Summary

Housing market at the heart of Great Recession  after sustained boom, housing prices collapsed =
triggering financial crisis + fall in household expenditures + macro frictions  slump in employment

Approach: structural equilibrium  use both cross-sectional micro data and macro time-series
Model: OLG incomplete market models of US economy + housing finance sector + realistic household
consumption behaviour:
Uninsurable idiosyncratic shock + financial constaints + housing can be used as collateral
(housing as the illiquid savings instrument)  precautionary savings (as is common to literature)
Possibility to refinance creates financial accelerator through HELOC

Here: Develop a fully stochastic model with aggregate shocks

 stochastic equilibrium dynamics for housing investment, house process, rents, mortgages and risk
spreads  paper extends aggregate fluctuations research to housing sector
Explain aggregate fluctuations
 BUT also sufficient hh heterogeneity to allow tight mapping to hh earnings + asset portfolios

Standard household specifications for endowment but include both bonds b and houses (rent or own)
as savings  model houses by quantitative measure of size
 Renting generates consumption housing services one for one with size of house 𝑠 = ℎ wth
no transaction cost
 Owning generates additional utility from home ownership 𝑠 = 𝑠ℎ, 𝑠 ≥ 1 with
maintenance/tax cost
Purchase of housing financed by mortgages  defaultable/financed by repaying origination cost;
long-term, subject to fixed origination cost and amortised over remaining life of buyer at common real
interest rate
 2 constraints: max loan-to-value limit (initial mortgage value must be less than some fraction
of collateral value of house being purchased)
 Max payment-to-income ratio limit (minimum mortgage payment must be fraction of income
per period)
′
 Law of motion for mortgage balance: 𝑠 = 𝑠(1 + 𝑠𝑠 ) − 𝑠
Mortgages long term since after origination there is no requirement that the principal outstanding
𝑠
of mortgage be less than 𝑠 of current value of the house
 Only requirement for a borrow to not be in default is that it makes minimum payment on
outstanding balance of home.
 If house prices decline  could be in negative equity but as long as it meets minimum
payments  not forced to deleverage (similar to ST debt held period-by-period)
Mortgage owners can always refinance by repaying residual principal balance and originating new
mortgage  cash-out refinancing only since IR fixed
𝑠
 Possible HELOC: one-period non-defautable contracts: hh can borrow up to fraction 𝑠 of
value of their house at IR = 𝑠𝑠 (1 + 𝑠)
 HELOC can be refinanced each period and thus subject to period-by-period constraint on
𝑠
balance relative to current home value: −𝑠 ≤ 𝑠 𝑠ℎ (Ω)ℎ
PAYG system

Choice of renter: can choose to rent again or buy house subject to LTV, PTI  but no collateral so
cannot borrow
 Owning = extra U flow; mortgage interest payment tax deductible while rents are not; housing
can be used as collateral for borrowing through HELOCs  houses insure HH against
fluctuations in rent prices but exposes us to K gain and losses from movements in house
prices
Choice of owners: can keep current house-mortgage and make min required payment; refinance
mortgage; sell house or default
 Possible to borrow against housing collateral through HELOCs
 Refinancing only useful as means to extract equity (not IR financing)  optimal if housing
prices rise so LTV relaxed + individuals income grows so PTI constraint loosened = depends
on which constraint binding at origination

Role of financial intermediairies: risk neutral agents with deep pockets

 Prices the mortgage
 Intuitively, if household sells or refinances, it must repay the balance remaining of the
mortgage, so the financial intermediary receives full principal + interest. If household defaults,
the intermediairy forecloses, sells the house and recovers market value of the depreciated
home. If the household continues on payment, the value of the contract to the intermediairy is
the value of the mortgage payment 𝑠 + continuation value
  Possible to solve the mortgage pricing function recursively

Rental sector: owns housing units and rents them out competitively; frictionless acquisition and sale of
units on the housing market, operating cost for each unit of housing rented out.
 Optimization implies eqm rental rate = user cost of housing
Production: 2 sectors  final goods + housing; perfectly mobile labor across sectors
 Final good sector: CRS technology, so wage = aggregate labour technology level
 Construction sector: Cobb-Douglas across land and labour; assume permits sold at
competitive price to developers  zero profit

Aggregative Risk and Computation of Equilibrium:

= news shock about fundamental parameters  can use standard techniques for computing
3 shocks: Ω (vector of exogenous and endogenous aggregate states)  each shocks follows a
Markov process-
1. Change in household income generated by shocks to aggregate productivity
2. Change in household financing conditions generated by shocks to subset of model
parameters that determine mortgage debt limits and borrowing costs
3. Change in beliefs about future housing demand generated by stochastic fluctuations between
2 regimes that differ only in likelihood of transiting to 3rd regime in which all households have
a stronger preference for housing services
+ presence of time-varying parameters that characterize credit conditions in mortgage market:
𝑠 𝑠
maximum LTV ratio at origination 𝑠 , maximum PTI level at origination 𝑠 ; mortgage origination
𝑠 𝑠
cost 𝑠 and mortgage origination wedge 𝑠  index 𝑠 = perfectly correlated
+ aggregate uncertainty over future preferences for housing services: 𝑠 (relative share betw.
∗
consumption and housing)  3 state Markov process with 3 states 𝑠𝑠 , 𝑠𝑠 , 𝑠𝑠 where 𝑠𝑠 > 𝑠𝑠 =
∗
𝑠𝑠  so two lower states are the same but probability of transiting to higher state is different  so
∗ ∗
shifting from 𝑠𝑠 to 𝑠𝑠 is a news/belief shock; between 𝑠𝑠 or 𝑠𝑠 to 𝑠𝑠 is a preference shock 
allows to construct eqm paths in which there are changes in beliefs about future preferences
for housing but those changes in preferences themselves may not realise
 Vector of exogenous aggregate shocks (Θ, 𝑠, 𝑠) as 𝑠
 Distribution of households across individual states is a state variable: 𝑠
 So overall aggregate state vector is Ω
Computational strategy: Krusell-Smith, cannot keep track of full distribution of 𝑠, replace it with a
lower dimensional vector that provides sufficient information for agents to make accurate price
forecasts
 Caveat: in every period, there is only one price that hh need to know and forecast = price of
owner-occupied housing.
′
 So knowing 𝑠ℎ in this period and how to forecast 𝑠ℎ next period conditional on the
realisation of the vector of exogenous states 𝑠′ is sufficient to pin down both the full
mortgage pricing schedule (financial intermediaries) and the rental rate (rental sector).
 Assumptions of perfect competition and linear objectives in both financial and rental sectors
allow us to reduce the dimension of the price vector to be forecasted from three to one.

Consider approximate equilibrium in which households use a conditional one-period ahead forecast
rule for house prices that is a function of the current price, current exogenous states and next period
exogenous states: has promise because housing investment entirely pinned down by the price of
housing:
′ ′ ′ ′
 law of motion for 𝑠ℎ of the form: log 𝑠ℎ (𝑠ℎ , 𝑠, 𝑠 ) = 𝑠0 (𝑠, 𝑠 ) + 𝑠1 (𝑠, 𝑠 ) log 𝑠ℎ
 iterate using actual market-clearing prices at each step: achieve convergence on the vector of
′ ′
coefficients {𝑠0 (𝑠, 𝑠 ) , 𝑠1 (𝑠, 𝑠 )}
equilibrium in incomplete markets models with aggregate shocks
 only difference with KS is that the total stock is not exogenous/given, but needs to be pinned
down in equilibrium to clear the housing market every period

Parametrisation. parametrize model to match salient life-cycle + cross-sectional dimensions (use

1998 as a reference year)  standard parametrisations:
 parameters not determined in eqm are assigned externally
 parameters determined in equilibrium are chosen to minimize the distance between a number
of eqm moments from the stationary ergodic distribution implied by the model’s stochastic
steady state + data counterparts
 parameters for preferences, endowment, financial instruments and government are defined in
a standard way to match model and moments; parameters for housing are chosen in the
same way
Shocks.
 𝑠 𝑠 𝑠 𝑠
3 types of shock. 𝑠: labour income; 𝑠 = (𝑠 , 𝑠 , 𝑠 , 𝑠 ): credit conditions and 𝑠: utilty
over housing services.
 Aggregate labour income: two point Markov chain that is obtained as discrete approximation
to AR(1) process estimated from the linearly de-trended series for total labour productivity for
the US.
 Credit conditions: capture 2 important consequences of the transformation in housing finance
that occurred in early 2000s  rise in securisation of private-label mortaages = (1) increased
their appeal with investors and enhance liquidity, reduce origination costs of underlying
mortgages for lenders + (2) reduced originator’s incentives to verify borrower’s documentation
and led to a deterioration of lenders’ screening practices  allowed many buyers to purchase
houses with virtually no downpayment and other buyers to borrow large amounts that would
have been previously possible, given their income
= Once in a generation regime shift in the credit conditions
∗
 Expectations about Future Housing Demand: three-state Markov with (𝑠𝑠 , 𝑠𝑠 , 𝑠𝑠 ) and a
𝑠
transition matrix 𝑠𝑠𝑠  2 symmetry restrictions + rows add to unity = 6 parameters to
estimate (preference parameters and transition probabilities)  defined to match data in
housing.
= simulate model  simulate a particular joint realization of these stochastic processes in order
to engineer a boom-bust episode that accurately describes the household earnings, housing
finance and house price expectations conditions during the house price boom and bust.
 create tail event (p=0.05%) where all shocks hit (increase aggregate income, relax credit
constraints, all agents believe demand for housing increase in near future),
 then shocks reversed (decrease aggregate income, tighten credit constraints, backtrack
optimistic forecast of future demand for housing)
at no point is there an actual shift in preference for housing services  about expectations 
expectations/beliefs shock

 parametrization leads to life cycle profiles for labor income, pension income, non-durable
consumption and housing consumption  match incomplete market models and empirical
counterparts
 home ownership rises with age (10% at age 25; 80% at age 55): takes time for hh to
accumulate enough savings to overcome downpayment and PTI constraints, some succeed
earlier than others because of income and wealth heterogeneity + CCRA means the optimal
portfolio allocation implies a roughly constant share of risky asset = housing so as wealth
grows, amount of owner-occupied housing grows
 among home owners with mortgages, leverage decline with age: leverage = ratio of debt to
house value  debt decreases steeply during the working life because of the retirement
savings motive + decrease in retirement because mortgage interest deduction become less
valuable as income and the relevant marginal rate fall.

Wealth distribution in the model closely reproduces the wealth distribution in the SCF: we
miss the high degree of wealth concentration among the rich that is observed in the data for
the top 10%  household only represents only one-quarter of their net worth, and thus one
would expect these households not to play a major role in the dynamics of aggregate house
prices and consumption.
 In literature: usually only focus on role of low- and middle-income homeowners.

Results. Analysis  use orthogonal nature of shocks to decompose the dynamics of different
aggregate time series into the effects of labour income, credit conditions and beliefs.
 Simulate eqm dynamics the occur when each shock hits the economy in isolation:
decompositions + counterfactuals  isolate effect of each shock on patterns of boom-bust
data (possible that there are strong interactions in the economy’s response, so that in
general, the three components do not sum to the equilibrium dynamics that occur when all
shocks hit the economy simultaneously)

Question: 2 driving forces: credit constraints + expectations: which one is more important? 
mixed empirical evidence
3 questions:
Benchmark model generates a 30% increase in house prices followed by a similar-size decline:
 Only shock that generates substantial fluctuations in house prices is the shift in beliefs about
future house price appreciation: changes in labor productivity generate very small deviations
in house prices + change in credit conditions trivial
1. Sources of boom and bust in housing market?  measured in prices
 a. Shift in beliefs: dominant force behind observed growth in house prices + change in
rent-price ration change (more than half of the fall generated by model) 
decomposition demonstrates that this is entirely due to belief shock
b. BUT Change in credit conditions  no effect on prices/rents but key factor in
dynamics of leverage and home ownership
Belief shifts  increase in house prices w/o corresponding change in debt
 counterfactual falls in leverage during boom
 (from rental rate eqm: current price increases, rents increase too, so without any change in beliefs,
the rent-price ratio would remain roughly stable; under belief shock, there is an increase in expected
future house price growth, which pushes down rents and aligns the rent-price ratio in the model with
its empirical counterpart) rent/price ratio falls  pushing marginal households out of home ownership
into renting
 we might expect counterfactual decline in home ownership since the rent-price ratio falls in the
boom (renting more attractive relative to owning)  2 factors against increase in home ownership
when only belief shock at work: rents cheaper relative to prices (marginal shift towards renting),
large increase in prices induced by shift in beliefs make downpayment constraint binding for more
households.
BUT productivity shock (rise in agg. Income relaxing PTI so more renters can now buy a house = 3%
increase in house ownership) and looser credit constraints  expand borrowing capacity (given fixed
price), loosen LTV contratint and lower cost of originating mortgages = 3% increase in households 
counterfactual increase in home ownership
BUT summing the 3 effects does not generate the dynamics of the benchmark model: interaction
between belief shock and credit relaxation: taking advantage of looser PTI, LTV requires sacrificing
current consumption, which is more acceptable when house prices expected to grow: belief shock
makes people want to own, credit conditions make people able to own.
 This rise and fall in home ownership is driven by young households: since LTV and PTI
more likely to bind and belief and credit effects more salient.

Relaxation of credit constraints affect home ownership but not house prices.
 Change in home ownership is due to households switching tenure status from renters to
owners
 Change in house prcies are driven by changes in combined aggregate quantity of housing
services demanded by both renters and owners
 So when renters become owners by buying similar sized houses to the one they were
previously renting, they affect home ownership but not aggregate housing demands/prices
Most surprising insight: house prices almost completely decoupled from credit constraints  due
to presence of rental markets + LT defaultable mortgages (usually omitted in most literature)
 For credit constraints to affect house prices: necessary for many hh to be constrained in
quantity of housing services they desire to consume and for changes in credit constraints
loosen/tighten these constraints
 Doesn’t happen in this model: (1) possibility to rent rather than own (too few households
constrainted this way)  households who cannot buy house of desired size end up renting,
do not buy small houses instead  when credit constraint relaxed: some renters become
home owners due to change in desired mix of housing equity and mortgage debt BUT does
not mean they wish to substantially increase consumption of housing services  houses buy
similar size houses than that which they have been renting  increases home ownership
rate but not aggregate demand from housing/house prices (constrained households)
 Why are renters buying when these constraints relaxed if they are already consuming
the optimal amount of housing?  housing both consumption good and asset, so
expected house appreciation tilts optimal asset portfolio towards housing.
 So mostly unconstrained owners drive movement in house prices  due to change in beliefs
 belief shock’s role is here: expectations are optimistic: homeowners upsize to take
advantage of higher expected future house price (increase house prices without changing
home ownerships)

Finding not affected by the way credit conditions chosen (from literature)
- Houses as ATMs: possible rise in approval rates for piggyback second liens and looser limits
on HELOCs  model effects through increase in the max HELOC limit
𝑠
- Adjustable rate mortgages: changes in the amortization rate 𝑠
- Movements in risk-free rate: model credit relaxation as a fall in risk-free rate over 20 year
period
Other literature attempt to explain changes in credit conditions can indeed explain a substantial
fraction of the boom and bust in house prices:
 Models in which house prices are sensitive to credit conditions are typically models in which
(1) housing commands a large collateral value due to binding collateral constraint and in
which credit conditions improve the ability to collaterise the asset or (2) housing commands a
 large risk premium, and in which changes in credit conditions have a large size of the
premium
 In class (2), Favilukis find that the house prices respond strongly to changes in the maximum
LTV  need a large housing risk premium that is sensitive to LTV limits  comes from short-
term non-defaultable mortgages, absence of rental market, high risk aversion.
 Absence of rental market means that all households are home owners, including a large
fraction of household who prefer to consume more housing but are constrained by the max
LTV limit  so when credit constraints are relaxed, these households mechanically increase
their demand for housing and so house prices increase. In addition, they increase their LTV
ratio, generating a counterfactual increase in aggregate leverage
 Short-term non-defautable mortgage debt: housing is a very risky asset for HH, any fall in
house prices requires them to cut consumption in order to rollover their housing debt  high
risk aversion = large risk premium to hold an asset that exposes them to so much risk SO size
of required risk premium responds endogenously to changes in the maximum LTV limit; when
limit is loosened, household consumptions is less exposed to those fluctuations, so risk
premium falls and house price increase.
 Short-term mortgages and absence of rental market are key: without which risk premium is
much smaller and less sensitive; with LT defaultable mortgages, consumption less exposed to
movements in house prices + presence of rental markets means less constrainted HH
 Which set more convincing?: existence of rental market more realistic than absence (1/3 of
HH in US renters) + long-term more realistic: near universal in the US (median mortgager had
30y contract) than one-period non-defaultable debt
 Credit conditions in explaining the boom-bust in house prices only if one modifies the model in
directions that take it further from, rather than closer to reality.

Leverage. Model generates flat path for leverage (=debt/housing wealth) during the boom: 2 effects =
belief effect pushes up house prices so leverage falls + looser credit conditions lead to expansion in
debt, causing leverage to rise
 During bust: belief shock generates sharp rise in leverage because of large drop in prices
BUT stock of mortgage debt remains well above pre-boom level for over a decade 
households deleverage slower
 (2) presence of LT mortgages dampens channels: if debt ST then sudden credit constraint
forces all home owners to cut consumption to meet new limit  this risk factor leads to
sizable/volatile housing risk premiums  change in prices proportional to debt; if debt LT then
constraints only bind at origination and irrelevant for remainder of contract  reduces
strength of risk premium channel since credit tightening does not force home owners into
costly consumption fluctuations to quickly deleverage
 Households who do not want to default can slowly reduce their debt by sticking to their
existing amortization schedule  avoid large swings in consumption
Foreclosures. Model spike matches spike in foreclosures at start of bust in data:
 hh with high levels of mortgage debt at peak, dragged into negative equity after shifts in
beliefs decrease prices
 not due to tightening of credit conditions: does not move prices + in environment with LT debt,
tightening of LTV and PTI constraints is only relevant at origin
 strong interactions between credit conditions and beliefs: credit relaxation amplifies the effect
of belief shifts on foreclosures because during the boom, it enables optimistic buyers to obtain
larger and cheaper mortgages, which in the bust find themselves underwater due to price fall.

Matching data?
= realigns model with data (no real change in demand for homes)
Matches cross-sectional observations: uniform expansion of mortgage debt during boom across
income level + increasing share of default during bust (subprime borrowers) + crucial role of young
households in accounting for dynamics of home ownership

2. To what extent/through what channels did movement in house prices influence consumption?
Half of corresponding boom-bust in non-durable expenditures comes from the boom-bust in house
prices (belief shock);
 Credit conditions has virtually no effect since it does not affect prices  half the boom-bust in
consumption can be accounted for by change in house prices
the other half from labour income
 wealth effect (drop in consumption during bust + initial share of housing wealth in total wealth)
 the larger the share of housing wealth in total wealth, the bigger is the impact of the fall in
house prices on consumption (elasticity close to 1)
 negative wealth effect from fall in house prices  fall in aggregate consumption
 o counteracting force: drop in house prices leads to a positive wealth effect for renters
who plan to become homeowners + positive wealth effect for some existing
homeowners who plan to upsize
o sign and size of aggregate wealth effect on non-durables depends on the joint
distribution across households of expected future changes in housing units and
marginal propensities to consume non-durables
o by age 40-45, both extensive (home ownership) and intensive (housing units)
margins have levelled off  expect to climb down, rather than up, the housing ladder
 75% of households experience a negative wealth effect from fall in house prices
(expect to downsize)
o despite having slightly smaller MPCs on average than the remaining 25% of
households, their abundance means that the aggregate effect is a fall in consumption
 the aggregate elasticity of consumption to house prices in our model is broadly
consistent with principle that wealth effect is operative

 emphasis on heterogeneity in household balance sheets as key factor in consumption dynamics

around Recession
 aggregate non-zero wealth effect: depends on shape of life cycle profile of home-ownership in the
model: homeowners who expect to downsize (bust loser) control greater share of aggregate
consumption than those households who expect to increase demand for housing (bust winners)
 substitution, collateral effect: less important channel

3. Role for debt forgiveness policies at height of crisis?

Home Affordable Modification Program (HAMP) + Home Affordable Refinance Program (HARP) 
cushion collapse of house prices + aggregate demand but limited success (too complex, scope too
small)
 model: policy intervention in which all homeowners with LTV ratios above 95% in 2009 have a
fraction of their mortgage debt forgiven so that their LTV ratio is brought down to 95%.
 = ¼ of homeowners in data  more generous than HARP or HAMP + financing does not
induce additional distortions from higher taxes
 Reducing underwater households, the intervention is successful at reducing foreclosures. 
to extent foreclosures lead to utility losses and higher depreciation, there are welfare gains for
affected households and for corresponding banks whose assets more valuable  so
mechanical smaller increase in aggregate leverage
 Does not affect decline in house prices nor decline in consumption (disconnect between credit
and house prices, and here house prices are key driver of consumption dynamics in the
model)
 Foreclosure is also motor for consumption smoothing, so limiting foreclosures induces
households who would have otherwise defaulted to consume less (make mortgage payments
happily to avoid default)  where HAMP used most extensively, experience virtually no
change in non-durable consumption but did experience a reduction in foreclosures
 Intervention has trivial impact on fall in consumption during housing bust, but has small but
very persistent, positive effect on consumption during the subsequent recovery  forgiving
some debt for a large no. of hh = lower future mortgage payments + many hh are nearly
hand-to-mouth so they increase consumption slowly as the lower payments are ralised, rather
than immediately (agrees with Ganong-Noel)
= despite timing and scale, debt forgiveness program would not have prevented sharp drop in house
prices + aggregate expenditures  driven by very weak effects of change in leverage on house prices
BUT significantly dampened the spike in foreclosures by keeping many HH above water.
 program lowers mortgage payments for remaining life of mortgage contract  aggregate non-
durable consumption increases through most of post-bust recovery?

Generally, large empirical literature to explain causes/consequences of housing boom and bust by
exploiting cross-sectional variation (households/regions) in house prices, balance sheets, housing
market outcomes, financial conditions, consumer expenditures and labour market outcomes
 Early emphasis on subprime borrowers: households who were excluded from mortgage
markets before the credit relaxation of the early 2000’s opened the door for them to become
home owners  this group of low-income, high-risk households identified as the group most
responsible for the expansion in mortgage debt during the boom, and consequent
delinquencies and foreclosures during the bust.
 Evolved since: 2 challenges  (1) credit growth during boom years not concentrated among
low-income households, but rather was uniformly across the income distribution (model is
consistent with his observation). Here: shift in expectations about future house price growth is
key reason why the model generates an expansion of credit even for high-income households
 all household expect large capital gains from holding housing, but high-income, low-risk
 households are those in the best position to take advantage of the optimistic beliefs, since
they can access low-cost mortgages even in the absence of looser credit conditions.
 (2): shares of foreclosures for households in the lowest quartile of the FICO score distribution
at origination decreased during the bust  prime borrowers contributed to the dynamics of
the housing crisis at least as much as sub-prime ones; model here is consistent with a decline
in the share of foreclosures among subprime borrowers; with the role of expectations is key. A
sizable fraction of prime borrowers, who levered up to buy more housing expecting to soon
realise capital gains, find themselves with negative equity and choose to default.

Overall: central role for expectations about increase in future house prices = expectations about
future HH demand = expectations over future availability of buildable land  as long as shock to
future preferences.
= generate rational changes in beliefs about future house prices through exogenous stochastic
changes in conditional probability over future preferences for housing services: requires 3
assumptions, test by counterfactual experiments where we modify some aspects of the belief shock
(credit conditions, labour productivity shocks follow the same path as in benchmark)
1) Agents all share beliefs
 Need for all agents to share same beliefs? all HH must expect future house price growth 
for HH to demand more housing  push up prices
 Rent/price ratio to fall as in data  rental sector must also expect future price growth
 Lenders also believe that house prices likely to increase  rationally expect borrowers less
likely to default  reduce spreads on mortgage rates (esp. risky ones) as in boom
 Shifts in endogenous beliefs generate endogenous shifts in credit supply  expansion +
contraction of cheap finds to subprime borrowers.

In benchmark: assume all agents share same beliefs about future house prices and share the same
references over housing services and beliefs about those preferences. So when preferences switch
∗
from state 𝑠 = 𝑠𝑠 to 𝑠 = 𝑠𝑠 , there are four channels through which the resulting shifts in beliefs
could affect the economy
A: own beliefs / B: other households’ beliefs:
 individual hh believe that in the future they themselves are likely to desire more housing
services relative to non-durable consumption but housing adjustment costly, they might
increase their housing demand immediately, even in the absence of any current change in
house prices  direct effect
 individual HH believe that all other hhs are likely to desire more housing services in the future,
and thus rationally foresee that this future expansion in house demand will lead to higher
future prices  speculative motive may lead household to increase its housing demand
 direct effect vs speculative motive: each motive shown to be quite weak on its own, strong
interaction between the two  optimistic expectations about own preferences lead
households to want to purchase more housing + optimistic beliefs about other hh’s
preferences leads to optimistic beliefs about house price growth  induces them to move
those purchases to present, this drives up house prices and rationalized beliefs
C: lenders’ beliefs: mortgage lenders understand both the direct and speculative effect  rationally
expect future house prices to rise + expectations of lower default rates  so lenders optimally offer
more attractive mortgage contracts to households
 endogenous improvement in mortgage credit may lead households to demand more housing
 if only lenders optimistic: no boom-bust in house prices, and the movement in consumption
are severely dampened  little growth in house price, pessimistic household economy does
not generate a fall in the rent-price ratio  counterfactually high leverage during the boom
 the default probabilities forecasted by the banks + implied offered mortgage rates do not
embed the impact of belief shifts on expected future house price growth
 lender beliefs are relatively unimportant for the boom-bust in house prices and consumption
 shifts in lenders’ expectations are crucial for the dynamics of home ownership, leverage
and foreclosure
 without lenders: home ownership counterfactually falls when households become optimistic
during the boom, and leverage and foreclosures increase much less than in the data when
households’ optimisms reverses during the data  particularly stark effect on foreclosures
 lenders’ beliefs = when lenders also believe that future house price growth will increase,
expect lower default rates, the mortgage rate schedule flatten, leading to a fall in IR for highly-
levered borrowers  our model generates an endogenous credit supply shock = expansion of
cheap funds to risk borrowers  accounts for a rise in mortgage debt during the boom of
10%, or nearly 1/3 of total.
 without shift in lender beliefs about expected future house price growth, high risk borrowers
do not experience this endogenous fall in spreads, many of them then stay renters/buy but
take lower leveraged positions which are less prone to default in the bust
  only small effect on house prices, but critical to match joint dynamics of home ownership,
leverage and foreclosure
D: rental company beliefs
 in benchmark: during boom rental companies also rationally expect an increase in future
house prices  lowers the user cost of housing and keeps rental rate down
 allows the model to match sharp fall and subsequent rise in rent-price ratio
 without these, we would infer that belief shocks lead to excessive rise in home ownerships.

2) Change in beliefs about preferences rather than preferences themselves

Compare model with shock to beliefs (L to L*) vs shock to preferences (L to H)
 Similar boom-bust in prices
 When actual preferences change: consumption dynamics are counterfactual  since
household’s utility from housing services increases relative to non-durable consumption, they
substitute away from consumption, causing it drop sharply in the boom and to rise in the bust
 opposite of what’s in the data
 Joint dynamics of consumption and house prices are hence strong evidence against an actual
shift in housing demand and are, instead, consistent with a shift in beliefs (NO CHANGE IN
UTILITY  valid assumption?)

3) Changes in beliefs are over a parameter that affects housing demand rather than
housing supply
Possible to induce a boom-bust in house prices though a change in beliefs over any structural
parameter, provided that were that change to actually occur, equilibrium house prices would
qualitatively move in same way  not essential that the swing in beliefs affects housing demand
rather than housing supply.
 Can generate an increase in expected future house price growth by assuming that
households come to believe that fewer land permits will be available in the future
 Cannot be actual fall in number of available land permits would lead to an increase in prices,
as it would also lead to a counterfactual drop in housing investment (analogously to the
counterfactual drop in consumption)

Lack of direct evidence on role of expectations  evidence does point convincingly to (hh, interests,
lender) widely sharing beliefs
 Index of sentiment in housing market: tone of housing news  big swings
 Mid-level scrutinized finance managers did not sell off their personal housing assets during
boom  lenders share same beliefs as rest of market about future house prices growth
 Internal reports from major banks reveal that analysts were accurate in their forecast of gains
+ losses but optimistic in that extremely low probability assigned to collapse of home prices

Conclusion:
Optimistic beliefs  lenders to endogenous expand cheaper credit to risky borrowers + led to housing
boom and bust
Exogenous relaxation of credit conditions was essentially for explaining the dynamics of leverage,
home ownership and foreclosure, despite it having a negligible impact on house prices and
consumption
 Difficult to separately identify the effects of credit supply shocks and shifts in expectations
from micro data

Methodology: heterogeneous-agent incomplete-markets OLG with endogenous house prices, rent

and mortgage schedules + multiple aggregate shocks  seemingly intractable
 Helped to assume rental and credit markets are both frictionless and competititve: single law
of motion for house prices
 Computational complexity

3 discussion areas:
1. Exogenous belief shifts  conceivable that optimistic beliefs emerged due to a primitive
change in some fundamental such as low IR environment or the change in lenders’ behaviour
= learning models: fundamental shocks can lead to the formation of optimistic expectations 
low IR seems more promising
2. Low internal propagation of exog shocks to house prices and cannot generate house price
momentum  does generate propagation for home ownership/cons/leverage  more gradual
unraveling of initial shocks/diffusion of beliefs in population OR slow response due to frictions
3. No feedback from house prices to earnings: include nominal rigidities/demand externalities
and collapse in house prices could cause decline in aggregate labour demand that leads to
decline in earnings  part of fall in expenditures that we attributed to labour income would
actually be attributed to house prices  this model provides lower bound
Very robust