1. Economic Time Series Modeling of U.S. Housing Prices
Master of Engineering | Financial Engineering Project
May 11th, 2009
Vikas Garg Yasin Khan Shirley Lu Timothy Roberts Yi Fan Tang Wesley Tillu

2. Agenda Introduction State Results Modeling Methodology
Forecasting Methodology
Conclusion

3. What Happened? Housing Market Decline Mortgage Market Decline
Broader Economic Meltdown
State Results
Introduction
Methodology
Forecasting
Conclusion

4. Purpose and Objectives
Housing prices drive Mortgage-Backed Securities (MBS) valuation
Devise profitable trading strategies
Objectives
Develop efficient and robust models to understand the housing price process
Forecast the evolution of housing prices over medium to long term horizon
Shirley
1 In billions of US dollars. 2 Assumes unrated MBS bonds are written down completely and ABX prices are applied to the respective outstanding MBS volumes. 3 As a percentage of par
Sources: ABSNET.net; JPMorgan Chase; UBS; BIS calculations.
State Results
Introduction
Methodology
Forecasting
Conclusion

5. Housing Price Index (HPI)
What is it?
Measures movement of single-family house prices
Quarterly, state-level data from 1975 – 2008
Updated based on repeat-transaction methodology
Published by the US Office of Federal Housing Enterprise Oversight (OFHEO) based on Fannie / Freddie data
Advantages
Government collected, state-level
Publicly available
Disadvantages
Jumbo loans >$417,000 and risky subprime sales not included
Continually updated with every repeat-transaction
vikas
State Results
Introduction
Methodology
Forecasting
Conclusion

6. Effect on Americans Number of Negative Equity Mortgages by State
More than 20% of American homeowners owe more on their mortgage debt than they can sell their homes for, according to an industry report released Wednesday.
The real estate Web site Zillow.com reported that 21.8% of all U.S. homes, representing more than 20 million residences, were in a “negative equity” or “underwater” position after prices dropped more than 14% nationally in the year ended March 31.
Implications
Falling prices make it more difficult for homeowners in financial trouble to refinance or sell their homes
Inability to take advantage of lower interest rates
Many borrowers took out home-equity lines of credit at the time of purchase and have fully tapped the amount they can borrow
Effects of Downward Home Prices
20% of U.S. homeowners owe more on a mortgage than their homes are worth
~78.2 million owner-occupied single-family homes
Increased difficulty for homeowners in financial trouble to refinance or sell homes
State Results
Introduction
Methodology
Forecasting
Conclusion

7. Distribution of States
States by foreclosure rate
vikas
Source: HotPads.com, 05/09
Introduction
State Results
Methodology
Forecasting
Conclusion

8. Selected States for Discussion
High foreclosure rate (>1 in 150)
Arizona
California
Florida
Mid-level foreclosure rate (>1 in 600)
Texas
Low foreclosure rate (>1 in 20,000)
Montana
Vikas
Qualitative analysis of states before quantitative methods
Introduction
State Results
Methodology
Forecasting
Conclusion

9. Arizona
yasin
Phoenix and its suburbs are among the hardest hit by the foreclosure crisis. Avondale, just west of Phoenix, doubled in residents between 2000 and Now residents are unable to sell their houses except at fire-sale prices. - WSJ
Introduction
State Results
Methodology
Forecasting
Conclusion

10. Arizona
Appraisers valued the house at $132,000 for a $103,000 mortgage before the housing bubble burst. Under foreclosure the house sold for $18, WSJ
Introduction
State Results
Methodology
Forecasting
Conclusion

11. Arizona – HPI Introduction State Results Methodology Forecasting
Conclusion

12. Arizona – Outlook
Town of Maricopa: 75% of all homeowners owe more on their mortgages than the current value of their homes (national estimate of 18%)
Government will only help you refinance your mortgage if you owe between 80% and 105% of your home's value
Arizona is one of the worst hit areas where homes prices have plummeted far below the level of many mortgages
Federal stimulus plan up to $8,000 for anyone who hasn't owned a home for at least three years
wes
Introduction
State Results
Methodology
Forecasting
Conclusion

13. Arizona – Outlook
"The foreclosure problem in Arizona is only going to get worse," - Fred Karnas, the new director of the Arizona Department of Housing
Currently the state with the nation's third-highest foreclosure rate
Recent drop in building permits for homes across metropolitan Phoenix
State Economy has dependence on three industries that led the nation into recession: construction, housing and financial services
Increase in home prices and sales likely won't happen until 2012
Source: AZCentral.com. Real Estate Outlook
Introduction
State Results
Methodology
Forecasting
Conclusion

14. Arizona Model Forecasts Introduction State Results Methodology
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

15. Arizona – Forecast (Simple Returns)
HPI simple return
Forecasted Simple Return
95% Prediction Interval
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

16. Arizona - Forecast Introduction State Results Methodology Forecasting
HPI
Forecasted Index
2008 Q3 and Q4 Observed HPI
Introduction
State Results
Methodology
Forecasting
Conclusion

17. California Introduction State Results Methodology Forecasting
wes
Introduction
State Results
Methodology
Forecasting
Conclusion

18. California – Current Climate
Poor homeowner quality
Bad Sign: Pre-foreclosure notices up 80% in Q1’09
Delinquencies on dues to homeowner associations on the rise, some areas as high as 15%
Defaults up by ~35% in LA, San Francisco, San Louis Obispo this year
New home construction at 25-year low
Economic effects
Unemployment 11.2%, one of highest in nation
wes
Introduction
State Results
Methodology
Forecasting
Conclusion

19. California – Future Outlook
Signs of recovery slowly
64% increase in sales of single-family homes over prior year
Federal stimulus plan of $8,000 for first-time home buyers
State tax credit of $10,000 to purchase new unoccupied home
Homebuilder sales rose around 15% in the West
Continued concerns
Ending of moratoriums on foreclosures by Fannie / Freddie
Continued 11.2% unemployment
wes
Introduction
State Results
Methodology
Forecasting
Conclusion

20. California – Future Outlook (cont’d)
State tax credit discourages purchase of existing foreclosed homes
… although housing inventory remains stable
Construction is rising …
Total Residential Permits in California
Months Needed to Sell Existing Inventory at Current Sales Prices
Source: WSJ, 04/24/09.
Introduction
State Results
Methodology
Forecasting
Conclusion

21. California Model Forecasts Introduction State Results Methodology
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

22. California – Forecast (simple returns)
HPI simple return
Forecasted Simple Return
95% Prediction Interval
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

23. California – Forecast (HPI)
Forecasted Index
2008 Q3 and Q4 Observed HPI
Introduction
State Results
Methodology
Forecasting
Conclusion

24. Florida Housing Market
Florida HPI
Florida HPI Simple Returns
Introduction
State Results
Methodology
Forecasting
Conclusion

25. Florida – “It’s June in Miami!”
History repeats itself
1920’s housing bubble
Increasing demand
Booming tourism
Skyrocketing land prices
Exponential population growth
Plentiful credit and capital
Speculative investors
Starting in 1920, many Americans became enamored by the materialistic and prosperous lifestyle of the time. During this time, the stock market was moving forward at an extremely fast pace. Many investors were becoming quite wealthy. Florida became a hot spot for these newly rich people, who didn’t enjoy the cold. Many whole families took vacations to Florida. It was at this point that tourism started booming and land prices were skyrocketing. Many astute investors took notice and started buying Florida real estate. The population in Florida was growing exponentially and housing couldn’t meet the demand. Florida became the “playground of the rich and famous”. Illegal casinos and drinking parlors became widespread in Miami.
At this point, almost anybody could invest in Florida, even without much money. Credit was plentiful and soon everybody in Florida was either a real estate investor or a real estate agent. In 1922, the Miami Herald became the heaviest newspaper in the world as a result of its humongous real estate advertisements. People in the North heard about the real estate prices “doubling and tripling”, causing a snowball effect. Capital was rapidly pumped into the real estate market. Whole golf communities were developed, such as Temple Terrace. Resorts and retirement communities were developed almost overnight. Mansions were sprawling in every area, as were swimming pools. As always, waterfront property was the most desirable. Florida was seen as a veritable Utopia.
History repeats itself: Florida’s 1920’s housing bubble
Setting the stage for the Florida housing bubble
1920’s time of great prosperity and wealth
Hot vacation spot for newly rich
Booming tourism and skyrocketing land prices => Speculative investors
Exponential population growth => growing housing demand
Florida became the “playground of the rich and famous”
Plentiful credit
Everybody in Florida was either a real estate investor or a real estate agent
1922: Miami Herald became the heaviest newspaper in the world as a result of its humongous real estate advertisements
People in the North heard about the real estate prices “doubling and tripling”, causing a snowball effect
Capital was rapidly pumped into the real estate market
Entire communities were developed
Introduction
State Results
Methodology
Forecasting
Conclusion

26. Florida – early 2000’s
Prices there are up 33% in one year, up 105% over the five years, and 180.7% over the past decade
Speculative investors craze
Increasing adjustable-rate mortgages that allow speculative borrowers to delay the cost of their loans
Few markets have boomed like Florida. Prices there are up 33% in the past year, up 105% over the past five years, and have risen a honking 180.7% over the past decade.
examined the speculative craze in investing during the late-1990s dot-com boom and subsequent stock market crash. His second edition of Irrational Exuberance, published last month, shifts its focus to the current real estate market.
Shiller argues the runup in housing prices was spurred in part by people who got rich in the dot-com boom. The same national psychology of speculation that fed the stock market craze is now feeding the real estate boom.
He also warned of the flurry of adjustable-rate mortgage products that allow speculative borrowers to delay the cost of their loans.
"If a buyer can get an adjustable-rate loan that cuts in half the initial cost of today's traditional 6-percent, 30-year fixed mortgage, it means twice as many people can afford a house at the start," Rosenberg said. But eventually, those loans have to be paid of.
Introduction
State Results
Methodology
Forecasting
Conclusion

27. What’s Different About Florida?
(1) International investors
(2) Luxury million $ homes
Source: National Association of Realtors
(3) Vacation homes
Source: Harris’ Annual Poll
Introduction
State Results
Methodology
Forecasting
Conclusion

28. Florida Outlook – “First In, First Out”
Population growth
By 2010, FL forecasted to be the 3rd most populated state in the nation
FL’s population expected to increase about 75% by 2030
Estimated that 900 people move to FL every day
Unemployment
4th largest labor force
Unemployment rate historically below national average for past decade
Median Income
FL no state income tax
Extra boost to resident’s income
Signs of shrinking inventory
Although hundreds of foreclosed properties go on the market each month, the inventory of houses for sale is shrinking
Current 12-month supply of inventory, down from a peak of 30 months (3-6 months supply is typical of a healthy housing market
Investors
Steep price declines have encouraged first-time homebuyers with good credit, and lots of investors
Cash-rich international investors who are taking advantage of the low prices
In recent months, prices have begun to flatten out
"First in, first out" => 2005 was peak. is a popular phrase in the Florida real estate industry. In a state hit early by the economic downturn, there's hope it will be on the leading edge of a recovery. Construction and real estate play big roles in the state's economy. Emerging Signs Of Hope But those who follow the real estate market here are starting to see some signs of hope. The increase in sales has been especially notable in Cape Coral. The Florida Association of Realtors reported a 43 percent increase in January over last year's sales in Lee County. After a devastating 2007, prices began dropping in 2008 and the market responded. Steep price declines have encouraged first-time homebuyers with good credit, and lots of investors. In recent months, prices have begun to flatten out. Signs of A Shrinking Inventory And here's a positive sign: Even with hundreds of foreclosed properties coming on the market each month, the inventory of houses that are for sale is shrinking. At current rates, there's now just over a 12-month supply of houses for sale in Cape Coral. That's still higher than the three- to six-month inventory typical of a healthy housing market. But Jeff Tumbarello, who studies the market for the Southwest Florida Real Estate Investment Association, says it's becoming harder to find real bargains. Attracting Investors Matt Zacharias, a general contractor from Lancaster, Pa., had some luck. He recently bought a three-bedroom, two-bath home with a pool and its own boat dock. He paid $182,000. Sixteen months ago, the house sold for more than $500,000. "The market down here, the prices are just too good to be true," Zacharias says. "I had to come down and check it out for myself. One thing led to another and here we are — I'm actually going to look at another property later as another investment property." Affordable Housing Returns "Affordable housing is back. That also attracts business. It also attracts residents to come into our area. We should take advantage of that and we will." says Mayor Jim Burch Burch isn't quite ready to say that Cape Coral has reached the bottom of the housing market, but he believes it's probably not far away. "We do believe in the first in, first out theory: We were the first in, without a doubt and we think, right now, we're getting ready to be the first one to come out of this," he says. A Cautionary Tale Economist Hank Fishkind, like others who watch the housing market, notes it's not possible to know you've hit the bottom except in hindsight, after prices have begun to rise. He agrees though that Cape Coral is probably close to the bottom.
Introduction
State Results
Methodology
Forecasting
Conclusion

29. Florida Model Forecasts Introduction State Results Methodology
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

30. Florida – Forecast (simple returns)
HPI simple return
Forecasted Simple Return
95% Prediction Interval
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

31. Florida – Forecast (HPI)
Forecasted Index
2008 Q3 and Q4 Observed HPI
Introduction
State Results
Methodology
Forecasting
Conclusion

32. “Don’t Mess With Texas”
Texas HPI
Tim
“Don’t Mess With Texas”
Introduction
State Results
Methodology
Forecasting
Conclusion

33. Historical Mortgage Foreclosure Rate Recent Mortgage Foreclosure Rate
Texas
Texas Learned Their Lesson
“Flatland” vs. “Zoned Zone”
“The Two Americas” Paul Krugman
Historical Mortgage Foreclosure Rate
Recent Mortgage Foreclosure Rate
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) -> elastic housing supply -> prices determined by const of contruction, not cost of credit->less bubble prone
Introduction
State Results
Methodology
Forecasting
Conclusion

34. Texas Model Forecasts Introduction State Results Methodology
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

35. Texas Forecasts Introduction State Results Methodology Forecasting
HPI simple return
Forecasted Simple Return
95% Prediction Interval
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

36. Texas Forecast Introduction State Results Methodology Forecasting
HPI
Forecasted Index
2008 Q3 and Q4 Observed HPI
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

37. Montana Leads 2009 Top Housing Market Forecast
Introduction
State Results
Methodology
Forecasting
Conclusion

38. Montana – Outlook
Has one of the highest employment rates in the nation due to growth in the energy industry.
Most of Montana have seen a growing population and is now forecast to deflate a marginal 3.7% in
Unemployment rate is 5.6% in January, which is well below January's national average of 7.6%.
Bozeman has one of the highest numbers of home businesses in the country.
wes
Introduction
State Results
Methodology
Forecasting
Conclusion

39. Montana Model II. State Results Introduction State Results Methodology
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

40. Montana Forecasts II. State Results Introduction State Results
HPI simple return
Forecasted Simple Return
Standard Error
Introduction
State Results
Methodology
Forecasting
Conclusion

41. Montana Forecasts (cont’d)
II. State Results
Montana Forecasts (cont’d)
HPI
Forecasted Index
2008 Q3 and Q4 Observed HPI
Tim
Real estate crash in 80s triggered by
High population density + zoning restrictions (restricted land use) ->
Introduction
State Results
Methodology
Forecasting
Conclusion

42. Montana (ACF of Residuals)
Introduction
State Results
Methodology
Forecasting
Conclusion

43. Variable Relationships
Tommy
Negative
Positive
Neutral
Introduction
State Results
Methodology
Forecasting
Conclusion

44. Aggregate Empirical Results
Strength of variable relationships across states
Average variable effect on total variance
Direction of relationships consistent with theoretical values
Tommy
Introduction
State Results
Methodology
Forecasting
Conclusion

45. Modeling Approach Drift Model Volatility Model
Multiple linear regression
Predictors: Macroeconomic factors on both the supply and demand side
Response: Housing Price Index (HPI)
Requirements: Unit-root stationarity of the residuals
Equation:
Volatility Model
Model autocorrelation of residuals
Time series modeling using ARFIMA / GARCH
Equations:
Vikas
Introduction
State Results
Methodology
Forecasting
Conclusion

46. Drift Modeling Step 1 Step 2 Step 3
Vikas
Determine lags of predictor variables
Robust multiple linear regression
Stationarity & autocorrelation of residuals
Introduction
State Results
Methodology
Forecasting
Conclusion

47. Volatility Modeling Purpose Univariate Time Series Modeling
To model serial autocorrelations in residuals
To stabilize variance and volatility clustering from drift model
Original residuals should already be stationary
Univariate Time Series Modeling
Auto-Regressive, Moving Average (ARMA)
General Auto-Regressive Conditional Heteroscedasticity (GARCH)
Yasin
Introduction
State Results
Methodology
Forecasting
Conclusion

48. Fit ARFIMA/GARCH models Residuals are uncorrelated
Volatility Modeling
Step 1
Step 2
Step 3
Yasin
Correlated residuals
Fit ARFIMA/GARCH models
Residuals are uncorrelated
Introduction
State Results
Methodology
Forecasting
Conclusion

49. Modeling Challenges Data Regression Coefficients
Influential observations in the pre-1990 HPI Data
Accuracy of the data in this time period is questionable
Hindered stationarity of residuals
Removed using the Grubb’s test for influential data points
Frequency and Temporal Relationships
Linear interpolation used to obtain monthly estimates
Multiple “lags” of each variable tested
Regression Coefficients
T statistic may not be accurate indicator of statistical significance
Predictor may still explain a substantial amount of variation in the HPI
wes
Introduction
State Results
Methodology
Forecasting
Conclusion

50. Modeling Challenges (cont’d)
Constraints
Parsimonious model: single lag for each predictor variable
Coefficients make economic sense
Stationarity of residuals
wes
Introduction
State Results
Methodology
Forecasting
Conclusion

51. Forecasting HPI Drift Forecasting Volatility Forecasting
Examine trends in each of 7 macroeconomic variables
Determine key assumptions for each variable
Project out to 60 months
Assume relationships are the same as before
Volatility Forecasting
Based on ARMA/GARCH models built
Explains additional model volatility
Tim
Introduction
State Results
Methodology
Forecasting
Conclusion

52. Variable Forecasts / Assumptions
Unemployment Rate
Population Size
Median Income
Mortgage Originations
30-Year Commitment Rate
Building Permits
Foreclosure
Periodic trend
Linear growth
Grows at long term inflation
Zero for 24 months
Analyst forecasts (Global Insight)
Move about pre-bubble mean
Based on total foreclosures (CSFB)
Demand
Tim
Supply
Introduction
State Results
Methodology
Forecasting
Conclusion

53. Sample Forecasts - California
Tim
Introduction
State Results
Methodology
Forecasting
Conclusion

54. Conclusion Drivers of housing prices – summary
The bigger they are, the harder they fall
Introduction
State Results
Methodology
Forecasting
Conclusion

55. Acknowledgements Sponsor: Dr. Paul Thurston Advisors Program
Dr. David Matteson
Dr. David Ruppert
Program
Dr. Kathryn Caggiano
Victoria Averbukh
Selene Cammer

56. Questions

57. Appendix I: Model Coefficients

58. Appendix II: HPI - OFHEO vs. Case-Shiller
Purchase Price and Refinance Appraisals
Mortgage data from Fannie Mae and Freddie Mac
Equally weights all prices
Covers all states; important when some areas are growing rapidly
Purchase Price
Mortgage data comes from county records
Value-weighted, giving more weight to higher priced homes
No data from 13 states

59. Appendix III: HPI with S&P500

60. Appendix IV: Sources FOR CALIFORNIA
FOR Jumbo Mortgages
FOR Pictures