THE IMPACT OF HOUSING FINANCE REFORM ON MORTGAGE RATES, HOME BUYERS AND THE ECONOMY

Transcription

1 THE IMPACT OF HOUSING FINANCE REFORM ON MORTGAGE RATES, HOME BUYERS AND THE ECONOMY February 14, 2014 Kent W. Colton, PhD Senior Fellow, Harvard Joint Center for Housing Studies President, The Colton Housing Group, LLC Michael Carliner Economic Consultant Fellow, Harvard Joint Center for Housing Studies

2 EXECUTIVE SUMMARY Background Over the past year, a number of proposals have been made to reform the housing finance system. One of the first came in February, 2013 with the release of the report of the Bipartisan Policy Center (BPC) Housing Commission, and in June, 2013, Senators Corker (R- Tennessee) and Warner (D- Virginia) introduced a bipartisan reform bill in the Senate. This legislation would eventually phase out and replace Fannie Mae and Freddie Mac with a redesigned housing finance system with the Federal Mortgage Insurance Corporation (FMIC), providing for a limited and explicit government guarantee for catastrophic risk. An effort is now underway in the Senate Banking Committee, lead by Chairman Johnson (D- South Dakota) and Ranking Member Crapo (R- Idaho), to mark up a revised version of this bill and bring it to the Senate floor later this year. However, this new system will also bring an increase in mortgage rates because the private sector will be required to set aside capital to cover the first loss risk, and the government will establish a fund for catastrophic risk. The basic questions are how much will mortgage rates rise, and what will be the impact on homebuyers and the economy? Studies of the Impact of Housing Finance Reform on Mortgages Rates Several studies have tried to identify the impact of housing finance reform on mortgage rates. In a paper prepared for the BPC Housing Commission, Andrew Davidson & Co. 1 found that the additional interest rate (or Guarantee Fee - could be in the range of basis points (bps), assuming that FICO scores were greater than 750 and Loan to Value (LTV) ratios were 80% or below. After subtracting the current g- fee paid for mortgages guaranteed by Fannie Mae and Freddie Mac, which is in the range of 52 bps, mortgage rates would increase by approximately 25 bps. However, the Davidson study also pointed out that these numbers were based on homebuyers with excellent credit, and if mortgage credit was extended beyond these pristine levels t example, the credit cost for loans with FICO greater than 750 and LTV below 80% would be less than 25 bps a year, while the credit costs for loans with FICO below 700 and LTV greater than 90% would be more than 10 times higher and exceed Andrew Davidso 2013, Prepared for the Bipartisan Policy Center Housing Commission and Referenced in their final report, 2 Andrew Davidson & Co. Inc, ibid., p. 2. This paper was prepared for the Leading Builders of America (LBA) i

3 In a second study by Mark Zandi and Christian deritis, the authors reviewed the mortgage rate implications of four proposals including the Corker- Warner legislation. 3. Their estimated rise in mortgage rates for the legislation was in the range of 94 to 119 bps, which - after subtracting the current g- fee of 52 bps - results in an increase of at least 42 to 67 basis points. Similar to the Davidson & Co. study, the Zandi/deRitis study assumes a FICO score of 750 and a LTV of 80 % - - both high 4 Impact of Mortgage Rate Increases The base case estimates of these two studies both rely on assumptions of high credit standards environment is holding back the housing market and the economy. Recognizing the importance of more realistic, consumer friendly assumptions where FICO scores are closer to the historical range of %, this paper identifies a range of four possible interest rate increases - above the current g- fee of 52 bps - that may result from housing finance reform: (1) 25 basis points (the lowest scenario from the Davidson & Co. Study); (2) 42 basis points (the minimum mortgage rate rise in the Zandi/deRitis study); (3) 100 basis point (a middle estimate based on a more likely distribution of FICO credit scores and LTVs); and (4) 150 basis points (a higher, but still realistic, scenario where FICO scores are closer to the historical range of % range). The impacts of these potential increases in mortgage rates are summarized in the chart below, assuming a base mortgage rate of 4.5%. 5 Impact of Housing Finance Reform on Mortgage Rates, Monthly Payments, Buyers Disqualified by a Rise in Rates, Housing Starts, and the Economy No Change 25 BPS Increase 42 BPS Increase 100 BPS Increase 150 BPS Increase Mortgage Rate Impact 4.5% 4.75% 4.92% 5.5% 6.0% Monthly Payments $1, $1, $1, $1, $1, Change in Monthly Payments --- +$ $ $ $ Mark Zandi and Christian deritis, Cost of Housing Finance Reform Consumer Credit Analytics, November, Mark Zandi and Christian deritis, ibid., p It is worth noting that the range of mortgage credit costs could be even higher. As stated above by Davidson & higher (than 25 bps per year) and exceed 250 bps a yea impacts on the home buyer, housing starts, and the economy will also rise accordingly. ii

4 Share of home buyers who would no longer qualify No Change 25 BPS Increase 42 BPS Increase 100 BPS Increase 150 BPS Increase % -2.6% -5.6% -9.4% Loss in Housing Starts ,006-42,771-94, ,893 Impact on Economy/GDP --- -$8.0 Billion Loss Impacts on Home Buyers and the Economy -$12.2 Billion Loss -$26.8 Billion Loss -$44.4 Billion Loss The chart above outlines the effects of rising mortgage rates on monthly payments and on the share of prospective home buyers qualifying for mortgages on a median- priced home, as well as the calculated effects on housing starts and on GDP. The details of the calculations are described in the full report. An increase in mortgage rates of 25 or 42 bps would raise the monthly payments for a mortgage on a median- priced existing home by less than $40. Even such modest increases would raise the household income needed to qualify by as much as $1,700 and would reduce construction activity and GDP. Rate increases of 100 bps or 150 bps would have a more substantial negative effect, disqualifying 5.6 percent to 9.4 percent of households who would be looking to buy homes and who would qualify under the current housing finance system. Housing starts would be reduced by an estimated 94,042 to 155,893. The negative effect on GDP could be in the range of $27 billion to $44 billion. The effects of the rate increases would be greatest on the ability of younger households and on minority households to buy homes, since those groups have incomes clustered closer to the minimums required to qualify for mortgages. Thus, the effect could be to widen the gap between homeownership rates and housing quality for white middle- aged households and minority and younger households. Conclusions If the rise in interest rates is at the low end of the range or 42 basis points - - then the impact of housing finance reform will be relatively modest with monthly payments rising between $23 - $40, the share of borrowers who would no longer qualify in the - 1.7% to - 2.9% range, housings starts dropping by 28,000 to 42,000, and GDP losses in the range of $5.4 - $8.2 billion. However, if more realistic credit assumptions are applied and mortgage rates rise at the middle or higher end of the possible range ( basis points), then the impact would be substantial, with monthly payments increasing by $96 - $146, the share of borrowers who would no longer qualify in the - 5.6% to % range, housing starts dropping by 94, ,000 per year, and GDP losses of $27 $44 billion. iii

5 THE IMPACT OF HOUSING FINANCE REFORM ON MORTGAGE RATES, HOME BUYERS AND THE ECONOMY Background housing finance system. One of the first came in February, 2013 with the release of the report of the Bipartisan Policy Center (BPC) Housing Commission. It called for the elimination - phased out over an appropriate period of time - of Fannie Mae and Freddie Mac, and a far greater role for the private sector with private institutions withstand significant losses. In addition, the Commission called for a limited role for the government to explicitly guarantee mortgage backed securities (MBS), but only to bear catastrophic risk for the MBS, and not to guarantee the equity or debt of the entities that issue or insure the MBS. The Commission also outlined a possible structure for a redesigned housing finance system that would meet these principles. In June of 2013 Senators Corker of Tennessee and Warner of Virginia introduced a bipartisan bill in the Senate Banking Committee designed to meet the same principles outlined above, and a bipartisan effort is now under way led by the Chairman of the Senate Banking Committee, Senator Johnson, and the Ranking Member of the Committee, Senator Crapo, working with Senators Corker, Warner and many others, to craft legislation in the Senate that will eventually phase out and replace the GSE model (Fannie Mae and Freddie Mac) with a redesigned housing and a new regulator, the Federal Mortgage Insurance Corporation (FMIC), with the power to oversee the system and to provide for a limited and explicit guarantee for catastrophic risk. However, this new system will also bring an increase in mortgage rates because the private sector will be required to set aside capital to cover the first loss risk, and the government will establish a fund for catastrophic risk. These added costs are - against credit- related losses, and for operating costs such as bundling, issuing, and reporting on mortgage backed securities. The basic questions are how much will the rates rise, and what will be the impact on homebuyers and the economy? Studies of the Impact of Housing Finance Reform on Mortgage Rates Several studies have tried to identify the impact of housing finance reform most Andrew Davidson & Co., Inc, January 2013 (Prepared for the Bipartisan Policy Center Housing Commission and referenced in their final report, Future: New Directions for National Policy, pp 62-63); and (2) Cost of Housing Finance Reform This paper was prepared for the Leading Builders of America (LBA) 1

6 and Consumer Credit Analytics, November, The second study not only analyzed the impact of the Corker- Warner proposal, but it also examined the mortgage cost implications of three other proposals: the current system, a nationalized system, and legislation passed by the House Financial Services Committee entitled the PATH Act - Protecting American Taxpayers and Homeowners (PATH) Act. The purpose of this paper is to examine these studies more closely, particularly as they relate to the Corker- Warner legislation and the BPC Housing Commission proposal, first to develop a range of cost estimates based on different assumptions, and then to discuss the impact of these costs on homebuyers and the economy. The Davidson & Co. Study The Davidson & Co. study provides a range of estimates of the possible costs of implementing the housing finance recommendations of the Bipartisan Policy Center (BPC) Housing Commission. The summary of the study included in the BPC Housing Commission final report, Policy, suggests that the additional g- fees paid by a borrower with no mortgage insurance under the housing finance system proposed by the BPC Housing Commission would be in the range from 59 to 81 basis points including: (1) Annual Credit Charges of basis points for the private sector to set aside capital to cover possible losses and a risk adjusted return; (2) 8 basis points set aside for catastrophic risk to cover the limited government guarantee; and (3) 6 basis points to pay for the operating costs of the Public Guarantor. Davidson describes these g- Costs By comparison, g- fees for mortgages currently guaranteed by Fannie Mae and Freddie Mac are now in the range of 52 basis points (including a 10 basis point charge for the Treasury in order to pay for the payroll tax deduction.) 6 After subtracting the current g- fees, the range of estimates in the Davidson & Co. study suggests that based on positive assumptions of high FICO scores (FICO standing for Fair Isaac Co.) and low Loan to Value (LTV) ratios, mortgage rates would increase by approximately 25 basis points. (59 to 81 bps minus 52 = 7 29 bps.) However, these estimates assume a relatively stable housing market with modest growth in house prices, and depending on market conditions and the credit quality of the mortgage pool, the impact on mortgage rates will vary significantly. then the increase in mortgage rates will be modest (in the range of 25 bps). However, if mortgage credit is extended beyond these pristine levels to homebuyers then the mortgage price increase will be significantly higher. Indeed, the Davidson 6 The current g- fees charged by Fannie Mae and Freddie are used to cover operating costs for the two GSEs and ay- 2

7 & Co. study point credit FICO scores and LTV ratios. For example, the credit cost for loans with FICO greater than 750 and LTV below 80 % would be less than 25 basis points a year, while the credit costs for loans with FICO below 700 and LTV greater than 90% 7 In making these calculations, the Andrew Davidson & Co. study examined a pool of Freddie Mac eligible mortgages originated in Using two different approaches to identify the amount of capital needed to absorb the first loss credit risk for these mortgages, they calculated the Annual Credit Charges (ACC) - - the annual portion of the mortgage rate (in basis points) - - required to set aside enough capital to bear the losses for the loans under different economic scenarios and different dimensions of credit risk. Based on their company model, they analyzed two alternative approaches to identify the capital required: (1) the Senior/Sub Approach, and the (2) Capital Requirement Approach (privately insuring the mortgage pool). (For a more detailed description of these approaches see the Davidson & Co. paper.) To show the range of possible mortgage rate impacts based on the credit quality of the mortgage pool, Exhibit 1 excerpted from the Davidson & Co study, p provides a matrix of the Annual Credit Charges (ACC) for the Senior/Sub approach and for the Capital Requirement approach with three different ranges of FICO scores across the top ( , , and ), and three different ranges of LTV ratios down the side (80-85, 85-90, and 90-95). In the 80-85% LTV bucket in the FICO score range, the ACC ranges from 63 to 99 basis points (bps). In the 85-90% bucket with FICO scores, the range is 90 bps to 142, and in the 90-95% bucket, again with FICO scores, the range is 121 bps to 192. Furthermore, if a FICO score bucket of was selected with LTVs, the range would be 171 to 276 bps. As noted earlier, the Davidson & Co. study found that for loans with FICO scores greater than 750 and LTVs below 80% mortgage rates would increase by approximately 25 basis points. However, based on the ranges of Annual Credit Charges in Exhibit 1, and adding 14 bps to calculate Annual Credit Costs, it is reasonable to surmise that they could also be as high as basis points for the different more realistic - assumptions where FICO scores are in the % range. With this in mind, we concluded that based on the Davidson and Co. study, it would be reasonable and in fact important to examine a range of possible mortgage rate increases starting as low as 25 basis points and going as high as 150 basis points. 7 Modeling the Impact of Housing Finance Reform on Mortgage Rates, Andrew Davidson & Co., Inc., January, 2013, p 2. 3

8 Exhibit 1 Annual Credit Charges Calculated for Selected FICO and Loan to Value (LTV) Ratios These charges are calculated by Andrew Davidson & Co. for Selected FICO Scores and Loan to Value the Impact of Housing Finance Reform on FICO Scores Loan to Value Ranges ACC 1 (bps) Using Sr Sub Model ACC(bps) Using Capital Requirement Model ACC(bps) Sr Sub Model ACC(bps) Capital Requirement Model ACC(bps) Sr Sub Model ACC(bps) Capital Requirement Model ACC = Annual Credit Charge 2. Sr Sub=Senior Subordinated Securitization Structure Model The Zandi and deritis Study The study by Zandi and deritis also outlines a range of possible impacts. This paper will focus primarily on the mortgage rate impact of the Corker- Warner legislation in the Senate Banking Committee. However, Zandi and deritis also review the mortgage rate implications of three other proposals: the current system, a nationalized system, and the PATH (Protecting American Taxpayers and Homeowners) Act passed by the House Financial Services Committee. (Appendix 1 provides a summary of the mortgage rate impacts of all the proposals.) Regarding the Corker- Warner legislation, Zandi and deritis estimate a rise in g- fees to the range of 94 to 119 basis points including: (1) Cost of capital of 75 basis points for the first 5% capitalization and 20% before- tax return on equity (ROE); (2) Cost of capital of 5 20 basis points for the next 5% capitalization from capital markets; (3) Administrative costs of 10 basis points; (4) Expected losses of 9 basis points; (5) Mortgage Insurance Fund 0-5 basis points; (6) Market Access Fund 5-10 basis points; and (7) Benefits of a more efficient system reducing g- fees by 10 basis points. 4

9 By comparison, g- fees for mortgages currently guaranteed by Fannie Mae and Freddie Mac are in the range of 52 basis points (including a 10 basis point charge for the Treasury in order to pay for the payroll tax deduction.) Subtracting the current g- fee, the range of estimates in the Zandi/deRitis study suggests that mortgage rates will need to increase by at least 42 to 67 basis points if the Corker- Warner legislation is implemented. (94 to 119 minus 52 = basis points.) In their analysis Zandi and deritis identify a Mortgage Insurance Fund (MIF) of 0-5 bps to cover the cost to the government to insure catastrophic risk, but note that the MIF would be bps in the first 15 years as the fund is built up an increase for the first 15 years of 10 bps over their original estimate. They also subtract 10 bps for a more efficient system something that could be done without the passage of Corker Warner. If these 20 bps are added to the higher end of their cost estimate, then the possible mortgage rate increase would go to 87 bps. In addition, the Zandi/deRitis study assumes FICO scores of 754 and a LTV of 80 %, both high assumptions (that illustrate the tight credit of our current mortgage market). The study therefore points out that the rise in g- fees of their low estimate be greater for loans to less creditworthy borrowers who are still eligible for a government guarantee, particularly during recessions when investors demand points above, if more realistic, consumer friendly assumptions are made where FICO scores are in the range of and LTVs are between 85-95%, we concluded from the Zandi/deRitis study that it would be reasonable to examine possible mortgage rate rises from 42 to 150 basis points. Estimates Made in an Environment of Tight Credit It is important to remember that the base case estimates from the two studies on the impact of housing finance reform cited above both rely on assumptions of high credit standards FICO credit environment is holding back the housing market and the economy. Hopefully, it will begin to ease soon, and mortgage credit will be extended beyond these pristine levels to homebuyers whose FICO scores are closer to the historical range before the Great Recession of This historical range is demonstrated in Exhibit 2 which displays FICO scores from Family Loan Whereas FICO scores from , before the Great Recession, averaged 714, after the recession they were above 760 for In addition, in each of the years between , over 35% of the loans had FICO scores below The dataset was released by Freddie Mac at the direction of its regulator, the Federal Housing Finance Agency (FHFA) in December, It includes data on a portion of Single Family mortgages acquired by Freddie Mac and includes approximately 17 million loans. 5

10 Exhibit 2 Some of the consequences of this environment of tight mortgage credit are highlighted in a paper by Lewis Ranieri, Kenneth Rosen, Mark Goldhaber and broad cross- section of borrowers. Since the introduction of credit scores in the economic downturn, a 620 FICO with 5 % down was an insurable prime loan. It used to be that below 620 was considered a non- 9 Additional problems caused by this current tight credit environment were pointed out by Elizabeth A. Duke, a Member of the Board of Governors of the Federal Reserve System, in a talk given on May 9, She noted that the drop in mortgage originations between 2007 and 2012 has been most pronounced among borrowers with lower credit scores prime purchase mortgages fell about 30 percent for borrowers with credit scores greater than 780, compared with a drop of about 90 percent for borrowers with credit scores between 620 and 680. Originations are virtually non existent for borrowers with credit scores below 620. The distribution of credit scores in these purchase origination data tells the same story in a different way: The median credit score on these originations rose from 730 in 2007 to 770 in 2013, whereas scores for mortgages at the 10 th 10 9 Ranieri, Lewis S., Rosen, Kenneth T., Goldhaber, Mark, & Lepcio, Andrea. 10 Elizabeth A. Duke, Member, Board of Governors of the Federal Reserve System, May 9, 2013, Housing Policy Executive Council, Washington, D.C., p. 2. The calculations in Exhibit 3 are based on data provided to the Federal Reserve Board by McDash Analytics, LLC, a wholly owned subsidiary of Lender Processing Services, Inc. The. hose eligible for sale to the government- sponsored enterprises (GSEs) as well as 6

11 Exhibit 3 Credit Scores on New Prime Mortgages This exhibit is found in Conditions, May 9, 2013, Housing Policy Executive Council, Washington, D.C., Figure 7, Page 7. A Range of Cost Impacts In light of the discussion above, this paper will identify a range of possible mortgage rate increases that may come with housing finance reform, and use this range to examine the potential impact of housing finance reform on homebuyers and on the economy. Based on the Davidson & Co. Study, the Zandi and deritis Study, and considerations related to alternative FICO scores and LTVs (with high FICO scores of assumptions with FICO scores in the %), we concluded that the range of mortgage rate increases could vary from as little as 25 basis points to as high as 150 basis points, and the range of estimates should include: (1) 25 basis points (the lowest scenario from the Davidson & Co. Study); (2) 42 basis points (the minimum mortgage rate rise in the Zandi/deRitis study); (3) 100 basis point (a middle estimate based on a more likely distribution of FICO credit scores and LTVs); and 7

12 (4) 150 basis points (a higher, but still realistic, scenario where FICO scores are closer to the historical range of % range). It is worth noting that the range of mortgage credit costs could be even higher. As LTV greater than 90% would be more than 10 times higher (than 25 bps per year) impacts on home buyers, housing starts and the economy will also rise accordingly. Impact on Home Buyers An increase in mortgage rates would mean that fewer prospective home buyers would qualify for mortgages. Although other key criteria are used in mortgage underwriting, the extent of the effect of higher rates on the number of buyers can be seen by comparing monthly payments for principal, interest, taxes and insurance (PITI) to household incomes. Typically, the qualification standard for conventional loans is a ratio of 28 percent PITI to gross income. Comparing the required payments to the distribution of incomes among all households and among those buying homes provides a measure of the extent to which potential homebuyers would be disqualified. The possible effect on construction and on the overall U.S. economy can then be calculated as described below. The disproportionate social impact is discussed as well. The 2013 median existing single- family home price was $197,400. With a down payment of 20 percent, current mortgage rates of 4.5 percent, and property tax plus insurance of 1.5 percent of home value, the monthly payment would be $1,047, as shown in the first column of Exhibit American Community Survey, 55.5 percent of all households had sufficient income to make that payment without exceeding the 28 percent standard. Clearly, not all of those 64.3 million households sought to buy homes. Of the 3.50 million who did buy homes in the 12 months before they were surveyed, 68.3 percent would qualify to buy the 2013 median- priced home The 2012 ACS data indicate that there were only 3.5 million buyers. This is lower than the number of existing and new homes sold in either 2011 or 2012 (over 4 million). The difference reflects the fact that not all home purchasers are owner- occupants, especially purchasers of distressed properties. The demographic characteristics of the 2012 home buyers are presumably somewhat similar, however, to the larger number of buyers in other years. 8

13 Exhibit 4 Estimated Effect of Increases in Mortgage Rates on Qualification, Construction, and GDP Mortgage Rate No Change +25 bps +42 bps +100 bps +150 bps (4.5%) (4.75%) (4.92%) (5.5%) (6.0%) House Price $197,400 $197,400 $197,400 $197,400 $197,400 LTV 80% 80% 80% 80% 80% Monthly Payment $1, $1, $1, $1, $1, PITI $ $ $ $ $ Tax $ $ $ $ $ Insurance $65.80 $65.80 $65.80 $65.80 $65.80 Monthly Payment Change $23.63 $39.88 $96.49 $ Total Households (2012 ACS) 115,969, ,969, ,969, ,969, ,969,580 Number of Households Qualifying 64,340,421 63,020,811 62,345,029 59,949,493 57,390,534 Percent Qualifying 55.5% 54.3% 53.8% 51.7% 49.5% Disqualified by Payment Increase -1,319,610-1,995,392-4,390,928-6,949,887 Percent of Currently Qualified -2.1% -3.1% -6.8% -10.8% Home Buyers in ,504,391 3,504,391 3,504,391 3,504,391 3,504,391 Number of Buyers Qualifying 2,392,135 2,351,943 2,330,752 2,257,172 2,168,406 Percent Qualifying 68.3% 67.1% 66.5% 64.4% 61.9% Disqualified by Payment Increase -40,192-61, , ,729 Percent of Currently Qualified -1.7% -2.6% -5.6% -9.4% 2013 Single-Family Home Purchases 5,051,000 5,051,000 5,051,000 5,051,000 5,051,000 New Homes Completed 568, , , , ,000 Existing Homes Sold 4,483,000 4,483,000 4,483,000 4,483,000 4,483,000 Reduction from disqualified buyers -84, , , ,404 SF Starts Impact (.33 x disqualified) -28,006-42,771-94, ,893 Impact on GDP ($billions) Number of Households, Buyers qualifying based on whether monthly payment exceeds 28% of income Buyers are households who own homes and reported moving in during 12 months prior to survey. Sources: Census Bureau, 2012 American Community Survey; New Residential Construction National Association of Realtors, Existing Home Sales Impact of Higher Rates on Qualifying Households and Buyers An increase of 25 bps - bringing mortgage rates to 4.75 % - would raise the monthly payment by $23.63 to $1,071 and reduce the number of 2012 households who would qualify by 1.3 million, or 2.1 percent. Among those who bought in and who would have qualified to buy the median- priced home before the rate change, 40,192, or 1.7 percent, would become disqualified. Greater increases in mortgage rates would obviously disqualify more prospective home buyers. An increase of 42 bps (bringing mortgage rates to 4.92%) would raise the monthly payment to $1,087, disqualifying 61,383, or 2.6 %, of the 2012 homebuyers who would have qualified at the initial rate. An increase of 100 bps (with mortgage rates at 5.5 %) would mean monthly payments of $1,143, disqualifying 4.4 million total households, including 134,963 or 5.6 % of

14 homebuyers. And an increase of 150 bps (with mortgage rates at 6.0 %) would disqualify 6.95 million or 10.8 percent of the households and 223,729 or 9.4 percent of the 2012 homebuyers who would qualify at the initial mortgage rate. Impacts on Different Population Groups Looking at income distributions by race and age provides further perspective on the effects of rate increases from a restructured housing finance system. African- Americans and Hispanics already had substantially lower homeownership rates than non- Hispanic whites, with the gap widening during the Great Recession. Whether or not they are currently homeowners, the share of minority households who would qualify for a mortgage on the median- priced home at current mortgage rates is lower than the share among all households (Exhibit 5). Only 38.8 percent of black households and 44.9 percent of Hispanic households had the income needed to qualify, compared to 59.7 percent of whites and 60.2 percent of other race (largely Asian) households. An increase of 150 bps would disqualify 10.8 percent of all households who would qualify at current rates, but among blacks and Hispanics, 15.0 percent and 14.5 percent, respectively, would no longer qualify. Exhibit 5 Mortgage Rates and Qualifying Households by Race and Ethnicity Non-Hispanic Hispanic All White Black Other Race Total Households 80,837,096 13,791,324 7,339,740 14,001, ,969,580 Buyers 2,677, , , ,899 3,504,391 Other Owners 55,145,040 5,714,668 3,820,148 6,042,942 70,722,798 Renters 23,014,299 7,874,381 3,274,132 7,579,579 41,742,391 Ownership Rate 71.5% 42.9% 55.4% 45.9% 64.0% Qualify at 4.5% Number 48,292,999 5,351,802 4,415,114 6,280,506 64,340,421 Percent 59.7% 38.8% 60.2% 44.9% 55.5% Disqualified by Higher Rate +25 bps -903, ,919-73, ,027-1,319, bps -1,375, , , ,375-1,995, bps -3,031, , , ,649-4,390, bps -4,832, , , ,084-6,949,887 Share of 4.5% +25 bps -1.9% -3.0% -1.7% -2.9% -2.1% +42 bps -2.8% -4.4% -2.5% -4.4% -3.1% +100 bps -6.3% -9.6% -5.6% -9.6% -6.8% +150 bps -10.0% -15.0% -9.1% -14.5% -10.8% Based on 2012 American Community Survey Assumes house price of $197,400, LTV=80%, Taxes and Insurance=1.5% of home value 10

15 A similar pattern is found when ability to qualify by age is considered (Exhibit 6). Among young households, fewer would qualify at current rates compared to middle- aged households, and the share of those currently qualified who would no longer qualify after an increase in rates is also greater among young households. The effect on the 25 to 34 year old group is especially noteworthy, since they are the households most likely to buy homes. The 2012 data show 5.5 percent of households in that age group buying homes in the 12 months preceding the survey, compared to only 3.0 percent of all households. Exhibit 6 Mortgage Rates and Qualifying Households by Age of Householder Under &Up All Ages Total Households 4,653,930 17,701,340 21,011,532 24,090,320 22,025,254 26,487, ,969,580 Buyers 195, , , , , ,064 3,504,391 Other Owners 392,163 5,763,553 11,535,286 16,255,761 16,330,809 20,445,226 70,722,798 Renters 4,066,426 10,963,016 8,697,222 7,190,072 5,191,741 5,633,914 41,742,391 Ownership Rate 12.6% 38.1% 58.6% 70.2% 76.4% 78.7% 64.0% Qualify at 4.5% Number 1,107,910 9,622,707 13,508,725 15,819,895 13,450,905 10,830,279 64,340,421 Percent 23.8% 54.4% 64.3% 65.7% 61.1% 40.9% 55.5% Disqualified by Higher Rate +25 bps -57, , , , , ,746-1,319, bps -82, , , , , ,736-1,995, bps -172, , , , ,116-1,015,133-4,390, bps -267,003-1,269,709-1,279,168-1,364,468-1,261,774-1,507,765-6,949,887 Share of 4.5% +25 bps -5.2% -2.7% -1.9% -1.6% -1.7% -2.4% -2.1% +42 bps -7.5% -3.9% -2.8% -2.5% -2.6% -3.9% -3.1% +100 bps -15.6% -8.2% -5.8% -5.3% -5.9% -9.4% -6.8% +150 bps -24.1% -13.2% -9.5% -8.6% -9.4% -13.9% -10.8% Based on 2012 American Community Survey Assumes house price of $197,400, LTV=80%, Taxes and Insurance=1.5% of home value These calculations are based on an assumption that rates would increase equally for all prospective buyers. Depending on how mortgage market reform is structured, however, rates may increase more for those closer to the margin for qualification than for those with ample income and wealth, and with higher credit ratings. If there is an increased differential between the rates charged to those who easily qualify compared to those with less exalted profiles, the number disqualified as a result of changes in the housing finance system would be greater. 11

16 Impact on Housing Sales and Starts Housing sales and starts are sensitive to mortgage rates. Exhibit 7 illustrates the strong negative relationship between mortgage rates and new home sales from June 2010 to December In 2011, as mortgage rates declined, new home sales finally started to recover from record- low volumes. In mid- 2013, rates on 30- year mortgages jumped, rising from 3.35 percent in the first week of May to 4.51 percent in mid- July. That was quickly reflected in new home sales. When mortgage rates temporarily dipped in October, new home sales again responded by increasing. Exhibit 7 Mortgage Rates and New Home Sales Year Mortgage Rates New Home Sales Jun- 10 Sep- 10 Dec- 10 Mar- 11 Jun- 11 Sep- 11 Dec- 11 Mar- 12 Jun- 12 Sep- 12 Dec- 12 Mar- 13 Jun- 13 Sep- 13 Dec- 13 Although most housing is supplied by the existing stock, the size of the existing housing stock is fixed, so changes in housing demand are disproportionately reflected in the volume of new construction and in new home sales. In 2005, 21 percent of the homes that were acquired were new homes statistics on new home sales). From 2005 to 2011, the decline in purchases of new homes accounted for 33 percent of the decline in total home purchases, and new homes as a share of total purchases fell to about 11 percent. As sales rebounded from 2011 to 2013, new home sales increased by 40 percent, compared to an increase of 18 percent in existing home sales. 12

17 For all of 2013, total purchases of single- family homes were over 5.05 million (Exhibit 4). If an increase of 100 bps in mortgage rates due to changes in the housing finance system were to disqualify 5.6 percent of aspiring home buyers who would qualify under the current system, as the calculations above suggest, and that produced a proportionate change in total home purchases, it would mean 285,000 fewer purchases. Translating differences in mortgage rates and in qualifying households into changes in home sales and construction is imperfect and would depend on many other factors, but for purposes of illustration, at least, it would be reasonable to expect that the effect on construction of new single- family homes would be about a third of the projected decline in total sales. The resulting estimated change in housing starts of 94,000 units from a 100 bps change in rates is in line with estimates from econometric models. 12 Exhibit 4 outlines the full range of possible impacts of housing finance reform on housing starts. If mortgage rates rise by only 25 or 42 bps, the impact could be a drop in housing starts of 28,000 to 42,000. However, if mortgage rates rise by 100 to 150 bps, housing starts could drop by as much as 94,000 to 150,000. Impact on the Economy In 2013, the average value of construction per new single- family unit was $284,557, according to Census Bureau estimates used in calculating GDP. Thus a change of 94,000 in the number of new single- family units, brought about by a 100 bps rise in mortgage rates, translates into a direct impact of about a $27 billion drop in the nation's output (Exhibit 4). In addition, the full range of impacts would be a drop in GDP of $8.0 billion for a 25 bps rise, a drop of $12.2 billion for a 42 bps rise, and a drop of $44.4 billion for a 150 bps rise. It is also worth noting that changes in other economic activity associated with new housing, such as production of furniture and appliances and a wide range of services, would occur as well. As Exhibit 8 shows, the level of new home construction during was atypical. For the 50- year period from 1964 to 2013, single- family starts averaged 1.04 million, with total housing starts averaging 1.46 million. In 2013, there were only 618,000 single family homes started, out of total housing starts of 923,000. As the Great Recession began, a glut of homes- - reflected in high vacancy rates, depressed prices, and foreclosures- - limited the need and demand for construction. Moreover, despite the lowest mortgage rates in more than 50 years, 12 Advisers typically indicate that a 100 bps increase in mortgage rates would cut housing starts by 100,000 to 200,000 units, with a lag of several quarters before the maximum impact occurs. The models suggest that the effect of a permanent change in mortgage rates dissipates over time, but that occurs only after several years. The simplified analysis based on qualification may also imply that eventually the effects would dissipate. Households who are unable to qualify immediately may qualify later. Moreover, unless the number of households is sensitive to mortgage rates, over time the lack of supply will lead to pressures from tight markets, stimulating production. 13

18 the availability of credit was limited, further reducing new home demand. But the excess supply of housing is being steadily absorbed, and gradual improvements in employment and output have helped support the beginnings of a recovery. If production recovers over the next few years to levels closer to the historical average, as is likely to occur given underlying demographic trends and even modest replacement demand, the differential effect of higher mortgage rates would be correspondingly greater as well. 2,500 Exhibit 8 Housing Starts ,000 Multifamily Single-Family 1,500 1, Those households who are unable to qualify for mortgages would be likely to rent apartments or homes instead, suggesting increased demand for construction of new rental housing that might offset the decline in demand for homes for owner- occupancy. Changes in the mortgage market that would raise financing costs for home buyers would raise financing costs for rental housing as well, however, probably reducing, rather than increasing, construction in that segment. Moreover, to the extent that failure to qualify for a mortgage means living in multifamily housing rather than single- family, construction output would be lower. The contribution per housing unit of multifamily construction to GDP is much smaller than for single- family which averaged $98,922 in Lower volumes of new home construction would mean less employment in the construction industry, which has already seen a severe contraction. Payroll employment in residential builders and subcontractors fell from 3.4 million in

19 to 2.1 million in Indeed, the decline in residential construction employment represents more than half of the decline in total nonfarm employment since the pre- recession peak. An even greater number of jobs would be affected in industries that supply construction. About two- thirds of the cost of building homes is the cost of materials and purchased services, and most of that ultimately goes to labor in building materials, distribution, and other industries. Conclusion This paper has examined the impacts of a range of possible increases in the cost of housing finance due to ch finance system. If the impact on mortgage rates is 25 bps or 42 bps as suggested in the base case of the Davidson & Co. and Zandi/deRitis studies using assumptions of borrowers having very high credit ratings and low loan- to- value ratios the impact of higher rates would be relatively modest in terms of monthly mortgage payment, numbers of buyers that would be disqualified, and the impacts on housing starts and GDP. However, if the effect is to raise rates by 100 bps or 150 bps which may occur if more realistic credit assumptions are applied and financing is provided to borrowers with good but not outstanding credit (i.e. FICO scores between and LTVs of percent), the impact on housing, the economy, and the ability of households to become homeowners would be substantial. 15

20 Appendix 1 Chart taken from Cost of Housing Finance Reform by Mark Analytics, Economic and Consumer Credit Analytics, November, 2013, p.2. 16

21 About the Authors Kent Colton is a Senior Fellow at the Joint Center for Housing Studies of Harvard University. He is also the President of The Colton Housing Group, a housing research and consulting company. He is the author of a number of books and articles on housing finance, housing policy, and a range of management issues. He is the former CEO of the National Association of Home Builders where he served from 1984 l999. He was an Executive Vice President at Freddie Mac from In Housing. Colton was a also a professor of public management and finance at Brigham Young University from l978 l981, and an Associate Professor at the Massachusetts Institute of Technology (MIT) from Colton received an M.P.A. from Syracuse University in 1968 and a Ph.D. in Urban Studies and Planning from MIT in In 1974, he was chosen as a White House Fellow. Colton and his wife, Kathryn, have five children. They live in McLean, Virginia. Michael Carliner is an economic consultant specializing in areas related to construction and housing markets. He has done extensive work involving economic measurement and econometric analysis, industrial organization, and the relationship between construction and macroeconomic activity. His principal focus during the past few years has been on affordable housing, transportation, and energy. He has been responsible for reports on rental housing and energy for the Harvard Joint Center for Housing Studies, where he is a Fellow, and has been a consultant to government agencies and companies regarding housing policy, building materials, demographics, and construction activity. From 1984 to 2007, Michael was Staff Vice President for Economics at the National Association of Home Builders, responsible for forecasting, survey research, and analysis of government policies. Before joining NAHB, he was Director of Real Estate and Construction Economics at Chase Econometrics, Vice President of Regional Data Associates, Senior Economist at Dynamics Associates, and a Research Fellow at the University of Pennsylvania. 17