《FHFA：2025-2027年可負擔企業獨棟產業抵押貸款市場規模預測報告（英文版）（35頁）.pdf》由會員分享，可在線閱讀，更多相關《FHFA：2025-2027年可負擔企業獨棟產業抵押貸款市場規模預測報告（英文版）（35頁）.pdf（35頁珍藏版）》請在三個皮匠報告上搜索。

1、 THE SIZE OF THE AFFORDABLE MORTGAGE MARKET:2025-2027 ENTERPRISE SINGLE-FAMILY HOUSING GOALS December 19,2024 2 This Federal Housing Finance Agency(FHFA)technical report documents the statistical forecast models that the modeling team has developed as part of the process for establishing the afforda

2、ble housing goal benchmark levels for Fannie Mae and Freddie Mac for 2025 through 2027.The report was prepared by Omena Ubogu and reviewed by Jay Schultz,Samson Ramasamy,and Padmasini Raman.3 Introduction The Federal Housing Enterprises Financial Safety and Soundness Act of 1992(the Safety and Sound

3、ness Act),as amended,mandates that the Federal Housing Finance Agency(FHFA)establish annual housing goals for Fannie Mae and Freddie Mac(the Enterprises).1 Since 2010,FHFA has adopted a two-part approach to establishing and measuring the Enterprise housing goals.The“benchmark”level is set prospectiv

4、ely by rulemaking based on various factors set out in the statute,including FHFAs forecast of the goals-qualifying market based on the econometric models described in this paper.The actual market level is determined retrospectively by FHFA based on the Home Mortgage Disclosure Act(HMDA)data for the

5、year when it becomes available.Both the benchmark market and the retrospective market levels are determined at the national level and for a full calendar year.In any given year,an Enterprise is deemed to have met the goal if it meets or exceeds either the benchmark level or the retrospective market

6、level.Typically,HMDA data for a given calendar year is released in the second quarter of the following year so that FHFAs retrospective market level determination of the Enterprises housing goals performance is made the following year.2 The benchmark level is based on the market forecast model(and o

7、ther factors)and is set in advance for the goal period to provide a planning target for Enterprise activities.The market forecast model referred to here is the national level statistical model that is estimated using monthly goal-qualifying share data from HMDA and the resulting monthly forecasts ar

8、e then averaged into an annual forecast for each of the three years in the goal period.The retrospective market level is based on FHFAs determination of the goal qualifying market for each year based on HMDA data.This is not a statistical modeling exercise but rather an aggregation based on applying

9、 counting rules to HMDA data.The Safety and Soundness Act sets out seven factors that FHFA is expected to consider when setting the benchmark level.3 FHFAs approach has been to incorporate as many of these factors into the statistical forecast model as possible,generating model forecasts for each of

10、 the goal years along with confidence intervals.For instance,four of the seven factors(national housing needs;economic,housing,and demographic conditions;other mortgage data;and the size of the conventional purchase money or refinance mortgage segment)are explicitly modeled in the statistical foreca

11、st models.Three factors(performance and effort of the Enterprises to lead the industry in making mortgage credit available;the ability of the Enterprises to do so;and the need to maintain sound financial condition of the Enterprises)are not readily quantifiable and there are no public data on these

12、factors.As a result,they are not explicitly modeled in the statistical forecast models.FHFA incorporates these factors into the benchmark setting process while picking the specific point estimate within the model-generated confidence intervals for a given goal year.That process is documented in the

13、preamble to the final rulemaking and is beyond the scope of this report.This report focuses on documenting the statistical models and the associated confidence intervals of the estimates.The single-family goals are limited to conventional conforming mortgages on owner-occupied housing with a total o

14、f one to four units.Therefore,jumbo mortgages(with loan amounts above the 1 12 U.S.C.4561(a).2 Typically,FHFA will issue a preliminary determination of each Enterprises housing goals performance in a given calendar year,in the following October.The Enterprises will have 30 days to respond to the det

15、ermination and FHFA typically issues a final determination in December.3 12 U.S.C.4562(e)(2).4 conforming loan limit),mortgage loans to investors,mortgages on second homes,and non-conventional loans(loans with some form of government insurance on them)are all excluded.FHFA is required to establish t

16、hree single-family home purchase goals and one refinance goal.FHFA has also established two additional single-family home purchase subgoals for low-income census tracts and minority census tracts.The single-family home purchase goals and subgoals and the single-family refinance goal are defined as f

17、ollows:Low-Income Home Purchase(LIP)Goal:This goal measures the share of each Enterprises goal-qualifying purchase loans made to families with incomes no greater than 80 percent of Area Median Income(AMI).Very Low-Income Home Purchase(VLIP)Goal:This goal measures the share of each Enterprises goal-q

18、ualifying purchase loans made to families with incomes no greater than 50 percent of AMI.Minority Census Tracts Purchase(MCT)Subgoal:This goal measures the share of each Enterprises goal-qualifying purchase loans made to families with incomes no greater than 100 percent of AMI in minority census tra

19、cts.4 Low-Income Census Tracts Purchase(LCT)Subgoal:This goal measures the share of each Enterprises goal-qualifying purchase loans made to two subgroups:(1)families(regardless of income)in low-income census tracts5 that are not minority census tracts,and(2)families with incomes greater than 100 per

20、cent of AMI in low-income census tracts that are also minority census tracts.Low-Income Areas Home Purchase(LIA)Goal:This goal measures the shares of each Enterprises goal-qualifying purchase loans that are included in the minority census tracts and low-income census tracts subgoals,plus purchase mo

21、rtgages made to families with incomes no greater than 100 percent of AMI living in a federally-declared disaster area.Low-Income Refinance(LIR)Goal:This goal measures the share of each Enterprises goal-qualifying refinance loans made to families with incomes no greater than 80 percent of AMI.FHFA se

22、ts the low-income areas home purchase goal each year based on the sum of the minority census tracts subgoal and low-income census tracts subgoal benchmark levels,plus an additional increment based on federally-declared disaster areas over the past three years.As a result,FHFA does not create a separ

23、ate statistical forecast model for the low-income areas home purchase goal.The current set of statistical forecast models all use outcome variables(i.e.,market share estimates for the five housing goals and subgoals)that are derived from the HMDA data.We rely on twenty years of HMDA data:data from 2

24、004 until 2023.As we will discuss in the next section,although HMDA data prior to 2004 is available,those datasets do not contain key variables needed to define the market shares for the outcome variables.The current goal cycle continues the practice established in the 2018-2020 rulemaking cycle of

25、using Moodys Analytics as the primary data source for the independent or driver variables.6 This practice streamlined the data collection process and permitted FHFA to rely on Moodys Analytics forecasts.There are some exceptions.For some of the variables,Moodys 4 Census tracts that have a minority p

26、opulation of at least 30 percent and a median income of less than 100 percent of AMI.5 Census tracts where the median income is no greater than 80 percent of AMI.6 The variables on the right-hand-side of a forecast model are often referred to as the driver variables,while they are also called indepe

27、ndent variables and covariates in statistics and economics.5 forecasts were not available.For such cases,we use FHFAs own forecasts.The goal of FHFAs statistical forecast models is to provide our best estimate of various affordable market segments for the next housing goal period.This naturally reli

28、es on forecasts of the key driver variables for that period.FHFAs models include the best-fit model specifications and key driver variables for all goal-qualifying shares while following generally accepted professional practices and standards adopted by economists at other federal agencies.The model

29、s not only fit the historical data well but are also robust(as indicated by the out-of-sample tests).However,as is the case with any forecasting model,whether the models yield“accurate”forecasts is crucially dependent on the accuracy of the forecasts for the driver variables that are inputs to the m

30、odels.Moreover,the length of the forecast period is important:the longer out the forecast period,the wider the confidence bands around the forecasts.This report provides technical documentation of the market models used to generate the single-family housing goal forecasts for the 2025-2027 period.It

31、 assumes familiarity with econometric methods and forecasting practices that are commonly used by economists.The report is organized as follows:Section 1 describes the mortgage market and economic forecast data used to construct the econometric models.Section 2 identifies the model driver variables

32、and uses them to provide an overview of the housing and macro-economic environments that shape the mortgage market.Section 3 and Section 4 present the models for each of the four housing goals.Finally,concluding remarks are provided in Section 5.Technical appendices on sensitivity analysis and data

33、sources are attached at the end.Section 1:Sources of Data The historical monthly time-series data used in estimating the Enterprise housing goals forecast models are produced by a variety of sources.We use HMDA data to construct the outcome variables that is,the estimates for the goal-qualifying mar

34、ket shares for the two home purchase goals,two home purchase subgoals,and one refinance goal.Our Home Mortgage Disclosure Act(HMDA)dataset is provided by the Consumer Financial Protection Bureau(CFPB).7 The dataset contains loan-level records of mortgage originations that occurred during a calendar

35、year,including the month of mortgage origination.HMDA data is considered to be broadly representative of the mortgage market in the United States.For the purpose of estimating the single-family mortgage market for goal-qualifying loans,we limit the HMDA records to originations of conventional confor

36、ming first lien,prime home purchase(or refinance)mortgages.8 We further limit the data to originations since January 2004 because HMDA records from the pre-2004 time period do not include a number of variables that are critical in identifying the originations that are relevant to the housing goals.I

37、n particular,the pre-2004 HMDA data do not identify property type,lien status,Home Ownership Equity Protection Act(HOEPA)status,and the Average Prime Offer Rate(APOR)rate spread.The pre-2004 data were also less precise in identifying manufactured housing loans and high-priced mortgages.Since 2004,7

38、Prior to 2018,HMDA data was processed and released by the Federal Reserve Board on behalf of Federal Financial Institutions Examination Council(FFIEC),an interagency body empowered to administer HMDA.Beginning with the 2018 HMDA data,it has been processed and released by CFPB;in addition,new data el

39、ements on loan,borrower,and property characteristics were added as part of the reporting requirements.8 For the purpose of this analysis,prime mortgage loans are defined as mortgage originations that are not high-priced.In HMDA data,we identify high-priced loans as those with a spread(difference)bet

40、ween the Annual Percentage Rate(APR)of the loan and the applicable Average Prime Offer Rate(APOR)of 1.5 percentage points or greater.6 HMDA data began including:(1)rate-spread information for high-priced loans,(2)an indicator for manufactured housing loans,and(3)an identifier for first-lien mortgage

41、s.The rate-spread and manufactured housing information help to better identify subprime and chattel loans.HMDA data through December 2023 are used in these models.Historical and forecast values of the model driver variables were downloaded from Moodys Analytics web site.Moodys Analytics obtains the

42、historical values of the variables from various government agencies and industry trade groups and then generates forecasts for the variables using statistical models.Specifically,the unemployment rate,labor force participation rate,consumer price index,and new housing sales come from the Census Bure

43、au and the Bureau of Labor Statistics.Constant maturity interest rates on Government notes and bonds are generated by the U.S.Department of the Treasury,while mortgage interest rates are obtained from Freddie Macs Primary Mortgage Market Survey.The Housing Affordability Index(HAI)is provided by the

44、National Association of Realtors(NAR)to Moodys.To measure house price changes,we use FHFAs House Price Index(HPI)(for all transactions and for home purchase loans,separately).The refinance rate and the government market share information are calculated from the HMDA data.The volume of refinance appl

45、ications is captured by the refinance application index released by the Mortgage Bankers Association.The household debt service burden variable is collected by the Federal Reserve System.Per capita income information is released by the Bureau of Economic Analysis.Finally,a measure of the tightness o

46、f underwriting standards is constructed from data from the Senior Loan Officer Opinion Survey on Bank Lending Practices.The survey is conducted by the Federal Reserve System.For a complete list of data sources,see Appendix B.The modeling team used Moodys forecasts published in August 2024(specifical

47、ly the Baseline Forecast scenario)for the macroeconomic drivers where available.The Baseline Forecast scenario is designed by Moodys such that the probability that the economy will perform better than this projection is equal to 50 percent,as is the probability that it will perform worse.Model resul

48、ts based on other alternative forecast scenarios are reported in the Appendix section of the report.In cases where Moodys forecasts were not available(in particular,for the government share of home purchases and refinances),the team generated and tested its own forecasts.Section 2:Housing and Mortga

49、ge Market Driver Variables This section describes the historical and forecasts of key driver variables and the macroeconomic environment that has shaped the mortgage market for the forecast period.Macroeconomic Outlook Embedded in the Models There are many factors that impact the affordable segments

50、 of the housing market.Interest rates are arguably one of the most important variables in determining the trajectory of the mortgage market.The Federal Open Market Committee(FOMC)of the Federal Reserve,at its September 2024 meeting,reiterated its commitment to seeking maximum employment and inflatio

51、n at the rate of 2 percent in the long run,by lowering its target range for the federal funds rate by 0.5 percentage points to 4.75 percent to 5.0 percent.9 In its November 2024 meeting,the FOMC lowered the target range by an 9See https:/www.federalreserve.gov/newsevents/pressreleases/monetary202409

52、18a.htm.7 additional 0.25 percentage points to 4.5 percent to 4.75 percent.10 Moodys August 2024 Baseline forecast does not include these cuts but assumes that the FOMC will cut rates by 0.25 percentage points in September 2024 and December 2024.The forecast projects that mortgage interest rates par

53、ticularly the 30-year fixed rate,which is closely tied to the federal funds rate and the 10-year Treasury note yield will remain elevated,and only decline 0.4 percent from 6.4 percent in 2025 to 6.0 percent in 2027.Moodys forecast also projects that the unemployment rate will stay around 4 percent t

54、hrough 2027.Moodys also forecasts an increase in per capita disposable nominal income growth from$64,800 in 2025 to$70,100 in 2027.Inflation is expected be below 2.5 percent for the 2025-2027 period.House prices rose by 51.0 percent in the fourth quarter of 2023 relative to the same time in 2019,acc

55、ording to FHFAs purchase-only House Price Index(HPI).11 The rise was 5.7 percent in the second quarter of 2024,relative to the same time the previous year.12 FHFA noted that the market showed signs of softening as house prices saw the third consecutive slowdown in quarterly growth.13 Moodys predicts

56、 that home price appreciation will slow down even more in 2024.Its August 2024 forecast of the same HPI index expects the annual rates of house price growth to be 1.0,1.5,and 2.1 percent in 2025,2026,and 2027,respectively.Everything else being equal,the continued elevation in mortgage interest rates

57、 and house prices will likely affect the ability of low-and very low-income households to purchase homes.Housing affordability,as measured by Moodys forecast of the National Association of Realtors Housing Affordability Index(HAI),is projected to modestly rise from 105.9 in 2025 to 111.6 in 2027.(Hi

58、gher values of the HAI imply that affordability has improved).Affordable housing supply the third main factor in housing affordability has not kept pace with the growth of demographic demand even prior to the onset of the COVID-19 pandemic.14 High interest rates led to a steep drop in refinance acti

59、vity in 2023.The refinance share of overall mortgage originations declined from a 62.4 percent in 2020 to 19.2 percent in 2023.Moodys forecasts this share to decline further to 18.0 percent in 2024 but rise to 19.2,25.7 and 33.8 percent in 2025,2026 and 2027 respectively.The economic outlook from Mo

60、odys described above is largely consistent with the outlook provided by other forecasters.According to the most recent estimate published by the Congressional Budget Office(CBO)in June 2024,GDP is projected to grow by 2 percent in 2024,then remain in the 1.7-1.8 percent range through 2028.15 CBO pro

61、jects the unemployment rate to be 3.9 percent in 2024 and be in the 4.0-4.4 percent range through 2028.16 Exhibit 1 provides summary statistics on key macroeconomic indicators and the driver variables that are used in the forecast models.Variables that are forecasted by Moodys are presented in the f

62、irst 10 See https:/www.federalreserve.gov/newsevents/pressreleases/monetary20241107a.htm.11 See https:/www.fhfa.gov/sites/default/files/2024-08/HPI_2024Q2.pdf 12 Ibid 13 Ibid 14 See https:/www.urban.org/research/publication/place-blame-where-it-belongs 15 An Update to the Budget and Economic Outlook

63、:2024 to 2034.Published in June 2024.https:/www.cbo.gov/publication/60419 16 Congressional Budget Office(CBO),An Update to the Budget and Economic Outlook:2024 to 2034.Published in August 2024.https:/www.cbo.gov/publication/59946 8 panel of the Exhibit,while those that are forecasted by FHFA are pre

64、sented in the second panel of the Exhibit.Exhibit 1:Historical and Projected Trends of Key Macroeconomic Driver Variables Historical Trends Projected Trends 2019 2020 2021 2022 2023 2024 2025 2026 2027 Real Gross Domestic Product(Billions 2009$)$20,694$20,236$21,410$21,823$22,379$22,952$23,348$23,77

65、4$24,294 Real GDP Growth Rate 2.5-2.2 5.8 1.9 2.5 2.6 1.7 1.8 2.2 Unemployment Rate 3.7 8.1 5.3 3.6 3.6 4.0 4.1 4.0 4.0 Labor Force Participation Rate 63.1 61.7 61.7 62.2 62.6 62.6 62.7 62.5 62.5 Inflation Rate(Change in CPI)1.8 1.2 4.7 8.0 4.1 3.0 2.5 2.4 2.3 Consumer Confidence Index 128.3 101.0 1

66、12.7 104.5 105.4 103.1 109.2 106.8 105.6 10-Year Treasury Yield 2.1 0.9 1.4 3.0 4.0 4.3 4.2 4.1 4.1 Yield Curve 0.1 0.5 1.3 0.1-1.1-0.6-0.1 0.6 0.9 30-Year Mortgage Fixed Rate 3.9 3.1 3.0 5.3 6.8 6.7 6.4 6.1 6.0 30-Year FHA Mortgage Fixed Rate 4.3 3.4 3.2 5.7 7.0 6.9 6.5 6.3 6.2 Per Capita Disposabl

67、e Income(1000s$)$48.9$52.4$56.1$56.1$60.3$62.4$64.8$67.3$70.1 Household Debt Service Rati 4.2 3.9 3.7 4.0 4.0 4.1 4.2 4.3 4.3 Existing Home Sales(1000s)4,746 5,060 5,425 4,529 3,673 3,755 4,185 4,411 4,577 Net Percent of Banks Tightening Standards-0.6 18.1-8.0-2.3 8.5 0.6-0.9-1.8-1.2 Refinance Mortg

68、age Application Share 48.9 62.9 58.7 30.6 20.7 22.5 19.8 26.4 34.5 Housing Affordability Index 166.2 167.0 141.0 101.9 94.5 99.5 105.9 110.1 111.6 Percent Change in House Prices(PO)1 5.5 11.2 17.7 8.2 6.5 3.3 1.0 1.5 2.1 Percent Change in House Prices(ALL)2 5.1 6.5 18.1 11.5 5.7 3.4 1.4 1.9 2.3 Refi

69、nance Mortgage Share 42.8 62.4 57.4 27.1 19.2 18.0 19.2 25.7 33.8 Percent Gov.Insured Home Purchase Loans 28.6 28.6 25.0 23.9 27.0 27.4 27.1 28.2 29.3 Percent Gov.Insured Refinance Loans 21.1 17.9 15.9 18.2 21.2 20.8 21.2 21.8 22.4 Note:Historical values and projected trends are provided by Moodys A

70、nalytics.Government shares of the home purchase and refinance market are forecasted by FHFA.1 Purchase transactions only(Q4/Q4%Change)2 All transactions(Q4/Q4%Change)Expectations Regarding Key Driver Variables Interest Rates Interest rates are affected by many demand and supply factors.Trends in int

71、erest rates on longer-term financial instruments such as mortgages typically follow the fluctuations of the 10-Year Treasury note yield.There is a-30 to 90 basis point spread between the 1-year and 10-year Treasury yields forecast for the 2025-2027 period,reflecting the expected differences in liqui

72、dity and credit risk during that period.This is following an expected two years of negative spreads,with the lowest point of-139 basis points in June 2023.Interest rates are heavily influenced by the monetary policies of the 9 Federal Reserve Boards Federal Open Market Committee(FOMC).Since mid-2008

73、,the FOMC has maintained an accommodative monetary policy in support of its dual mandate of fostering maximum employment and price stability.While near-term risks to the economic outlook appear roughly balanced,the FOMC monitors the inflation rates closely.Exhibit 2 shows the historical and forecast

74、 values of the 30-year fixed mortgage interest rate(FRM)and the 10-year Treasury yield.The 30-year fixed-rate mortgage interest rate reached a historic low of 2.7 percent in November 2020.The rate rose to 3.2 percent in December 2021 then rapidly rose to a peak of 7.4 percent in October 2023.For the

75、 forecast period,that rate is expected to decline slightly from 6.4 percent in 2025 to 6.0 percent in 2027.Exhibit 2:Historical and Projected Trends of Mortgage Rates and 10-Year Treasury Unemployment,Labor Force Participation,and Inflation In addition to being a general indicator of the economys he

76、alth,changes in employment also directly affect the housing market because buying a house is,for many households,the single largest investment and a long-term commitment that requires stable employment.The unemployment rate had fallen steadily from 9.1 percent in August 2011 to around 3.5 percent in

77、 last quarter of 2019.The rate then rose sharply in the first quarter of 2020 and peaked at 14.8 percent in April 2020 due to the pandemic and related economic shutdowns.Due to the aggressive fiscal and monetary policies,the unemployment rate fell to 3.9 percent in January 2022 and is forecast to be

78、 in the 4.0 4.1 range from 2025 to 2027 The labor force participation rate was rising gradually in the past few years prior to the pandemic and reached 63.1 percent in 2019.As expected,the rate dropped noticeably in 2020 to 61.7 percent then stayed at that level in 2021,before rising to 62.6 in 2023

79、.It is expected to be in the 62.5-62.7 percent range from 2025 to 2027.One of the stated objectives of the FOMCs interest rate policy is fostering maximum employment.Given the foreseeable monetary policy and the continued growth in the economy,the unemployment rate is expected to average about 4 per

80、cent for the 2025 to 2027 period according to Moodys forecasts.Its impact on the affordable home purchase market will depend on the composition of the 10 unemployment rate and could be expected to be greater if the unemployment rate of lower-wage earners substantially drives change in the overall un

81、employment rate.The second stated objective of the FOMC in determining its interest rate policy is maintaining price stability.The inflation rate declined from 8.0 percent in 2022 to 4.1 percent in 2023 and Moodys forecasts that inflation will average 2.3 to 2.5 in the 2025 to 2027 period.Exhibit 3

82、plots the history and forecasts of the unemployment rate and the annual growth rate of the consumer price index(CPI).Exhibit 3:Historical and Projected Trends of Unemployment Rate and Annual Growth Rate of Consumer Price Index(CPI)House Prices and Affordability Trends in house prices influence the h

83、ousing and mortgage markets.In periods of house price appreciation,home sales and mortgage originations may increase as the expected return on investment rises.In periods of price depreciation and/or price uncertainty,home sales and mortgage originations tend to decrease as risk-averse homebuyers ar

84、e reluctant to enter the market.To measure house price appreciation,we use FHFAs purchase-only House Price Index(HPI).In the left-hand-side panel of Exhibit 4,we plot the level of the HPI(red line,represented by the left axis)and the year-over-year growth rate of the HPI(grey line,represented by the

85、 right axis)over the 2004-2027 period.As the graph shows,house prices fell dramatically during the 2007 financial crisis but have recovered and increased since 2012 with price increases in excess of 5.0 percent per year.The growth rate began to accelerate in the beginning of 2020 and reached 10+perc

86、ent on a year-over-year basis during the fall quarter of 2020 and staying at two-digit percent growth through September 2022.Although price increases decelerated,growth rate remained strong in 2023.According to Moodys,house price growth will continue to decelerate in 2024.Then house price appreciati

87、on is expected to be 1.0 percent,1.5 percent and 2.1 percent in 2025,2026 and 2027.The expected slight declines in interest rates and house prices in 2024 and 2025 are expected to moderately improve housing affordability from the 2023 low.To measure housing affordability,we use the Housing Affordabi

88、lity Index(HAI)published by the National Association of Realtors and provided by Moodys.This index takes general house prices,mortgage interest rates,and borrower incomes into consideration.Higher HAI values means that housing is more affordable.The right 11 panel of Exhibit 4 presents the history a

89、nd trends of the index.It shows that housing affordability had increased from early 2019,largely due to falling interest rates.Despite the economic downturns caused by the pandemic,housing affordability remained high during 2020 due to the historic low interest rates.However,affordability declined s

90、harply in 2021-2023 due to high home price appreciation,rising and elevated interest rates and supply shortage.Exhibit 4:Historical and Projected Trends of House Prices and Affordability Refinance Share of Mortgage Applications The size of the refinance mortgage market has an impact on the affordabl

91、e share of refinance mortgages.Historically,refinance mortgage volume increases when the refinancing of mortgages is motivated by low interest rates(“rate-and-term”refinances).Higher-income borrowers tend to make up a greater share of this increased volume.As a result,in periods of low interest rate

92、s,the share of lower-income borrowers among all refinancers tends to decrease.Likewise,refinancing that occurred when interest rates were high tends to have a higher proportion of lower-income homeowners who refinance to consolidate their debts or draw equity out of their homes for other uses.Exhibi

93、t 5 plots the historical and projected values of the 30-year fixed mortgage rate and refinance share of mortgage applications(measured by MBAs refinance index).As expected,it shows a roughly negative correlation between the refinance share and mortgage rate.The refinance share had consistently excee

94、ded 50 percent of the mortgage applications for the period between 2009 and 2016 because of the low interest rates.2020 and 2021 represented two solid“refinance boom”years due to the historic low mortgage rates.Refinance shares shrank in 2022 and 2023 due to the high interest rates,declining to as l

95、ow as 19.2 percent in 2023.Moodys forecasts that the refinance share will rise but stay well below 50 percent through 2027.12 Exhibit 5:Historical and Projected Trends of Refinance Share and 30-Year Fixed Mortgage Rate Underwriting Standards Mortgage underwriting standards clearly have an impact on

96、mortgage originations.One measure of the tightness of underwriting standards is the net percentage of lenders who reported a tightening of underwriting standards in the Federal Reserve Boards Senior Loan Officer Opinion Survey on Bank Lending Practices.If the net percentage is greater than zero,it m

97、eans that there is a higher percentage of lenders reporting a tightening of underwriting standards than lenders reporting a loosening of underwriting standards.Conversely,a negative net percentage implies that more lenders are reporting a loosening of underwriting standards.The variable is a qualita

98、tive assessment by nature but very useful in capturing underwriter attitudes as well as the use of credit score overlays and other mechanisms in which underwriting standards might tighten.Exhibit 6 plots the series over time.During the housing boom years,underwriting standards loosened.The graph sho

99、ws that,following the housing bust in 2007,lenders tightened underwriting standards considerably for a number of years.The underwriting standards in 2020 were almost as tight as those during the 2007-2009 period.Following the COVID-19 pandemic,underwriting standards loosened but began to tighten as

100、interest rates rose in 2022 and 2023.Underwriting standards have loosened in 2024 and Moodys forecasts that they will remain so through 2027.13 Exhibit 6:Historical and Projected Trends of Mortgage Underwriting Standards Share of Government-Insured and Guaranteed Mortgages Moodys does not publish fo

101、recasts for the government share of home purchase and refinance originations.The modeling team has developed basic forecast models to project these series for the 2025-2027 period.Government shares are modeled to be driven by trends in the unemployment rate,interest rates,the spread between FHA and

102、conventional mortgage rates,existing home sales,relative underwriting tightness,and house prices.Exhibit 7 plots the historical and projected paths of the two series.Both series,especially the refinance share,declined noticeably during the pandemic in 2020 but they have since rebounded.It is expecte

103、d that the government share of home purchases and refinances will reach 29.3 percent and 22.4 percent by 2027.Exhibit 7:Historical and Projected Trends of Government Shares of Mortgages 14 Section 3:Econometric Models of the Single-Family Housing Goals The purpose of the statistical models is to for

104、ecast the market share of the goal-qualifying mortgage originations in the mortgage market for the 2024-2027 period.The focus is on generating reliable model forecasts,rather than testing economic hypotheses or trying to explain the relationship between the variables.Moreover,the forecast equation i

105、tself is a reduced-form model for the affordable market.Separate models are developed and estimated for each of the five single-family housing goals.We use a type of time-series regression models called Autoregressive Integrated Moving Average(ARIMA)models.Widely used in the field of economic foreca

106、sting,ARIMA models are known for their flexibility and ability to generate good fit to various historical series.FHFA has employed these ARIMA models in past rulemaking cycles to generate market forecasts.We have updated the model specification of the ARIMA models in this goal cycle from the models

107、used in the 2022 rulemaking cycle as needed.The outcome variable for each of the models is expressed in percentage share format.It is constructed as a monthly time-series using the 2004-2023 HMDA data,as described earlier.All outcome variables and driver variables were first-differenced to ensure th

108、at the data series have the statistical property that they are stationary.The augmented Dickey-Fuller tests(ADF)have been conducted on each of the first-differenced series to confirm that they are stationary.Fitting an ARIMA model on first-differenced variables means that the reported regression coe

109、fficients can no longer be interpreted in the way the“beta”coefficients are interpreted in linear regression.Not only are the estimated equations non-linear(indicating the slope of the line is dependent on the specific point the reader is interested in),the coefficients themselves represent the chan

110、ge in the first difference of the dependent variable.Fortunately,since the goal of our models is to forecast rather than explain the reduced form equation of the various housing goal“markets,”we are able to rely on the“hold-out”samples and out-of-sample forecasts to test the robustness of the model

111、specifications as is common among forecasters.In order to model the non-contemporaneous nature of the relationship between macroeconomic variables,both lagged and un-lagged versions of each driver variable were tested and the appropriate form of the variable chosen.17 Moreover,to avoid multi-colline

112、arity(or correlation among driver variables),the modeling team has examined the correlation between potential driver variables and ensured that the final model specifications did not include driver variables that are highly correlated.Monthly indicator variables are included in the models to control

113、 for the seasonal/cyclic patterns of the mortgage activities.Finally,to allow for non-linearity between the outcome variable and driver variables,we tested and included some of the variables in natural logarithm scale.For example,the consumer confidence index variable is entered in logarithm scale i

114、n the model for the low-income home purchase market share because it provides a better fit to the data in that form.During the model development process,we grouped the factors that are expected by housing market economists to have an impact on the market share of affordable housing into seven broad

115、categories.They are:Demand-side factors such as per capita income and household debt service indicators;Interest rate environment including the 30-year FRM mortgage rate;17 Lagged variables are reported with a subscript such as t-1 in the tables for the regression coefficients.15 Expectation factors

116、 and indicators of the health of the economy such as the unemployment rate,labor force participation rate,the consumer confidence index and consumer confidence index;House price level including FHFAs house price indices and the Housing Affordability Index;Supply-side factors including indicators of

117、existing home sales;Underwriting standards including the Senior Loan Officer Opinion survey;and Other factors including the share of government insured or guaranteed mortgages.For each category of variables,many variables were tested but only retained when they exhibited predictive power.For each mo

118、del,we tested and succeeded in including at least one driver variable per category.We concentrated on the theoretical and behavioral implications when selecting the driver variables but since our goal was to develop robust forecasting models(rather than developing explanatory models),we retained dri

119、ver variables depending on the strength of the explanatory power.Market Forecast for Low-Income Home Purchase Goal(LIP)Exhibit 8 reports the final set of model driver variables that we have selected for the low-income home purchase goal.It also reports the regression coefficients,their level of stat

120、istical significance,and model diagnostics.The driver variables that were found to be statistically significant are:Unemployment rate Consumer confidence Housing affordability index FHFAs purchase-only House Price Index Sale of existing homes Share of government insured or guaranteed mortgages The C

121、hi-square statistics and other model fit measures reported at the bottom of the table are reasonable for forecasting models for this kind of data.The Pr(Chi-square)statistic,for instance,indicates the probability that,after including the relevant driver variables in the model,the regression residual

122、s are following a white noise distribution(therefore random).To test for reliability and accuracy of the models forecasts,we have conducted an out-of-sample test where we withhold the last 8 months(May 2023-December 2023)of the HMDA series and estimated the model using the remaining data series(that

123、 is,January 2004-April 2023).We then compared the model forecasts in the“hold-out”period(May 2023-December 2023)to the actual values in the same period(May 2023-December 2023).Results of this analysis are presented in Exhibit 9.The blue line represents the historical HMDA series,while the red line r

124、epresents the model forecast.Focusing on the out-of-sample forecast period(May 2023-December 2023)while the two lines do not track each other perfectly,the forecast model is able to capture the downturns and upturns in the actual data(blue line),as the forecast line and actual data line move in a si

125、milar fashion,albeit with a lag.This is an indication of the robustness of the model specification.16 Exhibit 8:Regression Coefficients of Market Forecast Model for the Low-Income Home Purchase Goal Outcome Variable(First Difference)Share of Borrowers with Low Income Driver Variable(First Difference

126、)Coefficient Estimate Unemployment Rate t-1 0.0016*(0.0009)Ln(Consumer Confidence)0.0183*(0.0097)Consumer Price Index(CPI)-0.0009 (0.0008)Housing Affordability Index 0.0006*(0.0002)FHFA House Price Index(HPI)-Purchase Only 0.0011*(0.0004)Sale of Existing Homes 0.0087*(0.0036)Senior Loan Officer Opin

127、ion Survey:Percent of lenders reported tightening of underwriting-0.0098 (0.0097)Share of Government-Insured or Guaranteed Mortgages 0.2285*(0.0444)AR(1)-0.0338 (0.0680)AR(2)0.0096 (0.0687)Model Diagnostics 2 13.73 Prob(2)0.01 AIC-1717.96 SBC-1644.95 Notes:All variables are in first-difference forma

128、t.Model includes monthly dummy variables that control for seasonal variations.Standard Errors are reported in paratheses.Significance levels:*p0.1,*p0.05,*p2)0.06 AIC-2117.71 SBC-2044.70 Notes:All variables are in first-difference format.Model includes monthly dummy variables that control for season

129、al variations.Standard Errors are reported in paratheses.Significance levels:*p0.1,*p0.05,*p2)0.06 AIC-2062.16 SBC-1996.11 Notes:All variables are in first-difference format.Model includes monthly dummy variables that control for seasonal variations.Standard Errors are reported in paratheses.Signifi

130、cance levels:*p0.1,*p0.05,*p2)0.23 AIC-2349.69 SBC-2266.25 Notes:All variables are in first-difference format.Model includes monthly dummy variables that control for seasonal variations.Standard Errors are reported in paratheses.Significance levels:*p0.1,*p0.05,*p2)0.34 AIC-1594.32 SBC-1524.79 Notes

131、:All variables are in first-difference format.Model includes monthly dummy variables that control for seasonal variations.Standard Errors are reported in paratheses.Significance levels:*p0.1,*p0.05,*p0.01.29 Exhibit 25:Robustness Test of Market Forecast Model for the Low-Income Refinance Goal Exhibi

132、t 26:Model Forecast for the Low-Income Refinance Goal Exhibit 27 presents a summary of the annual forecasts(and their confidence intervals),together with the historical market shares(HMDA)and the benchmark levels that FHFA set for the Enterprises in the past.As a reference,the graph also shows the m

133、odel forecasts for the 2021-2024 period that FHFA published in the Federal Register for the 2022 Final Rule.Forecasts from the new model are shown in the last row:38.1 percent for 2025,36.4 percent for 2026,and 34.3 percent for 2027.The average annual forecast across the 2025-2027 period is 36.3 per

134、cent.30 Exhibit 27:Historical Performance and Model Forecast for the Low-Income Refinance Goal Historical Performance Projected Performance 2019 2020 2021 2022 2023 2024 2025 2026 2027 Actual Market 24 21 26.1 37.3 40.3 Benchmark 21 21 21 26 26 26 2022 Final Rule Market Forecast 24.2 22.3 25.5 29.1

135、+/-+/-+/-+/-2.9 5.0 6.4 7.4 Current Market Forecast 38.5 38.1 36.4 34.3 +/-+/-+/-+/-3.1 5.5 7.0 8.3 Section 4:Sensitivity of Model Estimates The modeling team has also investigated the sensitivity of the model forecasts with respect to our assumption of future macroeconomic scenarios.Results are sum

136、marized in Appendix A.Overall,we find that the forecasts are somewhat sensitive to the use of alternate scenarios.Section 5:Concluding Remarks This paper provides the technical details of the econometric models that we have developed to forecast the market share of the goal-qualifying mortgage origi

137、nations in the mortgage market for each of the four single-family housing goals.Forecasts for the 2025-2027 period are presented.Model diagnostics and reliability of the forecasts have also been discussed.To examine the sensitivity of the model forecasts,we have also provided model forecasts using d

138、river variables embedded in two alternative scenarios published by Moodys.31 Appendix A:Sensitivity of Model Estimates to Alternative Macroeconomic Forecast Scenarios This section analyzes the extent to which our market share estimates are sensitive to the macroeconomic environments that we have ass

139、umed.The market share estimates we have reported so far are based on forecast values of the driver variables from Moodys Baseline scenario.To assess the sensitivity of our market share estimates,we use forecast values of the driver variable from two alternative macroeconomic scenarios from Moodys.Th

140、ey are:Stronger Near-Term Growth(S1)scenario.This scenario assumes that theres an increase in demand for US exports and manufacturing,and that the Feds efforts to resolve bank failures stabilizes consumer confidence.It assumes that unemployment declines below the baseline and with the stronger growt

141、h in the economy,inflation and interest rates are higher than in the Baseline.Business sentiment and consumer confidence increases,accelerating consumer spending and business investment growth.Strong gains in employment lift household incomes and spending.House prices rise by 5.0 percent and 2.7 per

142、cent in 2025 and 2026,compared to 1.0 percent and 1.5 percent in the Baseline scenario.This above-baseline scenario is designed such that the probability that the economy will perform better than this projection is 10 percent,and the probability that it will perform worse is 90 percent.Protracted Sl

143、ump(S4)scenario.This scenario assumes that the still-elevated interest rates cause credit-sensitive spending and consumer confidence to plummet.Bank failure fears further reduce consumer confidence and banks tighten their lending standards.The stock market falls sharply and a recession begins in the

144、 fourth quarter of 2024.GDP is-2.8 percent in 2025 and-0.4 percent in 2026 compared to 1.7 percent and 1.8 percent in the Baseline scenario.Unemployment rate reaches 8.2 percent in 2025 and house price growth is-10.7 percent in the same year.This below-baseline scenario is designed such that the pro

145、bability that the economy will perform better than this projection is 96 percent,and the probability that it will perform worse is 4 percent.The comparison of the driver variables for all scenarios is presented in Exhibit A-1 and,as shown,the differences in many of the variables themselves are more

146、subtle than stark.32 Exhibit A-1:Comparison of Forecast Scenarios 2024 2025 2026 2027 GDP Growth Rate Baseline 2.6 1.7 1.8 2.2 Stronger Near-Term Growth 2.7 3.3 2.4 2.2 Protracted Slump 2.2-2.8-0.4 2.7 Unemployment Rate Baseline 4.0 4.1 4.0 4.0 Stronger Near-Term Growth 3.9 3.1 3.3 3.4 Protracted Sl

147、ump 4.6 8.2 8.9 8.4 30-Yr Fixed Rate Baseline 6.7 6.4 6.1 6.0 Stronger Near-Term Growth 6.8 6.4 6.2 6.1 Protracted Slump 6.5 4.5 4.7 5.1 FHFA HPI Growth Rate Baseline 3.3 1.0 1.5 2.1 Stronger Near-Term Growth 4.8 5.0 2.7 3.1 Protracted Slump-4.7-10.7 2.6 2.8 Housing Affordability Index Baseline 99.5

148、 105.9 110.1 111.6 Stronger Near-Term Growth 97.6 98.9 102.1 103.8 Protracted Slump 119.6 164.8 147.7 143.8 Refinance Share Mortgage Applications Baseline 22.5 19.8 26.4 34.5 Stronger Near-Term Growth 22.5 18.4 23.1 30.0 Protracted Slump 22.8 39.1 64.4 54.1 To generate the alternative market share f

149、orecasts,we input the forecast values of the driver variables from each of the alternative scenarios into the regression equations.Monthly forecast results are then aggregated into yearly forecasts.We compare these to the market share forecasts based on the Baseline scenario.Exhibit A-2 provides a s

150、ummary of the results by year,while Exhibit A-3 presents the monthly forecast results.Results for the low-income purchase share are shown in the first panel of Exhibit A-2.The first row of each panel highlights the results based on the Baseline scenario,while the following rows show results from usi

151、ng the alternative scenarios.Average annual market performances over the 2025-2027 period are presented in the last column.The results show that the share forecasts for the low-income home purchase goal change slightly in response to alternative macroeconomic environments.33 Exhibit A-2:Summary of M

152、arket Model Forecasts Based on Different Scenarios 2025-2027 Average 2025 2026 2027 LIP Baseline 26.2%25.9%25.7%25.9%Stronger Near-Term Growth 27.9%28.0%28.2%28.1%Protracted Slump 20.8%19.8%20.4%20.4%VLIP Baseline 6.1%6.0%5.9%6.0%Stronger Near-Term Growth 6.3%6.3%6.2%6.3%Protracted Slump 5.6%5.0%4.8

153、%5.1%MCT Baseline 12.5%12.7%13.0%12.7%Stronger Near-Term Growth 12.9%13.3%13.7%13.3%Protracted Slump 10.9%11.1%11.5%11.2%LCT Baseline 10.1%10.1%10.1%10.1%Stronger Near-Term Growth 10.4%10.5%10.5%10.4%Protracted Slump 8.7%8.8%9.1%8.8%LIR Baseline 38.1%36.4%34.3%36.3%Stronger Near-Term Growth 38.7%37.

154、4%35.5%37.2%Protracted Slump 29.2%25.2%30.0%28.1%Note:LIP=Low-Income Purchase Goal VLIP=Very Low-Income Purchase Goal MCT=Minority Census Tracts Subgoal LCT=Low-Income Census Tracts Subgoal LIR=Low-Income Refinance Goal 34 Exhibit A-3:Market Model Forecasts Based on Alternative Scenarios Low-Income

155、Home Purchase Goal Very Low-Income Home Purchase Goal Minority Census Tracts Subgoal Low-Income Census Tracts Subgoal Low-Income Refinance Goal 35 Appendix B:List of Data Sources Moodys Analytics https:/ Federal Housing Finance Agency:House Price Index https:/www.fhfa.gov/data/hpi Congressional Budget Office https:/www.cbo.gov