這種環(huán)境下，傳統(tǒng)的房產(chǎn)抵押品評(píng)估方法和新的自動(dòng)評(píng)估模型(Automated Valuation Models, AVMs)正在進(jìn)行一場(chǎng)無(wú)聲的戰(zhàn)斗?？雌饋?lái)新技術(shù)似乎贏了這場(chǎng)斗爭(zhēng)，但它對(duì)抵押貸款行業(yè)可能構(gòu)成重大風(fēng)險(xiǎn)，因此值得研究。

　　評(píng)估業(yè)的“紅字”(Scarlet Letter)

　　The housing appraisal industry bears a stigma to this day for egregious breakdowns in the appraisal process during the boom years preceding the mortgage crisis. Well-documented practices of appraisal units embedded with loan production departments led to substantial over-inflation of property values that were exacerbated by lax credit standards on such critical risk factors as loan-to-value (LTV) ratio. Indeed, in the race for market share, lenders of the time fought for dominance in part by jettisoning best practices of appraisal independence.

　　In the years since the crisis, new standards regarding appraisal independence were introduced and, over time, appraisal management companies (AMCs) proliferated as intermediaries in the appraisal process. However, while these middlemen have improved the independence issue, they have simultaneously increased appraisal timelines and costs.

　　Moreover, the appraisal industry has not adapted to the times. The appraisal process remains cumbersome and largely manual, with a good dose of subjectivity, limited training on a national level and over-reliance on worn-out methodologies.

　　自動(dòng)評(píng)估模型的優(yōu)與劣

　　These deficiencies have ushered in a golden age of collateral valuation technology that has evolved in the years since the crisis. AVM models are not new to the mortgage industry.

　　In some cases, AVMs were used as a cost-efficiency tool at the expense of accuracy. As a result, a model's hit rate – or the instances in which a model estimate could be produced from the AVM – won out over accurate depictions of property value.

　　The statistical models behind AVMs rely, like all other models, on feeding them the right data and appropriate econometric specification. In the early going, AVMs were criticized for their inability to discern the marketability and condition of a subject property, as well as their ineffectiveness in assessing different properties with heterogeneous characteristics. Though these issues remain today, AVMs have become much more accurate over the years, as data and techniques have evolved and developers have begun tapping AI methods to augment their traditional data sources.

　　Advances in data and modeling – combined with the appraisal industry's crisis of confidence and ability to adapt to a technology-driven mortgage market – have naturally positioned AVMs as the future of mortgage collateral valuation.

　　自動(dòng)評(píng)估模型： 評(píng)估的靈丹妙藥還是長(zhǎng)期風(fēng)險(xiǎn)?

　　While the benefits are clear, this shift toward automation comes at a price. As AVMs continue to find favor among the Government-sponsored enterprise (GSEs), it puts enormous pressure on just two AVM models for the accuracy of industry collateral valuation for a significant number of properties.

　　When manual appraisals dominated the stage, a benefit of the process was that valuation errors were spread across thousands of appraisers. Unfortunately, this diversification effect was upended by the process integrity issues noted earlier. Moreover, we must now ask ourselves whether we have traded appraisal process deficiencies for AVM model risk in the long-run.

　　Eventually, all models deteriorate, and AVMs are no exception. What's more, the vast majority of loans originated today wind up in a GSE mortgage-backed security that may rely on either of the two GSE-developed AVM models. Taking all of these factors into account, we have the potential to create a systemic risk for the mortgage sector from over-reliance on just a couple of analytical tools.

　　Let me be clear: the risk today from GSE use of AVMs is small, but could grow over time as market conditions and collateral and credit policies change. One of the touted advantages of AVM models is their reliance on updated market data from local offices throughout the country. But in an abnormal housing market where prices are rising more quickly than expected (due to such forces as inventory constraints), an AVM might build abnormal pricing into its estimates of value, reinforcing market outcomes rather than reflecting long-term market fundamentals.

　　一些其他想法

　　AVMs have come into their own as accepted methodologies embraced by the dominant mortgage market participants, and that is unlikely to change. However, the broken appraisal process must also be fixed.

　　Improved oversight of the appraisal process is required, along with a focus on improving appraisal methods, data and training. The collateral valuation process cannot become essentially a two-AVM world.

　　If we have learned one thing from the last crisis, it is that the best processes fuse the human element with technology, providing proper checks and balances while leveraging the best of both man and machine.

　　Clifford Rossi博士是馬里蘭大學(xué)Robert H. Smith商學(xué)院的實(shí)務(wù)教授和駐校執(zhí)行官，也是切薩皮克風(fēng)險(xiǎn)顧問(wèn)公司(Chesapeake Risk Advisors, LLC)的負(fù)責(zé)人。 他擁有近25年的金融風(fēng)險(xiǎn)管理經(jīng)驗(yàn)，曾在幾個(gè)主要銀行機(jī)構(gòu)擔(dān)任過(guò)多個(gè)C級(jí)高管職位。 在擔(dān)任現(xiàn)職之前，他是花旗集團(tuán)北美消費(fèi)者貸款部門(mén)的首席風(fēng)險(xiǎn)官。本文有刪減。

　　金程推薦，FRM一級(jí)智能***計(jì)劃體驗(yàn)班 0元體驗(yàn)。學(xué)習(xí)素材+人工智能完美結(jié)合，為FRM學(xué)員通過(guò)保駕護(hù)航;權(quán)威名師嘔心瀝血，傾情打造;小班制管理，班主任督學(xué)，1對(duì)1私教。

FRM課程詳情在線咨詢

相關(guān)推薦：FRM考試科目 FRM科目占比 FRM成績(jī) FRM是什么

2019年FRM備考群 835405115 FRM資訊&資料隨時(shí)分享，與眾多FRM持證人交流考試經(jīng)驗(yàn)。

返回首頁(yè)