2008 Financial Crisis: A Ten-Year Review conference. Panel on Housing and Financial Crises
Articles,  Blog

2008 Financial Crisis: A Ten-Year Review conference. Panel on Housing and Financial Crises

All right, if you
look at your program, you’ll notice that this is
the session that is standing between you and refreshments. So among other
reasons, I’m going to try to keep us on track. We have two speakers. Our topic is Housing
and Housing Finance. And as was brought out
by Andy Lowe’s comments earlier today, we’ve
got lots of narratives, but one narrative
that doesn’t go away is that there were issues
about Housing Finance. And so, we’ve got two speakers
today to address that. Each of you have 20 minutes. I’m going to sit down there. And if you see me waving
my hand– you know, I’ll put up five. I’ll put up three– you’re all witnesses
to this, and we’ll try to get you all out
to refreshments on time. We’ve got our two
speakers, Chris Foote, who is a senior economist
and advisor, a policy advisor at the Federal
Reserve Bank of Boston. And we’ve got Antoinette Schoar,
who is the Michael M Kroner professor of Finance
and Entrepreneurship at the MIT Sloan
School of Management. And so, Chris is going to start. Chris, you’ve got 20 minutes. I’m going to be tough. And take it away. OK, thanks very much. So the most important– I want to also point out this
is a joint paper with Paul Willen, who is seated
safely in the back row and can help me answer
any difficult questions. It’s very important, though,
that I point out Paul and I are both employees of the
Federal Reserve Bank of Boston. These views on mortgage
default and other things are our own views. They do not represent– well, they may, but
they don’t necessarily represent the views of
our boss Eric Rosengren or the big boss Jay Powell. So it is not surprising that
the editors of the ARFE for 2018 would want a paper
on mortgage default, because we certainly
saw, as you well know, a large increase in
mortgage defaults during the recent recession. 90-day defaults,
90-day delinquencies, which typically do not move
very much in recessions, rose with a vengeance in
the most recent recession. And a natural question
that economists have is, why were there so many defaults? But when you frame the
question in that way, you’re really asking,
why did we have a recession that was so closely
tied to the housing cycle? Why did we have a housing
cycle to begin with? And there are common
explanations for that. We’ll talk more
about them later. Were there a lot of unaffordable
loans, that were originated, that later went bad? Were there distorted beliefs,
or overoptimistic beliefs, about housing prices
that encouraged lenders to make a lot of
loans that went bad? That’s actually
not the right way to frame this issue,
to frame this question, if you’re standing from the
vantage point of a mortgage default researcher. If you’re coming at it
and ask, what influence does this green line,
these increase in defaults, have on the mortgage
default literature? There, the right
question is really, why we had so few defaults? And by few defaults,
I mean so few defaults conditional on the
collapse of housing prices that we saw before and
during the Great Recession. So on the left panel,
the blue and red lines show you HPA, House
Price Appreciation– the rate of change in
housing prices over time. That typically
declines in recessions, but it declined with a vengeance
in the most recent recession. On the right, you see
the fraction of homes, according to CoreLogic,
that had negative equity or near-negative equity, LTV
ratios equal to or greater than 100% or greater
than or equal to 95%. And if you can’t see the numbers
in 2009, we had about a quarter to a third of mortgaged
households in the United States, that owed
more on their homes than their homes were worth,
that had negative equity. And as you know, negative
equity is typically a precursor of
default, but a third or a quarter of mortgages in the
United States did not default. The number of defaults was
much, much lower than that, as you saw in the
previous graph. Now theoretically,
as it turns out, the benchmark model that we
use in the Mortgage Default Literature, the
Frictionless Option Model, is qualitatively
consistent with the fact that a lot of borrowers with
negative equity do not default. And this model, which
has been developed over the last several
decades, comes out of applying standard
option pricing techniques to the
borrower’s default decision. Now the graph I have up
here is from the paper. We’re very proud of this
graph and the MATLAB program that generated it. I’m not going to go through
it in detail mainly because I don’t really want to explain
option pricing when Robert Merton is sitting
in the front row, but the main, for
my purposes, it’s enough to really give
you the intuition of what this Frictionless Option
Model is all about. Now it starts with the
assumption, as you might guess, of frictionless
capital markets– of someone who has a mortgage
can also freely borrow and lend, outside of that
mortgage arrangement, at the risk-free rate. There’s no limits to
short and long positions, and we sort of have
the common treatment of stochastic processes for
house prices and interest rates. They’re volatile, but there’s
a no arbitrage condition that sort of pins them down. Now what’s the
implication of this setup? Well, the implication
is that once borrowers get a little bit of
negative equity– if housing prices
fall a little bit below the value of
their mortgage– they don’t instantly
default. Why? Because there’s an
option value of waiting to see if the housing prices
recover, and default only occurs when the amount
of negative equity reaches some level that’s so
deep that, in a precise sense, it’s really not worth
continuing making the payments, in the hope that the price
will go back up again. Now in 2008, along with Chris
Gerardi and Paul Willen, the three of us wrote a paper
that used Massachusetts data from the early 1990s, which
had gone through a house price decline at that time, to show
that we could actually measure, using deed registry
data, how many mortgage borrowers
in Massachusetts had the negative equity. And we showed that, over
time, only about 10% of those home owners
lost their homes. So we were, you know, score
one for the Frictionless Option Model. We also sort of wrote down
a simple two-period model to express some
of this intuition. Now that’s not the
end of the story though, because
problems come when you move from qualitative
statements, generated by the FOM about default,
to quantitative statements. What happens when you actually
calibrate the Frictionless Option Model with a reasonable
variance of housing prices? How big is that region
of negative equity, where people aren’t going to default? And I’m going to illustrate
this with a graph here, so this is kind of pairing
the default function, as generated by the FOM. I don’t if I can
really point here. But as you move to the
right into the blue region of the graph, you’re moving
to a higher LTV ratio, and once you pass 100% you’re
in negative equity land. You don’t default, according
to the FOM, right at 100% negative equity. You default at, you know, some
positive or some further value of negative equity. That’s where it says
FOM Default Threshold. But at that level, boom,
your probability of default goes all the way up to 1,
and at very high levels of negative equity you default. Now how big is that
FOM default threshold with reasonable parameters? You’re typically looking
at something like 110% LTV, 115%, 120%. It’s hard to get that
to be 150%, all right? So that’s sort of how we think
about things quantitatively. Now, the blue and
the red lines there are sort of empirical
default functions, with borrowers of different
types of credit scores. That’s not so important,
the difference between the blue and the red. The important point is that,
empirically, the default function looks like
there’s a lot of people, who have very high levels of
negative equity, that are not defaulting. And in the literature,
those are attributed to transactions costs. And transactions costs are, to
mortgage default researchers, what dark matter are or
is to astronomers, right? It’s all the stuff we really
can’t quantify and really don’t know about. It might be the stigma that
you get from defaulting, the hit that you get
from your credit score, having to move your
kids into a new school. Whatever it is, it’s a cost that
depresses the empirical default function below what you
would get from the FOM. That’s problem number one. And we’ve known about
these transactions costs going way back to the 1980s. A nice paper in the
Journal of Finance, by Buddha, shows that
they’re relevant for the current crisis, as well. Another problem is that
borrow-level characteristics seem to matter for default.
And I illustrate here with sort of characteristics
of a borrower’s credit score, but an easier way
to think about it is think about a borrower
who loses his or her job, or who gets divorce,
or who gets sick– who suffers some sort of adverse
life event that makes it hard for them to keep up with
paying their mortgage. In the Frictionless
Model, remember those borrowers
can borrow and lend freely even outside their
mortgage arrangement. So if the borrower
loses the job, the borrower just borrows and
takes a loan, and that’s that. If the FOM says they
should continue to pay, they continue to pay. In the real world, however,
borrowing constraints limit the ability of mortgage
borrowers, who face or suffer adverse life events, from
taking those types of loans. So what happens? Borrowers who suffer adverse
life events often default. Now when we think about the
effect of adverse life events on default, we’re led to
sort of a different approach to thinking about
default, and this is the famous
double-trigger model. So on the left panel there, you
have a very simple, or basic, treatment of the
double-trigger model, which says default occurs
when two triggers are pulled. First, the borrower has
to be in the left column, so that they have
negative equity. Borrowers who have
positive equity, don’t default. They just
sell their house no matter what happens and pocket
what’s left over. The top row, though– the
second trigger is– the top row says the borrower also has to
suffer an adverse life event. So if both of those triggers
get pulled, according to the double-trigger model, the
borrower is going to default. The other panel there, panel
B, shows a more sophisticated– or what we call the modified
double-trigger model. And in that model,
you might think that there’s a little
bit of optimization going on, such that if you’re
in the upper right corner of that triangle, you’ve
got low, negative equity. You basically have
almost positive equity, and so it’s going to take a
very big income shock for you to default. If you’re
in the other corner, you’ve got very deep,
negative equity, so it may take only
a small adverse life event for you to default. And so, when we start
thinking about the modified double-trigger model, what we’re
trying to do is really kind of blend ideas from the
Frictionless Option model– which sort of really stresses
only negative equity– to adverse life events, and
that really takes us to the– I want to actually skip this. That actually takes us
to the research frontier. What are the papers
and mortgage defaults, that are appearing in
journals, trying to do today? Well, they’re
trying to write down models in which expectations
about future house prices matter, as implied by the FOM. They may try to sort
of adjust the way that expectations are
formed, so that they might be non-rational, in some sense. They might sort of
expect mean reversion. So what do I mean by that? Well, here are some graphs
of house prices for the US, as a whole, and three of
the so-called sand states. Arizona, Florida, Nevada, and
California is not on there. So if you look at the
upper right panel, that’s for Arizona. You see that Arizona
had a big decline, but then there was a non-trivial
recovery in prices after that. What would a borrower do if
they were sitting in the trough, and they sort of knew– really what happens
after crashes is that the prices crash,
but typically there’s a period at which they are more
likely to rise than to decline. They would therefore
be more likely to hold on to their homes at any
given level of negative equity that would be consistent
with what we find– is that the default
is less prevalent than a pure Frictionless
Option Model would suggest. Third sort of thing
that people are trying to do in these
models is that they want to show that they are
borrowing constraints, or at least allow for
borrowing constraints, so that adverse
life events matter– and that most
borrowers are going to remain current on their loans
even when negative equity is deep and/or
liquidity constraints bind, because of these
amorphous transactions costs– that we don’t
know much about, but that we’d like
to know more about. Now that’s sort of
the whirlwind tour of kind of how the mortgage
default research is now progressing, given what we’ve
sort of had in the crisis. I next want to talk
about– oh, before I leave that, I want to say
that, over time, the data to test this model is going
to probably get even better. Most people working in the
mortgage default literature, empirically, are
using loan-level data on millions of observations. Ideally though,
we’d like to have a data that is large like that,
but has information not on only how the mortgage is doing,
but what else is going on in the borrower’s life. It would be great to
know, for example, on a month-by-month basis, how
much income that borrower has– not too many data sets like
that, but some of those may come online later on. OK, so the organizers of the–
or the editors of the journal. I learned today that
these papers are supposed to be reviewed with attitude. We weren’t told about
the attitude part. We were just told
about the review part, but I think the next section has
a little bit of attitude in it. Because what we do in the paper,
in three areas of the paper, is talk about how lessons from
the mortgage default literature could be very relevant for
important questions related to policy in the
foreclosure crisis– and also relevant for sort
of why the crisis happened to begin with. So one question is
should the government have written down mortgage
principal more aggressively? And there’s also ones about
the patterns of defaults and the potential causality
between foreclosures and house prices. So the first one was
there was a lot of debate about how the government should
try to prevent foreclosures during the acute
phase of the crisis, and specifically whether
or not the government should try to reduce principal,
mass write-downs of principal, or to focus on reducing
the current monthly payment that borrowers have to pay. And what they eventually
settled on, with HAMP, was really policies that
were targeted mostly on the payment
reduction side, not mass reductions of the
principle that they borrowed. Now we think that’s a good
idea, because the key problem with principal reduction, as
an anti-foreclosure strategy, is it really doesn’t work
that well no matter what type of default you’re
trying to prevent. So think about a
double-trigger default. The whole point about
double-trigger defaults, or the whole thing that sort
of gets them into the model, are liquidity constraints– the idea that $1 today is worth
a heck of a lot more than $1 later on. Well, what does a mass
write-down of principle do? It basically means, hey,
when you sell this house, you’re going to get a lot more
money in your pocket later on. And that doesn’t help you
too much in the present, and there’s a great paper
in the Brookings series, by Emily and Christine
Murthy, which really draws this implication out. Well, what about
principal reduction, as a way of reducing just
pure strategic defaults that would be implied by the
Frictionless Option Model– defaults that occur just because
negative equity is very deep? Well, there, the
transactions costs come into play, because
the very low default rate among
non-stressed borrowers, or borrowers that
haven’t lost their job or suffered any
adverse life event, means that mass principal
reductions aren’t going to be cost effective
in preventing those defaults. There might be a lot of
people with negative equity, very deep levels of negative
equity, but empirically few of those people
actually default. The data show, in papers
that are coming out now– one written by Paul
and, a co-author, Andreas [? Fewster, ?] do show that
payment reductions that are sort of caused by,
for example, interest rate resets– if your interest
rate resets down, such that your payment is
reduced by a large amount, lo and behold, that has a
very significant impact on your default frequency. However, as a matter of
policy, the important thing to keep in mind is that any
anti-foreclosure program is going to have an imperfect
information problem. Everyone’s going to want
to say, yes, you know, I can’t pay my
mortgage this month. I need a reduction. Or yes, I’ve got
negative equity. I’m going to default. But
in reality, not everyone is actually going
to default, and that makes things a little tricky. So the second question
involves default patterns and the underwriting
standards as the ultimate sort of cause of the crisis. So there were a number of early
papers in the literature that used patterns of defaults
among marginal borrowers to sort of point to this
idea that the crisis resulted in large part from sort
of an exogenous increase in bad underwriting, through
securitization or whatever. And we talk about
these papers a lot, and sort of discuss some
alternative interpretations of their findings. I think now recent
work is showing that the problems in
the mortgage market were widespread throughout
the income distribution– and certainly this
is what Antoinette is going to talk
about– and that there’s a new narrative that links
the housing cycle to distorted beliefs about asset
pricing, rather than securitization
or bad underwriting. But a lot of this research that
has sort of– in thinking about defaults– can kind of help us with this. In 2012, again,
with Chris Gerardi, we pointed out that if
you look at the– people were talking about
subprime defaults, and the resets of subprime
defaults to higher interest rates were causing
a lot of defaults. We talk about how, in
the paper, 89% or 88% of people who
ultimately defaulted did so paying the exact
same payment they made at origination, and so
if you look at the paper you’ll see a lot of
discussion of defaults and the patterns of defaults–
and show how they may not be consistent with the
conventional wisdom, but could be convention is
consistent with the story about beliefs. So finally, there’s
a lot of talk about the effect of
foreclosures on house prices. Pretty much everything
I’ve said up to now talks about how declining prices
can lead to more foreclosures. Well, could there
be a causality, going from a wave
of foreclosures, causing house prices to decline? And we have a long section– well, not too long, but a
thorough treatment of that question in the paper. Now it is true, and
theoretically, foreclosures do reduce the set of
potential home buyers, but unless those folks who get
foreclosed on go live in caves foreclosures also raise the
set of potential renters and, therefore, rents. And so therefore,
with higher rents, that encourages landlords
to go buy those homes, and convert them to
rentals, and rent them out. Now a key question
here– and that is sort of having a dampening
on any decline in prices that would come from the
initial wave of foreclosures. Now a key question
here that’s also relevant for a lot
of other issues is how separate the
owner-occupied market and the rental
markets really are? How easy is it to
turn a house that is owned into a house that
is rented, and vise versa? But in any case,
in the data, you would actually see when housing
prices stabilized was precisely the time in which there were a
lot of completed foreclosures going out on the market. So our view is that
theoretically this is possible, that you may see
foreclosures affect housing prices, but the
existence of landlords and the data– landlords, in
theory, and the data suggests this is not
really a big channel. So the bottom lines
of our paper– just as the Great
Depression sort of stimulated macroeconomics,
we think the great recession has very significantly
stimulated research on mortgage default.
The basic patterns, though, were consistent with what
any default researcher would have told you circa 2002. You’re going to have a
big increase in defaults after a house price collapse. That’s not surprising. The vast majority of
people with negative equity are not going to
default, and as a result that has big implications for
policy to prevent foreclosures, making it difficult.
Going forward, researchers are
going to try to blend the FOM with the
double-trigger models, and here the treatment
of price expectations is going to be critical. Do borrowers kind of have
over-optimistic views either in the boom or
after a big trough? And empirical work
is really going to be helped by more
big data coming online. I would want to say,
though, that when you see a bunch of
defaults going on, it is a feature of risk
sharing in our economy. And we close our paper
with a discussion of very important papers that
say, under imperfect markets, if you can’t specify
outcomes for a mortgage payment at every point
in time given what’s happening with a
borrower, defaults may be the best way
we have to share risk. And with that, I
think I’m out of time. Thank you, Chris. Thank you for staying
on track and on time. By the way, I don’t think
I’ve introduced myself before. I’m Larry Boyd. I’m a professor of economics
here at the Stern School. All right, you heard
from Andy earlier today about Antoinette’s research. Antoinette, take over. OK, fantastic. Cool. Thank you very much. So what I want to talk about
is the role of housing markets in the run up to the crisis. And I want to exactly pick
up, where Chris left off, to show you a set of
results about housing market and mortgage markets in the
run-up, in the aftermath, of the crisis that might
actually, to some of you, be surprising, given
the narratives– that we have
learned to believe– about housing markets
in the Great Recession. And this work is a joint work
with Manuel Adelino and Felipe Severino, who are my
co-authors in a set of papers and also an RFE paper that
we wrote for this volume. So what I want to pick
up on is something that Chris said
at the end, which is that the narrative
about the crisis initially started
off only focusing on distorted incentives
in housing markets that then led to a bunch
of trigger on effects. But what I want
to show you today is a set of results,
empirical results, that much more suggests that what we
had at the core of the housing market was a classic
housing bubble that then led basically– or
driven by rising house prices and
over-optimistic beliefs, we saw that households
increased borrowing. And households also
increased participation in buying and housing markets. But at the same time
banks also bought into these rising housing
prices and expectations by lending against collateral,
if you want, unabashedly, and in some sense seem
to have underestimated the potential for
reversal in house prices and basically visit the
defaults that then Chris was talking about. And before I show
you the results, to me, the big takeaway
or the big, in a way, misnomer of this crises– except, we’re still calling
it the subprime crisis– what I will show
you is actually this was a much wider crisis
in the housing market. And in particular, middle
class and upper middle class borrowers were very much
part of the credit expansion and, in particular, the
defaults afterwards. And, you know, this
obviously matters, as we’ll discuss at
the end, because it affects how we want to think
about policy implications in preventing these kind
of crises going forward. So the way I see this is that
the financial sector acted as an amplification mechanism
by lending into these increased house prices, which
kind of picked up on the over-optimistic
expectations but what we do not
find and I show you now the data is that
there was somehow an unilateral shift or
an unprecedented shift say of lending to just marginal
or subprime borrowers again there was an over-arching shift. So the facts that we’ll
be looking at today is basically first
focusing on the run up to the financial crisis. What I’ll show
you is that credit expanded across the entire
income distribution. It didn’t just
expansive– for subprime or, you know, poor and
marginal borrowers. In particular, I’ll show you
that middle and high income households actually had
much larger contributions to the run up in the total
dollar of mortgage debt before the crisis,
and then I’ll also show you that, actually,
those households, afterwards, defaulted at the
very high level. But interestingly, what
I will also show you is that debt-to-income levels
expanded proportionately across all types of
borrowers– subprime and prime borrowers, low-income,
high-income borrowers. Surprisingly at
the same time, we see that the Loan-to-Value
distribution, the LTV distribution, actually stayed
stable over that time period, meaning that, in
particular, the increase in the debt-to-income
levels seemed to have been fueled by lending
against increased collateral values, but not a shift
towards higher LTVs. And I’ll show you
this in a second. Now on top of it, I’ll also
show you that home owners seemed to have been buying and
selling, or flipping housing, at a much higher rate in
the run up to the crisis, but surprisingly we do not find,
for example, that ownership rate increased. In fact, if anything,
home ownership rates decreased even
prior to the crisis, which is actually a result
that’s quite surprising. But as I’ll show you,
it’s heterogeneous between subprime and prime, or
poorer and richer, borrowers. And then in the
aftermath of the crisis, I will show you how there were
sharp increases in delinquency rates, but the biggest increase
in the dollar value of defaults actually came from
the middle class and the upper middle
class, who defaulted at an unprecedented rate. All right, so let’s
dive into the data. Ultimately, this is,
as Chris also said, an empirical question. So the first thing,
as I said, right– what I want to show
you is in the run up to the financial
crisis, there was no dislocation of
where credit was flowing, but what we see is that there
was a symmetric increase, if you want, of home owners
participating in the housing market and in the mortgage
market over that time period. So the first thing I’ll show
you– if you look from the left to the right, this
is mortgage dollars originated by income quantile. Right, the top is the
highest income quantile. The bottom is the
lowest income quantile, which we fixed as of 2000
and then held constant. And this is to show you that,
actually, from 2002 till 2006, year by year, the distribution
of mortgage dollars actually stayed constant. So if you look at this, the
white cells at the bottom, this would be typically
probably your lower income, subprime customers, but
you see that there was not a disproportionate increase of
credit flowing to these groups. No, instead we see that,
across the income distribution, people were participating
and, of course, expanding borrowing
over this time period. I can show you this also
in a different– oops, why is it doing it? OK, I can show you this
in a different way. Rather than binning by
income, I been by FICO score. So this is what we typically
would call a subprime rate below 660, between 660 and
620, and the top graph here– sorry, and 720. And the top graph is above 720. And what do you see again is
that, between 2000 and 2006– we actually see that the
dollar the fraction of mortgage dollars flowing to
subprime customers stayed constant, right? Of course, the mortgage
market expanded. There was much
more lending done, but the distribution across
the different FICO bins actually stayed constant, right? So again, this is saying
that there was not a disproportionate
shift of where mortgage dollars
were originated, or where they were flowing. Another way of
saying it, again, is that everybody seemed to
have been participating in that credit boom. All right, so of
course I already said– these graphs, while I’m saying
they proportionately stay flat, we know that underneath it–
total amount of dollars raised or originated– increased. And how can I see this? There’s two ways I
can show you that. Number one is– so this is from
the Survey of Consumer Finance. This is total mortgage dollar
at the household level, including– people call it CLTV, right– including second leans and
all mortgage dollars, what you see is proportionately– again, across all
income quantiles, debt-to-income level
at the household level increased, right. But it was, again, proportionate
across all income levels. There was not a kind of
disproportionate shift to, say, more marginal,
or subprime, borrowers. Or another way of
saying this, it seems everybody bought
into this housing bubble and, basically, into the
increased housing prices. Actually, surprisingly,
just to show you, in a way almost how mechanical
people were borrowing is when you look at the loan-to-value
distribution, the famous LTV, from the mid-1990s– and I’ll show you
this basically– till 2012, what you see is
that that distribution actually stayed very flat, right. So this is saying that the
fraction of households that had, say, a 95% LTV,
a 90% LTV, a 80% LTV– these are the bins here, right– again, stayed stable
over that time period. Now of course I showed
you the DTI went up. So what does this mean? What this means is that
because V, the Value of homes were going up, people
were able to take on more credit against those
now more valuable houses. Therefore, DTI went up, but
the loan-to-value distribution stayed stable, meaning that what
banks did is they just almost, you could say, mechanically lent
against these now higher house price values, right. Now why am I making
such a big deal out of is because that
actually suggests that, in some sense,
what went wrong here is really the expectations
about how believable and how stable these
house price values are, and how banks should have
treated them as collateral. And what this data shows
is that they basically believed what they saw
and did not become, say, more cautious after seeing
a huge run-up of prices in many areas. Yeah? All right, the final
thing I want to show you– because in the run-up part, lots
of people then have said, OK, maybe the intensity– meaning the fraction
of loans that people took on, as you said, and
is kind of this story. But what about the
extensive margin, right? I’m sure many of us have
heard this story that, oh, but what the banks
really got wrong is they started to lend to more
and more marginal borrowers. They kind of were digging
in the garbage can, as people sometimes
like to say, and get out basically loans that should
have never been made. Actually, if you thought
about it– if that was true– we should have actually seen
an expansion of home ownership, within the US, over
that pre-crisis period, but what this graph shows
you is that this is exactly not the case. In fact, it’s quite stunning. So what we do here is we
plot for you changes in home ownership rates against– again, by these income bins
that I showed you before. And blue here is the
lowest income bin, red is the second lowest, and
then so on, and so on. And what you see,
we normalize in 2000 by, basically, the rate of home
ownership for each bin in 2000. Now what do you see? What you see actually is that
the fraction of home ownership in the lowest two income bins
started going down already in 2002, 2003, and then
really accelerated, of course, after 2007, 2008, right. So this is to show you that,
actually, it looks much more like many low-income households
started being priced out of the housing market
way before 2007, so it was not the case that the
run-up to the financial crisis, or the boom period of
the 2000s, afforded low-income and subprime
people to enter home ownership at an unprecedented rate. If anything, these
numbers show you that, yes, for
richer people, there was an increase in
home ownership, right. These are the top three in the
graph, but for poorer people, actually, it started already
declining way before 2007, and then, you know, as
obviously this graph shows, very much accelerated post 2008. Great. So, you know, I mentioned this. I want to skip
over this quickly. What we also saw– and there is
a large literature that is now kind of coming back to
this– is that the flipping– or the speed at which
houses were bought and sold, leading up to 2008,
increased dramatically. And so, what this is showing you
is kind of data from Zillow– the fraction of homes
that were bought and sold within 12 months of a given
year, and what you see is kind of leading up from 1998. You see this big spike
in the speed of turning– and then of course post
2008, the big drop off. But this again,
right, is suggesting that households, across
the income distribution, were participating in the
housing market and, with it, in the mortgage market
at a very high rate. All right, excellent. So this is kind of the run up. And, in a way, I wanted
to show you these results, because they suggest that there
was a widespread participation across income groups,
across FICO scores, subprime and prime borrowers–
and that, in particular, it looks like even at the
extensive margin in the run up to the crisis, we didn’t see
an increase in participation, but we saw– of, say, marginal home owners. We saw a doubling up, or kind of
a more intensive participation, of the middle class and
the upper middle class. Now a final piece,
of information I want to show you
from the aftermath of the financial crisis,
is who defaulted? So Chris already showed you,
in a really nice presentation, the type of tradeoffs that
people make when they consider life events, and whether
they are under water or not, in the default decisions. I want to just show you one
very simple thing that actually [INAUDIBLE] have also
shown in their work, but didn’t maybe come out as
much kind of in what he showed you– which is that house price
movements mattered a lot in who defaulted. So the first thing
I want to show you is just, you know,
cross-sectionally, who starts defaulting post 2008? The first bar I’m
showing you here is very similar to what I
showed you about, kind of, origination dollars. We take every cohort– say, the most
leftward here is 2003. So mortgages originated in
2003 sorted by whether they are subprime– meaning below
660, between 660, 720, and above 720– and then we follow
them for three years and see what fraction defaults
within the first three years. And we take three years, because
very often the more the time period by which people who
will default start defaulting is often three years. That’s why we
follow three years, but we can do the same
for five years, six years. It would look exactly
the same thing, right? But what the first
bar here shows you is what you would
normally have expected, which is that most
of the mortgage dollars in default
after three years come from the subprime group– come from the white
part of this bar. 70% of dollars in default
are people with a FICO score below 660, and
only 9% are people with a FICO score
above 720, which is exactly what should be the
case when we have a functioning mortgage market. The people that are
good credit score default at a much lower
rate than the people that kind of are
a low FICO score. But now fast forward three
years, and what do you see? So this is mortgages
originated in 2006, which I now follow to 2009. And what you see is that
a big shift is happening. The mortgage dollars in
default increased the most in the medium and
the prime segment, meaning in the group that is
660 to 720 and above 720, right. And you see that the white
bars, so the below 660, are going down significantly
over that time period, so what’s going on there, right? What’s going on is
actually that the default rates of the middle class and
the upper middle class went up, and they went to
unprecedented rates in the US. But the rate itself– the increase was significant,
but the rate, of course, is not as high as subprime. Let me say this in
a different way. So if you look at 2003,
you would have had, even in normal times
in the housing market, the subprime group might
default at a rate of 6% or 7% even when there was
no housing crisis. That’s why they have
low FICO scores. But high FICO borrowers,
in normal times, default basically at
a zero likelihood, but what happened post 2008 is
that the high FICO borrowers are the ones that start
defaulting after 2008. And as Chris was saying,
it’s a combination of being underwater and
potentially having life events, but the important
thing to note is that the high income
and high FICO borrowers are the ones that have
the big mortgages. So when they went from
0% default to 4% or 5%, that’s when the market really
got into trouble, right, and that’s when we see these
big increases of dollars in default of the high prime
and medium prime customers. And that’s the big shift
that we hadn’t seen before in that housing market. And then as the last thing,
what we can do is cut, or kind of divide,
the sample by people who live in neighborhoods that
saw a rapid increase in house prices and then a crash, versus
those where there was not much of a difference. So on the left are the people
that live in neighborhoods, or zip codes, where there was
not as much as a price run-up, and on the right
part of the US where there was a big increase
in prices and then a drop. And look. What you see is that the big
increase in prime borrowers defaulting is
exactly in the areas where prices went up
and then crashed, right? Again, seeing that house price
movements not only played a huge role in who took
on mortgage leverage and participation in the
mortgage market, but also then who defaulted
afterwards, right. So all this is to show you that
house prices and house price expectations seem to have
played a very important role in the run up to the crisis– and then explaining,
also, default behavior of mortgage
borrowers in the crisis. And why this is so important
because is that this actually suggests that the way we want
to think about regulation in mortgage markets is very
much from a macroprudential perspective, right. This type of expectations
shift, it’s not something we can regulate at a
bank-by-bank basis or, you know, at a
microprudential level, but we really have
to think about this from a systemic perspective, as
you know many of the speakers just independent before have
also basically alluded to. Thank you very much.

Leave a Reply

Your email address will not be published. Required fields are marked *