Michael Sklarz Michael Sklarz Pr Predictiv edictive Methods Confer Methods Conference ence May 2010 May 2010
History of Predicting Home Prices � Nearly 25 years ago, Profe Nearly 25 years ago, Professo sor No r Norm Miller and I w rm Miller and I wrote ote a a pape paper on predicting home pric r on predicting home prices using various market-base es using various market-based indicators. indicato rs. � I am happy to report that I am happy to report that afte after all this time, it is now r all this time, it is now possible possible to do this for most loca to do this for most local U.S. housing markets. l U.S. housing markets.
Key Developments Enabling Better Predictions � Access to key data el Access to key data elements incl ements including uding ◦ Richer, and more timely, sales data Richer, and more timely, sales data ◦ Listing Prices Listing Prices ◦ Shadow inventory Shadow inventory � Expired an Expired and Withdrawn d Withdrawn Listings Listings ◦ Months of Inventory Remaining Months of Inventory Remaining ◦ Day Days o on M Mark rket et ◦ Active Listing Prices versus Active Listing Prices versus Sales Prices and SP/LP Ratio Sales Prices and SP/LP Ratio ◦ Improvement in Home Pr Improvement in Home Price Index methodologies ice Index methodologies ◦ Ability to do macro and micr Ability to do macro and micro level home price forecasts o level home price forecasts
Sales activity is the most basic leading indicator of Sales activity is the most basic leading indicator of home prices home prices Escondido CA Higher prices Lower prices Increasing sales Price bottom Decreasing sales Increasing sales
Housing market cycles have actually been quite orderly Peaks have been about every 10-12 years Note how Note how Sa San Diego n Diego si signifi gnificantly overshot antly overshot in in the most recent the most recent cycle cycle and and has has since corrected since corrected Data Source: Moody’s/Economy.com
Sold Market Time is an excellent leading indicator Plotted inversely
Months of Inventory Remaining is another powerful leading indicator which leads prices by anywhere from 6 to 18 months Price Percent Change Months of Inventory
San Diego Median Price and Sold Market Time Data Source: Moody’s/Economy.com and CAR
These leading indicators work on all geographic levels and property types Price Percent Change Sold Market Time
Combining supply with demand data provides much more insight to the overall and internal conditions of a particular real estate market Thousa Th sand Oa nd Oaks CA Sin ks CA Single le Fa Family Cu mily Curre rrent List t Listings ings a and Past Year Sa d Past Year Sales By Price/Livin les By Price/Living � Area Area High Liquidity Low Liquidity
Months of Inventory Remaining can also be used to stratify market condition by price range in a particular market Th Thousa sand Oa nd Oaks CA Sin ks CA Single le Fa Family mily � Very weak Ve ry weak ma marke rket Ab Above $350/S e $350/Sq Ft Ft
New Listing Price is the best measure of the current direction of home prices – widely followed HPIs are typically reported with a two month lag and really reflect prices from 3 to 4 months ago New list price declining New list price increasing Sold prices flat, but declines shortly thereafter Sold Prices rise shortly thereafter
There is a definite need for micro-market data and HPIs Sign Significan ificant Var Varian ances
REO and Foreclosure sales have a significant and diverse impact in many markets Sacramento Single Family REO Discount
It is possible to create very granular REO discount factors
Defining and identifying home price bubbles 150% increase over 5 years rule
Home price declines drive mortgage delinquency and default – start with state level HPIs for current home values declines in decline in home prices home prices lea lead t to g grea eater ter delinquencie delinquencies and and default default Data Source: Lewtan ABSNet HomeVal
Better measures of home prices lead to better predictions of mortgage delinquency and default – CBSA level HPIs for current home values Data Source: Lewtan ABSNet HomeVal
Better measures of home prices Better measures of home prices lead to better predictions of lead to better predictions of mortgage delinquency and default – mortgage delinquency and default – Zip level HPIs ip level HPIs fo for current r current home home values values Data Source: Lewtan ABSNet HomeVal
Using Address Matchi Using Address Matching and AVMs to cal ng and AVMs to calcul ulate indivi ate individual home dual home values do values do the the best job fo best job for predicting mo r predicting mortgage rtgage delinquency and delinquency and default default Data Source: Lewtan ABSNet HomeVal
GSAMP 2007-NC1 Original and Current LTV Distributions of Active Loans Original LTV Distribution of Active Loans 2500 2000 s n a o f L 1500 r o e b 1000 m u N 500 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 1 1 1 1 1 2 2 2 2 2 3 3 3 3 3 4 LTV Range in Percent The significant move to the right, with higher LTV’s Is not good Data Source: Lewtan ABSNet HomeVal
A typical RMBS is quite complicat A typical RMBS is quite complicated - ed - note the wide variati ote the wide variation in on in constituent home price levels constituent home price levels and performances prior to and and performances prior to and after o afte r origination rigination It should not have been that difficult to see that late 2006 was a bad time to be originating risky mortgages • 4754 Zip Codes Originally • 3236 Zip Codes Currently
Very ba ry bad ti d time to to origi originat ate risky e risky secu securitie rities Data Source: Lewtan ABSNet HomeVal
Relationship between declining home prices and default is clear Data Source: Lewtan ABSNet HomeVal
Recommend
More recommend
