The Price of Homeowners:An Examination of the First-time Homebuyer Tax Credit

Erik Hembre∗

October 20, 2016

Abstract

A major policy response to the 2008 housing crisis was the First-time Homebuyer Tax Credit,worth up to $8,000. To estimate the tax credit effets on homeownership, I construct a quarterly first-time homebuyer time-series using American Housing Survey data. Using both an event-study and adifference-in-difference framework, I estimate the tax credit induced 301,900 first-time homeown-ers and calculate the government paid $24,180 per additional homeowner. I find no evidence thatfirst-time homebuyers bought more expensive houses or increased default rates. Estimating state-and MSA- level effects I find a strong correlation between effect size and average home values,with a doubling in average home values implying a drop in effect size by 19.7 percentage points.These local effects also reveal larger effects in areas with smaller housing busts, had lower mortgagedelinquency rates, and have higher housing supply elasticity.

JEL Classification Codes: R21, H5, I38Keywords: Homeownership, Policy, Public Assistance

∗Erik Hembre: University of Illinois at Chicago, ehembre@uic.edu.

1 Introduction

One of the largest policy responses to the 2008 housing bust was the First-time Homebuyer Tax Credit(FHTC). Between April 2008 and September 2010, this novel program offered $8,000 to first-timehomebuyers in an attempt to boost housing demand. Over 3.3 million households claimed the credit ata monetary cost of $21.1 billion. Scant prior evidence existed before the policy to predict to what extenthouseholds would respond to the tax credit and minimal investigation into its effects has been conductedsince. This paper examines the FHTC effects by first estimating the number of households induced intohomeownership at both a national and local level and then analyzing these results to determine whereand why the policy was most effective.

Prior to 2008, a tax credit targeting first-time homebuyers had never been offered at a federallevel, and very little experience existed with it at a local level. In theory, a tax credit targeting first-time homebuyers may be a useful tool for policymakers during a housing bust. If house prices dropand induce a rash of foreclosures, these newly vacant homes are a shock to housing supply and pushprices down further. To bring house prices back towards equilibrium, these vacant homes need tobe filled with current non-owners, demolished, turned into rental properties, or else forced off of themarket. Offering a tax-credit to first-time buyers helps induce more renters into homeownership, fillingthese foreclosed homes. A homeownership tax credit can additionally boost general housing demand,prompting first-time buyers to purchase larger homes, and in turn improving local economic conditions.A homeownership tax credit may mechanically raise home values, as Hilber and Turner (2014) findsthat the benefits of the mortgage-interest deduction are mostly capitalized into home values and couldsimilarly capitalize the value of the FHTC. These benefits are in addition the the positive externalitiesassociated with homeownership which are often cited to justify the mortgage-interest deduction.

Policymakers must also weigh negative aspects of a homeownership tax credit. As with anysubsidy, the FHTC creates a deadweight loss by inducing a suboptimal homeownership decision. Ad-ditionally, renters on the margin of homeownership may be more likely to eventually default on theirmortgage and could result in more foreclosures, which are costly to homeowners and the general pub-lic. If the tax credit raises house prices, the tax credit benefits are split between new and previoushomeowners. In general, homeowners are more wealthy than renters, so the tax credit works againstany redistributive goals of current tax policy. Similar to most housing programs, the FHTC also carriesa large price tag. Even if most of the credit is a wealth transfer, the FHTC benefits must be weightedagainst the deadweight loss of raising the funds to pay for it.

The effectiveness of the FHTC depends on the elasticity of homeownership, or the analogouslythe “price” of homeowners. By price of homeowners I am referring to cost per for each inducedhomeowner. The more elastic renter demand is towards homeownership, the cheaper it will be to buyhomeowners and fill vacant homes. Predicting the response of first-time homebuyers to the FHTC isdifficult both because it is a new program and because of problems in quantifying the relative sizeof the subsidy. On one hand, $8,000 is only about three percent of the cost of the average homepurchase, so the FHTC could be seen as a fairly trivial subsidy. However, the tax credit incentivizeshomeownership not necessarily housing, which costs far less than purchase price of a home. That is,a household could buy a new home, claim the FHTC, sell the home after three years, and move back

1

to renting. In this case, the $8,000 needs to be weighed against the “user cost” of homeownershipfor those three years which includes the cost of raising the downpayment amount, mortgage financingcosts, and the transaction costs of moving and selling the home. This homeownership cost is notdirectly observable and heterogeneous across households. Knowing the size of the homeownershipcost, the elasticity of homeownership, and the price of homeowner are crucial for evaluating the FHTC,predicting counterfactual FHTC policy effects, and for better understanding housing tenure choices.

This paper focuses primarily on the extensive margin FHTC effects on homeownership. Insteadof directly estimating the FHTC effects on macroeconomic variables, such as house prices and un-employment, I estimate the FHTC effects on homeownership and then indirectly check if this effecttranslated into changes in macroeconomic variables. As a secondary analysis, I consider potential in-tensive margin effects of the FHTC, such as size of home purchase, mortgage financing terms, andmortgage outcomes.

Several empirical challenges arise in identifying FHTC effects. One challenge is separating FHTCeffects from other housing-focused programs enacted during the Great Recession, such as the HomeAffordability Modification and Refinancing Programs, quantitative easing (which purchased mortgage-backed securities), and expansions to the federal conforming loan limit. These other programs largelyfocused on previous homeowners. To estimate the FHTC effects, I isolate the response of first-timehomebuyers, as opposed to previous owners, during the tax credit period. Using income eligibilitycutoffs to separate effects by eligible and non-eligible first-time homebuyers, helps me verify that FHTCeffects were concentrated only among the affected treatment group.

Another challenge is tracking first-time homebuyers as a group, which few datasources include ona systematic, nationwide basis. To overcome this, I construct a national, quarterly time-series of first-time homebuyer purchases. I do this by combining data from the American Housing Survey (AHS),which asks households of their homebuyer status and moving date, and the Home Mortgage DisclosureAct (HMDA) which tracks US mortgage originations. AHS data also allows me to investigate otherchanges to first-time homebuyers during the tax credit period, such as household income, downpaymentamount, reason for moving, and home size.

A last empirical challenge is identifying local FHTC effects. Local effects are important bothfor policy analysis and for understanding the homeownership decision. National-level FHTC effectsmay mask significant state- and MSA-level variation. If the FHTC was meant as a policy to counteractlocal housing busts we should be interested in whether FHTC effects were greater or smaller in areasexperiencing larger housing busts. For future potential FHTC policy we are also interested whetherother local housing market characteristics, such as the percent of renters, housing supply elasticity, andaverage home values, predict FHTC success. Average home value in particular is interesting because itprovides variation in the effective tax credit subsidy amount. That is $8,000 subsidizes a larger fractionof home purchases in Nebraska than in California, and so we would expect larger FHTC effect in areaswith lower home values. The relationship between tax credit percent of home purchase and effect sizeprovides an estimate of the elasticity of homeownership.

Unfortunately, the AHS data is too small to detect local FHTC effects, so instead I combine loan-level mortgage origination data from Fannie Mae, Freddie Mac, and the Federal Housing Administra-

2

tion which provide homebuyer status, origination date, and property location information. Prior to thehousing crisis, these three entities comprised less than fifty percent of the first-time homebuyer mort-gage originations. During and following the crisis their market share increased to above eighty percent,so a majority households claiming the FHTC are contained in this sample. The shifting first-time home-buyer market share during the housing crisis makes this data suboptimal for estimating national FHTCeffects, but alternatively provides a useful for measuring changes in first-time homebuyer originationsfollowing the FHTC expiration at the state and MSA levels.

Given these empirical challenges, this study utilizes both an event-study and a difference-in-difference framework to estimate FHTC effects. The event-study measures the total change in first-timehomebuyers during the tax credit time period relative to expectations. However, we may be worriedthat house price changes and a shifting structure of the mortgage origination market during the taxcredit period bias the event-study results. To counteract this, I split the first-time homebuyer group intoa high- and low- income group. Higher-income first-time homebuyers were not eligible for the FHTC,but were similarly affected by house price changes and the mortgage origination market, providing anappropriate control group for the difference-in-difference estimation. At the local level, I also use pre-vious owners as a control group for first-time homebuyers as they were also ineligible for the tax creditand because income information is not available for the local data.

I find that the FHTC increased first-time homebuyer purchases by 301,900 or 8.5 percent betweenApril 2008 and September 2010. A back-of-the-envelope deadweight loss estimate of $7.5 billion trans-lates into the government paying $24,180 per induced homeowner. The FHTC effect is concentrated inthe second two iterations of the tax credit, after it ceased requiring tax credit repayment and 10 percentof induced homeowners expedited their homeownership transition by a year or less. Induced homeown-ers were more likely to be younger and use a smaller downpayment, but did not spend more on housingor exhibit higher default or prepayment rates. State- and MSA-level analysis reveal a negative corre-lation between FHTC effect size and magnitude of housing bust. A main driver of FHTC effect sizevariation between states is average home values. More expensive states such as Hawaii and Californiadisplayed undetectable effects while less expensive states such as Nebraska or Oklahoma increasedfirst-time homebuyers by twenty percent or more. State FHTC effect size and utilization rates haveonly marginal correlations with changes in house prices, housing starts, vacancy rates, or employment,suggesting the tax credit had little stimulus effects on local housing markets. Overall these findingsindicate households clearly responded the tax credit while highlighting both the importance of policydesign to target specific areas and revealing a high price of boosting homeownership.

2 The First-time Homebuyer Tax Credit

In July 2008, Congress authorized the Housing Recovery Act that offered first-time homebuyers atax rebate. Initially set to expire in July 2009, Congress expanded the tax credit as part of the 2009American Recovery and Reinvestment Act and extended its deadline through November 2009. A finalversion of the FHTC was included in the Worker, Homeownership, and Business Assistance Act. Fulldetails of each iteration of the FHTC are found in Table 1. The initial rebate offered the lesser of

3

10% of the home purchase price or $7,500 to first-time homebuyers either making under $75,000 forsingle households or $150,000 for joint filers, repayable over a fifteen year period.1 The second FHTCiteration increased the maximum tax credit to $8,000 and changed the rebate to non-repayable taxcredit, significantly increasing its value. The third FHTC iteration loosened income limits for singlehouseholds to $125,000 and to $225,000 for joint filers, as well as offering the tax credit to previousowners in addition to first-time homebuyers.

A 2010 GAO report on the FHTC reports 3.3 million households claimed the federal first-timehomebuyer tax credit between 2008 and 2010.2 States with the highest utilization rates are locatedin the Mountain and Midwest regions.3 These areas have on average lower home values and higherhomeownership rates. However, separating the FHTC effect size from its utilization rate is an keyproblem this paper attempts to solve. Utilization is important for account where money from the FHTCwas distributed to, akin to how a wealth shock may affect a wide array of economic variables. But theFHTC effect on homeownership and the housing market was the target of the tax credit and providesnew information on household decisions regarding the rental to homeownership transition.

Open debate exists as to the effectiveness of the FHTC. In evaluating the FHTC, Baker (2012)states:

There can be little doubt the the first-time homebuyer tax credit had a large impact onthe country’s housing market. Sales took off immediately after the credit took effect...Theresult was that many people were persuaded to buy homes at bubble-inflated prices whowould have otherwise purchased them at prices that were more consistent with the longer-term trends in the housing market. This amounted to a substantial transfer of wealth fromnew homebuyers to home sellers.

Many housing indicators, including home sales, housing starts, housing permints, and vacancy rates,indeed began to recover during the FHTC eligibility. However separating out FHTC effects from otherhousing programs enacted during this time period remains difficult.

Several economic rationales could justify offering a tax credit to new homebuyers during a hous-ing bust. As house prices decline, more mortgages drop into negative equity which increases the like-lihood of delinquency and foreclosure. These foreclosed homes flood the supply of owner-occupiedhousing relative the number new home seekers, pushing home values down further and in turn induc-ing more foreclosures. To soak up this glut of vacant homes, they need to be filled by families notalready living in owner-occupied homes. A tax credit targeting first-time homeowners does exactlythat. Additionally, the FHTC may encourage prospective new homeowners to purchase a larger homeboth through a wealth effect and by relieving credit constraints. The FHTC could also be an effectivetool to distribute an economic stimulus. New homeowners often have little remaining savings after thepurchase and so may have a particularly high marginal propensity to consume.

1The tax credit is phased out in the $20,000 income range above each cutoff point for each iteration of the program.2Note these numbers differ somewhat from what the IRS reports, though IRS only reports for 2009 and 2010.3Utilization rates being defined as number of claimed tax credits divided by the number of renter households in the

state.

4

A federal homebuyer tax credit has not previously been implemented in the United States. Atthe local level, the only homebuyer tax credit I have located was offered in Washington D.C. worth$5,000 between 1997 and 2001. A policy brief on the tax credit reported an increase to DC houseprice appreciation of 4.9 percent compared to surrounding areas as a result of the tax credit, Tong(2005). This lack of previous evidence highlights the importance of a detailed investigation of thecurrent FHTC, especially if it is considered as a policy tool for future housing busts.

While the FHTC targets a relative unique population, tax credits or rebates are a common policytool used to affect consumer behavior. Using survey data, Shapiro and Slemrod (2001) examine the2001 tax rebate and find most households either saved or used the tax rebate to pay down existingdebt. However, Agarwal et al. (2007) use credit report data to find that while households saved andpaid down debt in the short term, their spending increased shortly after with 40% of the tax rebatebeing spent within nine months of received it. Ideally, to boost housing demand, households would usethe FHTC to either buy a home they would have otherwise rented or purchase a larger home. Thesestudies suggest that only a fraction of the tax credit would go towards increased spending, housing orotherwise. Since the FHTC is mostly appropriated as a lump-sum payment rather than as a percentageof the home value, we would expect a greater response on the extensive margin of buying the homerather than the intensive margin of the size of the new home.4

A recent tax credit similar to the FHTC is the “Cash for Clunkers” program of 2009, which offeredhouseholds $3,500 or $4,500 to subsidize new car purchases. Mian and Sufi (2010) find the programsignificantly boosted auto sales, though this boost was largely an inter-temporal shift in when cars werepurchased, and had negligible effects on employment, house prices, or home default rates. A similarworry may exist for the FHTC, whereby a jump in home sales may just be households who wouldhave become homeowners without the tax credit eventually, but just expedited this transition. Theseexpedited homeownership transitions offer only marginal benefits, as they simply shift the timing ofdecreased homeownership demand.

Existing literature examining the FHTC is limited. Dynan et al. (2013) evaluates the FHTC in twoways, first by comparing housing indicators to FHTC dates and second using state variation in homevalues within a difference-in-difference strategy. Using forecasting techniques, Dynan et al. (2013)finds a large positive effect of the tax credit, though forecasting during the turbulent 2008-2010 periodis difficult, and reports mixed results using the difference-in-difference approach. Dynan et al. (2013)then uses state variation of offered supplemental policies to find states which offered short-term loansor credits to have a positive effect on the housing market. While state-level supplemental policies areinteresting, there are drawbacks from using them in this analysis. Primarily, the state programs areof only small monetary value. Most just offered short-term, low-interest loans to finance the homepurchase. Only a few states offered additional tax credits. Georgia only offered $1,800, while Maineand Utah offered between $2,500 and $6,000 for less than a year each. California offered the up to$10,000, but was geared towards new homes instead of first-time buyers and also quickly ran out offunds for the credit. The short-term duration and varying eligibility requirements of these state plansraises also raises questions about their salience to potential homeowners, limiting their likely observed

4The FHTC is worth $8,000 as long as the home value is greater than $80,000. If less than $80,000 the tax credit isoffered at 10% of the home value.

5

impact.

3 Methodology

This section outlines the research methodology used to estimate FHTC effects. I employ both eventstudy and difference-in-difference approaches at both the national and local levels.

To estimate FHTC effects with an event study framework, I use the following equation:

Yt = β0 +δTreat+λX+ εt (1)

where Yt is the outcome of interest, the log of first-time homebuyer purchases at time t, X contains vari-ables controlling for time trends and εt is an iid error term. The coefficient δ , is the FHTC treatmenteffect size. Both Treat and δ are three-dimensional vectors in order to detect difference effect sizes byFHTC iteration. Specifically, δ = [δ1,δ2,δ3]. When estimating a single effect, δ1 = δ2 = δ3 = δ̄ andTreat is:

Treat =

{1 if 2008q2 ≤ t ≤ 2010q30 otherwise

When estimating separate effects by implementation, Treat is:

Treat =

[1,0,0] if 2008q2 ≤ t ≤ 2008q4[0,1,0] if 2009q1 ≤ t ≤ 2009q4[0,0,1] if 2010q1 ≤ t ≤ 2010q3[0,0,0] otherwise

The event study framework will measure the increase in first-time homebuyers during the FHTCeligibility period relative to what historical and proceeding trends would have predicted. Typically, postperiods are not used in estimating treatment effects since households may have changed their decisionsas a result of the earlier treatment. However in measuring the FHTC impacts, the post-period is crucialfor estimation because of the structural shifts occurring in the homebuyer market during the time period.I use two tactics to account for the possibility of households expediting their homeownership transition.First is to directly estimate the number of households which expedited their home purchase. I do thisby assuming this inter-temporal substitution effect decays linearly during the four quarters followingFHTC expiration, and including a “Substitution Effect” variable to reflect this. Alternatively, I also runregressions taking out the adjacent time periods to the FHTC eligibility which would have been mostaffected by these inter-temporal substitutions. These include the quarter prior to two quarters followingthe tax credit.

6

The biggest underlying assumption of the event study framework is that only difference betweentreatment and control time periods is that effect of the FHTC on first-time homebuyers. Given the largehousing market volatility and other housing programs going on during this time period, there is reasonto be weary of relying too hard on this assumption. To control for outside factors influencing householdhousing decisions, I use a difference-in-difference framework where eligible first-time homebuyers aretreated and the control group consists of non-treated households. Depending on the data source, thecontrol group are either higher-income first-time homebuyers or previous owners. These groups werenot eligible for the tax credit in the first two iterations of the program, making them appropriate controlgroups. In the third iteration of the tax credit, nearly all households were eligible for the tax credit,leaving no credible control group. In this case, the difference-in-difference effects for the third iterationresult in separate event-study FHTC effects by group.

The difference-in-difference estimation of FHTC effects uses the following equation:

Yi,t = β0 +β1 Group+β2 Treat+δ ( Treat x Group)+λX+ εi,t (2)

where Group is a dummy equal to one for the treated group and Treat is as defined above. Group isdefined either by income levels (high or low), or by homebuyer status (first-time homebuyer or previousowner). As in Equation (1), X contains time trend control variables and εi,t is an error term for eachgroup i in time t. δ is the outcome of interest, the effect of the FHTC on the treated group.

To estimate the price of homeowners I divide the total number of induced homeowners by the pro-gram cost. Induced homeowners is calculated by taking the difference in observed first-time homebuy-ers and predicted first-time homebuyers in the absence of the FHTC, given the regression coefficients.

4 Data and Empirical Implementation

Key to identifying FHTC effects are isolating the decisions of first-time homebuyers. Few datasourcestrack the behavior of first-time homebuyers, especially at a national scale. Perhaps the most commonlycited source is from the Survey Profile of Homebuyers and Sellers which is has been published bythe National Association of Realtors since the mid-1980s. The NAR only releases aggregated, annualstatistics from its survey of around 10,000 households, limiting its ability to estimate the FHTC effects,which was available for only portions of 2008 and 2010. Another source for tracking first-time home-buyers is the American Housing survey. Figure 1 compares the share of first-time homebuyers in theNAR versus the AHS. The two data sources report very similar levels and trends of first-time home-buyer shares between 2001 and 2012, beginning near forty percent, increasing to almost fifty percent in2009 and dropping back closer to forty percent in 2012. I focus on the AHS, which provides detailedsurvey data including a flag for first-time homebuyers and the moving date, allowing for a estimationof national first-time homebuyers at a quarterly rate. In the following sections I describe how to con-struct national and local level estimates of first-time (and previous owner) home purchases, and providesummary statistics of the data.

7

4.1 First-time Homebuyers Nationally

Data for the primary analysis of the FHTC comes from the American Housing Survey. The AHS beganin 1973 in order to survey housing and demographic characteristics and biennially surveys 55,000housing units nationally. The AHS asks respondents if (and when) they moved within the past twoyears, whether they ever owned a home before moving and their current tenure status. Between 2001through 2013, 23,162 AHS households reported moving, creating a sample of about 500 movers perquarter.5

One way to report the number of first-time homebuyers each quarter would be to count (or usea weighted count) of respondents claiming to be first-time buyers each quarter. This approach hastwo problems using AHS data. One is that the survey only asks about first-time buyer status for themost recent move. If a household moved more than once in the past two years and the first movewas a first-time purchase, then this would not be reflected in the data. This leads to an undercountingof first-time buyers for months further away from the survey date. Another problem is that the AHSsurveys households unevenly over the a four to five month period. This adds noise to the time series byweighting some time periods more than others, and is a difficult problem to correct properly given therelatively small number of movers each quarter.

I construct my quarterly, national estimate of first-time homebuyers using a three-step procedure.The basic idea is to get the quarter-by-homebuyer status- shares of home purchases from the AHS, andthen scale this by the total number home purchases each year. First, for each year I determine the shareof movers obtaining a mortgage for each quarter and homebuyer status, using AHS sample weightsto reflect a nationally representative sample. Using first-time homebuyer shares instead of counts re-duces the bias arising from the uneven staggering of interviews each quarter. Next, the quarterly buyershares are scaled by the total number of mortgages originated each year, according to Home MortgageDisclosure Act data.6 This gives an estimate of the total number mortgages originated for each quarter-homebuyer status. Lastly, to account for cash purchases, the quarterly mortgage originations by buyertype are divided by the percent of movers who used mortgage financing by buyer type and quarter inthe AHS. This same procedure is used to create quarterly home purchase time-series by income levelof first-time homebuyers, but splits the sample quarter-homebuyer status-income level buckets. I dividethe AHS sample into “high” and “low” income levels using a household income cutoff of $85,000.Income eligibility cutoffs for the FHTC vary over time and tax filer status. Initially, a single house-hold with an income under $75,000 can claim the full tax credit, but later on this threshold is raised to$125,000. Joint filers are initially fully eligible with incomes of $150,000 and then $225,000. The taxcredit is then phased out over the next $20,000 of income for all groups. I choose to split householdsinto high- and low-income groups at $85,000 because all households under this amount were alwayseligible for at least half of the tax credit in all period. Since I can not observe income filer status, I amunable to split the sample differently for joint and single households. Because some of the households

5Both FHTC eligibility and the other housing agency datasources consider households that have not owned a home inthe past three years to be “first-time homebuyers”. This is different than my definition based on the AHS, where first-timemeans “never have owned a home before”.

6Home Mortgage Disclosure Act mortgage origination counts are restricted to owner-occupied, first-lien, purchasemortgages. This data do not report first-time homebuyer status, and prior to 2003 do not report lien status.

8

in the high-income sample are eligible for the FHTC, my estimate will understate the true effect size.

Figure 2 displays national home purchases by homebuyer status at an annual rate, and the first-time homebuyer quarterly series, with vertical lines marking FHTC implementation dates. The left-hand panel shows that first-time homebuyer purchases peaked in 2004 before declining each year until2010. Home purchases by previous owners dropped significantly more than first-time buyers. Theright-hand panel zooms in on the seasonally-adjusted quarterly rate of first-time homebuyer purchasesbetween 2003 and 2012. While FHTC effects appear subtle in 2008, there is an increase in 2009and 2010 before the tax credit expired. The 2009 increase is not observed among previous owners,hinting at the role of the FHTC. Figure 3 shows first-time and previous owner home purchases splitby income level. Here we see that the 2009 increase in first-time homebuyers is concentrated amonglower-income households, while high income households home purchases increase in 2010 after theincome eligibility requirements are relaxed. This provides supporting evidence of the FHTC effectsince there is no similar pattern among previous owners.

In addition to counts of first-time and previous owner home purchases, AHS data provides insightwhat types of households the FHTC induced into homeownership and potential intensive margin effects.Figure 4 shows the share of mover households becoming renters or homeowners, split by whether thehousehold previously owned or rented their home. Here we see only a small bump in the share movingrenters becoming new homeowners during the FHTC eligibility period. This indicates that tax crediteffects can be attributed more to inducing the move for a renter to become homeowner as opposed toaltering the decision of households already planning on moving to become owners instead of renters.The AHS asks recent mover households what their primary reason for moving was. During the firstand second iterations of the FHTC, the percentage of first-time homebuyers cited “Changing TenureStatus” peaked at thirty percent, an increase of nearly fifty percent relative to the two years prior orfollowing the FHTC. Other reasons for moving such as “Establishing a New Household” or “Movingto a Bigger or Better House” did not display similar spikes during the FHTC eligibility period. Overthe past decade, the average age of first-time homebuyers has increased steadily by nearly three yearsto over thirty-five years old. However, the average first-time homebuyer age dropped by a full yearduring the second FHTC iteration, suggesting homeowners induced by the FHTC were younger thantypical first-time homebuyers.

4.2 First-time Homebuyers Locally

Equally as interesting as the national FHTC effects is the variation across local markets in FHTC effectsizes. The tax credit was created in direct response to the housing and financial crisis, but these crisishit some regions, such as California, Florida, Arizona, and Nevada, much harder than others. I estimatestate- and MSA- level effects and compare these effects to local housing market conditions, measuresof housing bust severity, and changes to local economic conditions.

As stated earlier, the AHS sample size includes around 500 movers per quarters which is toosmall to create state- or MSA- level first-time homebuyer purchase time-series. Instead I track state-and MSA- level home purchases from mortgages purchased or insured by the Federal National Mort-

9

gage Association, the Federal Home Loan Corporation, and the Federal Housing Administration (alsoknown as Fannie Mae, Freddie Mac, and the FHA respectively). These agencies account for a majorityof owner-occupied, first-lien, purchase mortgage originated since 2008, particularly among first-timehomebuyers.

Fannie Mae and Freddie Mac are government-sponsored enterprises which guarantee mortgagesmade by financial institutions. Both Fannie Mae and Freddie Mac report annual mortgage originationsto the Federal Housing Finance Authority which include property location information and a first-timehomebuyer indicator. This data is publicly available. To estimate the distribution among quarters fromthe annual data, I use a large sample of loan-level data provided by each company which includes thedate of first mortgage payment, property location, and an first-time homebuyer indicator. This dataalso tracks performance of the loan through 2013, including whether the loan has been prepaid ordelinquent.

The FHA insures mortgages typically targeted towards lower-income households with smallerdownpayments available. The agency has seen its role in mortgage financing increase dramaticallyfollowing the housing bust and the collapse of private subprime lending, playing a critical role for first-time homebuyers. Data on all FHA mortgage originations and their performance between 2003 and2013 was obtained from the Department of Housing and Urban Development through a freedom ofinformation act request. The dataset includes thirteen million mortgages insured by the FHA between2003 through 2013, and includes a flag for first-time homebuyer status, origination date, and propertylocation.

Combining data from the FHA and from Freddie Mac and Fannie Mae, which are Government-Sponsored Enterprises or GSEs, covers a large share of the US mortgage market. Figure 5 shows thequarterly market share of the GSE, the FHA, and other lenders by homebuyer status.7 The GSE/FHAmarket share changed significantly between 2007 and 2009, as the FHA relaxed its lending standardsand attracted many first-time homebuyers. Since the impact of this FHA expansion likely varied acrossstates, this makes using the pre-FHTC time period a poor control group to estimate local effects.

Instead I use the post-FHTC time period as a control group to measure local FHTC effects. AsFigure 5 shows that since 2009, the combined GSE/FHA market share consistently remained above 80percent for first-time homebuyer purchases. And while the FHA expansion, along with other govern-ment programs, began around the first two FHTC implementations, there was no corresponding changeof programs at the end of the FHTC.

5 Results

In this section I present FHTC effect estimates. I begin by discussing national-level estimation results.These provide an estimate of the total number of households the FHTC induced into homeownership,which are then used to calculate the total price paid per new homeowner. I then present intensive margin

7The “Other” share is calculated using AHS total quarterly home purchase counts by homebuyer status and subtractingoff the FHA and GSE origination counts.

10

effects, including effects on purchase price, downpayment amount, and mortgage outcome variables. Ilastly present state- and MSA- level FHTC estimates. The size and distribution of these effects are thencompared with various housing and economic measures to understand where and why the tax creditwas most effective.

5.1 National FHTC Effects

Table 2 reports regression results estimating both the total (above) and by iteration (below) FHTCeffects on first-time homebuyer purchases using the event-study framework outlined in Equation (1).Column (1), the main specification, finds that the FHTC increased first-time homebuyers by 8.5 percent,and using the average equating to 301,900 new homeowners. This effect is not statistically significant,but when estimating separately by iteration the FHTC effect is found to be concentrated among thesecond and third iterations and statistically significant. This is in line with expectations, as the taxcredit became much more valuable in the second and third iterations. The specification includes the“Substitution Effect” which estimates the density of households possibly shifting the the timing of theirhome purchase for the tax credit. This effect estimates that 10 percent of the increase in first-timehomebuyers was from households who would have made their first-time purchase within a year afterthe tax credit expired.8

Robustness checks adding control variables and removing the substitution effect in columns (2)and (3) respectively yield similar, though larger, results to the main specification. Given concerns abouthouseholds anticipating or altering moving decision prior to or following the tax credit, column (4)removes the “Adjacent Periods”. This specification lowers the total FHTC effect, however this strategystill finds positive effects for the second and third iterations while finding the first iteration has a strongnegative effect on purchases, which is an unlikely result. Due to the chaos and structural changes in themortgage market occurring before the initial FHTC implementation, Columns (5) through (7) estimatethe FHTC effect using only periods 2008q2-2012q4. This increases the magnitude of the FHTC effectsize to around 13 percent and statistically significant, while still being concentrated in the second andthird iterations. Focusing on this time period also increases the estimated share of shifting householdsto 15 percent.

Table 3 reports the difference-in-difference results for both the total and by iteration FHTC effectsusing Equation (2) with high income first-time homebuyers being the control group for the treated lowincome first-time homebuyers. As expected, these results show the FHTC effects were concentratedamong lower income households, with total effect sizes ranging from 18.5 to 23.1 depending on thespecification. Columns (1), (2), and (6) all also find that the substitution effect was mostly due to higherincome borrowers shifting their home purchase earlier, which may be due to the more limited time forwhich they were eligible for the credit.

The lower panel in Table 3 displays the FHTC effects by iteration. For the first and second

8To calculate this share of inter-temporal substituting households, I set the Substitution Effect equal to four in thequarter following the FHTC expiration, and dropping in value one each quarter over the next year. A regression coefficientof -0.015 means the quarter after expiration, first-time homebuyers dropped by six percent relative to expectations.

11

iterations, the FHTC X LowInc coefficient is the treatment effect. For the third iteration, both high andlow income groups were eligible for the tax credit and so the FHTC Third is the event study estimatefor the FHTC effect on high income first-time homebuyers. Across specifications, these estimatesshow larger FHTC effects in the second iteration relative to the first among lower-income first-timehomebuyers. The second iteration difference-in-difference estimates are also similar in magnitude tothe third period event study estimate of the FHTC effect on higher income borrowers.

Policymakers considering responses to housing busts must consider the cost effectiveness of theirchoices. I present two calculations of the “price of homeowners” paid by the government using theFHTC. Nationally, I estimate the program induced 301,900 new first-time homebuyers. With around3.3 million first-time homebuyers eligible to claim the credit, direct expenditures are roughly $21.1billion.9 This translates into paying $69,890 per new homeowner, though most of this is a directtransfer of wealth to first-time homebuyers.

When considering social welfare implications of the tax credit, the FHTC cost is better representedby the deadweight loss associated with the program. To estimate FHTC deadweight loss I combineliterature estimates of the marginal cost of raising revenue and the expected deadweight loss of alteringthe decisions of previous renters. Ballard et al. (1985) find the deadweight loss of raising tax revenueto be between seventeen and fifty-six cents on the dollar, which I approximate as thirty cents, meaning$21 billion costs $6.3 billion. To estimate of the deadweight loss from altering renters decisions, I usehalf the value of the claimed tax credit. The logic is as follows: assume all renters begin some $Xdistance from preferring to be homeowners. The tax credit induces those with an X less than 8,000 toswitch to homeownership. The deadweight loss is then the sum of $X across induced homeowners only.If we assume a uniform distribution across $X between $0 and $8,000 among induced homeowners,the deadweight loss is half the value of the claimed credit, totalling $1.2 billion. Combining these twosources of deadweight loss totals $7.5 billion and translates to a price per homeowner of $24,180 .

5.2 Intensive Margin Effects

Aside from measuring the extensive margin effects of the FHTC on inducing households into home-ownership, we also may expect the FHTC to affect intensive margin choices. In particular the size of thehome, mortgage financing choices, and mortgage outcomes. policymakers may care about outcomes ofthese new homeowners. For instance, if households on the margin between renting and owning are lessfinancially prepared to handle costs associated with homeownership or simply have a lower preferencefor homeownership they may more likely to default on their mortgage. Foreclosures are costly to localgovernments, banks, and residential neighbors and likely to negate any positive benefit that increasinghomeownership rates may have provided. Similarly, these new homeowners may limit FHTC benefitson increased homeownership if they quickly transition back to renting.

Using AHS data, figures 6 and 7 display the log of average home price and downpayment size

9Note that the first iteration of the tax credit is repayable over fifteen years. I discount claims during this iteration ata discount rate of 0.95. Consistent with the rest of the study, I exclude any effects or costs of the tax credit expansion toprevious owners in the third FHTC iteration.

12

by homebuyer status over time in relation to the FHTC. On average, first-time homebuyers did not buymore expensive houses during the FHTC eligibility period than the year before or after the tax credit,nor did the difference in purchase price between previous owners and first-time homebuyers differ fromits time trend. Similarly, Figure 7 shows first-time homebuyers were not more likely to put more moneydown on their home purchase during the tax credit. In fact, putting down five percent or less is morecommon during the eligibility period than after. This indicates the households did not affect intensivemargin housing decisions but instead resulted either increased non-housing consumption or savings.

To examine the outcomes of first-time homebuyers, I use the loan-level mortgage performancedata from Fannie Mae, Freddie Mac, and the Federal Housing Administration. Figure 8 shows the dif-ference between first-time homebuyers and previous owners in delinquency and prepayment rates aftertwo, three, and four years after origination. Previous owners are used to control both for changes inhouse prices and credit availability across cohorts. Overall, delinquency rates of first-time homebuyersare lower than that of previous owners during the FHTC eligibility period, reversing the trend prior to2008. Transitions across FHTC implementation dates are mostly smooth, suggesting limited effectsof the FHTC on delinquency. However two possibly troubling observations should be noting. One isthe increase in increase in first-time homebuyer delinquency relative to previous owners beginning inJanuary 2009 and the other is the jump in first-time homebuyer delinquency beginning in the fourthyear after origination in 2009. More time needs to pass before these effects could be more accuratelymeasured, but may raise some concerns if they persist. While neither effect is large in magnitude,even small changes in delinquency rates can impose significant costs to banks and governments. Pre-payment rate differences in Figure 8 are both too noisy and too short since origination to draw strongconclusions about FHTC effects. These initial outcomes do show large differences between the twoyear and the three and four year prepayment rates during the FHTC eligibility time period, with thefirst-time homebuyers difference on prepayment being much higher rate during the first two years afterorigination. The cause of this difference needs to be explored further, as the data do not track whetherprepayment is the result of refinancing or moving to a new home and most interestingly, if that newhome is owner-occupied or a rental. The two year prepayment difference is particularly puzzling sincehouseholds receiving the FHTC must repay the tax credit if they move within three years after buyingthe home.

5.3 State and MSA FHTC Effects

While I find the FHTC increased first-time homebuyers by around 8.5 percent overall, we are also in-terested in its distributional impacts across the US. For example, did areas hit harder by the housingcrisis have larger responses to the tax credit? What aspects of local housing markets predict a higherresponse and what can we infer from it? Do larger FHTC effects correlate positively with changes inemployment or house prices during the tax credit? Or negatively after the credit expired? To answerthese questions I use the Fannie/Freddie/FHA state- and MSA-level first-time homebuyer data to esti-mate Equation (2) using previous owners as a control group. I focus my analysis on the average FHTCeffect size during the second and third iterations of the FHTC, since national results indicate only minorto zero response during the first iteration.

13

Overall, Midwestern and Southern states experienced the largest impacts while Western andNortheastern states displayed milder effects. Figures 9 and 10 plots both state- and MSA-level ef-fects against the peak-to-trough house price drop during the housing crisis, as reported by the FHFAPurchase-Only House Price Index, and the 2009 mortgage delinquency rate.10 These graphs display anegative correlation between FHTC effect size and housing bust magnitude, and alternative estimationspecifications yield a similar trend. Among the four states hit the most hardest by the crisis, only Ari-zona displays an above median FHTC effect, and overall the expected FHTC effect among states withthe most mild housing busts was double the expected affect among states with the most severe housingbusts.

Next, I compare FHTC effects to housing market characteristics. The FHTC provided the lesserof $8,000 or 10% of the home purchase price. Average home values in the US are near $250,000meaning a large majority of households claiming the credit received the full $8,000. Since home valuesdiffer greatly across local housing markets, this means that lower home value areas such as Nebraskaand Iowa effectively received a higher treatment effect than higher value areas such as California andHawaii. Figure 11 plots FHTC effects against the log of average 2009 state and MSA home values.11

As expected, there is a strong negative correlation between FHTC effect and average home value.Regressing average home value on effect size yields a statistically significant coefficient of -0.197 atthe state level and -0.178 at the MSA level.12 This implies that moving from an area fifty percent moreexpensive, such as from Tennessee to Massachusetts, reduces the expected FHTC effect by almost tenpercentage points.

Figure 11 also provides an estimate of what results counterfactual FHTC policies would haveyielded. The current design of offering $8,000 on average US home values of $250,000, Figure 11predicts around a fourteen percent effect, the same as the effect found for the second and third iterationsof the FHTC. These results suggest that altering the tax credit to be a percentage of home value ratherthan a fixed amount and offering 5% of the home value would have increased the FHTC effect to23.3% or 494,000 total induced new homeowners. Offering a percentage of the home value wouldhave additionally eliminated the tilt of FHTC effects towards lower home value areas and in turn bettertargeted the housing bust states.

Since variation in average home values alters the treatment effect of the FHTC across areas, otherhousing market characteristics will be considered using the residual after regressing FHTC effect sizeon average home value. Figure 12 plots the FHTC effect residual against the percentage of householdswhich were renters according to the 2009 American Community Survey. Ex ante, one could argue thegroup of renters changes as a higher fraction of household rent in a way that would correlate with ahigher expected response to the FHTC. However, I observe mixed and weak evidence that areas with ahigher fraction of renters responded any differently than areas with a lower fraction of renters.

10This delinquency rate is calculated among a sample of 2 million mortgages managed by the Wells Fargo Trustee,originated in 2005 or 2006, often referred to as the Corporate Trust Services data. While this delinquency rate is taken froma group of subprime borrowers, a similar ordering of states is found among prime borrowers as well.

11Home value data provided by the Lincoln Land Institute, and can be obtained at http://www.lincolninst.edu/subcenters/land-values/. Effect size is the average from the second and third FHTC implementations only.

12MSA level correlation reported if looking at areas with populations greater than 1,000,000 (n=51). Restricting topopulation greater than 500,000 (n=102) estimate is -0.122.

14

http://www.lincolninst.edu/subcenters/land-values/http://www.lincolninst.edu/subcenters/land-values/

I also consider the role housing supply elasticity and land use regulation might have played inthe FHTC response. Figure 13 plots the FHTC effect residual against a measure of housing supplyelasticity reported by Saiz (2008) and the Wharton Land Use Regulation index as detailed in Gyourkoet al. (2008). While only a slight positive correlation is found between higher regulation and effectsize, a stronger positive correlation is observed with housing supply elasticity. Metro areas with theleast elastic housing supply such as Miami and Los Angeles experienced a ten percentage point smallereffect on average than areas with the most elastic housing supply such as Kansas City and Charlotte,even after accounting for differences in average home value. This indicates that potential first-timehomebuyers were more responsive in areas they could build new homes as opposed to moving intoexisting or vacant homes.

Lastly, I consider FHTC effects on house prices, employment, housing starts and owner-occupiedvacancy rates. Figures 14 and 15 plot the average monthly percentage change to these four variablesagainst the FHTC effect size and FHTC utilization rate by state respectively.13 The FHTC effect sizeis the same as calculated above and represents the extensive margin impacts of the additional newhomeowners induced by the FHTC, while the GAO utilization rate represents the intensive margin orstimulus FHTC effects by counting the number of households claiming the tax credit according to aSeptember 2010 GAO report as a percentage of the total population.14 Both Figures 14 and 15 showa small and statistically insignificant correlation between changes in house prices and FHTC effect orutilization and implies that increasing the utilization rate by a standard deviation would increase houseprices by 0.67% each month during the credit. Similarly weak and insignificant correlations are foundbetween the FHTC effects and changes to housing starts, employment, and owner-occupied housingvacancy rates. While is should be noted that on average state had a 1.1% increase in house prices andan average decrease in vacancy rates during the FHTC, since the variation of these state-level effectsare not significantly correlated with either the increase or level of first-time homebuyers it is difficult toattribute these changes to the FHTC.

6 Conclusion

This paper studies the effects of the First-time Homebuyer Tax Credit. At a national level, I find thetax credit increased first-time homebuyers by 8.5 percent for a total of 301,900 households inducedinto homeownership between April 2008 and September 2010. This effect was concentrated betweenJanuary 2009 and September 2010 after the tax credit was no longer repayable, and splitting the samplebased on income verifies only eligible households responded to the tax credit. Given total programexpenditures are roughly $21.1 billion, I approximate the deadweight loss of raising that revenue com-bined with altering households decisions to be $7.5 billion . This translates into paying $24,180 perinduced homeowner.

13House price data comes from the FHFA Purchase-Only house price index, while state employment and housing startsdata are provided by the Bureau of Labor Statistics and the Census Bureau respectively. Annual housing vacancy rates aretaken from the Census Bureau. The shock to each variable is calculated as the average monthly residual during the secondand third FHTC iterations from predicting the time-series using a quadratic time-trend.

14The GAO report can be found at:http://www.gao.gov/new.items/d101025r.pdf

15

http://www.gao.gov/new.items/d101025r.pdf

Survey data provide no evidence that first-time homebuyers receiving the tax credit bought big-ger houses or put more money in the downpayment relative to before or after the tax credit, but wereyounger and cited “Changing Tenure Status” more frequently as the primary reason for moving. State-and MSA-level analysis find the tax credit was more effective in areas with lower housing values andmore elastic housing supply, while land use regulation and a higher rental percentage were not predic-tors of effect size after accounting for home value.

Considering whether $24,180 is a reasonable price to pay for homeowners is a difficult quesiton toanswer. This paper only begins to answer welfare benefits from the FHTC by measuring the extensiveand intensive margin responses of households to the credit. Quantifying the benefits of dampening thedrop in home values or restoring consumer confidence is logical, though difficult, next step. One reasonhomeownership has long been a federal policy goal is the expected positive externalities stemmingfrom homeownership, such as better citizenship and home upkeep. Coulson and Li (2013) measuresthe annual positive externality from homeownership, capitalized into neighborhood home values, tobe $1,327 annually. If each induced FHTC induced homeowner is expected to be a homeowner fivemore years than without it, and using a discount factor of 0.95, this adds up to an increase of $6,600of value. This benefit alone is nearly half the price of buying a homeowner. Additionally, we shouldconsider what would happen to the vacant homes if they are occupied by new homeownerships. Arecent evaluation of the Neighborhood Stabilization Program by Spader et al. (2015) finds that onaverage it costs local governments around $11,000 to demolish vacant homes. Assuming all inducedhomeowners occupy a vacant home, additional benefits from stabilizing the housing market would needto be greater than $6,600 per induced homeowner to make the program cost-effective.

These findings highlight several important issues regrading the federal response to the US housingcrisis, in particular the importance of targeting and the high cost of homeowners. Households clearlyresponded to the credit, however given that for every induced first-time homebuyer the tax credit waspaid to six always-movers, considering more efficient ways to targeted households on the margin be-tween buying and renting or in housing bust areas would improve program performance. That inducedhomeowners were not more likely to prepay or default on their mortgage suggests that many currentrenters could handle the financial challenges of homeownership.

A future FHTC improvement could be better to target housing bust areas by altering the tax creditpayout structure. The 2008 housing busts states were in relatively high-cost areas and so could havebeen better targeted by raising the maximum award dollar amount but lowering the highest percentageof the purchased home value it could be. For instance, changing the tax credit to pay up to 5% of thehome value, uncapped, would have shifted a greater percentage of new homeowners to California andFlorida and away from the Midwest and the Southern states. Housing busts are typically accompaniedby high foreclosure and vacancy rates, so targeting could be improved by additionally requiring orincentivizing moving into a foreclosed or vacant home.

The FHTC has also lent insight into the decision homeowners face between purchasing or rentingtheir home. The own or rent housing decision is a research area in need of further exploration as the UShomeownership rate has receded to its lowest rate since 1995. The differential cost between owning andrenting housing is not well defined and certainly heterogeneous. A deeper look at the FHTC consideringcredit history, expected tenure duration, and income trajectory and uncertainty could feasibly provide a

16

better understanding and estimation of this cost.

For policy relevance, the mortgage interest deduction remains a hot political issue and one of thelargest US tax breaks costing $70 billion each year. For those believing that the $69,890 price forhomeowners is high, remember that this number cuts both ways. That is, the federal government couldreceive a similarly high price for “selling” homeowners if they reduce or eliminating the mortgageinterest deduction or other homeownership incentives. Moreover, if we wish to keep homeownershipincentives in place, replacing the mortgage interest deduction with a permanent first-time homebuyertax credit or an annual homeowner tax credit, as advocated in Green and Vandell (1999). This changecould be welfare-improving as the tax credit would directly subsidize homeownership, as opposed toindirectly through mortgage financing, which has the negative aspects of offering more benefit to highercredit risks, punishing cash buyers, and subsidizing cash-out refinancing.

While the housing bust of the Great Recession was the largest the nation had seen in eighty years,regional housing busts occur on a much more frequent basis. Often times, these regional housingbusts can devastate local economies and local leaders have little evidence of effective policy remedies.From a policy perspective, the Great Recession provided a treasure trove of potential new weaponsthat governments can utilize to combat housing busts. While disentangling FHTC effects from otherhousing programs during the same period is difficult, this paper has found evidence of the householdresponse to the program and quantified the costs associated with it. While additional FHTC benefits,such as its stimulus effects on consumer spending, not considered here could also be important, futurepolicymakers have better evidence now the value the FHTC provides.

ReferencesAgarwal, Sumit, Chunlin Liu, and Nicholas S Souleles, “The reaction of consumer spending and debt

to tax rebates–evidence from consumer credit data,” Technical Report, National Bureau of EconomicResearch 2007.

Baker, Dean, “First Time Underwater: The Impact of the First-time Homebuyer Tax Credit,” TechnicalReport, Center for Economic and Policy Research (CEPR) 2012.

Ballard, Charles L, John B Shoven, and John Whalley, “General equilibrium computations ofthe marginal welfare costs of taxes in the United States,” The American Economic Review, 1985,pp. 128–138.

Coulson, N Edward and Herman Li, “Measuring the external benefits of homeownership,” Journalof Urban Economics, 2013, 77, 57–67.

Dynan, Karen, Ted Gayer, and Natasha Plotkin, “An Evaluation of Federal and State HomebuyerTax Incentives,” Washington, DC: The Brookings Institution, 2013.

Green, Richard K and Kerry D Vandell, “Giving households credit: How changes in the US tax codecould promote homeownership,” Regional Science and Urban Economics, 1999, 29 (4), 419–444.

17

Gyourko, Joseph, Albert Saiz, and Anita Summers, “A new measure of the local regulatory environ-ment for housing markets: The Wharton Residential Land Use Regulatory Index,” Urban Studies,2008, 45 (3), 693–729.

Hilber, Christian AL and Tracy M Turner, “The mortgage interest deduction and its impact onhomeownership decisions,” Review of Economics and Statistics, 2014, 96 (4), 618–637.

Mian, Atif and Amir Sufi, “The effects of fiscal stimulus: Evidence from the 2009 Cash for clunkersprogram,” Technical Report, National Bureau of Economic Research 2010.

Saiz, Albert, “On Local Housing Supply Elasticity,” Technical Report 2008.

Shapiro, Matthew D and Joel Slemrod, “Consumer response to tax rebates,” Technical Report, Na-tional Bureau of Economic Research 2001.

Spader, Jonathan, Alvaro Cortes, Kimberly Burnett, Larry Buron, Michael DiDomenico, AnnaJefferson, Stephen Whitlow, Jennifer Lewis Buell, Christian Redfearn, and Jenny Schuetz,“The Evaluation of the Neighborhood Stabilization Program,” 2015.

Tong, Zhong Yi, “Washington, DCs First-Time Home-Buyer Tax Credit,” 2005.

18

Figure 1: First-Time Homebuyer Share, NAR vs. AHS0

1020

3040

5060

Firs

t-Tim

e H

omeb

uyer

Sha

re (%

)

2000 2005 2010

AHS NAR

Note: This figure compares the percent of homebuyers reporting to be first-time buyers in the AHSsurvey data and NAR Profile of Homebuyers and Sellers. AHS first-time homebuyer share based onyear of moving date.Source: American Housing Survey, National Association of Realtors Profile of Homebuyers andSellers.

19

Figure 2: Annual Home Purchases by Homeowner Status and Quarterly First-time Homebuyer Pur-chases

0.5

11.

52

2.5

33.

54

Mor

tgag

e O

rigin

atio

ns (M

illio

ns)

2001q1 2004q1 2007q1 2010q1 2013q1

First-time HomebuyersPrevious Owners

010

020

030

040

050

060

0Fi

rst-T

ime

Hom

ebuy

er O

rigin

atio

ns (0

00s)

2001q1 2004q1 2007q1 2010q1 2013q1

Notes: This figure displays annual home purchases by buyer status (left) and quarterly home purchasesby first-time homebuyers (right).Source: American Housing Survey and HMDA data.

20

Figure 3: Number of Homebuyers, by Homeowner Status and Income Level0

100

200

300

400

500

2000q1 2003q1 2006q1 2009q1 2012q1 2000q1 2003q1 2006q1 2009q1 2012q1

Previous Owners First-Time Homebuyers

Low Income High Income

Hom

e P

urch

ases

(000

s)

Notes: This figure displays estimates of the number of homebuyers each month by income level andhomebuyer status. Mortgage originations use log scale. High income defined as households reportingabove $85,000 annual income, low income below this threshold. AHS data extrapolated using HMDAannual counts of total first-lien, owner-occupied, purchase mortgages, by income categories. Solidvertical lines represent FHTC start and end dates and dashed lines represent each iteration date.Source: American Housing Survey

21

Figure 4: New Housing Tenure Shares, by Previous Tenure0

2040

6080

2001q3 2005q3 2009q3 2013q32001q3 2005q3 2009q3 2013q3

Previous Owners Previous Renters

New Owners New Renters

New

Hou

sing

Ten

ure

(%)

Notes: This figure displays new housing tenure shares among movers, by year and previous tenurestatus. Based on AHS data using sample weights and smoothed with a local polynomial. Solid verticallines represent FHTC start and end dates and dashed lines represent each iteration date.Source: American Housing Survey

22

Figure 5: Home Purchase Financing Shares of GSE, FHA, and Other Sources, by Buyer Status0

2040

6080

100

2004q1 2006q1 2008q1 2010q1 2012q1 2004q1 2006q1 2008q1 2010q1 2012q1

Previous Owners First-Time Homebuyers

GSE FHA Cash/Other

Mar

ket S

hare

(%)

Notes: This figure displays the home purchase financing share of the GSEs, FHA, and other sourcessuch as private label or cash-only purchases. Other share is calculated as the difference between theestimated total quarterly home purchases in the AHS and the sum of GSE and FHA origination.Source: Authors calculations using American Housing Survey data, Fannie Mae, Freddie Mac, andFHA loan-level mortgage origination data.

23

Figure 6: New Home Price, by Buyer Status

-.5-.4

-.3-.2

Diff

eren

ce

1212

.212

.412

.6Ln

(Hou

se P

rice)

2002q3 2005q1 2007q3 2010q1 2012q3

Previous Owners First-time BuyersDifference (right axis)

Notes: This figure displays the log average new house price of recent movers, by buyer status andlog difference between previous owners and first-time homebuyers house price. House prices are win-sorized at the five percent level. Solid vertical lines represent FHTC start and end dates and dashedlines represent each iteration date.Source: American Housing Survey

24

Figure 7: Size of Mortgage Downpayment by Buyer Status0

2040

6080

2005q1 2008q1 2011q1 2014q12005q1 2008q1 2011q1 2014q1

Previous Owners First-time Buyers

95-100% 80-94%

Figure 8: Mortgage Outcome Differences:First-time Homebuyers vs. Previous Owners

-.005

0.0

05.0

1.0

15

2003q1 2005q1 2007q1 2009q1 2011q1Date

2 Years 3 Years 4 Years

Delinquency Difference

-.12

-.1-.0

8-.0

6-.0

4-.0

2

2003q1 2005q1 2007q1 2009q1 2011q1Date

2 Years 3 Years 4 Years

Prepay Difference

Notes: This figure displays the difference of first-time homebuyer less previous owners rates of delin-quency and prepayment for each mortgage vintage quarter after various lengths in time. Solid verticallines represent FHTC start and end dates and dashed lines represent each iteration date.Source: Fannie Mae, Freddie Mac, and Federal Housing Administration loan-level mortgage perfor-mance data.

26

Figure 9: FHTC Effect Size and House Price Drop 2006-2009

KY

NY

NC

MI

FL

OHAK

NHCO

VA

ME

NE

ID

MN

DC

MA

ND

KS

WV

WA

AR

AZ

CT

TN

HI

NV

MS

UT

SD

IL

PA

LA

RIIN

WYSC

TX

WI

NJ

DE

IA

MD

NM

CA

OK

VT

GA

AL

MO

OR

MT

-.10

.1.2

.3.4

FHTC

Effe

ct

0 10 20 30 40 50House Price Drop (2006-2009)

CincinnatiRichmond

Austin

Hartford

Columbus (OH)

Denver

Rochester (NY)Houston

Kansas City

Sacramento

Milwaukee

TampaNew Orleans

Phoenix

Jacksonville

San Jose

Portland

Louisville

St. Louis

Atlanta

IndianapolisCharlotte

Pittsburgh

Raleigh

Minneapolis

Buffalo

San Diego

San Antonio

BaltimoreProvidence

Birmingham

Virginia Beach

Orlando

Salt Lake City

Las Vegas

Memphis

Oklahoma City

Cleveland

Riverside

Nashville

-.10

.1.2

.3.4

FHTC

Effe

ct

0 10 20 30 40 50House Price Drop (2006-2009)

Notes: This figure displays the estimated FHTC effect size against the peak-to-trough house price dropbetween 2006 and 2010 for each area. Effect size of the second iteration of the FHTC only. Houseprice data comes from FHFA purchase-only house price index. Dark line is a linear fit of the data.

27

Figure 10: FHTC Effect Size and House Price Drop 2006-2009

KY

NY

NC

MI

FL

OHAK

NHCO

VA

ME

NE

ID

MN

DC

MA

ND

KS

WV

WA

AR

AZ

CT

TN

HI

NV

MS

UT

SD

IL

PA

LA

RIIN

WY SC

TX

WI

NJ

DE

IA

MD

NM

CA

OK

VT

GA

AL

MO

OR

MT

-.10

.1.2

.3.4

FHTC

Effe

ct

0 .1 .2 .3Mortgage Delinquency %

CincinnatiRichmond

Austin

Hartford

Columbus (OH)

Denver

Rochester (NY)Houston

Kansas City

Sacramento

Milwaukee

TampaNew Orleans

Phoenix

Jacksonville

San Jose

Portland

Louisville

St. Louis

Atlanta

IndianapolisCharlotte

Pittsburgh

Raleigh

Minneapolis

Buffalo

San Diego

San Antonio

BaltimoreProvidence

Birmingham

Virginia Beach

Orlando

Salt Lake City

Las Vegas

Memphis

Oklahoma City

Cleveland

Riverside

Nashville

-.10

.1.2

.3.4

FHTC

Effe

ct

0 .1 .2 .3Mortgage Delinquency %

Notes: This figure displays the estimated FHTC effect size against the 2009 mortgage delinquencyrate of each area. Effect size of the second iteration of the FHTC only. Mortgage delinquency ratescome from 2009 Freddie Mac loan-level data, with delinquency defined as loans sixty days or morebehind on payments. Dark line is a linear fit of the data.

28

Figure 11: FHTC Effect Size and Average Home Values

KY

NY

NC

MI

FL

OH AK

NHCO

VA

ME

NE

ID

MN

DC

MA

ND

KS

WV

WA

AR

AZ

CT

TN

HI

NV

MS

UT

SD

IL

PA

LA

RIIN

WYSC

TX

WI

NJ

DE

IA

MD

NM

CA

OK

VT

GA

AL

MO

OR

MT

-.20

.2.4

FHTC

Effe

ct

11 12 13 14Ln(Average Home Value)

Cincinnati

Austin

Chicago

Columbus (OH)

Denver

Houston

Kansas City

Sacramento

Milwaukee

Tampa

Los AngelesNew York City

Phoenix

San Jose

Seattle

Portland

St. Louis

Atlanta

Detroit

IndianapolisCharlotte

PittsburghMinneapolis

San Diego

San Antonio

Baltimore

Dallas

Boston

Providence

Philadelphia

Virginia Beach

San Francisco

Orlando

Miami

Las Vegas

Washington DC

Cleveland

Riverside

Nashville

-.20

.2.4

FHTC

Effe

ct

11 12 13 14Ln(Average Home Value)

Notes: This figure displays the estimated FHTC effect size against the log of the average homevalue by state and MSA. Effect size of the second iteration of the FHTC only. Average home valuescome from data provided by the Lincoln Land Institute as of 2009:http://www.lincolninst.edu/subcenters/land-values/. Dark line is a linear fit of the data.

29

http://www.lincolninst.edu/subcenters/land-values/http://www.lincolninst.edu/subcenters/land-values/

Figure 12: FHTC Effect Size Residual and Rental Percentage

KY

NY

NC

MI

FL

OH

AK

NH

CO

VAME

NE

IDMN

MA

ND

KS

WV

WAAR

AZ

CTTN

HI NVMS

UT

SD

IL

PALA

RI

IN

WY

SC

TX

WINJ

DE

IA

MD

NM

CAOK

VT

GA

AL

MO

OR

MT

-.3-.2

-.10

.1.2

FHTC

Effe

ct R

esid

ual

.15 .2 .25 .3 .35 .4Rental Percentage

Cincinnati

Richmond

Austin

Hartford

Chicago

Columbus (OH)

DenverRochester (NY)Houston

Kansas City

Sacramento

Milwaukee

Tampa

Los Angeles

New Orleans

New York City

Phoenix

Jacksonville

San JoseSeattle

Portland

Louisville

St. Louis

Atlanta

Detroit

Indianapolis

Charlotte

Pittsburgh

Raleigh

Minneapolis

Buffalo

San Diego

San AntonioBaltimore

DallasBostonProvidence

Birmingham

Philadelphia

Virginia Beach

San Francisco

OrlandoMiami

Salt Lake City

Las Vegas

Memphis

Washington DCOklahoma City

Cleveland

Riverside

Nashville

-.3-.2

-.10

.1.2

FHTC

Effe

ct R

esid

ual

.15 .2 .25 .3 .35 .4Rental Percentage

Notes: This figure displays the renter percentage in each area against the residual after regressing theFHTC effect on average home value of the area. Effect size of the second iteration of the FHTC only.Rental percentage as of 2008, authors calculations from American Community Survey data. Dark lineis a linear fit of the data.

30

Figure 13: FHTC Effect Size versus Housing Supply Elasticity and Land Use Regulation

Cincinnati

Richmond

Austin

Hartford

Chicago

Columbus (OH)

DenverRochester (NY)Houston

Kansas City

Milwaukee

Tampa

Los Angeles

New Orleans

New York City

Phoenix

Jacksonville

San JoseSeattle

Portland

Louisville

St. Louis

Atlanta

Detroit

Indianapolis

Charlotte

Pittsburgh

Raleigh

Minneapolis

Buffalo

San Diego

San AntonioBaltimore

DallasBostonProvidence

Birmingham

Philadelphia

Virginia Beach

San Francisco

OrlandoMiami

Salt Lake City

Las Vegas

Memphis

Washington DC Oklahoma City

Cleveland

Riverside

Nashville

-.3-.2

-.10

.1.2

FHTC

Effe

ct R

esid

ual

0 1 2 3 4Supply Elasticity

Cincinnati

Hartford

Chicago

DenverHouston

Sacramento

Milwaukee

Los AngelesNew York City

San Jose Seattle

Portland

Detroit

Baltimore

Dallas BostonProvidence

PhiladelphiaSan Francisco

MiamiWashington DC

Cleveland

Riverside-.3

-.2-.1

0.1

.2FH

TC E

ffect

Res

idua

l

-.5 0 .5 1 1.5 2Land Use Regulation

Notes: This figure displays the housing supply elasticity and a land use regulation index against theresidual after regressing the FHTC effect on average home value of the MSA. Effect size of the seconditeration of the FHTC only. Housing supply elasticity estimates are taken from Saiz (2008). Land useregulation index taken from Gyourko et al. (2008). Dark line is a linear fit of the data.

31

Figure 14: FHTC Effect Size and Changes to Local Economic Variables between January 2009-September 2010

AL

AK

AZ

ARCA

CO

CT

DE

FL

GA

HI

ID

IL

INIA

KS

KYLA

MEMDMA

MIMN

MS

MOMT

NE

NV

NHNJ

NM

NY

NC

ND

OH OK

OR

PA RI

SC

SDTNTX

UT

VT VA

WA

WV

WI

WY

-.01

0.0

1.0

2.0

3.0

4

-.1 0 .1 .2 .3 .4First-time Buyer Effect

House Prices

ALAK

AZAR

CA

COCT

DE

FLGA

HI ID

ILIN

IAKS

KYLA

ME

MDMA

MI

MNMS

MOMT

NE

NVNH

NJ NMNY NC

ND

OH

OKOR PA RI SC

SDTNTXUT VT VAWA WVWI

WY

-.3-.2

-.10

.1

-.1 0 .1 .2 .3 .4First-time Buyer Effect

Housing Starts

AL

AK AZ

ARCA

COCT

DE

FLGA

HI

ID

IL

IN

IAKS

KY

LA

MEMDMA

MI

MN

MS

MO

MT NENV NH

NJ NMNY

NC

ND

OH

OK

OR

PA

RI

SC

SD

TN

TXUTVT

VA

WAWVWI

WY

-.015

-.01

-.005

0

-.1 0 .1 .2 .3 .4First-time Buyer Effect

Employed

AL

AK

AZ

AR

CA

CO

CT

DEFL

GA

HI

IDIL

IN

IA

KS

KYLAMEMD

MAMI

MNMSMO

MT

NE

NV NHNJ

NM

NY

NCNDOH

OK

OR

PA

RISCSDTN

TXUT

VTVA

WA

WVWI

WY

-.01-

.005

0.0

05.0

1.0

15

-.1 0 .1 .2 .3 .4First-time Buyer Effect

Vacancy Rate

Note: These panels display the FHTC effect size against shocks to local economic variables. Effectsize of the second iteration of the FHTC only. The shock is measured as the average monthly residualbetween January 2009 and September 2010 from predicting each series using a quadratic time trend bystate. Vacancy rate from owner-occupied housing, available only at annual rate. Dark line is a linear fitof the data.Source: Census Bureau, FHFA, Bureau of Labor Statistics.

32

Figure 15: FHTC Utilization Rate and Changes to Local Economic Variables between January 2009-September 2010

AL

AK

AZ

ARCA

CO

CT

DE

FL

GA

HI

ID

IL

IN IA

KS

KYLA

MEMDMA

MIMN

MS

MOMT

NE

NV

NHNJ

NM

NY

NC

ND

OH OK

OR

PARI

SC

SDTNTX

UT

VT VA

WA

WV

WI

WY

-.01

0.0

1.0

2.0

3.0

4

.4 .6 .8 1 1.2First-time Buyer Rate

HPI

ALAK

AZAR

CA

COCT

DE

FLGA

HI ID

ILIN

IAKS

KYLA

ME

MDMA

MI

MNMS

MOMT

NE

NVNH

NJ NMNY NC

ND

OH

OKORPARISC

SDTNTX UTVT VAWAWV WI

WY

-.3-.2

-.10

.1

.4 .6 .8 1 1.2First-time Buyer Rate

Housing Starts

AL

AK AZ

ARCA

COCT

DE

FLGA

HI

ID

IL

IN

IAKS

KY

LA

MEMDMA

MI

MN

MS

MO

MT NENVNH

NJ NMNY

NC

ND

OH

OK

OR

PA

RI

SC

SD

TN

TX UTVT

VA

WAWV

WIWY

-.015

-.01

-.005

0

.4 .6 .8 1 1.2First-time Buyer Rate

Employed

AL

AK

AZ

AR

CA

CO

CT

DEFL

GA

HI

IDIL

IN

IA

KS

KYLAMEMD

MAMI

MNMSMO

MT

NE

NVNHNJ

NM

NY

NC NDOH

OK

OR

PA

RISC SDTN

TXUT

VTVA

WA

WV WI

WY

-.01-

.005

0.0

05.0

1.0

15

.4 .6 .8 1 1.2First-time Buyer Rate

Vacancy Rate

Note: These panels display the FHTC utilization rate against shocks to local economic variables. FHTCutilization rate calculated as percentage of households in each state claiming tax credit according toclaims according to a September 2010 GAO report. The shock is measured as the average monthlyresidual between January 2009 and September 2010 from predicting each series using a quadratic timetrend by state. Vacancy rate from owner-occupied housing, available only at annual rate. Dark line is alinear fit of the data.Sources: Census Bureau, FHFA, Bureau of Labor Statistics.

33

Table 1: First-time Homebuyer Tax Credit Details

GAO-10-1025R Tax Administration

Page 6

Enclosure I: Significant Differences among the Three Versions of the First-Time Homebuyer Credit Table 1: Major Distinctions among the Housing, Recovery, and Assistance Acts First-Time Homebuyer Credit Versions Housing and

Economic Recovery Act of 2008 (Housing Act)

American Recovery and Reinvestment Act of 2009 (Recovery Act)

Worker, Homeownership, and Business Assistance Act of 2009 (Assistance Act)a

Applicable dates April 9, 2008 – July 1, 2009

January 1, 2009 – November 30, 2009

November 7, 2009 – June 30, 2010 (buy, or enter into a binding contract to buy, by April 30, 2010, close by June 30, 2010)b

First-time homebuyer only?

Yes Yes No (includes long-term owners)

Maximum amount $7,500 $8,000 $8,000 Income phase outc Single: $75,000 -

$95,000 Joint: $150,000 - $170,000

Single: $75,000 - $95,000 Joint: $150,000 - $170,000

Single: $125,000 - $145,000 Joint: $225,000 - $245,000

Repayable Yes No (unless resold within 3 years at a gain)

No (unless resold within 3 years at a gain)

Documentation of purchase required?

No No Yesd

Maximum purchase price

No No $800,000

Source: GAO analysis of IRS data. aMembers of the Armed Forces and certain federal employees serving outside the U.S. have an

additional year to buy a principal residence in the U.S. and qualify using the Assistance Act. An eligible taxpayer must buy or enter into a binding contract to buy a home by April 30, 2011, and settle on the purchase by June 30, 2011. bThe Homebuyers Assistance and Improvement Act of 2010 extended the closing date for the credit through September 30, 2010. cIncome phase out amounts refer to modified adjusted gross income. dUnder the Assistance Act, claimants must attach a copy of the settlement statement to the tax return.

Note: This table is taken from the Government Accountability Office report on the FHTC: http://www.gao.gov/new.items/d101025r.pdf.

34

http://www.gao.gov/new.items/d101025r.pdfhttp://www.gao.gov/new.items/d101025r.pdf

Table 2: FHTC Effect Estimates, Event Study

(1) (2) (3) (4) (5) (6) (7)Total FHTC Effect 0.085 0.108 0.110* 0.023 0.128*** 0.140 0.125***

(0.069) (0.087) (0.055) (0.065) (0.041) (0.103) (0.036)Substitution Effect -0.015 -0.007 -0.023 -0.026

(0.025) (0.026) (0.018) (0.026)Controls x xAdjacent Periods x xFirst Iteration -0.006 0.052 0.008 -0.071 0.120** -0.067 0.091*

(0.075) (0.111) (0.066) (0.070) (0.054) (0.188) (0.051)Second Iteration 0.141* 0.192 0.159** 0.070 0.187*** 0.071 0.158***

(0.076) (0.128) (0.064) (0.070) (0.049) (0.136) (0.047)Third Iteration 0.145* 0.159 0.162** 0.080 0.143** 0.050 0.114**

(0.078) (0.094) (0.066) (0.071) (0.054) (0.103) (0.051)* p

Table 3: FHTC Effect Estimates, Difference-in-Difference

(1) (2) (3) (4) (5) (6) (7)FHTC X LowInc 0.221*** 0.205** 0.211*** 0.231*** 0.185** 0.207** 0.193**

(0.079) (0.086) (0.077) (0.074) (0.077) (0.078) (0.075)FHTC -0.055 -0.029 -0.019 -0.127 -0.001 -0.056 -0.039

(0.091) (0.122) (0.079) (0.085) (0.067) (0.071) (0.067)Low Income 0.968*** 0.973*** 0.982*** 0.959*** 0.976*** 0.948*** 0.966***

(0.048) (0.050) (0.044) (0.044) (0.053) (0.059) (0.055)Substitution Effect -0.034 -0.029 -0.056*

(0.034) (0.036) (0.030)Substitution X LowInc 0.033 0.031 0.039

(0.042) (0.043) (0.040)Controls xAdjacent Periods x xFHTC X LowInc First 0.179 0.179 0.161 0.190 0.188 0.250* 0.210

(0.135) (0.139) (0.131) (0.126) (0.135) (0.140) (0.133)FHTC X LowInc Second 0.261** 0.261** 0.243** 0.272** 0.270** 0.332** 0.292**

(0.120) (0.123) (0.116) (0.112) (0.122) (0.128) (0.120)FHTC X LowInc Third -0.026 -0.026 -0.044 -0.015 -0.017 0.045 0.005

(0.135) (0.139) (0.131) (0.126) (0.135) (0.140) (0.133)FHTC First -0.091 -0.016 -0.061 -0.162 -0.015 -0.087 -0.059

(0.111) (0.144) (0.102) (0.105) (0.095) (0.099) (0.094)FHTC Second -0.009 0.090 0.025 -0.085 -0.004 -0.076 -0.048

(0.108) (0.158) (0.095) (0.101) (0.086) (0.090) (0.085)FHTC Third 0.184 0.221* 0.218** 0.115 0.152 0.081 0.108

(0.114) (0.128) (0.102) (0.105) (0.095) (0.099) (0.094)* p