hybrid suvs pros and cons

let me welcome you to the first cfpb researchconference. our director, rich cordray, is going to provide some opening remarks, andi'm going to introduce him now. prior to his current role as the cfpb's firstdirector, director cordray led the cfpb's enforcement office. before that, he servedon the front lines of consumer protection as ohio's attorney general. in this role,he recovered more than $2 billion for ohio's retirees, investors, and business owners,and took major steps to help protect its consumers from fraudulent foreclosures and financialpredators. before serving as attorney general, he also served as an ohio state representative,ohio treasurer, and franklin county treasurer. so, director cordray.

thank you, chris. it feels like we alreadyheard most of the things about myself. i'll just say about my tenure as ohio treasurer,i served as the treasurer in charge of safeguarding billions of public dollars, of public funds,in 2007 and 2008, so that was kind of a unique experience for me and for anyone in that position. i want to start by thanking chris and ronand david silberman and all of our office of research team for the work they have doneto prepare for this first research conference of the consumer financial protection bureau.this is a big day for us, a very good day, today and tomorrow, and our folks have beenlooking forward to this for some time, and really striving to give you all an opportunityto benefit by understanding work that we're

doing and sharing your work as well with thelarger community across the country that's deeply interested in these issues. so thankyou all for being here, for taking part, and for engaging in what we hope will be a richdialog about the latest research in consumer finance. to recap our brief history, congress establishedthe consumer bureau in the wake of the recent financial crisis as a new agency with a broadset of tools designed to address problems that arise in the realm of consumer finance.our tools include rulemaking, supervision, enforcement, consumer complaint response,consumer education, and i would include publication and disclosure of information. our authorityfocuses on household financial products and

service including mortgages, credit cards,student loans, auto loans, bank account products, debt collection, credit reporting, paydayloans, prepaid cards, prepaid accounts, international money transfers, domestic money transfers,consumer loans, and more. speaking from my own vantage point—and iwas talking with members of the arc, who are much more steeped in the field than i am—itdoes seem to me that the subject of consumer finance seems to be broadly under-appreciated.in the broader field of economics, most attention is devoted to either macroeconomics or tomicroeconomic analysis of how businesses operate. far less attention is devoted to analyzinghow households operate. that's often taken for granted in economic models and by economists.yet in every macroeconomic model that exists

today for the u.s. economy, consumer spendingmakes up about 70 percent of all economic activity. most of that activity depends onconsumers having access to practical and sustainable methods of household financing. quite simply,we can no long afford to take this important economic activity for granted. one of thosemethods of financing, mortgage financing of home purchases, surprised many economistsby causing the worst financial crisis of our time. these are compelling reasons why thefield of consumer finance deserves considerably more attention. i realize that i’m preaching to the choirhere but i think it’s important to say. consider also that this topic has become increasinglyinteresting in recent years. in the last two

generations, the markets for household creditand consumer finance have expanded to become much more complex for the average americanfamily. the mortgage market today stands at nearly $10 trillion, student loan debt hasgreatly accelerated in recent years to reach about $1.3 trillion, auto loan debt now exceeds$900 million, and credit card debt totals approximately $700 billion. large numbersof americans utilize small dollar loans and other consumer loans, and as we’ve beenemerging from the recession it is estimated that over 30 million americans have debtsin collection. as the volume of consumer finance has grown,so has the variety of financial products and services, with remarkable growth over thepast half century since i was born. mortgages

then were simpler and almost always were offeredas 30-year fixed products. students loans and general purpose credit cards were embryonic,at best. auto loans existed on a much smaller scale, in part because cars themselves wereso much less expensive in real dollar terms. most of the other products just mentionedwere either nonexistent or, at most, a marginal part of household finance. this dramatic expansionof nuanced and more complex financial products has been fueled by the rise of the creditreporting industry, with the three nationwide credit reporting companies each compilingcredit files now on over 200 million americans. at the same time, the other predominant householdfinancial issue of saving for retirement was once a much simpler affair, without the wealthof products now available. in that era, many

working americans could rely on a definedbenefit pension, offered through a single employer or through a multi-employer plan.think about that for a minute and consider how dramatically the world has changed. youngpeople now can expect to have more than a dozen jobs over the course of a lifetime.at most, those jobs are likely to offer a defined contribution plan or some other meansof personal retirement savings, especially in the expanded world of non-salaried, non-full-time,non-employer-employee work that is temporary or seasonal or contracted or intermittentor part-time in nature. one conclusion that flows from these observationsis that the field of consumer finance is not only central to understanding the entire economicpicture but it is increasingly complex and

interesting. where credit is available topeople, it affords opportunity. it permits forward-looking investments. it creates theability to time shift one’s access to available and expected resources. it allows people toplan ahead in ways that are not available to those more constrained by the straightenedlimitations of present circumstance. it allows people more latitude to leverage current assetsagainst prospects for the future. it literally affords the potential for people to lay deliberateplans for how to change their lives for the better. in a recent article in the washington post,tom durkin and todd zywicki accurately described the positive role of consumer finance thisway. i quote, “the growth of consumer credit

in the postwar era has been a tremendous boonto american families and the american economy. credit has expanded access to the conveniencesof modern life that used to be reserved to the few—cars, washing machines, and refrigerators.in turn, access to those products has reshaped the way we live and work.” yet, as we saw in the recent mortgage crisis,and in several other areas, consumer financial markets are not fool-proof. they do not alwaysfunction in a manner that serves either families or the economy, and when they slide off thetrack it can take considerable time before functional equilibrium can once again be restored. as mentioned a moment ago, the large creditreporting companies exist today in part as

cause and in part as consequence of this greatexpansion of household credit. credit reporting was uncommon and dispersed 50 years ago, whereasa substantial majority of americans today have a credit record that is accessible tomost financial institutions. credit bureaus, in turn, have facilitated the consumer creditboom by permitting underwriting and risk assessment at a more sophisticated level than ever before.further, they have become independently influential actors in their own right and are now affectingconsumers in direct and immediate ways. more and more, the information they gather andthe models they use can influence whether a particular consumer can get credit at all,or on what terms, and at what price. they also now affect people’s prospects for employment,rental housing, and even insurance, with credit

checks growing more common as part of thebackground checks that may factor into those decisions or, in the case of insurance, intothe models that are used for underwriting purposes. so for all these reasons i would submit thatthe role of consumer finance should be elevated in the field of economics, because of itsgrowing interest, growing complexity, and growing importance. and this is true not onlyas an academic or research subject but also as a basis for directing policy-making, thatcan make an enormous difference in the everyday life of people in this country and, indeed,around the world. with that discussion in mind, let us considerthe important role that the consumer financial

protection bureau can play in american life.the historic financial reform law that congress adopted just 5 years ago created this newagency and positioned it at the center of those issues. much creative and interestingwork in the field of consumer finance is now underway at the bureau and it’s being influencedby its singular focus on these issues—its authority to develop and analyze new datasets, and its ability to conduct innovative new research and policy-making that affectsdaily life in virtually every household in america. we’re please to see so much thoughtfulresearch on the agenda discussed earlier at the council meeting today, and also at this,our very first research conference over the rest of today and tomorrow.

as all of you here know, congress createdthe consumer bureau’s office of research as an integral part of the financial reformlaw. the office of research designs and is responsible for conducting foundation, policy-relevant,quantitative research on consumer finance and household behavior. it is helping us developnew data streams to repair blind spots from before, such as where the federal reservehas repeatedly noted. it had an enormous blind spot about the mortgage market’s evolutionand possible effects. the office comprises an interdisciplinary group of economics andresearch psychologists who have joined us from other government agencies, from academia,and from the private sector. we’re very proud of our office of research and we believewe’re continuing to assemble a staff that

is world class. you can test them, probe themfor yourselves today and tomorrow. i was going to say poke and prod them but maybe that’sgoing a bit too far. when we first set about designing the bureaunearly 5 years ago, we deliberately integrated our research economists and research scientistswith our regulation attorneys and market analysts. in this way, the office of research supportsboth formative policy analysis, as we develop our rules, and the evaluation of our rulesthrough benefit cost analysis, and eventually retrospective rule review. our research expertsalso provide analytical support for our fair lending and compliance missions, as well asour consumer engagement team and our project catalyst efforts, to foster innovation inconsumer financial products and service.

we intend for our research initiatives todrive the policy dialog across the entire landscape of consumer finance. our researchersshare their findings broadly through papers, presentations, and reports, as will occurat this conference. one of their newest innovations is the period publication of focused researchon these issue in smaller, bite-sized quantities that we call data points. these pithy publicationsare intended to stimulate debate and discussion, while furthering our objective of providingan evidence-based perspective on consumer financial markets, consumer behavior, andregulations to inform the public discourse. at the council meeting earlier today, we hearda presentation about the most recent cfpb data point on what we call “credit invisibles,”or consumers with limited or nonexistent credit

histories. it reports, for the first time,that 26 million americans do not have a credit record, and we were quantifying specificallythis, and an additional 19 million have a credit record that cannot be scored usingone of the most common commercially available credit scoring models. this data point offersnew grounds for reflection on how well the credit reporting system is working, and forwhom. it augments prior influential work our researchers had already produced on the roleof medical debt in credit scores, on checking account activity among low- and moderate-incomehouseholds, on how the behavior of both consumers and firms is changing in response to regulatorychanges, and on other consumer finance issues ranging from student loans to arbitration.

as the consumer bureau continues to mature,we hope and intend that we will serve a central role in developing the field of householdfinance. i hope that many of you will come and work with us at some point during yourcareers. the conference today and tomorrow should serve as a critical step forward infulfilling these goals. thank you for all your submissions, your presentations, andyour involvement here today. thank you very much, director cordray. it’s a pleasure for me to be speaking toyou today. as the first chief economist at cfpb, i think my most important role has beento build bridges between our office of research and the academic community, and this conferenceis the most visible manifestation of that.

let me begin my talk by just briefly describingresearch at the cfpb and the role of our researchers in policy-making. even before cfpb came intoexistence it was envisioned as a 21st-century government agency that would engage in evidence-basedpolicy-making. our mission and vision statements say that we will “use data purposefullyto enable informed decision-making in all internal and external functions.” we’vetaken that very seriously. everyone in this room, though, knows that even the best datain the world can’t speak for themselves. the tools that economists and other researchershave developed for distinguishing cause from effect, for teasing out reasons from correlations,for understanding how individual decisions result in aggregate outcomes—these toolsare used every day and for every policy decision

that we address. the extent to which these tools are vitalto the effective performance of our mission is reflected in the fact that researchersfrom the office of research are involved, from the very beginning, in every importantpolicy process that goes on at the bureau. we are deeply involved in everything thatthe bureau does, but we’re not alone. the rulemaking part of the cfpb is called, asrich said a minute ago, the division of research, markets, and regulations, or rmr for short.actually, the lawyers and the economists might disagree about which r is first in research,markets, and regulation, but one thing that we both agree on is that we are improved bythe presence of the m in the rmr. we have

a group of markets people whose careers priorto joining the cfpb had been as participants in the industries that we regulate, and neitherthe researchers nor the lawyers, necessarily, would be able to bring that perspective tothe policy-making process. so we have, as we work on policies, a diversityof perspectives. the extraordinary, deep knowledge that regulatory the regulatory team has ofthe legal framework and legal precedence, and the sort of ways in which courts willinterpret our regulations is a vital part, but not the only part of what goes into theformulation of the regulations. we have a phd staff that includes not just economistsbut people from psychology, people from decision scientists. we have expertise in a very broadrange of topics that all have a vital contribution

to make to our ability to understand the complexitiesof household decision-making and the ways that those decisions add up to market outcomes. likewise, the scientific committee that helpedus choose the papers for this conference is a diverse set of people with expertise ina very broad range of questions. the fact that as we make our policy decisions, we researchersare constantly having to interact with and come to a meeting of minds with people withsuch different backgrounds—with the lawyers from the regulations group, with the peoplefrom the markets group—is not unrelated to the fact that we have taken a very broadview of the kinds of research that are useful for us to see at this conference. the generalperspective, the general point is that diversity

in ways of thinking, diversity in frameworksfor trying to understand the world, is not just desirable but essential for our goalof making good policy and our goal of understanding that complicated world that we live in. so we need all of that knowledge, all of thatwisdom, all of that diversity in perspective if we’re going to be successful at definingwhat it means to be a 21st-century government agency—and so far it’s going well. weare thrilled with the quality and quantity of submissions that we received for this conference,and we hope that you will be similarly impressed by the quality of the program that resulted. with that as the introduction, i am goingto turn things over now to chuck romeo, who

is the moderator for the first of our panelson shocks. our first three panelists are ben keys, universityof chicago; clinton key, pew charitable trusts; and john mondragon, university of california,berkeley. each speaker will have approximately 20 minutes. alex is our timekeeper. then mycolleague, jon lanning, will provide discussion remarks. the authors will get to respond tojon’s comments, and then the floor will be opened to questions for 15 to 20 minutes.thank you. good afternoon, everyone. my name is ben keysfrom the university of chicago, harris school of public policy. i’m thrilled to be heretoday. i want to thank the organizers for including this paper on what really appearsto be a fascinating 2 days of research. this

is joint work with amit seru, tom piskorskiand vincent yao. vincent is here in the audience today, from fannie mae, and so the usual disclaimerapplies that what we talk about is our views alone and not those of fannie mae. this research is really motivated by the big-picturequestion of what was the broader impact on households of the period in the wake of thefinancial crisis, when interest rates fell to extremely low levels and have sort of remainedthere thereafter. what we’re trying to get at in this paper is how did that factor inthrough household balance sheets and the real economy, so the title is intended to be asdescriptive as possible. over this period we saw these extraordinaryefforts to keep interest rates low and to

keep mortgage rates low, in particular, whenwe’re thinking about the transmission of monetary policy and a really important placeto look for that at the household level is the housing market. so for most households,the house is their largest asset, their mortgage is their largest liability, and when we’rethinking about the impacts of low interest rates we would think that the rate on themortgage is going to be especially crucial in terms of the household’s debt servicecosts. and in the context of this project, we’re going to be focused on adjustablerate mortgages, or arms, where we sort of can imagine that as a friction-less pass-throughof lower interest rates. so when monetary policy ramps up and interest rates fall, thoselow rates are passed through directly to arm

borrowers without any effort whatsoever. weheard in some of the morning conversations about the friction surrounding refinancingor inattentive refinancing. for adjustable rate mortgage holders there’s no frictionswhatsoever, regardless of whether they’re paying attention or not, or whether they’reunder water, so whether they need to qualify for a refinancing. instead, an adjustablerate mortgage reset is automatic. one of the challenges that i think comes upwhen thinking about the impact of interest rates on households is the identificationchallenge, and so if we’re thinking about a cross-section of homeowners who may havedifferent mortgage rates, while we are concerned that there are other aspects of those borrowersthat led them to receive a lower rate versus

the higher rate, similarly looking acrosstime we would worry the same thing, that individuals at a time period would take out a loan whenrates are low, may be very different than those who take out a loan when rates are high.so what we’re going to do this paper is try to disentangle those challenges in a differentcontext, focus on a hybrid arm—so these are adjustable rate mortgages that reset atdifferent dates. i’m going to show you the details of that in my remaining time, andthen we’re going to talk about a broader macro approach, if time permits. what we’re going to do in this paper issort of micro and a macro blend, where we’re going to look at both the impact on householdbalance sheets, using this unique data that’s

linked from one of the large mortgage marketparticipants, linked to credit records, and then we’re going to look at the impact ona broader economy. as well, we’re going to look at variation in exposure to interestrate regimes, based on the fraction of adjustable rate mortgages that are in that region ofthe country. i think this project really speaks to therelative magnitudes in terms of debt deleveraging hindering consumptions. this was a topic duringthe crisis—why aren’t consumers going out and consuming more now that the timesare getting better? one hypothesis is that, well, there’s this debt overhang or thisdebt deleveraging that consumers are facing and they’re paying that down before theycan go out and consume. i think we’re going

to find some pretty clear evidence of that,especially in the credit card context. but also getting back to this issue of automaticpass-through of interest rate policy and thinking about the successes or the limitations ofsome of the mortgage modifications and refinancing programs. i think what’s unique about lookingat adjustable rate mortgages is we see that this is a case in which all of these frictionswere removed, so we can think about this as sort of an upper bound. now we’re not going to be able to speakto the broader, general equilibrium consequences of lower mortgage rates. we’re not goingto be able to speak to the impact on investors who may be getting a smaller return. we’regoing to be focused on homeowners, and we’re

only going to be able to speak to this oneperiod of time. we wouldn’t want to assume that things will be necessarily symmetric,say, on the upside, when rates eventually go back up. with those caveats out of the way, i thinkour paper really links in nicely to a larger literature, some of which chris carroll hascontributed to, thinking about how households respond in terms of consumption to incomeshocks and thinking about the sort of income transfers through fiscal policy. i think there’sbeen quite a bit of work done on that dimension. you can think about what we’re doing asthinking about a change in mortgage service costs through changes in monetary policy.so much as people have tried to trace through

where the tax rebate check ends up after ithits people’s wallets or their bank accounts, we’re doing the same thing when they receivethis additional liquidity from lower mortgage payments. i also wanted to make sure to mention a relatedpaper that uses some similar identification with a different set of mortgages and findsbroadly consistent results as sort of complementary results. that’s written by marco di maggio,who is here today, along with his co-authors, which i think is a really nice complementand related work. our approach is going to be, as i said before,it’s difficult to disentangle these issues in just a cross-section or just a time series,so what we’re going to do is try to exploit

variation in adjustable rate mortgage contracttypes across households, and basically focus in not on the difference between fixed-rateborrowers and adjustable-rate borrowers but instead, within the group of people who aretaking out an adjustable rate mortgage, some of those folks take out a loan where it’sfixed for 5 years, others take out a loan where it’s fixed for 7 years. i’ll showyou that they are very, very similar on a range of observable characteristics, and we’vedone quite a bit in the robustness that i won’t have time to go into, to try to matchon a range of characteristics, including prepayment propensities, and look at variation acrossthose households in terms of their differential exposure to interest rates. and then, at theregional level, we’ll exploit variation

in the distribution of contract types acrosssimilar reasons. the micro data, which i mentioned brieflybefore, is coming from mortgage performance data from a large mortgage market participant,and that data is matched by social security number at the monthly level to individualcredit records. we have essentially a 100 percent match rate of this data and we knowquite a bit about these households’ balance sheets, so while previous research may havejust been able to focus on the mortgage side, we’re able to tie in the entire remainderof the household balance sheet. we have more than 350,000 borrowers and theseare agency borrowers, so these are people you can think of as being prime borrowers,much more comparable to the fixed-rate mortgage

market than, say, the sub-prime market, and350,000 of just these two types, of 5/1 arms and 7/1 arms that i’m going to describein just a moment. so here’s our sort of stylized version ofwhat’s going on and what we’re thinking about. think of two borrowers who are takingout a loan in early 2005. one of them takes out a 5/1 arm, the other one takes out a 7/1arm. what does that mean? well, it means that one borrower’s rate is fixed for 5 years,the other borrower is fixed for 7 years. you can see what happens here. there’s a pointafter 5 years where the 5/1 arm borrower’s rate drops dramatically—again, in the contextof this recent period, that wouldn’t necessarily have happened had interest rates gone a differentway. they divergent for these two years, and

then the 7/1 borrower rate drops as well andconverges at this point. what we’re going to be focused on in thispaper is really this 2-year period of differential exposure to lower interest rates. this periodhere is going to be our focus, this 2-year window. we’re going to look at loans thatwere originated across different points in time over this period and they’re goingto have different magnitudes in terms of the treatment. we’ve also looked at just 5/1arm borrowers who took out loans at different points in time, around their different resetdates as well. here’s what this looks like in the data.here we see, very similar to the stylized version i just drew you, the step function,where the rate resets after 5 years for the

5/1 borrower and then again after 6 years,and, on average, rates fell quite dramatically over this time period, by well over 100 basispoints, and that translated to a significant savings in terms of mortgage debt servicecosts. here’s in the picture in dollar terms, and once we quantify this in a simple difference-indifferenceframework, this is an average of $3,400 over these 2 years. so this is a really dramaticdifference, and when we think about comparing this to a tax rebate check, that’s a one-timecheck, which is generally not nearly this large. this is quite a large transfer andit continues to come with each month that you remain in the home. we can relate this change to mortgage defaults,and we observe that this is associated with

a relative decrease in mortgage defaults ofabout 36 percent, which is an extremely sizeable difference, and again, isn’t surprisingthat when you lower people’s mortgage payments they’re more able to make those payments. there’s a key identification assumptionwhich i need to get out there as well, which is that what we’re assuming throughout isthat in the absence of this rate reset that the outcomes of the 5/1 arm borrowers wouldhave behaved in the same way as the 7/1 borrowers, that there would’ve been these paralleltrends, and we can do quite a bit to try to push on that assumption and to look for paralleltrends in the pre-periods, as well as doing some re-weighting efforts, which are in thepaper.

the advantage of this data set that is linkedto credit records allows us to go beyond just the mortgage default outcome and to show youthings like what’s happening on the rest of the debt portfolio, and, in particular,the credit card picture may be most striking. there is a difference in average credit cardbalances prior to the first rate reset, which is the dotted line here, but they’re followingquite similar trends across the 5/1 group and the 7/1 group. at month 60 is that sharpdiscontinuity in the 5/1 borrower’s mortgage debt service costs. we don’t observe a sharpdiscontinuity in their debts. we observe, actually, if anything, a slight increase intheir debts for a few months, and then a sharp convergence over this time period.

this was actually a surprise to us in thatwe thought that homeowners might be anticipating this type of change. this is something that’squite foreseeable when you take out a 5/1 loan. you know exactly 60 months from todaymy rate is going to reset. we don’t observe much in the way of changes in behavior thatare occurring in a sharp way around this discontinuity. what we instead observe is sort of a gradualbuild-up of improvements or liquidity for these households. over this time period weestimate that about 20 percent of the additional liquidity that’s coming from lower mortgagepayments is going towards paying down credit card debts. what i’m going to show you injust a minute is that 20 percent number actually masks a lot of heterogeneity and that householdsthat were buried under a considerable amount

of credit card debt actually used a vast amountof their new-found liquidity to pay off credit card debt, which we can think of as sort oflimiting their ability to consume out of this new liquidity. here’s the picture that’s sort of on theflip side for auto debt, and what we observe is actually an increase in auto debt among5/1 arm borrowers. so the borrowers who have this additional liquidity, $3,400 of additionalliquidity, are actually using some of that as a down payment to go and buy cars. now,this picture is only for cars that are associated with a loan, but what we’re observing isnew auto loans—so not a new car but a new loan; it can be on a used car, as well—andthis really appears to diverge after the second

rate reset, so after households have beenreceiving this additional liquidity for a number of months. perhaps they’ve savedup a down payment. they’re actually going out and buying cars, and what we show is thatthis divergence in auto debt is essentially entirely driven by an increased propensityto take out a new car loan. so this is new car financing, and we see that this is reallythe big driving force of that difference. so households are rebalancing their portfolioaway from credit card debt and towards cars, but those are not necessarily the same householdswho are doing one or the other. to get at that issue, here’s the point thati just raised before, about how more liquidity-constrained borrowers are responding. if you look at justthe top quartile of credit card utilization

or the bottom quartile of credit scores, roughlytwo-thirds of the additional liquidity that they receive in the first year from theselower mortgage payments is going towards paying off their credit card debt. this makes sense.this is likely their highest interest rate, that at least we observe on the credit record.so what this means is that if we’re thinking about the marginal propensity to consume outof this new-found liquidity, we can sort of put a cap on that, that they can’t be consumingmore than about one-third of each dollar that they’re receiving, which i think is quiteclear evidence that these households are focusing on repaying debts over this time period, andthat could have been limiting the consumption response from this type of reaction to monetarypolicy.

what we also observe, and i wish i had moretime to share all the heterogeneity results with you but i just don’t, is that the wealth-constrainedhouseholds, those who are under water on their mortgages but who don’t have al to of creditcard debt, are the ones who are most likely to respond by purchasing cars. so there’sa consistency with some of the prior literature on tax rebates and other rebates that thelower-wealth households are the most likely to go out and spend this money, particularlyon the auto side, but not if they’re paying down a significant amount of credit card debt. i think that’s sort of the summary of themicro evidence, is that there’s a complicated interaction between wealth and liquidity constraints,that’s important to take into consideration

when we’re thinking about how we would expecthouseholds to respond to this type of reduction in debt service costs. in my remaining time, let me turn to the regionalanalysis, and what we’re going to do here is, just as we were looking at differentialexposure to interests rates across households, we’re going to now look at differentialexposure to interest rates across zip codes. what we’re going to do is first look ata sample of parts of the country with a high usage of adjustable rate mortgages. so we’renot going to just use the 5/1 arms and the 7/1 arms. we’re going to use all forms thathave an adjustable rate feature, and we’re going to use the arm share as of the secondquarter of 2007, which we’re going to show

predicts the treatment intensity, i think,quite well over the period, of how exposed they are to these interest rate changes. andthen we’re going to try to construct a sample of similar zip codes. now, of course, it’s the case that zip codesthat have more adjustable rate mortgages are going to differ on a range of other characteristics.we’re going to match on as many observables and pre-trends as we possibly can. we alsotried to match within various geographic regions, where possible. we also have an exercise inthe paper where we run an instrumental variables approach that uses all zip codes. so, again,we’re going to make a similar kind of parallel trends assumption here, which you can agreeor disagree with. maybe our discussant will

choose to disagree. we’ll see. he only hasmaybe 2-1/2 minutes with which to quibble. and then we’re going to investigate theimpact on a broader set of economic outcomes that we couldn’t get at with the creditrecords, so we’re going to look at house prices, a broader measure of auto sales—notjust those that are purchased with a loan—and then actually some employment effects as well. here’s the basic idea. the treatment linehere is the average mortgage rate in the 500 zip codes in the country with the largestarm shares, and the control group is matched as of 2007 q2, based on observable characteristicsin those zip codes, to have a very similar trend in interest rates prior, along withall the observables that we have at our disposal.

but these are less arm-intensive zip codes,so they’re going to be less exposed when interest rates fall, fewer loans are goingto reset, and the magnitude of the first stage is actually quite similar to what we foundin the micro evidence, which is that if arm share was 100 percent, we would observe adecrease in mortgage rates of 175 basis points. so what did we find? what’s the impact ofthis difference in—let me just go back for one second. this difference in mortgage rates could bedriven by a few different things. it could be driven by new buyers coming into the zipcode and taking out a loan at a lower rate, or it could be active refinancing of existingfixed-rate mortgages, and in the paper we

try to decompose this and actually show thatthis is really is an arm-share effect, rather than these other potential drivers, of whypeople in a zip code would have a lower mortgage rate. all right. so while my voice hangs in forat least a couple more minutes, this is house price growth over the time period, and wesee that house prices grow faster in the places that are more exposed to interest rate declines,and that this is really being driven by a reduction in foreclosures and foreclosure-relatedforced sales. here’s the auto growth, so here’s allauto sales in the a zip code, not just those that are being purchased through a loan, andwe see a faster growth in auto sales over

this time period, again, in the group of zipcodes that are most exposed to lower interest rates, to interest rate policy through adjustablerate mortgages. and then the employment response, so we observefaster employment growth in the treated zip codes as well. and what i can show you inthis next—i don’t have it in this table here; i thought i did—is that all of theemployment response is actually coming from the non-tradable sector, and, in particular,coming from restaurants and grocery stores. so our interpretation of this is that as householdsare having to put less towards their mortgage, not only are they paying down their creditcards, not only are they going out and using this as a down payment for an auto loan, butsome of that money is also coming back to

the local community in the form of purchasesat very local establishments, like the grocery store or the local restaurant. so we see noeffect in the tradable sector, where we wouldn’t expect to see an effect of interest rates. all right. so i’ve just thrown a lot atyou, and i lost my voice midway through, so let’s pause for a second to regroup anddigest some of this. these low interest rates policies had a quite meaningful impact onhousehold spending and in the broader economy, if you’re willing to buy our approach ofmatching zip codes in this way, and i think this supports the view that understandingshocks to household balance sheets are really important in understanding some of the otherbroader drivers of the economy, such as employment.

i think it’s an open question whether we’llsee a reversal when the stimulus is withdrawn, when we see rates go back up, and they willgo up at some point, some day, my former colleagues at the fed tell me. if you want to scale up our estimates moregenerally, i think they suggest that a 20 percent relative reduction in average mortgagerates in a region result in a 3.5 percent increase in the annual house price growthrate, a 5 percent increase in the annual auto purchase growth rate, and a 3 percent increasein the non-tradable employment growth rates. so these are not economically insignificantnumbers. and i mentioned this before. i think it’s important to say again that we reallycan’t quantify the overall impact of monetary

policy or qe or any of these kind of thingsusing this type of analysis. this is very much a partial estimate in a difference-indifferencesetting, so we don’t want to think about the impact on firms and others, we don’twant to think about the impact on exchange rates, and we don’t have data here on investors.so investors are obviously going to be hurt when individuals are making smaller debt servicepayments. we can make assumptions about their marginal propensity to consumer but we can’tspeak to that in this project. so just to wrap up, i think we find evidencethat household debt deleveraging significantly limited the ability to stimulate consumptionduring this period. for those households who had significant credit card debt burdens theyput a lot of their new-found liquidity towards

paying down those credit cards. i think oneof the things that this points us towards is this question of the high cost of creditcard debt, that credit card interest rates remained extremely high during the crisis,and there isn’t a natural or straightforward way to refinance credit card debt, althoughi think one of the unique things about social lending, in the kinds of propser.com typesof settings, is that the contracts that people set up on those types of websites are oftento consolidate and refinance credit card debt. so there may be ways in which social lendingis trying to take a step in this direction. one of the other conclusions of this workis that adjustable rate mortgage contracts make the transmission of monetary policy extremelysimple and straightforward. we avoid the institutional

frictions and the active decision-making that’srequired with refinancing, along with some of the other barriers, like servicer participationand other limitations of some of the other federal programs for modifications in refinancing.and so i think the adjustable rate mortgages are a nice feature in this regard, and whenwe’re thinking about international comparisons, which were discussed this morning, sort ofthe pros and cons of having a system that’s more reliant on arms versus more reliant onfixed-rate contracts, this is one advantage of adjustable rate mortgages, which again,there’s an asymmetric effect on the upside but we want to take into account the positivebenefits of adjustable rate mortgage contracts in the wake of a financial crisis.

thank you. thanks for having me. i think i was criminallyoptimistic in planning my slides, so we’ll see. this might be a bit of a sprint. the name of my paper is “household creditand employment in the great recession,” and the basic question that i’m trying tograpple with, going to try to give some answers about, is the following—how much did thecontraction in the supply of credit to households contribute to the decline in employment duringthe great recession? the basic intuition of this mechanism is that households in the lead-upto the crisis were using debt in part to fund their expenditures and that the financialcrisis led to a contraction in the supply

of debt to households, and this may have affectedhow much they could spend. so firms seeing lower demand may have responded by cuttingemployment. it helps to have a little bit of context inthinking about this in accounting for the great recession. one of the first shock that’sgotten a lot of attention was the collapse in house prices, with the idea that as houseprices fell this destroyed the collateral value of households, and also made them feelpoorer. as this happened, they had to reduce their expenditures and firms may have respondedto that. what’s important here to distinguish itfrom the channel i’m talking about is that this about the direct effect of house priceson household balance sheet, whereas i’m

going to be focusing on kind of holding houseprices constant. as the financial crisis happened, the supply of credit change—how did thataffect the households? this is similar to the first credit channel, which as the financialcrisis happened it’s led to a contraction supply of credit to firms, who then had cutinvestment and employment. i’m going to be talking about the householdcredit channel, and there’s been work along this, a lot theoretical, showing that thisseems to potentially be an important channel over this period, and there’s been someempirical work saying that it looks like there was a contraction in the supply of creditto households. what i’m going to be focusing on are essentially two empirical questions.first, how responsive was employment to this

kind of shock—so to try and get some sortof aggregated estimate of, as households lost access to credit, how much did their expendituresseem to respond to that and then how much did employment respond to that fall in expenditures?and then, second, can we try and get an idea of how large the shock was to households?so even if employment is very responsive to this kind of shock, if the shock is very smallthen it’s not going to be an important channel. quickly, what i’m going to be doing, togive you an overview of the talk—and hopefully i’ll get to the end—is i’m going tobe taking a cross-sectional approach, so using variation across u.s. counties and try toidentify some exogenous variation in the supply of credit to households across counties. andfor that i’m going to exploit the collapse

of a large, what i’m going to argue, wasa previously healthy lender, which was wachovia. wachovia is going to be useful for me becausei’m going to argue that the reason wachovia became distressed was due to its purchaseof a large toxic lender, golden west financial, and so that as wachovia fell apart, this wasan exogenous shock that had nothing to do with the quality of borrowers in the countiesthat traditionally depended on wachovia. then what i’m going to show you is thatcounties that did depend on wachovia suffered more. they suffered larger declines in theflow of household credit, as well as retail expenditures, and they suffered larger lossesin employment, primarily in residential construction and local non-tradable, so in the parts ofthe employment that you think would be responding

to changes in household demand. i’m goingto argue that this exposure to wachovia is really about exposure to wachovia in the householdcredit market, so this is going to be two parts, in part arguing that it’s not abouta shock coming from, say, the boom and bust in sub-prime lending or in house prices, andit’s also not about wachovia’s lending in the firm credit market. so this reallyseems to be about wachovia’s role in the household credit market. in doing that, i’m going to give some estimatesof the elasticity of local employment to supply-driven contractions in household credit, and thisis going to come out to be quite large, about 0.3. so a 10 percent decline in my measureof household credit driven by supply-side

shocks causes about 3 percent decline in totalemployment, which is a very large elasticity, considering that household credit declinedby about 40 percent over this period. and then, finally, given the size of thatelasticity, we’re interesting in knowing, is there a way of figuring out if the sizeof the shock was large, and to do that i’m going to construct a measure of the shockto each county, where the idealized measure of a shock to a county is simply the weightedsum of lender-specific shocks in that county, and then make some assumptions about how wecan sum that shock up across counties. when you do this in a couple of different waysyou’re going to get that the direct effect of these shocks—that is the losses impliedby the shock within each county, summed across

all counties—is on the order to 30 to 60percent of what we observed, depending on what kind of estimate you take. so together this is all suggesting that shocksto household credit supply do move around household expenditures and employment by quitea lot, and that the shocks seem to have been potentially quite large. the first part is convincing you that thatwachovia is a useful source of variation for shocks to household credit supply. what ihave here is a transcript from a wachovia second quarter 2007 earnings call, and it’sa conversation between an analyst and wachovia’s ceo, kennedy thompson. the analyst says, “ken,i need to ask this question. knowing what

you know now about the golden west deal andits impact on your stock market, would you still do the golden west deal?” ceos are always optimistic, and he says, “yeah,i think we’re going to be happy we did this deal, because of the experience we’re gettingin the west,” so they wanted to expand in the west coast of the united states, expandtheir branch footprint, and also because “on the mortgage side of the product, the pick-a-payproduct,” which was the golden west option arm product that wachovia was not participatingat all in the non-traditional mortgage market, “when things get better we’re going tobe very, very happy that we did this deal,” a year out, 2 years.

okay. so how did that turn out? i struggledwith whether or not to include this slide, but i figured you don’t get the opportunityto cite saturday night live very often in your career. this is a transcript from a saturdaynight live skit, after wachovia collapsed, the weekend after. what it is, it’s a fictionalconversation between nancy pelosi and the owners of golden west financial, and nancypelosi introduces them, says tell your story, and they say, “oh, it’s very sad. we hada great company that sold sub-prime mortgages to wachovia, and now it’s worthless.”she says, “oh, that’s too bad. did you sell it for anything?” he says, “yes,for a lot.” and she says, “oh, so you’re not actual victims,” and they say, “no,no, no. that would be wachovia bank.”

so there’s lots of actual factual inaccuraciesin this transcript, in that golden west financial did not securitize anything, and the reasonit was so toxic to wachovia is that they kept everything on their balance sheet, but kindof the content that i want you to take away, from both of these things, is that the reasonwachovia purchased golden west financial is because it was a very average lender, concentratedin the south and the east, for traditional reasons, and then the reason it failed wasbecause of all this attention and losses it was taking from golden west financial. so that’s the narrative evidence, as wesay in berkeley, but we have to show you some more than that.

what i’m going to show you is that wachoviawas distressed but that it was exceptionally distressed, and that this distress resultedin wachovia contracting household access to credit. and what i also then need to showyou is that this is going to matter for what we would like to learn about, because it’snot obvious in the household credit market that being exposed to wachovia locally shouldmatter for your access to credit. we don’t know that there are lots of frictions in thehousehold credit market, so my estimates are going to give you some sense that there are,but i’m also going to show you a little evidence that it looks like this is a reasonableapproach to take. and then, finally, i’m going to argue that exposure to wachovia isnot correlated with all these other shocks.

okay. for the first part i need some data.i’m not going to go through this a lot. essentially, all the data i use are public.to measure household credit i rely on the hmda data, so it’s the flow of home purchase,refinancing credit, and that’s useful for me because i can look in a county and i cansay how large a specific lender is in that county, and i’m going to see which countiesdepend on wachovia. i’m going to try and measure firm credit, small business credit,with the cra data. that’s going to be useful to see which channel this is operating through. the first thing is, who was wachovia and whodepended on it? this is just a map of wachovia’s market share in the home purchase and refinancemarkets in 2005-2006, from the hmda data,

and you see exactly in this picture that wachoviawas heavily concentrated in the south and the east, so the lighter red colors and orange,those are high market concentrations, and essentially we had very little, but was presentmarket share in the rest of the country, which was kind of their impetus for absorbing goldenwest financial. what’s also going to be an important takeawayfrom this picture is that i’m going to need to check if the variations that drive my resultsare also holding within states, because you can see there’s lots of state-specific correlationin wachovia’s market share, which makes sense given the changes in state regulationsover time. so was wachovia contracting household accessto credit? this is just a simple picture of

the total flow of originations coming fromwachovia and then the market in general, from the hmda data. you see a bit of a differencein 2006, when wachovia was absorbing golden west financial. they seemed to be similarin 2007, and then after that there’s a huge divergence, where that spike you see in theblue dotted line is coming from the boom in refinancings as interest rates hit zero. now,this has at least suggested that wachovia was behaving differently. you don’t seea boom in refinancing coming from wachovia, but this could be driven entirely by the factthat wachovia is concentrating in specific markets and things like that. so we need to do a bit more than this, andwhat i do is i run a simple regression, wherein

on the left-hand side, in the hmda data, ihave the probability that a loan is originated. on the right-hand side i have full set ofcounty-fixed effects, and then i have a dummy for whether or not that application was submittedto wachovia, and then i have some controls. the county fixed effects are taking out everythingthat’s common to all the applications at that county, and that dummy is going to tellme the probability of origination for a loan submitted to wachovia relative to the averagenon-wachovia lender in this period. and i do this for three different income groups,to see if there are different patterns there that will be interesting. on the axis you’reseeing this different in probability of origination. what you see is that leading up to the crisis,wachovia was almost extraordinarily average

in this market. economically, these are verysmall differences in the probability of origination, essentially 2 or 3 percentage points. as wachoviabecame very distressed, these things collapsed. so in 2008, all households, of each incometype, were about 20 percentage points less likely to get a loan at wachovia relativeto the average non-wachovia lender in that same county. in 2009, after wachovia had actuallybeen taken over by wells fargo, you see that these things collapse even further. high-incomehouseholds have about the same difference whereas low- and middle-income householdsare essentially not getting loans from wachovia anymore. and then, in 2010, these differencescollapse back to zero. i’m not going to show it to you here butthese things could be driven by differences

in the composition of borrowers and so on,all that sort of thing, and i do a lot of checking to look, that these borrowers lookessentially identical leading up to the crisis, and then, as you can see here, wachovia isselecting to lending to higher quality borrowers, and really stopped lending to low- and middle-incomeborrowers. that’s interesting and it’s very suggestivethat wachovia is contracting household access to credit. i’m going to show you more resultsalong that line, but do we have any reason to think that this might actually matter?one simple friction that would suggest that this should matter would be if distance mattersin the household credit market. if you see that places near a wachovia branch are morelikely to depend on wachovia, this would suggest

that there’s some friction limiting householdsearch or something encouraging those households to depend on wachovia that’s going to limittheir substitution away. what i show you here, which is not in thepaper, is difference in wachovia’s market share for a census tract relative to the meanin some circle as a function of that census tract’s distance from wachovia. if you comparethe census tracts that are less than a mile away from the census tracts that are about3 to 4 miles away, you’re seeing about a 1/2 percentage point difference in wachovia’smarket share, which is very, very large. if you look at refinancing those home purchases,you see you’re getting close to almost a percentage point difference in wachovia’smarket share. this is all just to say that

it seems like there are some spatial frictionsin the household credit market which would suggest that looking at places that dependon wachovia might be a useful way of learning about these shocks. finally, i’m going to check if exposureto wachovia is correlated with other shocks and whether or not it seems to matter forthe outcomes that we’re interested in. i do a lot of regressions in the paper, checkingthat in these correlations there’s nothing really systematic going on in terms of whodepended on wachovia, they didn’t seem to have higher house price growth, and that sortof thing, but it’s nice to look at pictures. what i’m going to draw here is the growthrate in home purchase credit, at the county

level, between a year t and a base year of2007, so before wachovia’s contraction was apparently in the hmda data, and i’m goingto regress this on my measure of wachovia exposure, which is wachovia’s market sharein the household credit market. what we see is quite striking. leading upto the crisis in 2007, there’s essentially very little evidence of a trend in home purchasegrowth, whereas if you look at this relationship with, say, mortgage leverage or share of sub-primeyou see massive home purchase growth in these counties leading up to the crisis and verylittle afterwards. the standard errors are massive leading up to 2007, then after 2007there’s a large and significant decline, persisting through to 2011 and 2010. so, togive you a magnitude, this is on the order

of a 1 percentage point increase in the exposureto wachovia, causing about a 2 percent decline in home purchase growth by 2010, 2011. there’s a question about how to scale thatexactly, but the takeaway from here is there’s very little evidence of a pre-trend and significantevidence of large declines in household credit growth after wachovia begins contracting.if you look at house prices, you again see no evidence of a boom or a preceding bustin the house price market, but then following, essentially with a year lag, house pricesstart to decline significantly, although the standard errors are quite large here. what we also see is broad declines in housesales, and these things are robust, generally,

to including fixed effects and that sort ofthing, so you see a decline, essentially, in household expenditures in the housing market.and we also see a decline in household expenditures in non-housing, so a constructed measure ofretail expenditures from the nielsen retail scanner data, and this also declined significantlyrobustly with exposure to wachovia. given that, that we see large declines inhousing and non-housing expenditures, we’re curious if we see employment results, andparticularly in the areas where we expect them. what i show here is the same kind ofgraph of non-tradable employment, similar to what ben showed before of, essentially,retail and local restaurant employment, and you see essentially no trend leading up to2007, and large declines starting in 2009,

essentially, again, with a year lag with theresults you saw on the household credit market. the order of this is about, again, a 1 percentagepoint increase in exposure. wachovia causes about a 1 percent decline in non-tradableemployment here. that’s a very large responses. this diminishes somewhat when you’re includingstate fixed effects. you can see that here. the first column is just controlling for mortgageleverage. the second column is including state fixed effects, and you see a pretty similarestimate but a loss of precision. and then the second two columns are payrolls, and againwe see robust declines in all of these things in local non-tradable, correlated with exposureto wachovia. what we also see, that i’m not going toshow you here, is that the other place where

you see employment losses are in local residentialconstruction. non-residential construction almost doesn’t move with exposure to wachovia,so all of the losses are essentially being driven by local non-tradable and residentialconstruction, consistent with this operating through household expenditures. that’s all fine and consistent but whati haven’t really nailed is this about wachovia’s role in the local household credit marketor in the local firm credit market. again, just to give a suggested picture, and thenwe’ll look at some regressions, this is plotting household credit growth, the samepicture that i showed you before, but then adding onto it small business originationsfrom the cra data. wachovia is in the red

and then the market is in the dotted blueline. what’s striking is that wachovia was contracting exceptionally in the householdcredit market but not so much in the first credit market. i think this is surprisingbut potentially interesting if you think that the relationship value to firm credit relationshipsis very different from those in the household credit market. but given that, this is suggestive, but whati can also do is look within the counties and see which counties depend on wachoviain the household credit market and which counties depend on wachovia more in the firm creditmarket, and see which of those is driving the results.

here i’m just regressing the change in non-tradableemployment on exposure to wachovia in the household credit market from the hmda data,and then exposure to wachovia in the firm credit market, from the cra data. the firstcolumn is the baseline that i reported before, and the second column includes a dummy forwhether or not a county is very much exposed to wachovia in the firm credit market, andyou see essentially no loading on that second coefficient. if you include both as dummiesyou see it again all loading on exposure in the household credit market, and if you includeexposure continuously in both markets it is again all loading in the household creditmarket. what’s surprise is you think, well, maybethis firm credit exposure is just garbage.

it is very predictive of declines in the firmcredit market, so it does have some predictive content for movements. but all of the employmentsresults and movements in the household credit market are really being driven by knowingthat wachovia is important in the household credit market, and that’s robust acrosslots of different outcomes. that seems to be showing that these relationships i’mfinding with wachovia are really about wachovia’s role in the household credit market. given that, i can then estimate these elasticities,and i’ll show them here, of total employment in the county with respect to supply-drivenchanges in my measure of household credit, which is the flow of non-refinanced credit.you can measure in lots of different ways

and you get kind of similar estimates. ifyou estimated directly, you’re getting an elasticity estimate of about 0.15. once youinstrument for the flow in household credit with exposure to wachovia you’re gettingestimates more on the order of 0.3, with the final column aggregating to the commutingzone level. so these are, again, very large estimates,suggesting 10 percent decline in the flow of household credit driven by supply shockscause about 3 percent decline in total employment. and i’m doing total employment here becauseyou would be concerned if other sectors were growing to undue the shocks to local non-tradablein residential construction and that sort of thing.

to summarize that part, and i have 5 minutesleft for the rest of the paper, which should be okay, exposure to wachovia, what i’mtrying to have you take away here is that experience to wachovia is really about a shockto the household credit market, and that it seems to have very large effects on the flowof local credit and on household expenditures, say in the housing market and in the non-housingmarket, and that employment responds very strongly to these kinds of shocks, so theco-movement of things is significant. this is all going to suggest that this kind ofshock could be very important over this period, but we don’t really have an idea of howlarge the shock was to the household credit market. at the very end here, we don’t reallyknow how much of the decline in credit was

driven by supply as opposed to the demand. what i’m going to try and give you a flavorof, for the second part of the paper, is the following. what we’d like to be able tocalculate, in this limited way, is what i’m calling the aggregate direct contribution.if you see a shock in a county, what’s the average direct effect of that shock in thecounty, the losses in employment in that county implied by the shock to that county. thisis the average effect times the sum of shocks across all these counties. what i’m going to be able to get at is somethingrelated to that under some assumptions. essentially, i construct a measure of the shock, whichis related to the true shock plus some error

term, and we’re going to have to make assumptionsabout the error term. and with this you’re going to be able to estimate the effects ofthese measured shocks on employment. under some assumptions, i’m going to be able tocalculate the aggregate direct contribution plus some error, where this error is goingto have good properties that i’m going to be able to get rid of in a nice way, and i’mjust going to say that for the moment and show you some pictures. but the way to construct that shock is verysimple. the whole takeaway from wachovia is that the shock to a county is really a functionof the shocks from each of the lenders in that county. to construct a measure of theshock in that county i just need to construct

a measure of the shocks from each of thoselenders. to do that, i run a regression, where on the left-hand side i have the changes incredit flows between lenders and counties, and on the right-hand side you have a setof lender fixed effects and county fixed effects. those lender fixed effects are going to giveyou measures of the shocks from each of those lenders, net of the borrower fixed effects,and those are going to be measures essentially of the shock from that lender. so if you think about wachovia, across allthe counties where wachovia operated there was a common shock due to wachovia’s kindof internal distress. this is going to be a measure of the size of that shock throughits effect on observed lender quantities.

so when you weight these lender fixed effectsfor each county you’re going to have a measure of the shock to that county, and the advantagehere is that you’re agnostic about what’s driving these shocks. you don’t have tomake a claim about it’s being driven by your relationship to lehman brothers, or theholdings of a cdo obligation, that sort of thing. i estimate this using the hmda data, but onceyou do that you can construct these shocks and you get this distribution of shocks, andusually i do this with a laser point and with my hands but we’ll try it. it’ll workhere. the problem with this distribution is youdon’t know if it should be kind of shifted

to the right or to the left, and so to correctfor that, this procedure allows me to say that the error or the dispersion of this distributionis being driven entirely by supply side shocks. so if i just subtract the average shock atthe top of the distribution from each of the other shocks, i’m going to be taking outthe error and then pushing this to make it less negative, make it look less contractioning.that may be too much fort his discussion, but basically when you do this, you can calculatewhat was the partial equilibrium contribution of these shocks to the total losses in employment,and you can do it in a couple of different ways, either using direct estimates of theseeffects or instrumented estimates, and this gives you these different numbers of 30 percentof 60 percent of the total decline was driven

by these kinds of shocks. now, this is not the full general equilibriumeffect of this kind of shock. that has the same kind of limitations that ben was talkingabout. but it’s the direct effect of these shocks. that is, if you thought this was drivenby household demand then you would be surprised that you don’t observe this in the countieswhere household demand is falling, if that makes sense. that may have been a bit fast, but basicallythe takeaways are two. the first, that these shocks to household credit supply seemed tohave mattered significantly, that there are frictions in the household credit market thatmade it such that the collapse of one lender

here, being wachovia, meant that householdsbecame unwilling to access credit at other lenders, and then this resulted in large declinesin housing and non-housing expenditures. and that these declines in household expenditures,then, resulted in employment losses, concentrated in residential construction local non-tradable.and then, finally, the co-movement of these things is quite large. and the second part is that imposing a littlebit of structure on the problem, i’m able to come up with an estimate for how largethe shock was. this estimate has some assumptions but the takeaway there is that with theseassumptions the shock seems to have been quite large, such that the household expendituredependence on credit and then employment is

dependent on those expenditures were an importantpart of the recession. good afternoon. i’d like to thank the organizers,the cfpb, and the office of research for allowing me to be with you today. although if you lookhistorically at me giving talks about shocks to household balance sheets, it’s neverreally been much of a favor to me. oftentimes when i come and talk about this topic—unexpectedlosses of income, unexpected expenses in households—i have one of my own to give a little bit ofcolor and detail. i have arrived with you all today unscathed and i’m happy for that,so i get to tell you about the misfortune that other households in the economy havefaced this time, as opposed to my own. i’m from the pew charitable trust, our projecton financial security and mobility. we do

research that looks broadly across householdbalance sheets, trying to understand the challenges that households are facing in the aftermathof the great recession, how they’re responding to those challenges, and the degree to whichthose challenges are susceptible to policy change grounded in the experiences of householdsthat might make households more financially secure, more in control of their own finances,and better able to seize financial opportunities to present themselves. one of the founding observations of our projectis that though there is recovery in the economy broadly after the recession, many householdsin our economy aren't feeling that recovery. about half of households perceive themselvestoday as financially insecure, and that includes

households who, on paper, based on their characteristics,we'd expect to be doing fairly well, households with relatively high educational attainmentwith relatively high income, and with relatively high stocks of wealth and assets. all, inlarge proportion, are feeling financially insecure. and we don’t think those householdsare wrong. we think that households are good judges of where they are and how they’refaring in today’s economy, and we want, through our research, to better understandthe challenges that are driving these perceptions. to that end, we’ve recently fielded a large-scalesurvey. we used the gfk knowledge network panel. you may be familiar with it. it’sthe same panel that the federal reserve board used for their survey of household economicdecision-making. we talked to about 7,800

households, collecting detailed data on whathouseholds have, what they owe, the decisions, choices, and tradeoffs that they make as theymanage their day-to-day financial lives. it’s important at this stage for me to offera few caveats around these data. the analyses we’re doing are purely descriptive, tryingto figure out where households stand at this particular point in time when we interviewedthem. we’re limited to household’s perceptions of their own situations, how people perceivetheir holdings, how they think, themselves, about their financial lives. and particularlywith respect to this effort, looking at unexpected expenses, households looking retrospectivelyback over time, things that have happened to them, as opposed to looking prospectively,and bringing all the biases and caveats that

come with that. it’s hard to talk about household financesbecause household finances are incredibly diverse and incredibly idiosyncratic. householdsdon’t think about their money the way that economists think about money, the way thatproject managers of financial products think about money. the more time you spend withhouseholds, talking about the finances and looking at the choices that they’re making,the more bizarre, strange, and sometimes ridiculous choices you see people doing. people, forinstance, who, when we asked them, say that they don’t have savings but have savingsaccounts and have retirement accounts; people who manage all of their household financesusing cash. so we want to get down to where

households are, where households think abouttheir money, and particularly to understand how money flows through households. my background is in the evaluation of savingsprograms, programs that help households to build more savings, and in that world there’sa tendency to think of savings as a bauble. you show up, you intervene, you help people,you get them a little modicum of savings, and it’s good. they put it up on the shelf,you walk away, you’re done, you declare victory. but in the actual lives of households,that’s now how savings works. households build up savings, they diminish them, buildthem up, diminish them, and constantly have to rebuild them over time.

so as we start to think forward through thelives of household finances towards policy, we want to start with the things that disrupthousehold financial wellbeing, what the scale, magnitude, and frequency of financial disruptionsare that households face, to give a sense of what a policy solution might look likethat would be appropriate to the actual needs of households events. we start by looking at what we call financialshocks, but what we really mean is out-of-the-ordinary expenses. if you think across your household’sbudget, you have the parts of your budget you expect to pay, month to month you payyour rent, you pay your transportation costs, you maybe buy some food. you get all the sortsof things that are normal. but also households

have things that are less normal that happento them month to month, the sorts of things i have been so fortunate as to avoid thisweek—a car breaks down, a visit to the doctor, things that are always going to happen. they’reinevitable. we can’t say that they’re unexpected, but no one ever really expectsthem to happen today. no one expects that today is the day that they have to shell out$100, $500, $1,000 that they didn’t expect to. to give you a sense of what we asked, we askedpeople to tell us—and you guys can all raise your hands or jot it down in your notebookif these have happened—in the past 12 months, has someone suffered a loss of income, anillness or an injury, a separation, divorce,

a need to repair or replace a car, or makea major repair to a home or appliance? and then we asked people for things that we missedin our cavalcade of misfortunes, places where big, unexpected, idiosyncratic events hitthe household. people told us about all sorts of things that were unfortunate—funeralexpenses, legal expenses, and also some things that were good—being able to pay money toseize opportunities for additional education for a child, to move to a better neighborhood,to take advantage of opportunities with respect to a small business. so these unexpected expenses happen, and wewant to get a sense of how common they are in people’s lives, so we asked, over thepast 12 months, have any of these things happened

to you. in 6 of 10 of the households thatwe talked to it had at least one of these unexpected expenses in the past year. themost common was a car repair, which hit 30 percent of the households who we spoke to.but troublingly, a quarter of the households told us that they had had a significant lossof income. a quarter said they had had a major medical event that required hospitalization,and a quarter said that they had to engage in major home repair. these out-of-the-ordinaryexpenses really aren’t that out of the ordinary in the financial lives of households. they,in fact, are quite common and they don’t happen independently. about a third of householdsexperience two or more types of these expenses in the course of the 12 months prior to thesurvey.

now, we thought that these wouldn’t be distributedequally across the entirety of the population so we looked to see if there was differentialexposure to these out-of-the-ordinary expenses by demographic groups, by income, by education,and what we found was that they actually aren’t systematically distributed by those factors.we thought maybe young people are reckless and get themselves into trouble and that leadsto these expenses; we didn’t find that. maybe older people get sick more often; wedidn’t find that. but what we did find is that my daughter causes almost all of theunexpected expenses that happen in the world. in our modeling, the presence of children,the presence of additional adults, owning a car, basically having more people and morestuff to have things happen to are the predictors

of whether or not a household has a financialshock. we then asked about one shock in particularfor each of the households, the most expensive event that they had had in the past year,to get a sense of the scale of the problem. it’s a really different policy responseif we’re talking about a few hundred dollars versus a few thousand. and for those in oursample who experienced unexpected expenses, the typical expense cost them $2,000, butthat exists in a fairly wide range. some households got off fairly easy. the 75th percentile ofhouseholds were spending $6,000 on just the most expensive of the shocks that they encounteredover the course of the year. we know that that has a different impact acrossbudgets so we segregated it by household income,

and not surprisingly we find that householdswith higher incomes spent more on their most expensive shock. this stands to reason. householdswith higher income have nicer stuff. nicer stuff costs more to fix. they also have moremoney to expend on a given shock. you can imagine a case where you can do a cheap repairor a more expensive repair. higher income households are more likely to engage the moreexpensive, spend a little bit more money. their shocks cost more. by the same token, lower income householdshocks cost them more as well. this is the exact same figure but rather than in dollarswe denominate the cost of the most expensive shock in days of the household’s income,and for the typical household whose income

is under $25,000, the one shock that theyexperienced that was most expensive cost them the equivalent of one month’s income—anenormous amount. for middle income households we see it taking about a half a month’sincome, and even for relatively well-off households, households making $85,000 a year or more,we see their most expensive shock taking up a third of a month’s income. so these are big numbers. they’re big numbersin the context of a household’s budget. but it doesn’t necessarily follow from thesenumbers that households who experience expenditure shocks like this are going to necessarilybe financially worse off. people could actually think about their money like economists. theyknow their car is going to break down, and

so they amortize the expected cost of thatrepair over time, through their savings, so they’ll be prepared when these things thatthey know are going to happen, happen. it turns out that no one actually does that. about half of the people in our sample whoexperienced a financial shock said that that shock made it hard for them to make ends meet.they experienced it as detrimental to their financial wellbeing, detrimental to theirfinancial security. and what’s more, the shocks that made it hard to make ends meetweren’t transitory. the effects of those shocks, households tell us, were felt in theirbudgets for a long time. seventeen percent said their finances returned to normal aftera month following the shock. almost half the

respondents said at the time of the surveytheir finances still hadn’t recovered to the point where they were before that unexpectedexpense, and most of those respondents were about 6 months or more out from their shock. we’re interested, then, not just in theshort-term impact of these shocks but how they translate to long-term measures of financialstability, financial control, financial well-being, which are tricky because we measure all ofthese things contemporaneously in our survey. to be perfectly clear, we’re not makinga causal argument here. we’re starting descriptively and exploratorily to look for what these relationshipsmight look like, suggest avenues for future research, that has a little bit more exogeneity,a little bit more exogenous from the shocks.

and we observed that households who had experiencedthese events are doing worse on measures of household financial wellbeing than those whoare able to avoid them. households who experienced a shock have about $3,900 less at the medianand liquid savings, which we define as money held in checking, savings accounts, cash savedat home. now, this could be evidence that household savings are working exactly as weexpect them to. you experience a shock, you pay down money from savings; everything isgoing to be okay. we see a similar pattern with respect to credit card debt. those whoexperienced a shock are significantly more likely to be rolling over a credit card debtmonth to month than our households who didn’t experience an unexpected expense.

where we start to get concerned, though, iswhen we look at other things that happened in the same year as the financial shock. whatshows is the incidence of financial shortfalls, households who were unable to meet core expenses.they missed a payment on their rent or their mortgage, they were unable to pay a majorutility bill, they were unable to afford food, they were unable to afford medical care—householdswho were materially poorly off over the course of the same year. now, we don’t know the relative timing ofthese shortfalls compared to the unexpected expense that we observe, but we see that householdswho experience an unexpected expense were significantly more likely to have experiencedone of these types of hardships. this is true

in general. this is also true of higher incomefamilies. among households who make $85,000 a year or more, a full third of householdswho had an unexpected expense reported one of the types of financial shortfall that welooked at. and finally, households who experienced financialshocks feel worse off at the time of the survey than those who didn’t. about 65 percentof households who did not experience shock say that they’re financially secure comparedto only 4 in 10 among the households who did experience an unexpected expense. these unexpectedexpenses, things that are out of the ordinary, turn out to be fairly ordinary and hugelyimpactful in the financial lives of households. the next phases of research in our projectlook at how households can be enabled to be

more resilient against these sorts of financialshocks, the degree to which households are able to accumulate credit, they’re ableto effectively use credit, accumulate savings to buffer them against these financial shocks,all ordering towards a set of policies that will make households more financially secure,better off, more resilient, and ultimately better able to seize opportunities. thanks for sharing some time with me today.i’m looking forward to the rest of the conference. hi, everyone. i’m jon lanning. i’m withthe cfpb. it’s my pleasure to be able to discuss these papers, and it’s always obligatoryto say, “i really enjoyed reading these papers. i learned a lot.” i’m in the luckyposition to actually be able to say that and

sincerely mean it in this case. these arereally good papers. one of the things that you will see throughoutthis conference is the disclaimer that the views expressed here are mine and mine alone.they do not necessarily represent those of either the cfpb or the united states. so ifyou are looking for the representative agent whose views do express those of the unitedstates, it’s not necessarily me. all right. so we learned a lot from thesethree papers. i also apologize at the outset. i’m going to kind of motor through thisbecause i am on a 10-minute timeline to discuss the three papers. these are three interestingand diverse takes on looking at how shocks impact household wellbeing. keys uses proprietarydata and a very clever identification strategy,

comparing 5/1 arms versus 7/1 arm reset times,draws implications to show that decreases in mortgage rates and payments lead to improvehousehold balance sheets, predominantly through deleveraging. they also increased durableconsumptions in terms of auto purchases. john mondragon’s paper looks at public usedata, and a lot of public use data. i don’t think that he did a good job of explainingto you the pain and suffering he must have gone through collecting and working with datafrom six very, very large and not the most easily worked with data sets, to get to theresponses that he found. he shows that decreases in the supply of credit drive significantdeclines in employment, and right where you would expect them, in the non-tradable andconstruction sectors that we think would be

most affected by the localized changes indemand, both for housing and for non-housing. and key from pew uses primarily survey data—it’sprimary data that’s collected specifically by them—and generates a descriptive exercise.i should note i am not one of those economists who uses the term “descriptive” in anysort of negative way. i think it’s actually very fascinating to look at the sort of findingsthat he has. it frames a lot of the problems that generates the stylized facts that a lotof the work that’s going to work on trying to figure out causality and disentangle thosethings that are going to rely on the future. he shows that shocks are a really common occurrencein virtually all households’ lives, that they are potentially very damaging, and thatthey can affect a lot of sort of different

measures of household wellbeing that may notbe initially obviously. i’m going to go ahead and motor throughthese really, really quickly. i don’t really have to talk too much about what we learnedfrom these papers. i think the authors did a great job of talking about their main findings.the questions are the types of things that i think could be added, if imperfectly, tothe current drafts of the papers. the extensions are projects that probably weren’t theirown exploration in the future. again, ben’s paper shows that shocks to disposable incomeincreased deleveraging, durable consumption. he talked at length about his findings. the biggest question i have is who are theseborrowers? we’re comparing 5/1 arm folks

to 7/1 arm folks. these are people who arestaying in these loans beyond the time that we might, as economists, think these peoplewould be looking to reset it. a lot of people purchase arms because they’re not expectingnecessarily to stay in it to the reset date, especially in the face of rates lowering throughtime. if i’m a 7/1 borrower, i see my neighbor who has a 5/1 arm, reset, get a lower rate,that rate is probably in the market for me if i’m wiling and able to refinance. sowhy aren’t they refinancing? is it an inattention story? is it a, i’m now underwater on mymortgage and i don’t qualify for a refinance? is it that my credit score has deteriorated?what’s going on to sort of keep this control group in the control group against which wecan compare it? i think that actually plays

a lot into how generalizable this is to otherforms of income shocks. if it’s an inattention story, for example, why is it that these peoplesort of look at their change in payments and immediately attribute it to deleveraging versuswhat we might expect from a more traditional stimulus payment, where they might sort ofsee it as a one-time shock to income, not really pay attention to the fact that i cansave this and use it for the future and use it today. so what is it about these borrowersthat is causing them to be there and what does that mean for the implications for othertypes of income shock, as well as other types of mortgages? would we see similar responsesif we magically decreased the rates in fixed-rate mortgages? there has been work that has lookedspecifically at this, and there’s a lot

of interconnectedness that could be made withthe literature here. in terms of extensions, i think—if i knowben he’s probably already done some of this—using similar framework but exploiting the monthlychanges in 5/1 arm reset rates. i count, in one of the major indices against which thesereset rates are targeted, i think i counted 16 instances where there was a month-over-monthchange of 25 basis points or more during the period in which he’s exploring. a 25 basispoint change from, you know, august 31st to september 2nd seems like you could get a prettyplausible source of quasi-exogenous variation there in the payments people are making. and then, also, if you could, for example,look at patterns amongst those who actually

are refinancing out of these 7/1 arms. arethese people refinancing out of them and experiencing similar deleveraging behavior once they’vedone that? i’m going to go ahead and pick it up evena little faster, so i’m going to skip all the way over john’s results. john’s resultsare great. he spoke to them at length. questions i have are, are there any additional application-levelcharacteristics that we might be able to observe? you’re looking at income for these underwritings.we might expect in a world where wachovia is facing a little bit more instability thefinancially borrowers or potential borrowers might actually avoid going to wachovia, knowingthat this is a little bit riskier, but might actually pursue other banks. we might seethe decrease in probability of origination

at wachovia be a selection thing. but again,you would need loan-level application data in order to figure that out. why are there so many frictions facing thesepotential borrowers when they search across lenders? why is it that if i get, for example,denied by wachovia it’s so hard for me to walk across the street to whatever other bankmight be in this town and look at that? are they inferring that mortgages are commoditizedproducts? if i got denied by one, that means i don’t qualify for a mortgage, as opposedto a mortgage at a specific bank. why is it that we’re observing these frictions? and then, lastly, if the supply of creditis at least partially responsible for the

unemployment we’re observing, why does theunemployment seem to so much more persistent? we see that the supply of credit spikes backup, but the employment rates in a lot of these areas have stayed a lot lower than they were.i think there is actually a potentially interesting story to be woven in with firm credit as well,seeing if they’re changing to whom they’re lending, if not the supply of firm credit,but i digress. in terms of extensions, i think—and thispains me to say because i come from a very reduced foreign background, and that’s mybread and butter—i do think a general equilibrium extension would be really fascinating here.i would require the imposition of a lot more structure than you did, which again, not myinclination to suggest. but i really think

there are some potentially fascinating thingsto be done there, as well as exploring the dynamics of other types of consumer credits.if you could look at helocs, if you could look at cash-out refis—are these amplifyingthe effects that we’re seeing or are they actually attenuating these effects? are peopletaking cash-out refis to sort of partially offset the decrease in demand that’s leadingto the decrease in employment we’re seeing? clinton’s paper, again, shocks are verycommon. they hit a lot of people. they have big impacts. people are missing core payments.you want to make sure you get that right. these aren’t “i faced a financial shockso i’m not going on vacation this year.” this is “i’m missing a mortgage payment.”“i’m missing some sort of meal.” “i’m

missing something major.” a question that’s bothered me for very longat this point is, if these are so common, why isn’t it the absence of shocks that’sshocking to people? why can’t we sort of reframe this so that every month my car doesn’tbreak down i feel like i won the lottery, and i go spend some extra money at the endof the month because it just worked out. why it is we can’t sort of reframe those thingshas always been a question to me. another thing is whether or not the liquidasset story. we saw that it impacts savings, but if we looked at those impacted savingsin terms of the number of days of work, how much time could i miss, for example, for amedical expense? would things look sort of

qualitatively better or worse if we sort ofthought of this as a buffer stock as opposed to just sort of a raw amount. in terms of extensions, i really think thatlinking this to administrative data, to figure out, we’ve seen in many studies the factthat some people perceive themselves to be much worse off than they actually are on paperand vice versa, trying to link that and figure out what the real effect on the balance sheetwould be there. and then trying to separate the truly exogenous shocks from the not-so-exogenousshocks. to draw a very, very quick example, up until recently i drove a 14-year-old fordsuv. it would not have been evenly remotely shocking if that car broke down at any pointin time. however, if i had been hit by another

motorist that would have been kind of a surprise.so if i’m driving around with my check engine light on and my car fails, that’s a shockbut it’s very different than the shock of something completely unexpected happens. andif we could sort of disentangle those, i think it would be fascinating. unfortunately—and this is sort of a themeof this discussion—most of the natural extensions of these might not be feasible with existingdata. for example, the suggestions i made for ben’s paper, there’s no real opportunityto target surveys to specific arm borrowers who are staying in there, to try to understandwhy are you staying in this mortgage when there’s cash left out on the table. it’sdifficult to link refinanced loans to the

loans that they’re actually replacing. soeven if we were able to sort of see refinanced loans and figure out what people are doingthere, trying to figure out what loans they’re replacing and whether or not they’re doingthis for cash-out reasons, whether or not they’re doing this just to take advantageof the lower interest rates, it’s hard to sort of identify that. in terms of john’s paper, the public hmdadata does not report all loan-level data relevant for underwriting and pricing, so you can’treally get at these things. data on other forms of consumer credit are also difficultto come by. in terms of clinton’s paper, sort of the real impact of perceived shocks,if we could link it to administrative records

it’s often difficult to link survey resultsto administrative records. the key takeaways here are that shocks tohousehold credit can have broad and deep impacts across a lot of things, both at the communitylevel and at the individual level. research in this area often requires significant entrepreneurship.you’ve got to hunt and peck for data, you’ve got to work hard to put it together, but thepayoffs can be really, really high. the way that i think about consumer finance a lotof the times is this isn’t labor. we can plausibly get most of what we think mattersto the outcomes and decisions to be part of our observable data, because mortgages don’tcare to whom they’re originated and they don’t care, typically, what property collateralizesthem, whereas labor tends to care where they

work and who they happen to work for. so youcan get to actual credible answer, i think, more consistently in this area if you canget the right data. however, the right data is hard to come by, and earlier this morningthe point was made that data is like cheese and economists are like mice, and we’redrawn towards it. my sincere hope is that if we get enough hungry mice together, lookingaround, we’re going to be able to find more cheese, and maybe we can even convince somebodyto provide some of that cheese for us. so that’s my brief discussion. i want tomake sure there’s plenty of time for questions and answers.okay. i guess first ben, responses to jon’s point about leaving money on the table.yes. first let me thank jon for his comments,

and i think he just was asking for governmentcheese. i’m not quite sure. the first question that he raised was theseinfeld critique, which is “who are these people?”there are two stages to this. the first is who takes out an arm, and then why take outa 5/1 versus a 7/1, and then conditional on having taken out that loan, why are you stillin that loan later on? looking at the characteristics of arm borrowers, we’ve already zoomed inon a very narrow range of the arm spectrum. if you’re comparing fixed rate and adjustablehouseholds, arm borrowers tend to have more liquidity constraints and they’re more dependenton hitting a certain dollar amount for the monthly payment, so they’re usually constrainedon that dimension more than any other. but

we’re looking at the agency market, so youcan think of these as being prime borrowers who generally put down a significant downpayment, but maybe a bit more constrained on the income side. in terms of who’s staying, part of the periodwe’re looking at is the period prior to the expansion of the hamp program, so a lotof the people who are staying in these loans, both in the 5/1s and in the 7/1s are actuallyunder water, so there’s quite a bit of high ltv borrowers who are still there. in termsof linking the data, one of the things that we’re looking into as a potential follow-upproject, with the data that’s available to one of my co-authors, if the loan has beenoriginated and backed by that same large market

participant they can actually link, over time,in terms of refinancing. so that’s something we’re looking to do, that, again, i thinkhasn’t been done elsewhere because it’s such a difficult thing to do with the databarriers. but thanks again. john, any responses to why people don’twalk across the street from wachovia? yeah. i think that is kind of central. i wasvery surprised with the size of the results that i got when i ran the regressions, soi was naturally thinking, what could be going on here? i think what we do know, to someextent already, is that households are very bad at search in the household credit market.they make lots of mistakes that are expensive

mistakes, and i think one of the mistakesthat households could be making is that they are not able to get a mortgage. wachovia wasa large retail lender that was kind of institutional, and most households had no idea that goldenwest financial existed, or world savings bank existed, so if they get denied i think it’svery unlikely they’d actually make the inference that it’s because wachovia is having liquidityproblems when they don’t even actually know what liquidity is. so i think that’s very likely, one thing,and the other frictions that i started to allude to, and i’m working on now, whichis, does it seem like space really is an important determiner in this market, and it really lookslike there are very large difference in market

shares according to space, and if householdsearch is as bad and as lazy as it seems to be from other papers that have been done,then that would make it kind of very natural. it’s hard to apply and if you do apply youget rejected, then you just kind of give up because you’re not very good at searchingin the market. but i think the other thing that could kindof explain why these shocks seem so big is that i think it is a bit of a traumatic shock,in that when you think about these elasticities and changes in the supply of credit, one changein supply of credit is going to have a change in the interest rate a little bit much onwhat you pay. another, the institution that has existed for the last 100 years—and 200years in some places—disappears. i think

that can elicit a very different responsefrom the households in terms of how they think the world is going to look in terms of accessto credit and that sort of thing. so that could be one difference, that economicallythere’s just actually a very real difference in the source of variation that i’m usingin how households think about it. i’d like to thank jonathan for his thoughtfulcomments and for his admirable efforts, given his time constraint, to get through everythinghe was able to. i actually used to live across the street from a wachovia.but i did my banking at the bank that was 100 yards closer, which actually speaks tomy response to jonathan’s first criticism, which is the feasibility of behavior changeand the difficulty for policy in changing

people’s orientations towards anything,whether we’re talking about weight loss, whether we’re talking about household finances.getting people to align their activity, particularly in cases where it requires effort and initiative,with the outcomes that they say that they desire, is one of the long-term challenges,one of the intractable challenges that we face with respect to policy. one avenue, or a set of products that removethose barriers through the use of behavioral techniques, through the use of channel factors, through the useof simplification, automation, and certainly all of those are on the table as we look towardssavings policy that could help households to be more financially secure.i agree, more cheese is always better. we’re

fundamentally interested in linking our observationaland survey data with the administrative data sources, both on the credit side but alsolooking at account records, the flow of funds through actual transaction accounts. so ifanyone has a pile of cheese that they’d like researchers to run through, i’d loveto speak with you afterwards. thank you very much.well, it’s already 2:45, so we’ll have a truncated question session. i was on the scientific committee. i thinkit’s very interesting that the macroeconomics and the household finance interact. so sometimesyou have research that’s motivated by the macroeconomics but it actually, i think, inmany ways, is more informative about the household

finance. so here, if you talk about the macroeconomics,a lot of it has to do with the monetary policy response, and yet, from this research, weget a picture of the households where, if they have a little more money, from the mortgagegoing down, they pay off their debt, we get a picture of the other things that they’redoing in response to things. they don’t want to go too far to a bank. and i thinkthat starts to paint a picture which is very relevant to the cfpb. on the other hand, there are things that aremotivated by household finance considerations that i think have huge importance for macroeconomicsstabilization. this morning we heard about pre-retirement withdrawals from 401(k)s—itused to be called leakage—but these pre-retirement

withdrawals are actually something that ifyou encourage them to happen in a recession that can have an important stabilizing impact.so you don’t want to miss that macroeconomic stabilization opportunity, which is somethingthat denmark used, that the u.s. used, to some extent, but that’s something that youwant to pay attention to and not break things into these silos. another one that i think is underappreciatedis the work that david laibson and many co-authors have done about the effects of defaults interms of what people end up saving in their 401(k)s, where if you automatically enrolledthem with relatively high contribution rates they save more, can have effects on the nationalsaving rate, which has huge effects on all

kinds of things—the level of investment,the trade balance. you know, people think that these trade deals have a big effect ontrade, but honestly, the saving rate in the u.s. has a much, much bigger effect on what’sgoing on in the world of trade than anything that people are doing in terms of these tradeagreements. so you don’t want to break these thingsinto the silos between household finance and the macroeconomics. sometimes something motivatedby the macroeconomics can tell you about the household finance, and vice versa.this is question for— can you state your name?brigitte madrian from harvard. this is a question for clinton. i thought one of the most interestingdescriptive tables you had was how long it

took households to recover from a financialshock, and i was wondering whether, as part of your survey, you had collected data onthe types of insurance that households had and on the types of financial assets theyhad, and whether those were mediating factors in any way for how long it takes householdsto bounce back, so we see households that have health insurance recover in less thana month and households that don’t are still taking 6 months to recover from a shock, orhouseholds that have a 401(k) that they can withdraw money from, or an ira, take out aloan or a credit card, recover quickly, and households that don’t take longer.absolutely. we have limited information from the survey on insurance. we have whether ornot the household has health insurance through

their employer. we collected detailed dataon households’ asset holdings by category. we didn’t look at those because we measuredit contemporaneously to the measurement of the shock, but fortunately we’re going backto the field in the fall, of the same sample, a follow-up survey, looking at the shocksthat have happened in intervening years, where we’ll have complete asset and liabilityinformation from the year prior, we can see the degree to which those savings, those resourceseither enabled or failed to enable households to recover differently in that second yearthan we observed in the first. i’d like to say something that picks upon something that jon said, and make it a little more concrete. he said that it wouldbe nice to have sort of sense, actually, about

the other john’s paper, of how the resultsmight fit into or look in the context of a general equilibrium model. i wouldn’t necessarilygo so far as to ask the question about general equilibrium but i think, actually, for allof the papers a key question, from the perspective of turning the kinds of work we’ve seenhere into wisdom that can be used for making better economic policy, is how to systematizethe results so that you might be able to guess what would happen, or that they tell you aboutwhat would happen in some different set of circumstances. as many people in the room have heard me saybefore, i think the best technology we have for trying to do that is to compare the actualresults we get in empirical projects like

this to the results that you would get frompeople who are solving some optical consumer financial choice model, and then see if wecan generalize the ways in which actual behavior deviates from the optimal, because if we cansay, okay, systematically people do such-and-such, they deviate from the optimal in the followingsets of says, in the predictable sets of circumstances, then that gives us the tools that we needto help at least get a better idea of how the world might change if we change some policy,because we at least have the ability to say, okay, we know behavior systematically lookskind of like this in many different contexts, so in a context of a world that got changedby one of our policies we’d have a better sense of how the world looks.

so what ought to be the optimal response tothe mortgage change that ben was looking at? well, the first responses there, of course,is the one that jon brought up, which is the optimal response is probably you want to refinance.all of these guys want to refinance first. okay. so then you might say, we’ll add afriction. the friction is an inattention, that people just don’t really notice thatthere’s the option to refinance, and if that’s the only friction in the model thenmaybe we can get the model to make a prediction about what their marginal propensity to consumewould be, with only that one friction in it, and then how does that compare to the numbersthat you get? that would be just an example, but if we can make a link because the empiricalresults in a sort of more structural framework,

that’s the outcome that is most valuableto us. i have a question for john, but first, thework on default originated with brigitte madrian, right there, in the room with us today. iwas a latecomer to that line of work. i just want to call out brigitte. a question for john, a two-part question.first, do we know whether wachovia focused their policy change on reducing credit throughthe kind of normal channels of not giving people credit card loans and things like that,or were they also working through other kind of more exotic channels, like changing theway or pace that they foreclosed homes, suggesting a very different kind of dynamic? and then,a kind of follow-up question, when the wells

fargo change in ownership occurs, or absorptionoccurs, we would i guess expect to see a sharp reversal in all of these effects, and do youobserve that too? okay. so i’ll start with the second. yeah,that was exactly what i was expecting with these original probabilities, is once wachoviais transferred to wells fargo the liquidity problem should be over—we know wells fargois relatively well capitalized—and that’s not at all what happens. what seems to havebeen going on is they were shutting down branches and they were restraining how much lendingwachovia was doing. i’ve been trying to find kind of evidence of them saying thatthis is what they were doing. they talk about kind of incrementally taking over wachovia’soperations, but i don’t think any lender

goes out and says that we’re denying everyonein this region credit. that’s not what you see, which i think isinteresting because we tend to think in these mechanical terms of there’s liquidity andif you have liquidity you make a loan, or you don’t. but i think here there was anintentional decision that wachovia had become very distressed and wells fargo wasn’t sureexactly why, and so they were kind of careful about what they did. and i’m lucky thatin the hmda data you can still differentiate between them. and then for the first question, no, not really.i mean, the only margin i really observe are these denials. i can also look at interestrates they’re paying but it’s a very censored

picture you see in the hmda data. i see branchclosures, lots of branch closures. but in terms of the other things of all the differentmargins—you can adjust credit prices, and limits on credit cards, and change in thesize of heloc—i’m not able to observe all that sort of thing. a lot of it is justshutting down branches. i mean, they closed lots and lots of branches, which is a verydifferent kind of shock to the supply of credit that i was alluding to before. but, no, i think it would be very interestingand i don’t have the data at the moment. one last question, if there is? okay. thenlet’s thank your speakers, and panel 1 is over.

Location:

Share :