Washington, DC – ARC Meeting on 05/02/2018


Good afternoon, everyone and welcome to the
2018 meeting of the Academic Research Council here at the bureau. I’m Tom Pahl. I’m the Policy Associate Director for the
Research, Markets and Regulations Division here at the bureau. I started at the bureau just eight days ago
after being the acting director of the Bureau for Consumer Protection at the FTC. I’m not a lawyer. I am a lawyer. Can’t disclaim that law degree. I am a lawyer, not an economist. However, having worked in consumer protection
for the last 30 years I’ve really gained a keen appreciation of the value of economists
and their work in a lot of consumer protection areas, especially when it comes to evaluating
the costs and benefits of various policy options. I’m pleased to have a chance to be joining
you today for a discussion about the bureau’s recent and ongoing economic research. As some of you may know the bureau created
the Academic Research Council in 2012 to provide the Office of Research with technical advice
and feedback on research methodologies, data collection strategies and methods of analysis. The council also provides feedback on the
Office of Research’s work and input into its strategic planning process and research agenda. Finally, the council performs critical peer
review of the Office of Research’s work. So we are very indebted to the council for
everything they do to help us with our work here at the bureau. The council meets on an annual basis and today
marks the seventh such annual gathering. Joining us today are Melvin Stephens who is
a professor of economics at the University of Michigan. He’s also a professor of public policy at
the University of Michigan’s Ford School of Public Policy, a research affiliate at the
Population Study Center and a faculty associate at the Survey Research Center. Also with us today is Brigitte Madrian who
is the Aetna Professor of Public Policy and Corporate Management at the Kennedy School
at Harvard University as well as co-chair of the NBER Working Group on Household Finance. We also have John Lynch who is the director
of the Center for Research on Consumer Financial Decision-making and senior associate dean
for faculty and research at the Leeds School of Business at the University of Colorado
Boulder. Joining us this year is Karen Dynan who is
professor of the practice in the economics department at Harvard University. She has served as assistant secretary for
economic policy and chief economist at the U.S. Department of Treasury, vice president
and co-director of the economic study program at the Brookings Institute. Dr. Dynan is the newest member of the council. The last member of the council is Ian Ayres,
the William K. Townsend Professor at Yale Law School and a professor at Yale’s School
of Management. Unfortunately Ian is unable to join us today. The council members have extensive experience
and expertise on economic issues relating to commercial financial goods and services. I look forward to hearing from them on the
many critical economic research topics which are on our program today. Before we jump into the program I wanted to
extend my heartfelt thanks to Emily Turner and Tangee Harrison from the bureau whose
hard work in handling coordination and logistics have made this event possible today. Let me now hand the program over to Ron Borzekowski
to kick off the meeting. Tom, thank you very, very much for the introduction. For those that don’t know me I’m Ron Borzekowski. I’m the research director here at the bureau. I just have a couple of remarks just to lay
out what’s going to happen for the next couple of hours. There are two sessions to our meeting today
and each one will be broken into sort of two subparts, some presentations and then a discussion,
some presentations and then another discussion. I will turn it over to the chairs of each
of those sessions in just a moment. But hopefully over the next couple of hours
we can have a lively and spirited discussion and exchange about some of the work the bureau
has done in the past year since the ARC last met and also some work that we are currently
doing. So with no further ado let me turn it over
to the chair of the first session, Ken Brevoort, one of the section chiefs in the Office of
Research. Good afternoon, everybody. At the bureau we continue to improve our data
resources relative to the functioning of markets for consumer financial products and services. And particularly within the Office of Research
we remain committed to using a variety of methodologies for analyzing and better understanding
these markets including the use of administrative data, field or lab trials and surveys. And what I’d like to do in today’s session
is give you some updates on exactly what we’ve been doing and some of the new products that
we’ve been issuing. And I will start things off myself by talking
about some of our new administrative data products. One of the data resources that we’ve developed
at the bureau is something we call our consumer credit panel. It’s a large panel of deidentified credit
records of about 5 million American consumers that we receive updated periodically. It contains nothing that directly identifies
these consumers, there’s no names, no Social Security numbers, no dates of birth, nothing
like that. But what it does give us is a comprehensive
view of sort of how the financial situation of these consumers is evolving. It’s very similar to data sources that are
used at other government regulators and in private industry. It’s something we’ve purchased commercially. But it’s something where I think we’ve been
doing some things with that have been really novel and interesting. We started using it for monitoring consumer
credit trends and developments. And a few years ago, or maybe a year and a
half ago or so we started producing what we call the Consumer Credit Trends which is a
website, a dashboard of developments in consumer financial product markets. And what I want to talk today is about a new
series of reports that we’ve started recently that’s designed to do a deeper dive when we
find topics that we’re interested in that are going on in the market that we want to
tell the public a little bit more about. And this is something we call the Quarterly
Consumer Credit Trends Report. We started it in November 2017. The next report was issued in February 2018. And we’re hoping to release roughly one of
these every quarter. So to give you an idea of what these reports
are about I’ll talk a little bit about the two and why we’ve done and what we think we’ve
learned from them. So the first report that we came out with
was devoted to looking at the growth of longer term auto loans. This is a trend that’s happened in the industry
that I think has probably not gotten as much attention as it probably should. And the focus is sort of highlighted by what’s
labeled on this graph as figure 1. This graph shows from 2009 to 2017 the distribution
of loan terms by origination year. And what you can see is there’s been a fairly
substantial change in the market in particular, that there’s been a movement from the use
of five-year auto loan terms to six-year auto loan terms. Between 2009 and 2017 six-year loans or longer
term went from about 26 percent of the market up to 42 percent of the market. So there’s been this real fundamental change. The driving cause of this we believe is an
attempt by consumers to deal with rising average loan balances. So if you look at the graph at the right this
shows what the average loan amount was from 2009 to 2017. To make things a little bit more comparable
we’ve indexed it to 100. So 100 is equal to the average loan amount
in 2019 which was about $18,000. And what you can see is that average loan
amounts over this eight-year period increased by about 16 percent. And because there’s been this large growth
in the use of longer term financing instruments the monthly payment required for these loans
has actually only gone up by about 7 percent. So this has been an effort by consumers we
think to help deal with a rising amount that they’re trying to finance while keeping the
minimum payments or the payments that are required at a more sustainable level. And the reason we’re a little bit concerned
about the growth of longer term auto loans is that if you compare the default risk to
the default experiences on these loans what you see is longer term auto financing loans
tend to have much higher default risks. So the graph that we’re showing right here,
the left graph is five-year loans, the right graph is six-year loans. And this shows the percentage of loans that
were 90 or more days past due as of a certain date. And so what you can see here is that six-year
loans default at a much higher rate pretty much consistently throughout time than do
five-year loans. Now, I want to be clear this does not mean
that it is the six-year loans that are causing the higher default rates. There is going to be some selection bias here. We haven’t attempted to control for this or
really to ascertain is it the longer term loans that lead to an increase in default
risk itself. But this does sort of highlight this possibility
that by seeing this movement to longer term loans we’re putting consumers or consumers
are putting themselves in a riskier position. They’re dealing with loans that are going
to be more expensive to finance even though the monthly payment is going to be lower. The interest that they’re going to wind up
paying over the life of the loan will be higher. And if they choose to try to trade their car
in to buy a new one after a few years their outstanding balance is going to be higher
as well which could cause them problems down the road. The second report that we issued in February
2018 had the rather uninformative title of Public Records. And what we were trying to look at here was
a change that had occurred in the industry, something called the National Consumer Assistance
Plan. This is a change that all three of the nationwide
credit reporting companies engaged in after they reached a settlement with 30 states attorneys
general. And it was a change in the composition of
credit records that occurred July 1, 2017 and it involved public records. Now public records basically come in three
varieties in terms of what you see in credit records. They can be bankruptcy filings, civil judgments
or tax liens. And this graph that we’re showing now shows
how many credit records of each of these different types we observed in our data in May, June
and July. And what you can see is that bankruptcies
over these three months were basically the same thing. They were relatively constant. Civil judgments which were the most common
type of public records totally disappeared in July. So it was the largest in May, the largest
in June, and July there’s just no more. And tax liens also, between May and June there
wasn’t much change. In July almost exactly half of them disappeared. So what you see is that because of this NCAP
program there was a dramatic change in the information contained in consumer’s credit
records. And this was causing some concern, particularly
in public press there were a lot of articles written about this. The fact that you had a lot of consumers who
had these public records and if you look at the credit scores of consumers with public
records which is what we’re showing in this graph here on the bottom the distribution
of consumers who have public records on their credit file tended to be relatively low. About 71 percent of consumers had a subprime
credit score who had these types of public records. And there was concern in these public articles
that now if you were taking this information away and just removing it off the credit records
altogether that was potentially going to increase the credit records of these consumers and
perhaps allow them to qualify for loans that they otherwise would not have been able to
do so. So we wanted to understand better what the
effect of removing this information was going to be for consumers. And so this graph shows our attempt to better
understand what the actual effect on scores was. Now, since we don’t have the underlying credit
scoring models that would go into the calculation of these scores we have to rely on more inferential
methods to figure out what’s going on. And so what we did here is we looked at the
change from one quarter to the next of consumers who had civil judgments or tax liens at the
earlier of the two quarters that we’re looking at the changeover. So the top three, look at the three quarters
immediately preceding the NCAP. And what you can see is that the score change
followed distributions that looked relatively constant over time. Then when you have the fourth graph at the
bottom, this is the change that occurred after the NCAP was instituted you can see that there’s
a dramatic change in the distribution of credit scores. And in particular there’s a dramatic increase
in the number of consumers whose credit scores went up by about 15 points. So it seems that the removal of this information
for most consumers to the extent it had an effect tended to increase consumer’s credit
scores by about 15 points. Now, when you think about the body of people
who have this type of information as I said most of them have subprime credit scores. Most of them generally — flip back to that
slide — most of them also were relatively low subprime credit scores. So they were around the 580 level or lower. And what we find is that when you increase
these scores by about 15 points it’s not generally enough to move somebody’s credit score from
one level to a level that would allow them to be really viewed by credit underwriters
in a dramatically different way. It wouldn’t necessarily improve their credit
standing all that much. So what we find is that as a result of this
change about 17 percent of people over this quarter moved to a higher score band. They moved from being subprime to prime or
perhaps from prime to super prime. But if you compare that rate of movement to
what we observed in the previous three quarters there was only a marginal improvement of about
4 percentage points. So in terms of the market as a whole it seems
that the NCAP had a relatively minor effect on consumer’s credit scores. And this is what we’ve concluded as a result
of the report. And this is the type of analysis that we’re
trying to get done in these reports, sort of very short deep dives on a topic that we
think people will be interested in. Hopefully the next report will be out within
the next few days or the coming weeks. And so with that I will take a short smirk
of victory over people who did not believe I could cover this in eight minutes. And I will turn it over to Cheryl Cooper who
will tell you about some of the improvements that we’ve been making. Thanks so much, Ken. So as we’ve spoken about at the Academic Research
Council meeting last year and in past years one of the major goals of the dynamics of
household balance sheet research agenda is capacity building, to improve the quality
of our data internally and to answer important — and allow us to answer basic research questions
to inform bureau policy in the future. As Ken just mentioned the Consumer Credit
Panel or CCP is an important resource we use to understand many of the markets that the
bureau oversees. But there’s important information for policy
research that’s not collected by the credit bureaus. We wanted to share some of the data improvements
that we’ve made this year to this resource to improve our ability to develop policy relevant
research in the future. Information we’ve added includes military
status as well as higher quality income information. Note that all of this information is constructed
to minimize privacy risks. So first I wanted to talk about our new SCRA
data match. At the bureau Dodd-Frank statutorily mandates
us to have an Office of Servicemember Affairs. This office is to support military families
in consumer financial markets. And these families have unique challenges
that they sometimes face. The Servicemembers Civil Relief Act or SCRA
is a program that provides certain protections in lending for servicemembers who are called
to active duty. The Department of Defense maintains a database
online for lenders to comply with this act. This year we directed our CCP contractor to
download active duty status of consumers from the SCRA website for December 2008 to 2017. Using this new information we are able to
better understand the credit experiences of military families. For example, we can compare servicemember
and veteran’s credit use to the civilian population. Some topics we are interested in exploring
in the future with this data includes military duty station moves, active duty mobilization,
and separation from active duty, and how all of these things relate to families and their
financial lives. We’re also interested in the effects of particular
servicemember programs like VA mortgages and home loan behavior. As you can see on the right side of this graph
the graph shows credit score by age broken down by SCRA status. This graph shows that active duty military
tend to have higher credit scores than their civilian peers, but for formerly active duty
military consumers tend to have lower scores. And to our knowledge this is the first time
that anybody has been able to produce this graph which we’re really excited about. So we’re excited in the Office of Research
to be able to use this new information to understand this important population better
and to allow us to be able to do policy relevant research on this population. The second improvement that we’ve made to
the CCP this year is the acquisition of improved income information. Income data is sparse but critical to understanding
different populations’ credit experiences. Powerlytics is a market intelligence business
platform that’s providing data derived from a variety of different sources. Our CCP contractor matched geographically
aggregated income data to the CCP this year. So it’s not individual level, it’s aggregated
information, but it’s a more accurate measure than what we have currently in house. And we’ve acquired this data historically
as well to be able to pursue longitudinal studies in the future. As the graph on the right shows Powerlytics
income data has a larger distribution than ACS census income estimates which is suggesting
a more accurate measure of income. So the benefit of having this improved income
estimate is that it allows the Office of Research to be able to improve its measures of for
example debt to income ratios or to evaluate more accurately things like income-driven
repayment plans in the student lending market. This information also better allows us to
measure retired households and households with children that are in the area. And this allows us to explore questions in
the future like how do these households’ differences vary over region and how do these differences
relate to financial well-being and credit use. So this type of basic research can inform
policy-making and other initiatives at the bureau. Some of the limitations of this information
is that this data is not at the individual level which minimizes privacy risks but means
we cannot measure household shocks which may be important to understand financial product
use. This data is also at the annual level so we’re
not seeing dynamic information, things like income volatility that may be related to financial
well-being and other credit outcomes. So to conclude we’re looking forward to ideas
from the Academic Research Council on how we can improve — how we can use these improvements
and others to support policy relevant basic research at the bureau. Thanks. Thank you, Cheryl. Now I will turn it over to Dustin Beckett
who’s going to talk a little bit about the improvements in products we’ve been developing
for field and lab trials. Thank you, Ken. Good afternoon, I am Dustin Beckett, an economist
here at the bureau. As Ken said today I will update you on the
Office of Research’s capacity to conduct laboratory studies. About five years ago the bureau began building
capacity for running laboratory studies. For the uninitiated at the bureau a laboratory
study or lab trial is a research methodology in which volunteers make choices in a controlled
environment. Like other empirical methodologies lab trials
are part of a virtuous cycle which inform and are informed by our models of the economy
and consumer decision-making. Lab studies in particular are prized for their
ability to address precise research questions that would otherwise be difficult to address
in the field or with observational data. The data collected in laboratory trials at
the bureau are all anonymized and stored without any direct identifying personal information. In the first phase of capacity building one
of our goals was to demonstrate that lab experimentation could be done well and cost efficiently at
the bureau. This guided our capacity building towards
replicating as much as possible laboratory capacity as it is traditionally implemented
in academic institutions. To this end the bureau contracted with a handful
of local universities to allow bureau staff to utilize their laboratories. The image you see is one of these laboratories. The key features of this laboratory are networked
computers, privacy screens which are down in the image but can be raised to provide
privacy between computers, presentation capabilities such as televisions and whiteboards, and of
course access to participants. Over the last couple of years the Office of
Research has had great success in both running studies on our own and in collaboration with
our university contractors. In one study ran with Middlebury College we
examined the effect of product complexity on participant’s ability to make real stakes
financial decisions and found strong evidence that greater product complexity leads participants
to make worse decisions on average. Like this work with Middlebury College at
the bureau lab studies have thus far been used to inform our disclosure research agenda
which you will hear about later today. This week the second stage in the Office of
Research is laboratory capacity building is being realized. The Office of Research is creating what we
are calling a mobile lab. The mobile lab is a collection of hardware
and software that allows researchers to be mobile, untethered from the traditional constraints
of university campuses. On the hardware side the mobile lab consists
of 24 iPad tablet computers and 2 very cool heavy duty carrying cases one of which you
— you can see an example of which on the slide. The tablets are connected to the bureau network
and to each other via cellular connections meaning whatever data is collected in the
field is stored securely at the bureau. Because the tablets are connected this also
means that the mobile lab is capable of handling almost anything that could be accomplished
in a traditional lab. On the software side we are using a platform
that utilizes a standard web browser for participant interactions. This means that any software programmed for
the mobile lab can also be implemented on PCs or in a traditional lab. The first benefit of a mobile lab is naturally
mobility. With the mobile lab the bureau now has the
ability to interact with and learn about consumers where it is most convenient for them. In particular we can now more easily study
diverse and varied populations. For example, we can seek out older Americans,
servicemembers, or economically vulnerable populations, all populations the bureau is
statutorily mandated to serve. In addition to the extent the mobile lab increases
the bureau’s ability to conduct research in house the mobile lab has the potential to
significantly reduce the cost of conducting lab studies at the bureau. I hope you agree these are exciting developments
for the Office of Research and I look forward to your comments and questions. Thank you, Dustin. With that as a reward for remaining quiet
longer than David thought you could we will throw it open to discussion. So I think it’s great to see all of the investments
that you’re making in data resources and research resources like the lab to try and better understand
the financial challenges that consumers are facing and enable better informed regulation
in those markets. I guess I have one question for Dustin. It seems like some of the experiments that
you might be running to field test disclosure policy or consumer decision-making either
in a standing lab or a mobile lab that not only can you write reports that people can
learn from but the methodology and the computer technology like the code is something that
maybe you could have be a public good and researchers at other institutions, say schools
and universities could use it in the classroom. Is that something that is possible and have
you thought about and if not could you do anything on that front? Leverage the investments. Thanks for the question. I love it. I think my personal vision has always included
things like this. My personal vision is not always the same
as the office’s vision. I think one thing that’s been holding us back
is we just haven’t had a lot of experiments that we’ve run and I think if we had a website
with just one link to one paltry experiment it might look a little bit sad. But as we run more experiments I think certainly
it’s in the realm of possibility. If there are people like you championing it
I think it makes it more possible. Perhaps Ron wants to weigh in on the feasibility
of this. But I personally love it. Yes, I think there are two parts to that. Thanks for the question. One is could we put the code base out there
and I think the answer is probably yes. We have the technology here for things like
a public gate hub site where we could possibly put the code that would run the experiment
up after they’re done. Which is also becoming state of the art and
the academic literature for replicability purposes. The other question is whether we could actually
put the raw underlying data up there, for example for others to look at and to ask different
questions of the same data that we hadn’t thought of. And I think there are possibilities there
as well. The bureau has actually recently established
something called the Disclosure Review Group so that were we to be able to make data public
we would do it in an anonymized privacy protected way. And once we overcome those procedures we’ll
be able to do it. We’re even willing to rent out one of Dustin’s
cool briefcases for the right price. I was going to say I think this is a really
exciting development and a terrific capability because if you’re using traditional economic
tool you can see whether something has an impact. But to understand where it will have an impact
you have to understand why it has an impact. And so one of the big advantages of what you’re
trying to do here is the ability to understand the process by which some intervention like
a disclosure would actually change consumer’s behavior. So I think it’s really, really good for that
purpose and will help you understand conditions where a given intervention will or will not
work. The other thing I think is really nice is
this idea of going to the mobile lab. I think that’s really terrific. This idea of understanding the differential
response of heterogenous populations. I know that there are some universities that
have a tool that’s sort of like this that I envy greatly like Carnegie Mellon has a
data truck to go out to speak with more vulnerable populations. And given the special focus on vulnerable
groups I think it’s a really terrific capability. I would just like to add one more thing though. You should add I-tracking to your lab because
of the fact that so much of what you’re doing has to do with whether people engage with
for example disclosure. Thank you. Thank you for these terrific read-outs. It’s a pleasure to be here and just to hear
about the truly interesting and important work that the CFPB is doing. A comment for Cheryl. I thought your presentation was great. I was excited about both of the things you
talked about. I thought the Powerlytics income data sounds
like a tremendous resource in terms of being able to do better research given the credit
data that you have. You did mention that not having income at
the individual level was a limitation. I’m sure you realize this but I wanted to
just note for everyone that there have been some just terrific studies done using aggregate
set different levels of geography. Particularly if you have zip plus four I think
you’re going to have a lot of interesting variation. And many times these studies really have to
do geographic aggregates because the covariates they want to merge in are only available at
that level. So I wanted to mention that comment and also
just ask a question which is do you think you’ll be able to put some of the geographic
aggregates out — or more of them out there publicly some day down the road because I
can imagine that there are a lot of researchers who would be interested in these data and
could help the bureau with its mission if they had access to this sort of data. Thanks so much for your comment. So I guess the data by itself is obviously
proprietary but I think in general we totally agree with you and this is something sort
of what Ron was just saying before where we want to ideally be trying to make a lot of
these things public goods so other researchers can use them. Within the agency we’ve been working on privacy
protections and other concerns in terms of being able to put out data publicly. But I think those are things we’re starting
to work through. So that’s definitely something that we’re
really interested in exploring in the future. To tack on a little bit, it’s also something
we’ve tried to do whenever we’ve had the ability to do so. So with a lot of these administrative data
sets it’s difficult because we’re buying them from somebody and so it’s proprietary. But for example we did an analysis using some
of our proprietary data about the number of people who didn’t have credit records, the
so-called credit invisibles and what the distribution of that is across the country. And we’ve been able to make maps available
to local governments to show them this is what your community looks like. And we’re trying to expand those resources
on an ongoing basis to make it public so researchers can have access to where are these pockets
of credit invisibility are and hopefully that will better inform debate. Thanks again for having us here. Another year. It’s been great to learn everything, see the
progress from last year as well as learn about things that are new. I actually had a similar comment to Karen
as we were talking and thinking about on the Census side there was the collection of unemployment
insurance information to understand what’s happening at the individual level and the
firm level. And then they have been producing these area
aggregates that have gained much more insight into the dynamics of firms and firm turnover
that we can learn a lot about. And I think being able to match that to what’s
happening with credit at a local level would be something that would be really helpful
for people to know. So I was seconding. I know that there’s a difference because there
they have the universe and here you have a sample so to be able to get the detail at
a very small level would be challenging. But thinking about the trade-off between what
is proprietary and what you’re able to make public and you’re obviously publishing a lot
of trends. It would be great to keep pushing on that
side of things. Have they made — has Census I guess or the
UI records made county level aggregates available or things like that? Because that’s something we haven’t to be
honest brought in house and tied to our records. If it’s the county level it’s something we
could easily do. I’m fairly certain they have been doing this
for a little while. I forget the name of the exact series. Something about local indicator. I’m forgetting the exact title of that series. Yes, if you aggregate it up to the number
of counties and the number of people you have I’m sure you’d be able to get pretty good
estimates at that level. Could I ask a question of Cheryl? So what’s your explanation of this result
that you find that servicemembers have better credit scores than their matched counterparts
during — while they’re in service and then worse afterwards? Yes, it’s interesting. So to be honest we just got this data in house
less than three weeks ago so that’s one of the first graphs we’ve been able to produce
which we’re really excited about. I think we’re going to have to spend some
resources to dive deeper into that question. I know staff are really excited to work with
the data. So yes, more to come. Can I follow up on that too? I thought that graph was just super interesting. I’m sure as you look at the different lines
you’re worried about the usual selection issues. And I was just wondering will you be able
to at some point look at changes for people as they leave active duty? Yes, definitely. I know that’s something we’re really interested
in. So we have 10 years of historical data. So we’ll be able to see both people entering
and exiting military active duty. So yes I think both of those transitions are
really interesting for us to explore. And one of the nice enhancements we have in
our consumer credit panel is that not only do we have the deidentified credit records
of the people who are in our sample, but we have deidentified records of everybody with
whom they share an account. We’ve also got the military records of that
population as well. So we’ll not only be able to see the effects
of active duty on the military person themselves, but potentially on the rest of the household. This isn’t something that you talked about
but given how big those differences are, 100 point difference in the credit score for being
in the military versus not it makes me wonder how many other things are out there in the
world that are having similarly large effects. And you’re able to merge the CCP data with
this data on who’s in the military. I’m wondering what other potential data sources
you could leverage in a similar way. So for example, could you get from FEMA everyone
who was impacted by the hurricanes in Houston last year and figure out which ones had insurance
and which ones didn’t and see what happens with their — what’s the impact of a natural
disaster on credit scores and how does that play out over time. I can see lots of potential here and I’m assuming
you’re just scratching the surface and next year we’ll hear more new and exciting things. You have all the ingredients to help us really
understand what are the circumstances that lead to household distress and how does household
credit respond to difficult situations. So we agree. And you share our instinct which is more data,
more questions, more interesting results. These I think I just want to be clear affect
largely — these are most likely selection effects of all kinds of sorts. And I should at least note because the data
are so new we reserve the right to revise and extend and shrink that a lot if we realize
we tabulated incorrectly. But we thought it was worth — we were quite
excited about this data largely because of the statutory mission of the bureau and a
lot of the consumer financial issues about military personnel both on the basic research
side but on the policy side, the cost and benefits of particular interventions, this
was a very natural place to start and I think one place we’ll mine very deeply. We have started looking at least at the geographic
level at some of the issues from natural disasters but not at the individual level that you had
described. But I think in general in terms of sort of
what you were encouraging us to do I think within the dynamics of household balance sheet
research agenda one of the things that the office is really committed to is making sure
we’re dedicating time and staff energy to be able to explore these types of matches
or other types of data improvements that we can make to be able to study these types of
questions better so that we can understand the markets that we regulate better. General comment for Ken. These quarterly consumer credit trends documents
are great. I read them both on the plane coming down
yesterday and I made a note to myself that I had to go back and find the earlier ones
you’ve done because these are both super interesting. So I’m a little disappointed to hear that
this is a new series but that also makes me excited about what’s to come. But I was just wondering if you could talk
a little bit about the process by which you’re kind of figuring out what topics to pursue. So that’s sort of still in development in
terms of who’s going to be identified to write these and what they want to write about. So by and large I think what we’re trying
to find are topics that are in the news or everyone is thinking about or you start to
hear rumors about. So the article that we wrote about the growth
in longer term financing loans I think really came out of a lot of the chatter that we were
hearing in industry that this was going on, that you had consumers that were taking out
these longer term auto loans that perhaps had larger balances when they tried to refinance
the car. They were rolling over some of these balances
from one loan to another. And in an environment where you’ve got interest
rates that are going down that’s sort of — you can do that. You can roll over balances from one loan to
another without increasing your monthly payment that much. But if you start to move into an environment
where things start to go up a little bit and interest rates are higher the music sort of
stops and everyone needs to grab a chair and that’s when things can really go wrong. So I think that one we started writing largely
because we were concerned that this was happening and we wanted to alert the world. Because I didn’t think there were as many
people in the policy community space thinking about what the implications of this longer
term financing instrument was. The NCAP program was a little bit easier. There were lots of newspaper articles written
about this saying that it was going to have all sorts of bad effects. There was one op-ed written by a researcher
who said — who likened this to the use of no income loans during the housing crisis,
that this was fundamentally going to lead to the second financial crisis that we were
seeing because it was going to allow a lot of people who really shouldn’t have qualified
for loans to qualify because their credit scores had been so sufficiently improved that
it was going to cause this. And so even though we’d heard from industry
that they didn’t think this was going to be a very big event we wanted to take our own
independent look and make sure we understood what the effect of this was going to be on
the market. And in this case it really seems like it’s
going to be a minor sort of event. So a lot of it is driven by what we’re thinking
about at the moment and what we’re hearing. Any suggestions, Karen as to things we should
be thinking about? I’d have to think more and get back to you
on that. I know you’re already doing a ton of work
around student loans but I think that’s a pressing issue and I know you all have resources
that other researchers don’t have. But I’ll let you know. All right, so with that we’ll resume the presentations. I will now turn things over to Sergey Kulaev
who’s going to talk about the national survey of mortgage originations. Thank you, Ken. I guess national survey of mortgage originations
doesn’t need that much of an introduction in this room but I’ll say a couple of words
for completeness’ sake. Every quarter about 6,000 of consumers who
recently got a mortgage get in the mail which is a survey of mortgage originations. That is about 100 questions asking them all
sorts of things about their experience with origination process, about their expectations,
about their just life circumstances overall. This survey offers a pretty rich view into
that moment in life of these consumers. We have response rate of about 30 percent
so for one year you get about four or five thousand, or five, six thousand responses. There are two primary advantages of this survey. One is that it is representative of recent
mortgagees and overall therefore of people who have a mortgage that they took not too
long ago. So therefore we can simply present some statistics
that would be applicable across the market. And I’ll speak about a couple of them in a
minute. And the second advantage as I mentioned the
breadth of questions that they ask consumers about. For example there are also questions about
recent life events that we haven’t explored enough yet. So change in family composition, moves and
stuff like that. So there are definitely lots of areas to explore. So we start scratching the surface by looking
at responses from 2013 and 2014 originations. And we produced a couple of papers from these
originations. The first paper is looking at mortgage experience
of borrowers who live in areas that are completely rural. It’s a designation by USDA. A completely rural county is a county where
it has fewer than 2,500 urban population. So there isn’t even a town that has at least
2,500 people. So these are very rural counties. They are mostly located in midwest. So far there has been really not enough evidence
on experience of rural borrowers simply because there’s just no data about that part of the
market. For example, best available data set on mortgages
that we have is HMDA and HMDA has serious issues with representing that market because
it excludes small lenders and also lenders who only operate in rural areas, lenders who
don’t have branches in MSAs they don’t have to report to HMDA. And so we’re missing a whole bunch of loans
by rural borrowers in HMDA. And that’s where this survey comes in because
it’s something based a set of criteria rather than HMDA. And so it’s able to reach out in pockets that
are underrepresented in HMDA. So this particular study looked at 2014 originations. We just did a bunch of comparisons around
satisfaction and knowledge and interest rates between rural consumers and non-rural consumers. And of course the general topics around originations
in rural areas is about whether there’s enough credit supply, whether these consumers are
being treated fairly, whether they are being charged fair interest rates and so on and
so forth. That’s why it was really interesting to look
at this issue. And what we found is actually fairly optimistic
maybe in some sense. For example one of the findings that are not
listed in the bullet points that I won’t read out is that when we looked at percentage of
consumers who were concerned about qualifying for a mortgage that doesn’t seem to vary a
lot with the metro status. However, one difference that we found is really
regarding the reliance on relationship banking. So there’s a question on the survey that asks
recent mortgagees about what lender characteristics were important for you when you took out the
loan. And consumers in rural areas were much more
likely to say that existing banking relationship was important. And on the other hand there is much less intermediation. They were also much less likely to say that
this lender was recommended by someone for example. They were also much more likely to directly
apply to a lender as opposed to through a broker. And so this kind of conforms to our expectations
a little bit. We looked at the effects of differences in
interest rates that consumers paid. We found a small but positive difference in
interest rate. So a rural consumer seemed to be paying a
little bit higher price controlling for some characteristics, for a fairly comprehensive
set of characteristics actually. So that’s definitely of interest although
the difference is not that large. I think it was between six and nine basis
points. It’s unlikely to make a big difference. There’s a battery of questions on the survey
that talk about confidence and knowledge and so on and so forth. And rural consumers seem to be less knowledgeable
about mortgages or at least less confident about their knowledge of mortgages. That actually brings me — so I will cover
the next paper right away and then we can talk about it. So that brings me to the next paper which
focuses on first-time homebuyers who receive pre-purchase counseling and comparing outcomes
of these homebuyers and basically their reports on the survey to similar first-time homebuyers
who did not receive counseling. You can’t see it from here but this particular
paper involves — the quoters on this paper involve three government agencies and two
government sponsored enterprises. The fact that the intersection of these five
entities is not empty is by itself a pretty amazing fact. These two papers are basically the first research
output out of the NMDB project which is another reason we’re excited about it. So let me give a little dive in to the homebuyer
counseling study. So we looked at first-time homebuyers with
credit records. You can identify those pretty reliably such
as these are people who actually never had a mortgage on their record as opposed to people
who say they are first-time homebuyers. There is a bit of a difference sometimes. And for the first time we can testify that
14 percent of first-time homebuyers received some sort of pre-purchase counseling. It could be different, there are very different
types of counseling available. There’s classroom education, there’s online,
there’s one to one counseling. So we don’t specify here but to me that number
seems surprisingly high. And it’s the first time we actually know how
much it is because all previous literature on the effects of counseling, and that’s where
this paper comes in, has focused on outcomes of recipients of a particular program. For example, Freddie Mac Affordable Home Program. There’s a good study that has required counseling. And there was a study look at the effect of
that. So all existing papers look at one program
at a time and only focusing on loan performance as outcomes. We look at first-time homebuyers across all
programs. So these are average results across all forms
of instructions. And we look at things other than loan performance. We look at mortgage knowledge, shopping and
so on. And so some of the highlights of possible
impacts of counseling. We do propensity score matching to control
for other observable characteristics and selection. It seems that folks who have gotten recently
some form of counseling report better mortgage knowledge or at least they are more confident
in their mortgage knowledge. We can’t really separate those two explanations. There is some research showing that sometimes
people are over confident in their knowledge when it comes to financial literacy. So that’s definitely something for future
research. However, one thing that recipients of counseling
do a lot more often is to compare the final cost to GFE which is a previous older version
of loan estimate. And that’s definitely a very useful exercise
that any mortgage borrower should do and maybe that’s what they’re taught during their counseling
sessions. We don’t find any effects on the extent of
shopping. However, folks who took out counseling seem
to put less emphasis on lender characteristics and particularly recommendation. They put less emphasis on recommendation by
real estate agent or friends or brokers when choosing the lender. Whether this is necessarily a good thing or
a bad thing that is an open question of course. But otherwise we also of course looked at
loan performance and unfortunately or maybe fortunately during this time the credit standards
were pretty tight so there were very, very few defaults. We didn’t find any effect on that. With that I will stop and talk about those
two studies. If anyone has any questions about these two
studies. So I have a question about the data. So can you — in the survey I’m assuming you’re
asking people what kind of a mortgage they think they have. Do you know the terms and conditions of their
actual mortgage from the — Yes, yes we do. So you know what their interest rate is, you
know whether it’s a fixed rate mortgage or an adjustable mortgage. You know the term of the mortgage. So you can compare what do consumers think
they have with what they actually got. So we prefer to rely on the administrative
data when it comes to this. Yes. So in the administrative data do you know
who the mortgage comes from? So can you see are there some mortgage lenders
that do a better job at communicating the terms and conditions in a way that consumers
can actually recall that information further down the road? Unfortunately that’s a limitation of NMDB
is it precludes that look. That’s one of the conditions of NMDB that
we can’t look at individual lenders in this analysis. It’s like across the market. It’s one of the conditions under which the
data was provided in the first place. A regulator might find that information potentially
useful. Yes, that’s exactly why. In achieving its mission. So part of the condition of the National Mortgage
Database is that we not use that for things like supervision or enforcement specifically. So we can learn about market trends. We can learn as you pointed out overall how
many people actually do understand or what types of people have very different expectations
about what their mortgage is or isn’t relative to what we see in the administrative records. But using that with any kind of identification
of the lender is something that we’re not supposed to do. One last comment. So Sergey characterized the 14 percent of
buyers receiving some sort of counseling, he thought that number was surprisingly high. I guess it’s probably high relative to what
I might have predicted it would be but it’s also shockingly low. When you think about these are individuals
who have never had a mortgage before and this is a really, the first time you buy a house
that’s a big financial decision. I mean for most households you’re taking on
a dramatic amount of debt relative to your income. It’s over a long time period. It’s a big financial commitment. And to think that only 14 percent of households
are getting any sort of — first-time homebuyers. Fourteen percent of people with no experience
doing this before are not getting some sort of help. Granted they could be getting help from friends
and family. But I don’t know. I would feel better if that number were higher. Well, they are not randomly distributed in
population obviously. The recipients of counseling is very specific
type — is very selected type of mortgage borrowers. So these rates, this is an average across
population. That includes prime borrowers, people who
are not required by any lending program to take counseling. A lot of it is coming from just requirements
coming from the lending program. So if you look at within segments of population
the counseling rate can be as high as 50 percent. So who needs counseling it’s a great question
and do they get it. One of the interesting things here, I think
it’s part of the theme of what we were talking about this morning is that the National Survey
of Mortgage Origination because it’s tied to the administrative record, the authors
are able to control for observable covariates pretty richly. So it’s propensity score matching, of course
enhanced matching, push as far as you can. Of course it’s still not a random selection
into those. But the sample sizes and the data on performance
of course are going to be richer than the RCTs that we’ve seen or people who have tried
to get at it in other ways. So we think this adds to the literature but
it can’t solve all of the selection problem to date. It will be interesting though as time evolves
of course and Sergey pointed out that we see performance on these borrowers for the first
24, 36 months of their life. We can see for example if anything changes
in the future because of the nature of the administrative record too. So I also looked at this on the plane. So the results about better mortgage knowledge,
a lot of it is really better self-assessed knowledge, correct. So it’s not actually showing that they had
objective knowledge, but rather they thought they knew what they were talking about. Yes. I wish we had actual literacy tests on this
survey but that idea was not favored by some. Was not considered to be a good idea on other
reasons. So yes, I will admit on that. However, if anything the selection should
go in the opposite direction because people who select themselves into counseling are
plausibly those who need it more. So the fact that we are getting positive results
actually tells us something I think. Even with all the selection issues that are
involved. There’s another literature that shows it’s
consistent with the direction you’re talking about. Where I was going with this is this self-assessed
knowledge versus objective knowledge. So there’s a terrific book by Sloman and Fernbach
called The Knowledge Illusion and it really has to do with our chronic over confidence
that we know what we’re talking about in whatever domain of life. And one of the main results they get in this
is that people — if people know people who know then they think they know. And so the question is whether these improvements
in subjective knowledge are just okay, now I’ve interacted with somebody who knows what
they’re talking about, therefore I know. Since I know some of you are coming to our
Boulder summer conference there’s a paper on the conference that shows a result very
consistent with what we’ve shown which is the effects of some sort of educational thing
on objective knowledge decline over time but subjective knowledge doesn’t. So you’ve forgotten but you think you still
know. And so I think it’s worthwhile in this stuff
and as you extend it to attend to subjective versus objective knowledge. So that’s actually, that’s very helpful. One thing we should ask and we can do it now
or later from the paper you mentioned that’s going to be at your conference or from other
papers if there are particular scales that you think would be useful for this. They have to be reasonably short because we’re
doing we think very well with the 30-35 percent response rate on these surveys and again we
can control for non-response we think pretty well. But the National Survey of Mortgage Origination
is fielded quarterly. So Sergey discussed this in the first paper. The way we were able to get the results about
completely rural borrowers is there was a deliberate over-sample in 2014 of folks in
that market. The bureau has a statutory mandate to look
at rural consumers to understand the costs and benefits of regulation in rural communities. That’s one reason that that was fielded. So we can do the same thing. We could either, if there are particular subpopulations
you think we should over-sample on this set of questions, subjective versus objective
knowledge, or there are particular scales that we can add that is something that with
enough planning and lead time we can do because we’re in the field every 13 or 14 weeks. Just a question on the rural borrowers. I thought this was all very interesting. Can you look at refi’s versus originations
for newly purchased homes? I was curious as to whether the rural borrowers
are less likely to take advantage of opportunities to lower their mortgage payment through refinancing. Whether they’re more likely to be in the money
because they haven’t had the opportunity to refinance. I think we could do that. We have separate set of analysis split up
by purchase and refi but your question is different. And if I could rephrase it we could look at
whether borrowers in these rural areas are aware of refinance opportunities and have
those opportunities. So those are definitely open questions that
with this data we can get at to some extent. Because again we don’t observe the lender
ID and so the lender supply is still an open question as you can see because again the
main census of lenders which is HMDA is not great for exactly these purposes. So that’s actually an open problem in terms
of measuring the supply side of credit in these areas is still a difficult thing. This is a question, maybe it’s for Ron. I think this survey is fantastic in terms
of getting at questions that we really want to know beyond what’s in the administrative
data. But as you move into survey methodologies
through the mail you have the response rate concern. Is there a systematic way that you’re thinking
about — if you had a contractor who was doing all of this you would probably require them
to produce a non-response report and do other sorts of things. Is that something that’s occurring here and
just didn’t come up? Yes, so we do have a contractor who gives
us a non-response report. One of the other things we have that I think
is a really attractive way of the way that we’re doing the NSMO is that because we’re
drawing this from administrative data that we have we actually observe a lot about the
people who respond and the people who don’t. So our ability to review non-response adjustments
I think is a little bit better and we’ve put a lot of time into that as well. Speaking of which we’ll go to our last presenter
in this session Scott Fulford who will talking about another survey, the Making Ends Meet
survey. Thank you. So last year for those who are new we actually
told you that we are actually going to be piloting a new survey off of our consumer
credit panel. And so one of the things that we’d like to
do is first give you a report back of what we’ve learned from that pilot, some of the
things that have changed and some of our thoughts about the future. And let me just begin with some very broad
overview of this. So our goal is really to understand the causes
and consequences of running out of money. And that’s in some sense a way of thinking
about this survey as complementing other surveys that exist in the survey literature, things
like the Survey of Consumer Finances, the board SHED, the Survey of Household and Economic
Decision-making and other things where we think that they may be missing important parts
about really getting at people when they’re having large marginal utility events, places
where there’s a real point where they’d like to spend more and are having difficulty doing
it. And I want to make sure I’m using my particular
— our economic jargon here, that we really do want to think about when are people having
problems. So the really nice thing about being able
to do this is we’re sampling from the consumer credit panel. And we’ve heard a lot about that but just
to remind you rich, long administrative panel data on credit and debts. And so that’s really exciting because it gives
you a great deal of useful information about people over time. But it’s lacking other things. It doesn’t tell you a lot about decision-making. You don’t see a lot necessarily about the
demographics. You don’t see much about their income. So the advantage of doing a survey off of
that is we can help fill in many of the things that are missing within the CCP so both richen
the CCP and also add to the general survey understanding of a very particular part, where
do we think that people may be having trouble. So to give you where we are on it we did a
pilot about a year ago starting in May for 12 weeks with 2,000 mailings and had about
a 20 percent response rate which is about what we’ve gotten for other surveys targeting
a population that may or may not want to respond to surveys all the time. We did a follow-up then asking the same people,
people who had responded to the first wave in October 2017. And had of the 390 initial respondents we
actually had a 65 percent response rate. So people who had responded in the first wave
about 65 percent of them responded in the second wave. And I just want to sort of pause on that. We were unsure in doing the pilot whether
we would actually get a good response rate when we followed up with people. If we got a 20 percent of a 20 percent that
wouldn’t be particularly useful. We wouldn’t really have much to say about
what we’re changing. Getting a large response rate from people
who have already responded really tells us that we can do panels and we can ask people
what’s changed in their lives. So let me give you a little bit of a sense
about some of the things we’ve learned. So first I think it’s worthwhile saying that
in piloting and in doing these studies we want to be continuously learning and improving
the way in which we approach doing surveys. So one of the things that we were doing within
this pilot was actually doing experiments on survey methodology. And partly just as a teaser I can talk more
about this. Brian Bucks and I who will be chairing the
next section have a paper which will be — it has a poster in tomorrow’s session if you
want to learn more about it. But a very brief sense about some of the things
that we wanted to learn. One of the big questions in particularly mail
survey methodology and I’m not sure — I have definitely gotten surveys like this in the
mail from other government agencies is how do you get people to answer your survey. And one of the things that one might want
to do is encourage people to take it online. So an online survey allows you to do better
skips, so people are only answering the questions that apply to them, not random questions that
might apply to somebody else. It also allows it to be a little bit more
flexible for them. So we actually tried to push people to try
to take it online. So we sent about half our sample just a paper
survey and an invitation to take it online and the other half just an invitation to take
it online with a secure link that they could use. And what we found is that people really like
their paper survey. So over the first five weeks at week five
we sent everybody who hadn’t responded a new paper survey. The response rates were much, much lower. So that’s the bottom green line on the graph. The people who only got an online survey were
much less likely to respond. Somehow just having that paper thing sitting
on your kitchen table seemed to make a big difference. Once we sent everybody a paper survey it caught
back up and at the end the response rates weren’t significantly different so that they’re
still a little bit lower. And we’ve done some cost analysis to try to
figure out what would be the best way about approaching — what’s the most cost effective
way of sending out surveys like this. So part of the fun of having a pilot is we
can now tell you some of the fun results. And I want to just caveat these at the very
beginning these are pilots. While we have weighted these to be nationally
representative they are still it’s a pilot with large standard errors and questions about
it may or may not be representative of the full population. That’s part of what doing a pilot is. So we found that one-third of all respondents
were unable to pay for either an infrequent major expense or a normal household expense
in the last 12 months. So one-third of people told us they couldn’t
pay for an expense in the past 12 months. What that means may be different for different
people and I’ll show you a little bit about what people report doing and really we’ll
suggest that people respond, think of running out of money in different ways. As you might expect running out of money is
a lot more common for people who have low credit scores. So nearly 40 percent of people with a low
credit score or below 650 ran out of money versus about 15 for people with higher credit
scores. For most people who ran out of money it was
because of some unexpected expense. When they say they ran out of money it was
largely because something unexpected happened. And to give you just a sense, infrequent major
expenses and normal household expenses are very high correlated. One of the things that we found is that asking
about these separately doesn’t — people who would answer one are very likely to say yes,
I also ran out — if I couldn’t pay for a normal household expense I also probably couldn’t
pay for an infrequent major expense. And so as we’re thinking about moving forward
we are putting somewhat less emphasis on that distinction because it doesn’t seem to be
important in the way people think about running out of money. So what do people do when they run out of
money. Well, how people deal with it actually depends
a great deal on their credit scores. So the graphs show how people with high credit
scores and low credit scores report what they did when they ran out of money. It’s from — options are from many different
options and you can select more than one. So the sum of all the options does not actually
have to add up to 100. You can do multiple things when you run out
of money. So people who have a high credit score when
they ran out of money mostly took money from savings or used a credit card. Notice an important part there is they may
have negotiated a lower or delayed payment. So somehow negotiating over what you’re paying
is an important part even for people who have high credit scores. People with low credit scores respond very
differently. The first option is actually to work overtime. And there were several other options that
didn’t quite make the top four that were also about acquiring additional income in one way
or another. And so taken together that was the most common
way of dealing with running out of money was to actually acquire in some way additional
income. But borrowing from friends and family was
also very important for those who have a low credit score. So one of the things that we — having had
a successful panel we wanted to add a little bit about what we learned from the panel. And I think Cheryl mentioned earlier we have
actually only from a similar data set — or a different data set. We’ve actually only had these data for about
a month so we haven’t actually had a lot of chance to think about all of the many things
that one can ask about a panel. But we wanted to highlight two results that
I think are important and that highlight why it’s both useful to go back and ask people
and you couldn’t learn unless you asked people a second time about something. So the first is it turns out that running
out of money is persistent. So of the people who reported that they were
having trouble paying for normal household expenses in May 74 percent reported that they
also had trouble in the next six months. When we went back to ask them they were very
likely to have trouble running out of money in the next six months. Moreover, if you report that you’re running
out of money you’re very likely to report that the most recent episode was within the
last one to two months. So what that also tells us is that if you
think about running out of money people who are running out of money are doing so very
frequently. That 63 percent of them have done so in the
last two months and to the extent which we can kind of pull that out that says that for
people who run out of money about 63 percent are running out every one to two months to
the extent that the cross section actually tells you that. So I want to highlight a little bit about
the limits of this is a pilot sample that we are — that it has small samples. You might have noticed I didn’t put standard
errors on there. There was some intentionality about that. It has limits to statistical significance
and precision. Things like a larger sample would allow us
to do an important comparison of more groups. For example, Sergey was talking about rural
populations. We don’t really have the ability to with the
sample we have to talk about whether people who are in rural areas are different. We don’t know for example whether they might
have different products or fewer products available to help smooth stocks. And the advantage of the CCP, combining this
kind of survey data with the CCP is it gives a sense about how people are using products
that are not in the CCP and how those might impact them going forward. So there are non-traditional credit sources
that may be very important for people and for their financial fragility and the extent
to which you can measure the impact of using those and what happens for people who are
using those, the advantage of the CCP is we can actually also measure them going forward
and think about what might be different about people who are using some products versus
others. With that we welcome any comments, questions
or anything else. I have kind of a specific question on your
results. When I look at the high credit score versus
the low credit score I was struck by how much shorter the borrowers are in terms of what
they did. No, the next one. Yes, exactly. If you look at the fact that the scales are
different on the side by side graphs then you see just people are — the numbers are
just so much lower considering they’re your top four categories. So what’s going on there? Do they just generally have fewer options,
or they’re just, I don’t know, not doing the thing that they wanted to do. What is going on there? I think there are several different answers
to that. The first is that I think one of the things
that we found, and this is something that we have both struggled with but it is also
the struggle of any financial survey is that people have very different options available
to them. And you want to be able to capture the experience
of people who are using — who have very different things available and what’s available is going
to be very different for different people. And so the idea of how to capture the lived
experience of somebody in a way that’s meaningful for them is difficult. One of the things I think we found is that
people with low credit scores just do more different things that sort of what a low credit
score means may be very different for many people. To misquote Anna Karenina happy families are
often alike. That somebody with a high credit score has
— part of having a high credit score is access to credit. And so there are advantages. The second one there is actually after savings
people use their credit cards. And when you don’t have those two things available
what next your option is in some sense is one of the things that we wanted to think
about is when you have a high marginal utility event what’s the equivalent interest rate
on your next best option and what’s the mapping of going from — so the way I like to put
it is borrowing from your brother-in-law has a pretty high implicit interest rate. Even if he’s not charging you interest he’s
going to bring it up every Thanksgiving for the next 20 years. That’s worth staying away. That may be less costly than working overtime,
for example. So in the earlier discussion about the income
data that’s being merged to the CCP one of the disadvantages is you don’t have individual
level income and you don’t have any sense of the volatility of income. And I don’t have the complete list of questions
here in front of me but one of the things I liked or at least think I remembered liking. Now John’s got me second guessing. Maybe I’m being over confident. Is that you do have some questions to try
and get at how much of this is being driven by income volatility as opposed to by expenditure
volatility. And I guess a third category would be just
poor management skills either in the face of volatility or not in the face of volatility. So I think that’s a lever that would be worth
pressing on. You’re doing this in the context of the survey
in the context of the CCP, that’s the data source you’re drawing from, but I could imagine
that this type of a survey would be an interesting and useful template for other entities to
merge onto other data sources where they have even better data on things like income and
spending volatility. So once again kudos to you for developing
an interesting study and exploit it to the extent you can with the data resources you
have in-house. But go out and evangelize and try and get
other data people who have other data that maybe you don’t have to try and do something
similar. And then this is more of a how to think about
things going forward, ties back to Karen’s question and the point you made, the big difference
between the high credit score and the low credit score is the high credit score individuals
have savings and they also have credit. And the CCP gives you really good data on
credit. It doesn’t give you any data on savings. And in terms of thinking about what’s the
role of public policy in helping these sorts of households I think it’s worth thinking
both on the credit side and on the savings side. Just as a more general matter, CFPB has lots
of great data on credit and the world in general has lots of great data on credit. And what we know about the savings side of
the household balance sheet is much more limited. And so I don’t know whether that’s something
that CFPB could invest resources in going forward. Yes, thanks. So your question sort of validates some of
what we’ve done. So we have the CCP. It has the sampling problems that it may have
which as you see people that have credit records. Some of them are unscored but the purely invisible
we can’t survey. But yet I think you see some pretty interesting
results here. And then we’ve managed to ask, we can ask
income means, we can ask income volatility questions. What I’m curious is relative to some other
data sources that are out there and particularly there are certain folks, I’m thinking about
the JPMorgan Chase Institute for example that has consumption data and can actually look
at consumption volatility. That’s something we don’t have. Nothing in here tells us what people are purchasing
or the like. How comfortable would you be, and this is
a question having seen that literature, John, when you ask people about stuff, of using
survey instruments to try to get at either consumption changes or effectively utility
changes. Subjective measures of people’s utility change. Because in the end that’s what you’d like
to measure. So if you’re thinking about access to credit. I live in a credit rich environment. Somebody equally that looks just like me in
every dimension lives in a let’s say credit sparser environment. Can we say something about the utility differences
in a dynamics sense or would you just discount that if we tried to get at that through these
questions. I think there is no perfect source of data. And what you hope is that you can triangulate
on these issues from different directions using different sources of data that all have
their problems. So the survey data alone allows you to ask
questions about both the income volatility and the expenditure volatility and you can
ask some questions about savings and credit. You’re not going to get as detailed data as
you would get with administrative data even in terms of responses, the length of the responses. And you’ve got the accuracy issues. But if you can take this survey that’s linked
to the CCP and that gives you a good sense of what’s going on on the credit side of the
balance sheet and how households respond to the survey. And then say you could get JPMorgan Chase
or Mint to do something like this with some of their consumers and then you get more information
on the income and the consumption side. And then you come at it from two different
directions. They both have problems. They’re both missing important pieces of information
but I think you put them together and the whole is more than the sum of the parts. So in that list we don’t see payday loans. So I assume it’s just low frequency in here. Yes. So payday loans are not particularly frequent
in the population. And so just in terms of the number of people
who are using them it’s hard for them to show up in any list. Because the thing that’s striking, this is
a theme that several people are bringing up, this figure and the fact that the high credit
score have all these sources, low credit score fewer sources. Just understanding why people are using one
thing, because they don’t have something else and then trying to use that to understand
part of the welfare effects of payday loans. Is it the case that when those are being used
it’s because all the other channels are shut down, et cetera, and when will a product be
— have more benefits to offset the costs than in other cases. So to expand on that one of the advantages
of using the CCP is for example it only gives you certain credit products, the formal ones. But we can check for example whether someone
who decided to borrow from friends and family had a credit card with an open line. And that at least tells us something about
that decision-making and in some sense the implicit interest rate that maybe borrowing
from friends and family is less expensive or my hypothesis would be more expensive,
that only people who have maxed out one go to the next. And again the ranking implicit interest rate
or however you want to put it is probably different for different people which makes
it difficult to sort of — but to the extent which we can sort of do that ranking, think
about when do people start using payday. What’s the thing they’ve exhausted before
that on average. Or a different kind of alternative financial
product. That’s I think an important part in why at
least observing all those formal parts is useful. So the other thing, John, I think you’re right
is that if we launch — we’re thinking about launching — the pilot proved that this works. You can survey people, we seem to get a response
rate that’s pretty robust and I think as Scott mentioned the panel response is actually quite
robust. So this notion of people understanding what’s
coming in their future and then checking what actually happened to them, the dynamic nature
of the shocks to the household is actually — I think we’ve proven it works. So now the question is going to be the sampling
frame. And to your point you could geographically
sample environments that have particular credit products be they payday loans, be they credit
builder products of some sort, high credit union penetrations, whatever you’d like to
think of, that’s going to be one of the difficult decisions and one we’d love to discuss further
with the ARC if we field this fully. There are pluses and minuses to over samples
and we should be very careful about that. But that’s something we’d love to continue
to engage in. I think some of it may depend on where you
want to go in the following sense. If you want to answer the question we’ve been
focusing on mainly one of the few things that you had time and space to write about what
do people do when they run out of money you can imagine there are existing panel studies
such as the American Life Panel where they have been collecting expenditure data and
income data at a monthly frequency. And if you wanted to put this question on
here. Obviously a large number of people are saying
they had a problem in the last 12 months. You’re going to capture a number of people
even though the panel only has maybe three or four thousand people, I forget the exact
number. And you can maybe dig down into that and see
what people are actually doing. Maybe that’s the third side of the triangulation
that Brigitte was talking about. If there are other questions I imagine that
are all about do I open a new line of credit or do those sorts of things. You could try to balance it out between what
do you want to ask there and what would you have here. So I guess part of it would also be what is
really necessary to use the Making Ends Meet survey and what’s really necessary about the
administrative data. Again if you were just going to cut it high
versus low credit score my guess would be is that you probably could use a limited number
of demographic characteristics plus some strategically asked questions to probably bin people as
a high credit person or a low credit person that you could then transport to these other
data sets even though there you don’t have all this administrative data. So again depending on what it is that you
want to do, if it’s questions like these maybe another data source could help answer that. If it’s some of the more rich detailed credit
bureau information which is — provides a lot of other useful information then maybe
thinking about designing this survey to go after those features would be the way to think
about the broader mission of where you want to take this particular research. Can I just jump in quickly and follow up on
what Brigitte and Mel were both saying which is on the issue of consumption — so first
of all I totally agree that having information about the asset side would be terrific as
well to the extent that you could do that. But on the consumption I would be a fan of
adding some simple consumption questions to the survey because I think we’ve gotten better
at asking those questions over time. My sense is we understand more about what
people can and can’t answer. They’re not very good at talking about how
much they spent on food over the past year. But on the other hand over the past week they
understand that better. Whereas you can ask about did you purchase
a car over the past year and people’s recall is better at that. So I think we’ve learned a lot about how to
ask about consumption on these sorts of surveys. And I think at the same time I think the administrative
data when it comes to consumption has its challenges. We know that information that’s based on what
you’re seeing in terms of people’s checking and saving accounts or card use is just a
very incomplete measure of that person’s consumption. And then on top of that you have all these
issues about the difference between that person’s consumption and the household’s consumption
and whatnot. So I just wanted to jump in there. Rand has these questions so that might be
the solution is to merge with that data set. But I just wanted to weigh in on that. All right with that I think we’ve reached
the end of our time so I’ll just say a brief thank you to everybody. I think we’re going to adjourn now for about
12 minutes and reconvene at 2:45. Thank you. Welcome back. Thank you very much. We look forward to the second session here. The first session focused as you recall primarily
on the resources that we’ve been building, the data resources and other resources, and
some of our regular reports. This session is going to turn now in more
detail to how we’re using these resources to ensure that we identify research questions
that are policy relevant and contribute to the bureau’s mission. The outline of this session that kind of follows
a common path for research projects. We’ll start by talking about how we identify
the research questions, then establishing the key facts that our analysis has got to
try to match or account for, and then finally the third part we’ll turn to the working papers
where we’ll highlight how we’ve gone into some of these questions in greater depth. In terms of the research agendas I want to
update you on the process by which we go about identifying and supporting research projects
that really take advantage not only of the data we have but also the expertise we have
throughout the bureau, and as you know the expertise of ARC members and others to identify
and answer these kind of policy relevant and mission relevant questions. Turn then to the data points which are publications
the Office of Research puts out periodically that use our unique data resources to establish
key facts, key patterns, stylized facts if you will about consumer finance, commercial
financial decision-making or consumer financial markets to expand our and others’ understanding
of consumer finance. Then we’ll turn to working paper series which
is our opportunity to disseminate our work and to get feedback on it and to tackle these
questions in greater depth. After some discussion of those first three
parts we’ll then hear about one of those research projects in greater depth. With that I turn the microphone over to Heidi
Johnson who is the program lead for the disclosure research agenda. Thank you, Brian. So I’m going to walk you through an update
to some processes we’ve developed over the last year or so. For those who are not intimately familiar
I’ll just very briefly review what our research agenda topics are. These are a few areas that we’ve chosen to
make sure to invest in, key topics that we think are important to explore. And so the first one is disclosure. Many of the regulations and statutes over
which the bureau has regulatory authority include some kind of disclosure requirement. And so we want to make sure that we know how
to do that well, that we know what is effective for consumers, also that we’re anticipating
the impact on firms and how firms are going to respond to disclosure and get both sides
of that picture. And throughout making sure we’re thinking
about the methodologies we can best use and how to leverage those for these different
research projects in these areas. The second key topic area is the dynamics
of household balance sheets which you’ve already heard a little bit about today. This is an effort really to focus holistically
on households, make sure we’re thinking about both sides of the balance sheet for a household. And looking at questions like the effect of
financial decisions that someone makes now on their future financial outcomes. Thinking about how long a shock that a household
might experience, like a spell of unemployment or an unexpected expense, how long that endures
for a household. Comparing differences for types of consumers
and digging into some of the questions we discussed earlier in the day about the Making
Ends Meet survey is one example. And also developing mathematical tools that
help us both identify optimal consumer behavior in certain conditions and how consumer behavior
in the real world might diverge from that. So these are the areas we’ve chosen to invest
in and already have a slate of research projects underway addressing them. But what we’ve focused on in the last year
is really cementing some of our processes around governing these agendas and how we
select projects. So we launched in this last year a new governance
structure really to make sure we’re incorporating input from folks around the bureau to make
sure this research is as Brian was saying policy relevant, hewing closely to our mission. And we’re doing that through developing these
two new groups, agenda working groups, one for each of the two key areas I mentioned
before which is really going to be at the staff level, making sure we’re getting input
from a wide variety of perspectives around the bureau. The second group will be an advisory panel
for each agenda topic and that will be more of the management and strategic decision-making
component of the governance. So we’re already actually underway in launching
each of these groups but I’ll dive in a little more into what each of them entails. The working groups as I mentioned is really
a staff level forum. The idea is to foster a dynamic culture in
which people are sharing their expertise, workshopping projects. We’re anticipating this will mostly include
staff from our Office of Research but we’re really excited to include experts such as
our ARC members as possible, our visiting scholars and folks from around the bureau
depending on the topic at hand. So a few of those topics we’re considering
are things like brainstorming what to do with our research assets and where we can best
deploy our time. Workshopping study design as folks run into
methodological challenges or anything else where it would be helpful to get input. And just sharing generally resources and best
practices. When it comes to the advisory panels this
will be more at the management level with the objective of getting input from folks
within the bureau as well as our external experts providing input to our deputy assistant
director. So this will be the body that’s helping us
review proposed projects and select from among them to decide what we’re doing in the next
year. Also to help scope out projects that are generated
by management. Sometimes needs arise and we want to develop
a new research project that is responsive to some policy challenge we’re facing. We’ll use these groups to help guide that
work. We’re also looking forward to leveraging the
scientific expertise and methodological input from these bodies and really help oversee
the projects as they all move forward. So as I mentioned this group is already getting
started helping us select projects for the next year. We do have a slew of research projects underway
currently and so I’m going to briefly highlight just a few of those under each agenda area. Within the dynamics of household balance sheets
you heard about the Making Ends Meet survey from Scott. That is part of this research agenda and was
developed to help answer key questions there. Our work in matching data to our existing
data resources and identifying military service members and gathering income data as Cheryl
discussed, that’s also part of the dynamics of household balance sheets. We also have several studies underway in the
area of student loan repayment that we’re really excited to share with you. This is a really important area for us to
be exploring and we have some unique data resources to deploy on those questions so
you’ll hear more about that later from Christa. And in the disclosure area I wanted to highlight
two lab studies. So this is leveraging the current capabilities
that Dustin shared with you earlier. Contracts that we have with researchers or
universities and other research institutions help us identify and isolate effects of changing
elements of a disclosure. So one such study is on tiered disclosure
where we’re going to be presenting a subset of information using two different techniques. And our researchers will be able to evaluate
the conditions under which those different approaches might be more or less useful for
consumers. Really help us understand how to best develop
them in the future. We also have a project coming up exploring
an issue of vague language. And this is really interesting for us with
disclosures. Often it’s very important for everything to
be legally accurate but that can also lead to a lot of uncertainty for consumers as they’re
trying to interpret the information in there. And we want to make sure we understand that
trade-off. So we need to understand consumers’ perceptions
of and processing of that information. Do they see it as important. Do they see it as relevant to them. What kind of impacts do we see from language
that is more or less vague in disclosure. So that’s a quick sampling. We’re really looking forward to getting feedback
on our new processes and getting your ideas for how to leverage these new groups that
we’re standing up. Thanks. Thank you, Heidi. As I said earlier the Office of Research data
points are a series of analyses of data that we put out to bring an empirical perspective
that improves our and others’ understanding of consumer finance, regulation of consumer
financial markets and such. So next I want to have Eva Nagypal and Christa
Gibbs who both authored data points give you a quick overview of some of those. Thank you, Brian. So the bureau has naturally a longstanding
interest in one of the most basic commercial financial products which is a checking account. And within that particularly in the case of
an overdraft which occurs when a consumer overspends the funds in their checking account. And the Office of Research for years now has
led the empirical aspiration of checking account overdraft. In 2014 we published our first data point
that established some basic findings about overdraft. And among other things we found that overdraft
is very concentrated. Most people don’t overdraw their accounts. Probably most people sitting in this room
do not overdraw their accounts. But there is a small fraction of account holders
who actually do a lot of overdrafting. And so we now refer to them as frequent overdrafters. And last summer we published a data point
where we were aiming to understand these frequent overdrafters. For the purposes of this data point we define
frequent overdrafters, those who have 10 or more overdraft or NSF in a year. This represents 9 percent of accounts but
these account holders are responsible for 79 percent of overdraft. So this is really where the action is which
is why we were interested in this population. So we set out to do several things. I’ll talk about two. First of all we were interested in how they
differ from non-overdrafters or light overdrafters. Probably not very surprisingly they have lower
balances, so they’re close to zero. That’s almost by definition. They also tended to use their debit cards
a lot more than non-overdrafters. And interestingly because checking account
behavior does not enter your credit report nonetheless they actually have lower credit
scores, significantly lower credit scores and have less access to credit, the frequent
overdrafters. Now there’s a lot more in the data point but
let me talk about sort of our next goal which is we ask how do these frequent overdrafters
differ from each other. And we sort of noticed that there was quite
a bit of variability in account usage and availability so we wanted to go into this
in detail. Moreover we’ve been hearing very different
stories about these frequent overdrafters. So if you talk to financial institutions they
would often tell you that these frequent overdrafters are well off consumers who have demand for
the service. And in fact to kind of support this point
they would show evidence that the more you overdraw your account the more deposits you
have. And that’s actually verifiable in the data
we were looking at too. On the other hand consumer groups would emphasize
how these overdrafters are vulnerable and low-income and they’re very different. So we decided to use something known as cluster
analysis which is a flexible data-driven statistical method to distinguish subgroups. It’s very flexible. It will tell you how many groups there should
be and how many number of people there should be in the different groups. So it kind of gives it some guiding input. So the input that we gave it is that we were
interested in clustering or making these groups by looking at similarity on credit scores,
monthly deposits, debit/deposit mismatch, and share of months with an overdraft. So are they using it a lot, frequently in
time or not. And so we went through all the algorithm. We had nice clustering scoring indices tell
us what was the best cluster. And we came up with five clusters. And I want to talk about these five. They’re in this pretty graph. And you see the distribution on the bottom. Of course we cannot cluster the no credit
score people because they have a credit score for the clustering to work but they also reported
not in the data point. So the first cluster has — these are the
people in green so I hope you all see this nicely in color. They have very low monthly deposits. They’re the most vulnerable — they seem to
be in most financial distress of any clusters. As I said lowest monthly deposits. The highest variability of deposits. Very high debit/deposit mismatch. So the needs for the funds and when the funds
arrive are kind of mismatched across months. And they have a very high charge-off rate
at 12.9 percent. Cluster two has slightly higher deposits,
has the lowest credit scores. So this is the cluster you’re looking at in
red. They still have higher than cluster one but
not particularly high deposits. So they actually still have a very high chance
of charging off their account at 11.4 percent. Cluster three, that’s in yellow, starts to
have more resources so they have more monthly deposits. This is a very interesting cluster because
with the more deposits come more activity on overdraft. These are actually the people who most frequently
and most regularly overdraw their accounts. But despite that fact they actually have a
3.8 percent charge-off rate. Cluster four which is the blue cluster still
not particularly high monthly deposits but significantly higher credit scores. And they have a charge-off rate very similar
to cluster three at 3.7 percent. Cluster five is very interesting. So this is actually the cluster that probably
the financial institutions mostly mean when they talk about overdrafting customers. They have very high monthly deposits. So their median monthly deposit is close to
$8,000 even though the median for all the other groups is $2,500 or less. The end of day balances are five to six times
those of cluster one through four. So they have the funds but they seem to be
occasionally getting into a situation where they’re just not paying attention and they’re
paying a price for not paying attention and maybe that’s perfectly optimal for them. The charge-off rates are the lowest of all
clusters and their credit scores are the second highest of any cluster. So with that I’d like to turn it over to Christa
to talk about student loans. Thank you, Eva. So actually I’m first going to talk about
becoming credit visible. Similar to the data point on overdrafters
three years ago the Office of Research released another data point highlighting people who
are credit invisible which Ken referred to earlier. So these are again consumers who don’t appear
— who don’t have a credit record at any of the three nationwide credit reporting agencies. And that earlier report found that about 11
percent of all adults in the U.S. are credit invisible. Now most people when they turn 18 are credit
invisible. They don’t have a credit account. But by the time they’re 30 91 percent of all
adults have a credit record. So there are a lot of young people transitioning
into credit visibility but there’s still many people who are credit invisible even above
the age of 30. So in part to fulfill the bureau’s mission
to ensure all consumers have access to consumer financial markets Ken Brevoort along with
Michelle Kambara produced a follow-up data point last year looking at this transition
from credit invisibility to having a credit record. They did this using the bureau’s consumer
credit panel and they looked at a variety of differences across groups, particularly
by age groups especially since young consumers are much more likely to be transitioning. They also looked across neighborhood income
levels and over time following about a 10-year period from 2006 to 2016. And then finally they also looked at entry
products. So what type of credit product is triggering
the creation of a credit record and causing somebody to show up in the data. So as I alluded to before most consumers who
are transitioning into credit visibility are doing so before they turn 25. And credit cards are the most common type
of credit product to do this. This is true across all age groups. And this has increased over time. So it’s now even more common to become visible
because of a credit card except for the younger group, consumers ages 18 to 25 where that’s
actually remained fairly flat or even decreased which I’ll get into a little bit more later. A large share of consumers owe their earlier
credit item to having an account with somebody else. So for about 15 percent of these consumers
it is because they have an account with a co-borrower typically on an installment loan. And an additional 9.5 percent have accounts
where they’re an authorized user typically with a credit card. So about a quarter of all consumers who become
visible are doing so through help from another consumer who typically already has a credit
account. In contrast there are a large number of consumers
who are becoming credit visible not because of a loan product or credit product but because
of a non-loan record. Typically these are negative. They are collections or public records. And so becoming visible for these consumers
has a very different implication than for consumers who are becoming visible because
of a loan or a credit card. And there are big differences by neighborhood
income in this distinction. So over a quarter of all consumers in lower
income neighborhoods become visible because of these non-loan products whereas less than
8 percent of consumers in higher income areas become visible because of these non-loan products. If we look more closely by neighborhood income
we can see in the second column this is the share of consumers who become visible in our
data between 2006 and 2016 by neighborhood income level. And as you can see consumers in higher income
areas are much more likely to become visible in our data and they’re much more likely to
do so while they’re younger. It’s noticeable how much less likely it is
for people in lower income areas to be visible by the time they’re 25. And overall this results in the share of people
who are invisible in lower income areas being much larger than it is in higher income areas
as was shown in the earlier data point. Overall this has important implications for
how we think about credit invisibility across different neighborhood markets and whether
that should affect how we’re thinking about how people become visible, especially given
how lower income consumers are more likely to become visible for negative credit records. So focusing on the type of entry product that
people have over time there’s more detail in the data point which shows that people
— most credit products are pretty stable over time with the exception of credit cards
and student loans as shown in these graphs. So credit cards as I alluded to before become
much more common for all consumer groups except for younger consumers but they’re still the
most common entry product for all consumers. Part of this difference for young consumers
seems to be driven by the striking increase in student loans for young consumers as shown
in the bottom right graph. And also some of it seems to be driven by
changes in the credit card market due to the Card Act. But as the data point goes into detail these
two things do not explain the change for young consumers alone and leaves it as an open question
to better understand what’s happening with younger consumers here. Overall this data point documents important
differences in the transitions to credit visibility and points to why we need to better understand
these differences in order to help ensure better credit access for these consumers. And related I will now get to student loans. So as you just saw student loans are a very
important entry product for young consumers. It’s also now the second largest form of consumer
debt behind mortgages in the U.S. with $1.4 trillion in outstanding student debt. At the same time we’ve had some big changes
in the student loan market over the last two decades. We have a lot more people borrowing as indicated
in the last data point. The loans that they’re taking out are a lot
larger. And we are also seeing a large increase in
the share of older Americans who are taking out student loans on behalf of other people
and not for themselves. At the same time we’ve had a very large expansion
of income driven repayment options and in the take-up of those options. So there are big differences in how people
might be repaying their loans at the same time that we’re seeing differences in who’s
taking out loans and how much they owe. So overall it’s unclear what might be happening
with student loan repayment and so this data point again using the CCP attempts to look
at how changes in repayment have evolved and also how delinquency has evolved for these
borrowers. We look at this in a variety of ways but the
main focus is on over time how do these change and how do these differences — how are these
different across loan size especially as loan sizes have become much bigger in the last
several years. Overall repayment rates are actually pretty
stable over time. Surprisingly the rate at which people fully
repay their loans hasn’t changed much over the last 20 years but there are very important
differences across people who haven’t fully repaid their loans. So to look at that here we have distributions
of balance ratios by cohort, repayment cohort for each year of repayment. So just focusing on that top left graph we
have the 2002 cohort, so people who entered repayment in 2002, what was their balance
ratio at the end of their first year in repayment, where everybody who has already repaid their
loan is not shown in this graph. You see there’s a mass right around one because
most people at the end of their first year still owe most of what they borrowed. But there are a number of people who owe actually
more than they originally borrowed and that’s that distribution you see between the third
and fourth tick of people who owe anywhere from 100 to 150 percent of what they initially
borrowed. As you move down that first column you see
that the distribution slowly shifts to the right fairly smoothly as people repay their
student loan balances pretty smoothly, but there’s this persistent group of borrowers
who continue to owe more than they borrowed and don’t really seem to be making much progress
in paying down their debt. If you move to the second and third columns
you see the 2005 and 2008 cohorts look remarkably similar. There’s not a lot of difference there. When you move to more recent cohorts, the
2011 and ’14 cohorts a lot more borrowers who haven’t fully repaid owe a lot more. Their balance ratios are a lot higher and
they don’t seem to be making much progress on paying down their balances. Now that might not be very surprising. We know income driven repayment has become
much more popular so we would expect people to not be borrowing — to not be repaying
as quickly as they once were. But we would also expect that those people
would be in good standing on their loans because they’ve been given payment options that mean
they don’t have to pay as much and they won’t get a delinquency on their account. But if we look at people who aren’t paying
their balances this graph shows of those borrowers who have non-decreasing balances what share
are in good standing on their loans, that is who do not have a delinquency on any of
their student loans. So just focusing on this first group of borrowers
who have smaller loans totaling less than $5,000 you can see that there hasn’t been
an improvement. We don’t see a growth in the share of borrowers
not paying down their balances who are in good standing. In fact they’re just as likely to have a delinquency
as before or even more likely in comparison to that 2003 cohort. And that’s true across each of these loan
sizes more or less with some variation. If you look across loan size, however, you
can see there’s a much bigger difference between people who have small loan amounts and large
loan amounts. People with smaller loans are much more likely
to have a delinquency on their student loan where they’re not paying down their loans
than higher balance borrowers. So overall I think ex ante most people would
expect that we would see a decrease in delinquencies over time but the data don’t suggest that
that’s what we’re finding. It really draws attention to these lower balance
borrowers who seem to be struggling much more than higher balance borrowers even though
the higher balance is also struggling more now it appears than they were in the past. And points to the need to look at this in
greater detail to understand whether this is having impacts on their ability to take
out credit in other markets as they get older especially since for so many of these borrowers
this is the first credit product that they have on their record. Thank you Eva and Christa. I’ll pass the mike back to Eva now to give
a quick overview of our working paper series which is a vehicle we’ve started in the last
couple of years to disseminate our research and to be able to get feedback on it. So please, Eva. Thank you. So from the bureau’s inception the Office
of Research has been growing and strengthening its ties to the research community. A big part of that is of course our fruitful
interaction with ARC members. Beyond that we have our seminar series. We have visiting appointments. We actively participate in conferences. And as you know we established our very own
research conference the third edition of which is going to be tomorrow and the day after
so I hope you all attend. In this vein of strengthening ties to the
research community the ARC has recommended to us that we launch a working paper series. We are very happy to say that we’ve done so. And in particular we’ve done so through the
Social Science Research Network and it’s up and live. And we have started publishing papers there. A variety of papers on consumer finance. I’m not going to go into the details of them
but I can say they really range in terms of methodology, some of them are structural. So the third paper on credit card utilization
is looking at a life cycle model of credit card use. Some are looking at natural experiments, debt
collection. The fourth one, the effect of debt collection
laws on access to credit is exploring a change in debt collection law. And some are lab experiments such as the fifth
one which is looking at prepaid card menus in the lab and how the variation of those
menus might impact comprehension. The papers also cover a wide range of markets. The first one is the shopping in the U.S.
mortgage market. Second is looking at the effect of interest
rate ceilings in the market for auto loans. The sixth one is trying to evaluate the benefit
from a form of small dollar credit, debit deposit advance products. And the last one is looking at credit card
services and the demand for these services. So really broad range of methodologies, topics
within the realm of consumer finance. So we believe that this series is actually
serving the purpose that you proposed that we establish which allows us to promote our
work and receive critical feedback to ensure that we can produce the highest quality research
that we’re able to. Thanks, everyone. Now I’d like to invite your questions and
comments. Please. Well, there’s nothing that makes someone feel
better than knowing that their suggestions in the past were taken. The SSRN Working Paper series is a good note
to end on for this section for that. I’m wondering have you been tracking, have
people — SSRN puts out little digests so the papers get featured but have people been
reading them and downloading them, and the people who are reading them and downloading
them, are they giving you feedback or is it just kind of a mechanism for visibility. So we’re starting to get downloads. Some of them have been downloaded quite a
number of times. We are one step short of full visibility in
the sense that we have not yet gotten them placed into SSRN’s weekly or monthly digest. So the household finance digest that Sulelos
and Tufano run is our next stop. And once our papers start getting broadcast
through that second mechanism. You can sign up actually at SSRN to receive
any updates to our working paper series. We recommend that everyone do that. There are six or eight thousand people that
have actually done that. But actually getting I think into the digest
is the next step. I’m impressed. If six or eight thousand people have done
that, like wow. There aren’t that many people that are signing
up to get my working papers. I only have one mom as Mel just pointed out. I had a specific — I mean this is all great
stuff. I had a specific question on the student loan
work which I thought was really interesting, Christa. I actually wanted to ask about you refer in
your last chart to loans with non-decreasing balances but your second to last chart really
shows that there is significant mass of loans with increasing balances. Am I getting that right? Yes. And so which is kind of concerning. You mentioned the income driven repayment
program story which seems plausible that people are getting some — newer cohorts are increasingly
participating in that program that don’t have to be paying down their loans although in
this case their loan balances are actually increasing. But I was trying to decide is this a good
news story about income driven repayment or is it a bad news story. In some senses I think the whole idea is that
you provide these borrowers with relief and as you were saying it means they don’t have
their credit records scarred by having delinquencies or default on them. On the other hand the fact that they need
to be in the program in the first place could be reflecting the fact that there are increasing
numbers of people who are getting educations that cost a lot and aren’t paying off. I was just wondering whether you were able
to kind of dig in on that front at all and learn anything about that. Excellent question. So right now we can’t speak to whether the
increase we’re seeing is because people need it more than they used to or it’s just because
it’s more available and they know more about it. That definitely seems to be a big driver of
this especially since we’re seeing that more people are taking it up who have been out
of school for a while and are taking it up later in their repayment perhaps as they’re
learning about it. And we’re actually trying to look into this
more specifically at people who are on income driven repayment plans. I think this first data point suggests that
while the idea is good maybe there is something happening in the execution. It’s unclear where in the process but there’s
something happening where people are at least some points not making payments and are still
seeing delinquencies and defaults. We’re looking actively to see more of what’s
going on there and hopefully we’ll have some more work to show on that soon. This is a question for Heidi. Could you say more about what you were thinking
in terms of firm response to new disclosures? Yes, absolutely, that’s a great question. So in looking at firm responses we want to
make sure we’re taking into account all the various ways that a firm might react when
a disclosure is implemented. So that might mean looking at holistically
beyond the disclosure are they changing things about the product itself for example in response
to the fact that there’s new disclosure requirements. We’re also thinking about the fact that even
if a consumer doesn’t read a disclosure, a given disclosure for example a firm because
of concerns about risk or competing with other providers might again be changing something
about the product, maybe lowering a price to remain competitive if there’s something
new that’s been developed in a disclosure. So that’s kind of a small portrait of the
kinds of impacts we expect we might be able to see. John, by way of illustration our prepaid rule
which hasn’t taken effect yet, one of the things we require to be disclosed is the number
of fees you have that are not any of the core fees that are disclosed. So one might hypothesize that one doesn’t
want to show a big number for the number of ancillary fees and that will affect the number
of ancillary fees that will be charged. We also have the net disclosure, a dynamic
piece where everybody has to disclose fees A, B, C, D, and E and then each entity has
to disclose the top three other fees if they make up more than Y percent of their revenue. That also might affect — if you don’t want
to show more fees you might then try and make sure that your revenue is coming from the
disclosed fees and not from other fees. So whether we’ll have any of those effects
we don’t know, but that’s the kind of things we think about when we think about how disclosure
may be impacting firm behavior. I think it’s a great direction for that research. A question for Christa. I’m wondering what if anything you are planning
to do next with the credit invisibles. I think it’s a really interesting question
to look at because I think we really don’t know much about the process by which people
start obtaining credit. One question that this discussion raised in
my mind in part probably because I’m the mother of a 17-year-old and a 21-year-old who are
kind of squarely in this age ranges. Are there long run implications of becoming
credit visible at certain ages and at certain times. And you may not have good exogenous sources
of variation in it but it would be just interesting to understand like if you compare individuals
who got their first credit card at age 18 with those who didn’t get one until 25 do
you see differences say at age 30 and who’s more or less responsible in terms of using
credit. Or is it better to start off with a credit
product where you’re the sole user or are you better off if you have a cosigner or you’re
a joint account holder with your parents, say. So what could we learn that would help families
transition their children from being credit invisibles to credit visibles but help them
to be responsible credit visibles. So, a couple of things. I think that’s a very interesting path to
go down. I don’t think that’s one we’ve really started
on yet. The current course I think that people in
our office are working on with these credit invisibles and visibility is thinking more
about locations and where people do or don’t appear to have access. Some of these credit invisibles are highly
concentrated in certain areas so that’s where they’re currently looking into what’s going
on with that. But I think especially some other offices
at the bureau would really enjoy having more practical tips for how to help consumers transition
in the best way possible. And so that’s definitely something that we
should look into more. I have one for Eva. So how did you achieve those clusters. You had five variables that you used to cluster. And I guess my question is was one of them
any measure of variability or volatility in either the income or expense stream. So we used four variables to cluster. It’s K means clustering. Sorry they’re not on the slide. So it’s credit score, monthly deposits, the
debit to deposit mismatch. So that’s essentially variability — so construct
the debit/deposit ratio and then look at the variability of that over time. So that is intended to capture how much your
resources are in line with your expenses. So that’s what that’s meant to capture, I
think what you are mentioning. And last was the number of months out of the
year that you are overdrafting. So one year span. So that’s intended to capture are your overdrafts
coming in infrequent bursts or just you’re more regularly overdrafting. So those are the four clustering variables
and then we used the K means clustering algorithm. And so the group that was the one that all
the bankers were looking for, so what explained that group again? So they had much higher — honestly what makes
them stand out is the deposits. Their credit scores are not the highest of
all the groups such as group four, cluster four had higher credit scores. But it’s really the deposits that made them
stand out. Where I was going was just that if somebody
had a lot of volatility in either income or expenses then even despite the fact that their
mean income could be high they could still wind up overdrafting. Right. Their monthly debit/deposit mismatch wasn’t
particularly high compared to the other groups. The variability actually was the highest — the
variability of deposits was highest for cluster one, for the most distressed. So I think what’s happening to them is that
they make large transactions and occasionally there’s mistakes. They’re the ones that are probably easiest
to claim neoclassical framework of it’s just costly to pay attention and this is the price
they pay for not having to pay attention. So just following up. So I missed the data source for the overdrafts. Is it also linked to some measure of you mentioned
expenditures, also say other pre-committed expenditures, say mortgages, vehicle loans,
et cetera. So essentially this is checking account data
matched with credit records. The reason I ask is that there’s this more
recent view of we’ll call the wealthy hand to mouth. So people that have high mortgages. So they look more like credit constrained
people although there may be — they have these consumption commitments which I’d imagine. So you’re already close to the margin. So the high income doesn’t tell you much about
how close to the margin you are every month or how much disposable income you may have
at the end of every single month. I was just curious if you were able to dig
a little bit deeper on that front because I imagine that some of that may be from that
sort of population. I don’t know if there’s any way to get at
that with other sources through the CCP or anything else. I guess perhaps not. It would be interesting to follow up if you’re
able to link costs to different data sources. For this data point we focused on available
liquid resources or available credit cards. We could look at their mortgages or other
obligations. I think the wealthy hand to mouth is probably
a better term than what we’ve done internally. So I think this discussion and these questions
have already highlighted just how rich our data resources are and I think illustrated
how the research agenda questions actually are very apt ones. I think many of the questions you’re asking
fall into those topic areas. I want to wrap up this session by having Ryan
Sandler discuss one of the working papers in greater depth, namely his joint working
paper with Chuck Romeo, The Effect of Debt Collection Laws on Access to Credit. I should note that as you probably recall
this study of debt collection follows on a lot of work that the bureau has already done
in the form of two surveys on debt collection topics. One of these surveys was the first nationally
representative surveys of consumers and their experience with debt collection. The other was a qualitative survey that the
Office of Research led talking to and doing written questionnaire with debt collection
firms to learn about their operational costs and potential compliance costs. So Ryan’s work with Chuck I think takes us
several steps further in understanding this market as well. Thanks, Brian. So debt collection is a pretty important part
of the credit system although it’s not one that’s been heavily studied by economists. At any given time there are many billions
of dollars of credit card debt and other kinds of unsecured debt that are severely delinquent. And frequently creditors respond to this by
turning to third party collectors to collect on those debts. Whether more directly by contracting with
contingency collectors who work on contingency or by selling the debts for less than face
value to debt buyers who then collect on their own account. According to one of the collection trade groups
about $55 billion of all kinds of debt not just credit card are recovered annually by
collectors. So it’s important to the creditor side of
things. On the consumer side there are, however, some
substantial consumer protection issues. As Brian mentioned in our survey we found
that about one-third of consumers in the past year had been contacted by a debt collector
and many of them also who had been contacted reported various problems with that experience. And it’s routinely one of the most frequent
sources of complaints to the bureau’s complaint database. So there is federal law governing debt collectors
through the Fair Debt Collection Practices Act or FDCPA which there are some specific
things that are required but there’s other things that are just sort of behave well. In addition there is a whole other layer or
several layers of state requirements that govern debt collectors. So there’s a lot of stuff going on with the
regulation of debt collection. And the reason why this is sort of an interesting
economic question and of interest to the bureau is that we imagine that creditors’ decisions
of whether to extend credit in the first place is going to be a function of the expected
value of that extension of credit which is going to be a function in turn of how much
will the consumer pay back with interest if they pay it back and how much will the creditor
if anything if the consumer does not pay it back on time. And of course if you make collection harder,
either the collectors are able to return less or it’s more expensive for them to do so that
will reduce the lender’s expected value in the first place. And so we show in a little simple model in
the paper that there’s at least three levers that lenders can pull on to adjust for this
change in expected value which can include the extensive margin, do you extend credit
in the first place, a sort of intensive margin, how much credit do you extend, the credit
limit, or interest rates. We’re specifically focused on credit card
debt for our empirical work but for other kinds of debt you might imagine they’d adjust
terms as well. So we leveraged two of the bureau’s large
administrative data sets to get at this question using several recent state law and regulation
changes. We used both the CCP which you’ve now heard
a lot about. We also use our CCDB or credit card database
which is an anonymized data set of deidentified credit card records from several large banks
where we have a monthly panel of performance data on those accounts. But for this paper we’re only using one record
per account for new open accounts. And so we’re looking at those three margins
that I mentioned a moment ago. Are consumers able to open accounts. If they open an account how big is their credit
limit. And if they open an account what kind of interest
rates are they able to get. So what we find in brief and I’ll show you
some of our general results on the next couple of slides is basically when you look at all
consumers pretty much nothing changes. The theory says there should be an effect
but empirically on average it’s not very large. However, when we focus in on subprime borrowers
where you’d be more likely to see an effect. If creditors are trying to hedge against the
state of the world where the borrower defaults you’re going to see effects more on borrowers
that the creditor thinks are likely to default. And we do see some effects on subprime borrowers
specifically on the credit limit margin and interest rates specifically through a decrease
in the prevalence of at least when we look at the data like zero interest rate credit
card offers, promotional rates. We also observe that there’s heterogeneity
in creditors over exactly which margin they decide to respond on. So here’s kind of the key table in the paper
for our CCP result. The thing with the CCP is we do observe — because
we have the whole credit record we do observe inquiries so we can tell if a consumer applied
for an account and then did or didn’t get one. And we observe credit limits of course. But we don’t observe interest rates. That’s one limitation in the CCP. So the first two columns there show our results. Basically it’s a difference in difference
model with lots of fixed effects including quadratic trends by states. We only have a small number of events here
and so we want to account for the fact that the states that happen to do these laws may
have had sort of different preexisting trends. The first two columns show the inquiry success
margin. So an observation is a hard inquiry for a
credit card and the outcome is was an account opened within 14 days. And so whether you’re looking at prime borrowers
or subprime borrowers basically there’s no effect. The treatment indicator is statistically insignificant
and the coefficient is just small in absolute value. We can reject a change of sort of moderate
size. One way to get a sense of the scale of this
change because a 0.4 percentage point change in success rate for increase isn’t that informative,
but if you look at our coefficient on credit score you can see okay, well, one point of
credit score increases the chance of a successful inquiry by what you see there. And so our treatment effect although not statistically
different from zero is also equivalent to like a 3 credit score point decrease for subprime
borrowers which is not. Looking at initial credit limits conditional
on getting an account. So here an observation is a new credit card
account and the outcome variable is the initial credit limit. We do see declines in our point estimates
for both prime and subprime borrowers. For prime borrowers you have to keep in mind
the average credit limit is much higher than for subprime borrowers so that’s still actually
small in magnitude in a relative sense. But for subprime borrowers we do get a negative
and statistically significant effect. And if you scale by credit scores about an
8.5 point change which is not nothing. It’s also not a lot. This graph shows you the heterogeneity across
which margins creditors adjust on. So this is interacting our treatment effect
with indicators for the 29 largest issuers. And the blue dots are the 10 largest issuers
among them. You should probably focus more on those because
even looking at the 29 largest credit card issuers after you get out of the top 10 the
sample size starts getting kind of small so it’s noisy. And you observe that although most of the
dots are in the negative quadrant so they’re cutting back on both margins, both inquiry
success and credit limit, there’s heterogeneity over which one they respond. Some do a little bit of both. Some do mostly one. Some do mostly the other. So turning now to our results from the CCDB. So here we can look at credit limits again
to have a confirming source. We can also look at interest rates because
we do have that in the issuer data. And this is specifically for subprime borrowers. It’s cut off. So for subprime borrowers we do have a negative
coefficient. It’s not statistically significant partly
because it’s a little bit noisy. I think you can’t actually reject that our
CCP and CCDB results on credit limit are different from each other. So it’s a moderate effect. We get some interesting results on APR though. We don’t scale by credit score here. One of the things is that credit limit and
inquiry success increases monotonically with credit score. Initial APR doesn’t in the data. Partly because the factors that determine
what interest rate consumers will accept differ as you get into introductory rates and rewards
cards and things like that. So it doesn’t scale properly. So we control for it flexibly in the regression
but there’s not an easy interpretation. So we do observe for subprime borrowers that
on average they get higher initial APRs but this is all coming from a decrease in zero
APR occurrences which at least according to the data occur 17 percent of the time I want
to say, something like that. Whereas for non-zero APR it’s actually if
anything a slight decrease. We’re talking 10 basis points so it’s not
large. One other thing that we do in the paper is
we try to get a little bit at the mechanism. So, so far all we’ve done is looking at the
creditor side but if there’s an effect on creditors unless they’re just anticipating
effects on the collectors something should happen to the collectors. It’s not easy to look at this because there’s
not data as much on collectors but we have some data in the CCP. Because one kind of data that is in the CCP
is collections trade lines. So this is where a debt collector is collecting
on a debt and they report that debt to the credit bureau. Now this data is not as good as for other
kinds of debt. Credit card trade lines are much more reliably
reported. But it gives us something to look at. So the outcomes we’re able to look at are
disputes. So there’s a flag in the bureau data for whether
the consumer had disputed the debt and also some indication of payments. Again payments are not recorded very reliably
but if we sort of combine all the different ways a payment could be recorded we get some
indication of whether or not the collections trade line was ever paid. One sort of nice thing about this is we get
an extra level of variation. The CCP for collections trade lines allows
us to distinguish between debt collectors and third party collectors versus buyers. And some of the specific laws we’re looking
at affected only buyers. So we can turn it from a difference and difference
into — it’s not quite a triple difference because it’s only some of the states, it’s
sort of like a two and a half difference. So what we find a small increase in the number
of credit bureau disputes. I think in our most recent results it’s not
statistically significant but the magnitude and the precision is not such that we could
say there’s no effect there. But we do feel fairly confident we can say
there’s no change in the likelihood of payments. We can reject a moderate decrease in the payment
rate. So that’s the paper as it is. The exact numbers you saw here are perhaps
subject to change. We’re actually in the process of updating
some things. But the basic message even in our more recent
results are not any different than that. So I turn it over to your thoughts. So I thought this was super interesting. I thought your results were great. You didn’t talk a lot about the specifics
of the changes they made and I was just curious whether — how I should think about them. Are they kind of changes that were viewed
as kind of moving states in line with the best practices that would be suggested by
people who worry about consumer protection or were they just small steps in the right
direction. Do you have a sense for that at all? Yes. I should have mentioned on the first or second
slide and forgot to so thank you. Yes. So one of the things that I think is actually
important about the laws we’re looking at is they were aimed at consumer protections
and things that would affect collectors’ costs, marginal costs in a way that actually would
matter. Whereas there’s one other paper or a couple
of other papers in the literature that are trying to look at this. They’re using a lot of changes in things like
debt collector licensing fees. So it’s going to affect fixed costs. It’s probably not going to change collector
practices at all. Certainly not from a consumer protection perspective. The laws that we’re looking at are all aimed
at consumer protection things. The Arkansas one just adds penalties for violations
of the FDCPA. The North Carolina one required a lot more
documentation before you could collect on a debt. So you had to make sure it was the right person
before you went forward. And it’s things of that nature. I don’t want to get too much into what the
right policy is given our role. Yes, of course. Another way to think about it is just were
they covered in the media as a big deal. Do you have any sense? I’m just trying to get at whether I should
think about these as big changes or not big changes or what. So according to our market folks who are — some
of whom are former debt collectors and talk with people who are in the market these are
things that the collectors were worried about and were going to change their practices. How much they actually did is a little hard
to know. It’s always a challenge. But yes as far as we know these are things
that the industry was expecting were going to have some kind of an effect. To varying degrees. There are some differences between the laws
and we’re being a little loose in lumping them all together. We do in the paper break them out by specific
law. There’s a little bit of heterogeneity though
I don’t remember if there’s much of a pattern into which one’s which. It may just be a power thing as you get down
to just one event. Harder to disentangle. It seems like the results maybe aren’t — they
have this nice consistency. There’s no impact on payments and since there’s
no impact on payments there’s really not much of an impact on credit availability. So that’s a nice result. I guess it kind of lends credence to the idea
that if debt collectors are harassing people and they’re like way up on the curve that
if you cut back on the margin it’s not going to have much of an effect because they were
already way over the top. You didn’t want to put an interpretation on
it but if I put that interpretation on it would that be misstating the results? Certainly possible. Certainly we know that there is a long tail
in the debt collection market. There are a lot of large players who are generally
well behaved and there’s a lot of small players who are very badly behaved. As is reflected by both the bureau and the
FTC going after some really shady characters. Then it’s a question of whether these laws
are affecting them or affecting the better players and I really don’t know that. But to push on where you’re going one reason
that we put this on the program is to think about the following question. So we now have the debt collection survey
which you guys heard about last year which looks at the consumer side of this, the 32
percent of folks that are facing the debt collection system, how often they may be called,
how they view that experience. There’s some evidence on firm costs. And now you have sort of effectively comparative
statics at least in these limited number of experiments from the diff in diff. So if we’re going to think about welfare analysis
which is where you were heading. I’m not ready to stipulate where we are on
the curve. Are there other pieces we need. And I’m thinking this is something where John
might get, are there ways to quantify, to monetize anything on the benefit side of this. So imagine you are up at the curve and imagine
that these laws actually cut back on some number of contacts of some sort how do we
as a research office figure out the value to consumer X of contact Y if we’re going
to try and do a welfare analysis. And the reason for the question is that we
have a statutory responsibility to be doing benefit cost analysis and try to push the
welfare calculations as far as we can. I don’t know if I have a good answer to that
but I was struck when you cited the statistics that 32 percent have been contacted about
a debt in collection. I don’t know exactly what the question was,
but I had my first experience with being contacted by a debt collector in the last few months
not for a debt that I didn’t pay but for a debt that a niece of my husband’s did not
pay. And they were trying to track her down. And we probably got I don’t know 30 or 40
phone calls from more than one, I don’t know how many more than one but we could track
the phone numbers on the called ID from multiple debt collection agencies trying to track down
this niece who had never been to our house and never had our phone number. Hadn’t even come within states of our house. Lived on the other side of the country. So if you’re trying to think about the benefits
I would think expansively. They might be more than just the benefits
to the person who has the debt. At least it wasn’t your 21-year-old son that
they were looking after. John, what’s the question that we ask Brigitte
for how much she would pay not to be contacted. What I was going to say on this is that there’s
some work that you did and then I and my coauthors have done some extensions developing in these
measures of consumer financial well-being. In our work we show there’s a distinction
between measures of perceptions of future financial security versus current money management
stress. So this latter one is the one that I think
you could — there’s a small set of items you could actually use as a dependent variable
but it would require new data collection. I have a question then a comment. My question first is for Brigitte. Did you file a complaint with the bureau? The answer is yes I did. As a matter of fact there was one phone call
where the person on the other end was very belligerent and wouldn’t leave me alone until
I said stop calling me, I am not the person you are trying to find, that individual does
not live here and I am going to file a complaint with the Consumer Protection Bureau. And she tried to argue with me and I said
I know my rights. I’m filing a complaint. And I did. That was not a setup by the way. Second, Ron, just on the firm costs, you mentioned
something about firm costs. I don’t know if you want to talk about this
more. But from how this would affect, my understanding
is a lot of debt is sold on the secondary market. And I guess my question would be how much
of this secondary market debt is state-specific. I just don’t know the answer but if I’m thinking
about rights, this is all credit card debt and I can’t harass people in Arkansas in a
particular way anymore but Arkansas’s debt is mixed in with every other state I mean
that would just I imagine have sort of a minor effect on the total debt. So the fact that it then affects prices in
Arkansas. I was starting to think through the whole
market equilibrium on this one. But you mentioned you had information on firm
costs. I don’t know if resale market, how available
this is. I’ve seen recently in the media that anyone
can buy this debt. So I just didn’t know. Yes. So there are —
so your question was — so there are sites out there with resale prices and I believe
we put our markets division that focuses on the debt collection market put out a report
that had looked at one of these markets. Partly looking at what information is being
sold. But also I think looking at prices. You’re interested in state variation. State variation is an important thing. Again I’m just parroting what our markets
people told me. Because there’s so much variation in state
law I believe things are kept separate because the collector has to follow state law or they’ll
get in a lot of trouble. And a lot of states have some level of regulation. Their regulation may just be restating the
Fair Debt Collection Practices Act, the federal one sometimes with no additional penalties
which really just allows a plaintiff to bring a case in that state. But yes, there definitely is state specific. To the point that New York I believe did something
that made it really, really hard to collect debts in New York either for a while. Certainly it hasn’t been for very long. I’m not sure if it’s still ongoing. To the point that collectors were just not
outsourcing debt in New York for a while. It might still be ongoing. I don’t remember if they’ve changed the law
to address that or not. It’s interesting. I don’t think we’ve looked at some kind of
delta on prices. But that still gets you at — it’ll get you
the expected value of the collection presumably and what somebody will pay for it. So that’s the transfer. And these guys still give me the comparative
statics on price and availability of credit and the access to credit issues. I like John’s suggestion of using even potentially
our own scale as modified by Lynch et al to look at distress and whether that might change
in the presence of particular behaviors. Is there any way to push even further. You can see where I was going with this. The consumption question earlier, the question
here. One of the things that I think that we’ve
been struggling with, not struggling with but thinking hard about is measurement of
benefits in the work that we do in the consumer protection context. This may be an ongoing conversation with the
Academic Research Council as we try to build out these methodologies going forward and
these just happen to be two particular cases, you know, can you tell me how much utility
is greater for folks with access to certain products. Now with access to certain products can you
tell me the benefits of removing what may be considered a bad practice in the
market. I don’t know if you have — we’re getting
close to the end of our time. We hadn’t planned on this but since we have
three minutes closing thoughts from any of the ARC members? I just threw you on the spot because I didn’t
tell you this was going to happen. My closing thought is I think that the research
office has really done a fantastic job of kind of creating itself from scratch over
the past several years and building up an amazing capacity to help us better understand
what’s going on in consumer markets and the market trends and understanding consumer behavior
and improving disclosure. I think we know a lot of things today that
we didn’t know 10 years ago because you’ve got a good research team asking good questions
and putting together the data sets to help do a better job in regulating these markets. So kudos to you and keep up the good work. So I second what Brigitte said. And I think continuing what has been done
collecting these data, creating new data sources and then researching these many questions
I think is fantastic. One thing that may be follow-up from the earlier
session is there was this question about the new quarterly trends that are out there and
part of it was finding what people are talking about and then putting out trends to try to
bolster the public discussion which is I think great. I think also just continuing some of these
may also be very useful over time. That is you have the mortgage — some of the
trends we saw going through this year a repeat in maybe every other year or something like
that. It may not be new and sexy anymore but still
just understanding those longer term trends and how they’re evolving may be very helpful
just to keep these patterns up. So over time the broader knowledge in the
public about consumer financial affairs. And I would say one of the things that’s really
striking is I really like the year to year continuity. This last paper is an example of in a prior
project you look at the consumer side or the demand side and then this paper looking at
the supply side. That’s an incredible strength. I really admire that. Thank you for inviting me. I really am impressed by the good rigorous
analysis that you all are doing. There’s really no substitute. And I think about where we were 12-15 years
ago when we just had kind of anecdotal speculative evidence to think about. And you guys have just managed to answer so
many questions. And I look forward to seeing where you go
in the future. And I also just wanted to bring up something
that I forgot to mention earlier which is as I was reading the materials you gave us
in preparation for this meeting big praise to all for making it so accessible. This is complicated stuff. It seems like you’ve just put a lot of effort
into making these studies easy to read for people who are not deep experts in these areas. And just making terrific use of kind of trying
to illustrate things graphically. So appreciate that. So I was waiting for — unfortunately Tom
Pahl wasn’t able to get here so I will close this out. Many thanks to all of you actually for reading
the materials, for giving us the feedback, for your feedback today at this session. As we described we’ve built some of our processes. I think you asked earlier how we developed
the QCCTs and what the topics are. And we have been working on that for the two
research agendas we showed you. We’ve now built a much more formal process,
an almost internal commissioning of which you guys are all now a part. So you can help us pick our projects, help
us direct the data resources that we have. Let me take another minute to thank all the
presenters who put together fantastic slides and fantastic presentations in both panels. It was yeoman’s work on a short time frame
so I’m very grateful to them. I’m very grateful to the ARC members and look
forward to continuing to interact with all of you. And with that I think we’ll call ourselves
adjourned for the day pretty much on time. So thank you all very, very much.

Leave a Reply

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