to begin on the topic of retirement. we take you to that event, now. [captioning performed by the national captioning institute, which is responsible for its caption content and accuracy. visit ncicap.org] [captions copyright national cable satellite corp. 2017]

>> thank you very much for coming. i think we will get started. welcome to the american enterprise institute, i am andrew biggs. i have the privilege of moderating today's session, what can irs data tell us about retirement inces? d-gradeagine a sort of thriller movie that said what if you knew just what if everything you thought you knew were wrong? the tagle might be what are many of the things if you don't were wrote and what if many were about -- what if the things you thought you knew were wrong? knowof what we think we about retirement income, the statistics regarding social security may in fact be wrote. think we know what retirement

is an was the poverty rates are and how dependent they are on social security and how much they receive from private plans. i can pull some statistics from my hand for the poverty rate for americans over 65 is 10%. about one third of retirees receive almost all of their income from social security benefits. those are what i know from the data currently used. we will find out if some of the statistics are wrong because some of the data they are bas on is incorrect. the household survey dat gathered by the government that asks people about soces of incomen timent uerat omritemeris-1(we may come to skd retireesns of how well are and we may believe retirees are daerously dependent also

security benefits, private retirement plans have failed and unable to remain the standard. the two studies presented try to come to a more accurate view by using more accurate data. in the process, they tell a story for a policy is close to astonishing. tax records,rnment the income for typical household is nearly one third higher than wehought and poverty isn't scheduling lower. has an incomeree similar to was a the enjoyed previously. consider an abject failure in a . we do not know what -- in a new light. -- we do n what e future hos to defined contribution -- likeke for one day 401(k)s.

more accurate data do seem to show that americans' retirement income woun't need to drop further before we could consider a system to be in crisis. i am happy to welcome r prison terms for our first will the peter brady, a senior economist and the research division following peter will be joshua mitchell,ho is a senior onomist at the american census bureau. thank you for coming in welcome to aei. both pete and josh are presenting a group work. co-authored.is withs paper is co-authored the census bureau. discussants who will offer discussions for our .irst is kevin moore our second willoughby bruce

mer -- our second willoughby bruce meyer -- our second will be bruce meyer. thank you for coming. following, we wilalw r brief interchange between the participants in turn to your questions. i would like to begin with our presentations, pete we have between 20 and 25 minutes and then we will go to josh. peter brady: thank you. as andrew mentioned, this a joint work and my co-authors are in front to work. i want to mention this paper is a product of the joint statistical program. our paper examines to questions. oirst, our workers able maintain their sndard after they claim social security benefit? will look at changes in spendable income. getndly, where do workers their income after they claim

social security> will lookwer that, we at the amount of income from different sources. the reason we took this research is that the social security system and tax referral for retirement plans are an important, of the u.s. tax and transfer system. they generate most of inco that retirees rely on. it is very importantor policymakers andor administrators to understand how these programs a working because there many proposals to change one or both. a lot of these prosals are premised tt there's retirement crisis that workers e not accumulatingnough resources while they are working. what did we find? we find that most are able maintain their inflation adjusted net income after they retire and the individuals who are lowest income uld we first observeeem to replace the

highest percentage. where do they get to their income? in addition to social security, it turns out most we observed have income from psions, annuities and iras, retirement income. sample aret in the going to rely on a mix of soci security benefits. the plan today is i will describe a little of the data. i am going to look at how spendable income changes in rerement. after claing social security and where do they get income discussion onef we can discern where the pension and annuity and, is generating tt income. the day that we use is a panel data set, 1999 to 2010. all taxpayers in 1999. we are going to look at on individuals either a primary or

secondary taxyer on joint return or non-joint and we will follow them through 2010. we want to obsve them before they claim social security, observe them claiming a look at what happens after that. what we are going to focus on its people aged 55 to 61 and 19, early rerement claing age working and not yet receiving any social security. thewhat we want to look that is what happens after they claim social security. one nice thing about focusing on thisarticular group is at the test data appears to be representative of the group we are studying. you look at all individuals aged 55 to 61, over 95% file a tax return. when you look at what were looking debt, which ones are still working, the percentage is

closer to 100%. most of the people we want to 1999 and theye in do not need to file a tax return, you do not set the file in later years to stay in the sample. we can measure your income, we grabbed wages and salaries from the w-2, social security 1099 and pension and are a distribution. -- and i are a distribution -- ira distribution. the first year receiving social security as claimant your -- as claiming yr, year t. when we rank individuals, we will rank them by 99 income. when they are wking and before

any of the claims. the income we may be measuring in at the tables where showing could be from as late as 2010 but we will rank individuals by the 1999 income. there is 12.5 million people under the test data that meet our criteria, 55 to 61, wking and not having social security. betweenalyze the 59% 2000 and007, we needed to observe the data three years after they claim. 19% that excludes are about who are going to claim, observed claimant in 2008. about 5% claim social curity disability before social security retirement. and hat -- 13%s of life and had not yet cimed and about 4% who died before the received it.

notice thoseho have not receiv odied before, we do not know if there were eligible, they met been covered by a government plan or did not have enough credits to qualify. i went through a lot of slice very quickly -- slides very quickly in this summarizes everything you need to kw, namely looking at people close to the early retirement age, andy claiming age working do not have social security income. we will follow anyone. we considered our first question, what happens to endable income? ny studies of retirement will focus on how much people's income or peoples earnings they can replace in retirement for economic very suggests people do not have a goal of replang come. in english, they want to maintain their standard of living. what that means is they want to

maintain how much spending they have in real terms. possible theys may be able to maintain their standard of living as spending the clients. we cannot measure spending but we will look at spendable income. -- as spending declines. spending after they saved and paid taxes. the test data do not allow us to asure savings as a whole but taxesn msure payroll and federal income taxes. the results i'going to show you today is work-related income. work-related income includes labor income, wages and salaries and income from two her sources that are generated by working, social security benefits and distributions from pensions and annuities and iras. taxableexclude other

income investment, rinse, royaies, etc. since we're focused on work-related income, our measure is bendable income and will be net work-related income. the next few slides are what happens to mean spendable income around the time claimant. the year in which we obsve claim and anywhere from 2000 to 2007. we look a years around that. it is inflation adjusted and per capita which means for america on a joint return, we split the income equally. mostthis shows is that people are able to maintain their spendable income per person after retirement. total work-related declines by $4400 between t-1 and t+3.

on a net, spendable income, the blue portion stays constant. the next is how this differs by individuals' 1999 income. what we see is for those in the lowest 20%, ttom quintile, 1999ncome, they experienced, on average, a 29% increase in spendable income. for those in theiddle quintile -- o.s . this is t fl.you n store oervie clnt unt higr come invidual a bottom half of the individuals in the simpler still working to three years after claiming social security. the next two slides weather work status -- whether work status of fixing these results. ad a 43%was quintile

-- the lowest quintile had 43% increase. at thetheecli top, pretty modest. ife lk at whapps the o are workg ree yes latere lorncome sthas irease spenda ince. s mu me mode. we see income in the middle step income at thend top declines significantly for those who do not work. that shows the main average to show you what is happening. there is aide variation. to look at how things verand what the diffent experience are, we calculate for every individual in the sample a replacement rate. the replacement rate is a replacement of spendable income, not gross income. we're focused onet worth.

in all of these replacement in, i will show you and we will go quickly, the denominator is going to be spendable income any year t minus onend we will look at how that relates to income in later years. the next few will show the distribution among everyone in the sample. thblue b will be the median. among everybody in our cell phone in the first year they claim social security, the median replacement is 106%. means 25% and a replacement rate of 80% and lower. we also show the 10th and ninth. it stes from3, the constant but the slight increase in the spread of results. what i want to focus on your t+3, and ii will --

will look at it by income. the bottom 60% of to the mide quintile, replacement rates on average, it is over 100%. the other thing is a general online, 9replacement rate a whole distribution seems to shift down. again about half of individuals are working to three years later so we want to seef they are working or not a sex results. working or not of sex results. you can see ey have -- if they are working orot affects results. you look at those wking, it tends to shift up. you see the pattern where the highesseems to be at the bottom and is slowly goes do. those were not working three years later tentative have lower

placement a and you see that because of the distribution of the rates. lowerneral pattern is 1999 income seems to have higher lowest is oveut 100%. we find that most pele, on average, will remain and the lower and those who were working three years ter tend to have gher replacement rates. we rno the second the question about where do people get income from? we will look at different sources. this illustrates retirement is perda transitional instead of a pointn time. they claimefore social security, abo 50% in the sample have stopped working. untilworking declis

three years later although nearly half of working three years later. 61% working ohave a spouse who working, a slow transitional period. the amount of people working declines, the amount of pele from pensions and annuitiesncrease. even the year before they claim social surity, nearly half have ince from these sources tot goes of to 72 -- up 72% aftethree years. those who claim in the first in allbout 92% who claim three years after claiming social security. the next few slides, the three years after claiming, t - 1 through t + 3 and by 1999 iome and see what the pattern is. what we see here is that for a large portion of the middle or upper parts from the 40th percentile to the 95th year,tile, by the 30

nearly 85% has to this income. the lowest incidence is at top 5% the lower 40%. if we extend this analys out to the full five year and say how many people receive income in one of the five years ported a three years later 81% receive income from these urs. in addition, we have evidence that some of theeople who never reived pension have some of these resources. in particular i'm a we observe someone -- in particular, we observed someone west and 99 [indiscernible] and that is like a rollover and they rolled itver and did not increase. they have an ra but not drawing down income. what you can see is the higher is likely to have resource but not drawing them down.

the upshot is if you are5 to 61 and not at the bottom quintile which means you have per capita income of $17,000 about $24,000, you have a good chance of having these resources. nearly everybody at the top has it. that tells you they have them and how important it is. what i want to look at is for everybody none of labor income, i want to rip will decide what to remove labor income-- i want to remove labor income. will do it by the 1999 income. social security benefits for people with lower incomand much smaller portion at the high end. this is the design of the benefit, the benefit formula designed to replace at the low d and lower percentages as income goes up. pictured from annuities have the opposite -- pensions from annuities have the opposite,

that it tends to increase in importance going up. for the ople in the middle quintile, the magtude is all part with associate -- on par th social security benefits and we do t really talk about the other income, but the wealthier, ty have no labor of different types. data shows aaverage where people are getting the income. there are lots of vartions. we will show y what we calculatis for each individual , we will calculate their income shares. this is including all sources of inco. what perceage it com from different sources? what you see as wle, looking those lebanonign, claimed social security, the median they get is a zero. -- those who have not claim social security, the median they get is zero.

medium is about 43% but a wide variation. 10% of the sample only had soal security in a year t + 3. andook at retirement income we still see as the median, very little the year before they claim social surity but half ofeople are getting the income and some of the higher percentage from that. as you go out to t + 3, the dian is about 33%. what that shows a we will look 43%, 33% of bars and look how it varies by income. of 1999 in the, on average, they get a lot of income from cial security. 25% only get social security. 75th percente at 100%. ofy few will get a median

zero. go up the distribution and i do not include the highest income, you see the pattern where social security is less important and retirement income more important. the medium quintile, again, the median amount from social security is retirement income or very similar i went one slide too many. these results were the most surprising and not what we expected and not what i think i knew. i thought i knew where people were gettinghe income. this income is widesprd. it is persistent intends to the substantial on par for the middle income people with what they are getting from social security although there's a

gradient reflected in the design of the social security system. the lower income will rely more on social security in higher income will rely more on pension. those are the 2 questio for our research attended to prompt others to ask us questions we are not willing climbed to answer but will take it . the question we often get is pension and anity, where is it coming from? rt of e answer is in the plants in the private sector. there is a big concern of how much is private secto the well designed answer from a tax perspective is really not relevant where thencome is, it has the same tax treatment. in the form 1040, we have 2 categories. additional ira therefrom pension or annuity and that is all we know. to identify the

original source of income, not only because in general we do not know if is pension or d.c. because we do not know where the ira money is coming from. it com fm all sources of plants, lump sum -- source of plans, lump s from government or private sector. we do not know the actual origin of that data. we used 2 methods that we do nothing either provides estimates but it helps, how much do we think is the ximum amount we could say is coming from the private sector? joshua, enter their paper, they do another attempt to g added. -- in their paper, they do another attempt to get at it. nel dataat our pa set.

chart, what i am going to show is pension income in both the 30 year after claimant and second-year after claimant. the supposition is it people tend to have about the same amount of about 100% in both years, that might indicate it is pension because they pay the same amount each year. ,f it shows a slight increase then that might be an dication that it is a government plan. have of feder plans cost-oliving increases every year. when we look at what we found, as far as percentage of individual and dollars, about of annuityincome falls in the category of 100%. to 105%.er into 101%

to test it, we looked about what about people receiving are a income -- ira income in two years? this may be a od way to identify people getting government db's that people have control of their accounts. sometimes getting the same amount each year. we are not completelyure if the 33% is all private stor d.c.. if it was, about one third of pension annuity and about 1/4 all we observing in t + 3 is coming from that source. another way to look at its to look at aggregate data this is for research we are in the process of writing off and analysis using cross-sectional data. this is 2010 data. what we know is from the taxes

data is in 2010, we measured total distribution at about $1.2 trillion. nearly 400 billion of that is transfers from one account to another. a littleleaves about over $800 billion in net. we identify what percentage of the $827 billion is coming from private pensions? is by lookingit at gross distributionsnd we knowrom the tax data at about 24% is iras and the rest is pensions and annuities. ando to the aggregate data we can see that about 61% of the pensions of this that we can identify from a db or dc. tsplans, private sector and and private sector db,

governmental db. what is in the other 15%? we don't know. if we did, we would tell you. what we can tell by looking at other data is a substantial proportion is distributions from nonqualified annuities, reported on the same line. we can take some the transfers tt and specific ite, versions are usually going to come out of iras and rollovers from pensions, but we do not know what kind of pensions and get down to thes numbers. what is the bottom line? 800ant to know what other $27 billion could come from traditional db plans? we start witprivate sector db but we cannot use that for number because about $54 billion comment from cash balance plans or hybri-- coming from cash hybrid.plans or

if all that income was ken out , none of it was rolled through an ira, we would get 40% of net distribution from the private sector. again, this is something that wi forcefully, very much of interestn the policy community and not for filing taxes or termining x liability. we can try to put some balance on it. to summari, our ma findings is that most people e able to maintain their standard of living, their spendable income after they cla social security. those were observed in the 1990 th the lowest had the highest replacement rates. we find a surprising amount of retirement inme thais widespread and persistent d tends to be subsntial. : thank you very

much and right on time. our next presenter will be josh mitchell presenting joint work he conducted with adam bee. : thank you, the views today are mine and adam 's. >> i forgot to mention that. joshua mitchell: the motivation is the most important question in retirement policy are americans saved enough for retirement? or arehey in danger of rolling -- running out of money? deborah iting studies they use different assumptions and methyl -- and met this they come to stkly different conclusions. what they pret much all have in common is they rely on one of a few popular household surveys. there are long-standing quality concerns with income data in

household surveys in particular if we take a measure of income and compare it to one of a few outside sources such as the national income or tax data or other administrative records theh i labeled as ad-rec, survey falls short of the outside target. a key point is the discrepancies cannot tell us a whole lot and could on the one hand a few household who are missing from the survey and not reporting income in which case the distributional statistics are not affected. if you put it backn the survey , it would alter our derstanding of the standers and the older population. the purpose of this paper is to answer that question. we do that by linking our survey at the mic data level to compare the records and a survey responses on a person by person basis.

this will low us do distribution of analysis and develop estimates of the median household income and the population 65 and older through 2012. we will step back and ask which of the administrative record data sources we use our most important foour findings? this will have implications for e importance of db versus dc, social security and we will do a little bit of transition to , asrement analysis well. the u.s. census bureau's 2013 current population survey assessment and this is in 2013 13, they askedhey your income for the previous year. the source for the official income and poverty report. household -- we had about 75,000 households with 15,000

households 65 and older. we were able to link from the survey to other administrative sources.ta i am bored to focus on validating five types of income important for older americans in the first is earnings fr wages and self-employment, social security benefits, ssi, dividends and interest, and retirementncome, separate from social security, retirement, survivor, disability. cps's long-standing focus on money income concept which could be defined as stream of regular payments and assess nicely with the traditional benefit plans that paid an annuity upon retirement. manyshift from db to dc retirees take their withdrawals needed or regular

basis. that led to many outde analysts to raise concerns that the cps was missing lots of dc income and lots ofomplicated other withdrawals like lump sumsollovers that pose a challenge d how should we look at measuring these. in 2014, cps underwent a major redesign to answer some of these questions anadd a specific questions about account withdrawals and some of the other income and to my colleagues evaluated the redesign and found a modest boost in income for the 2014 data. we do not have the administrative records from 2014 , all wi be through the 2013 survey and the traditional cps questions. in main cps question asks the previous year, did your anybody in your household receive pension fr a previous employer or any otr type of income apart from social

security or v.a. benefits? if you say yes, who received it, the amounts and at the source. if you look a oion number seven, you see the language of regular payments from iras, led many analysts to raise questions that most of the dc in, what not be based on this information. cpsoen linked to this several different record sources , three of which we obtained from the social security administration. earnings information from your w-2 and self-employment from your schedule sees a your monthly -- c's and deductions to medicare premiums and monthly benefits, federal and state. irs , we obtained2 data

sources and one is the 1040 and web limite yields so today, -- and have limited in yields. a return a new have to make an assumption about who is getting the income and what to do about non-fowlers -- non-fi lers. weggregate back up to the family level. we use the survey values for interest dividends. lastly bubbles consequentially, 1099r, the gross distribution from pensions or profit sharing, iras, etc. it captures all of the defined benefit and information return so you got one even if you do not file a 1099. what we obtained essences thatde certain 1099s-r

excludes others. we received a early distribution, normal distribution, death and we exclude certain distributions we would not wish to consider income including direct rollovers which really is a more about moving money from what -- one account to another. we want to aim on a measuring with draws. -- we want to aim for measuring withdrawals. we merge these different administrative data resources and measure income at the personal level. , weou're one of e people use your survey values and outside of these five sources where validating, we use cps values and accurate data bacp to the household level as needed. the first result, we compare the cps, the administrative record data source for 2 age groups for 18 to 64 and 65 and over.

ps capturing, nearly 100% for both groups. yo social security, a high a 65 andespecially over. as an aside, things are more mplex for younger workers and their often reporting as a side onbehalf of- s s i -- ssi behalf of their children. when we get to dividends, only capturing 16% for the older population into retirement income, only about 45%. when you put that these five sources together for working age people, about 95% ofhe total income because wages dominate. when you look in age 65 and older, we are missing about a 1/4 of income. that is the aggregate. the key point r a paper is we can look at dismissing income. we will produce statistics

similar tohe annual poverty report. for persons 65 and over, poverty is family income compared to the relevant thshold of the mily size. i will show results forhree groups. the whole sample. the subsewho are pik and the adc sple. 2012, the official income was $33,800. essentiallyith pik, the same. when we replace with the administrative recds, household income was44,000, $10,000 more, quite substantial. torrent -- i do not have time to will all the results but i flip through and shoulder are differces among most of the demographics.

65-74, a at how so dierence. aneven for the oldest households from a larger difference in percentage terms. if you look at family household, large differences, nonfamily, large differences. college graduate, large differences for so with -- larger differences. even something for those with less than a high school degree. we see it for a whiteousehold, black, asian, hispanic. very persuasive across socioeconomic stribution. we did comparisons to outside of data sources which isn the survey of consumer finances and that survey as it focuses on a detailed was information in more substances definition of withdrawals. it generally shows highe incomes thanps.

we did a comparison, we find the administrative record about 50% ander than scf 65 and older differ for 75 and older. that is the middle of the distribution. could these results of fact the poverty rate? the official poverty rate for the 65 aldas.1%. en wilplacvaes wit adminitiveecds it s byrois. esfiin a pvave ro sgrps fse lgeifree ov.4,5 t4 d a aigueyyose aggrie oport wi t aintriv ror, r e unstf e e d eldt autneoi. th inasiepci a

ofhedmisate cos on. ic othrerdatsoce isryg furou rul? wl ame iscrs e usolinmeisibio wleccute th swsheasin wn rsald t rngsnd ple thdmisate cos,e s t r oto ofhe be. nt apousoalecit anmo othdiriti los ryoo a meve rertg othboomndre tuedutoheonsi amg spdeseeesoal cutyndsi wewaouas se d erpoingoswa-- s d erpoingowa we s abierivgee thmilendasy hewe efe rlaetemt,e

id dmac cte 3 o dfencdar thugmo othhoehd co dtrutn. a dcu pontl su tt ulbe is. erhi wl'veonha tan e misttie co asf e utancoar t thsuey yoca rseuatybo t rerd tnk t da ofigquitannorely ienveorndidlso erepto e bau th wldrierddiol xiali. thescto otr suis atheuatyf eanag beeethsuey - oth lauaetenheurysnd u t se ocs r bwer -inge thercende e e me ocs r l e cos.

thla iues u na knlee ereurs- i yocackwlgehe socere trendhais fa pntuteot wre a tnsiongro dene befs din coritis,f u n cot t.c y wl t stteicrendt ao t intoisngsheten o owhsoly veocl sury co a noininneassoal cuty iomndubanal gur thaws om d d u el cfoabs asfyg ems vi t sa lin sndd? i lleat o e xtxanendrertg t sveanadnirave ecds fr e ffeninme src. tethinntf rg, op rorr setngls whhanoinhe mistti ror wesefalytrg rratn

r l e taouesnd time iomisotad rert tsoalecit inme in lk et gd. heerthk pelereayg eyavan ty o icoorocl cutya" rpodean bh alo creonnt s oy %ayth he y t c wre1%avit cointohedmisate cos. y bakow t 6, % vempyebad stbuon wk whloof e peasngha elas endrrorng tou eryemraic vaabe ul'on cld pln otharti i fae-gave -wecod onxpin% t viaon

inal-nate te aton splyoer ane pln ryite he vaatn uerpoinac onhe viaes -09r a th -its rdo n dn setngpefi weotheostrti i tes t nurofhe 10-r osemoy-bedy ghor stbuon ouava rgamntf time iom yr feo po -yoarlilyo po sothg. diritis twn eur th nt e ssiklyo po setng otr suth c braed

t rireninmes onti entn icas mit dtoeduricur oftaarofivg. we lk tsehoeciv rerentncen 12 aedhafrti oth al reid stbuonf e xt yr a wt s e gr ofltuiobeee t 2 yrsboer emoynt snsedndre. --ra fo yngeoe, aon ofevt. 100%.and erit fluctuationk at the in amount, those lover drop greater than 10% and those will hose whomall change -- have a drogreater than 10% and those who have a small change, if they have a larger change, they are just as lely to have an incase or decrease.

a very hot fraction have a very small change or increase. we look at our 80's again for 65 and older unlikely to haven distribution in 2013 if you had one in012. woulwe do look at the amount, considerable more volatility. 84, only about half have a small change in ra, if anything when they are large change likely to have an inease in the next yr. from thet our results presentation were so pervasive across the socioeconomic distribution and for the oldest tirees, we wanted to raise the question is the underreporting or ift simply t ira's it was missing from the cps? we did this analysis by looking at those with cps, a link to a

1099-r from a few government plans, traditional, by anybody's definition,ncome. we will look at those who have the federal civil service stem or calpers. you can see for those 65 and of any terms employer-based distribution, about 35% say i have nothing in cps. fractiona substantial reporting 0% in cps. 25% for the federal system, 25% for military and 15 percent for call purse. .- four calpers we want to examine how this boost will change our understanding of the relative importance of different income sources. we use o benchmark of a

publication put out by the based on thety current population survey data. they view its the unit of analis which they define as a person 65 or older or a married couple with a husband of 65 and older. in particular, if i take your age and ranking them by their adminirative record income and cuty rec ratbeeenn seeocl th surdeco crespon verylose. er of am, prt silaacrosst ofth distbuon. conas retirt ince, 7%ave soetiren -co wre thea qua bidierence when weak e ministve ubmponen we dalys le pete basehethe hink

diributis ira ok a t fifthntil out %ave ince a 35 ac havrectd sll ditributfrommpyer dc accot d most would have been rolled over and we would've seen it when it came out. of course, quite swed a you can see largamounts for the top, you can see average amount bo about $10,000whic $7,5s befiincome gra idone py bad on t sur th y can d atom basiy shinthe sh of me aosthe in disttionndhe bigrange is soci surity sure. s.

69% social security and the average share in the fifth is 72%. when we do the administrate records, fall into the fifth quinti. this is misclassification mostly between social security and that area got bigger. the fifth decile is income share went up. unlike the bottom decile of the middle, this is a misclassification of dollars of 30% higher income in the middle. social security share is out the se. the boost in the retirement at the expense of the earnings share. again, when we bak down the type oretirement income across the distribution, we see retirement income is about0% in the middle and we are getting th 22% of overall income isv

thchangethehas is the sstict rcengeof socsecuty curityat lstalf th inco 3 ly ono/ e sociecury th ou50%.raordsits whe lo athe 90 thresh the fctione in fal ihalf f36% to erytng shallwas in inmeata fr012 t li the coo my aves and were able to extend our analysis back a quarter of the century to look at trends. in particular, you can see even in 10, there is a discrepancy about 20% from the cps and the administrative workers in the

gross to 30% in 2012 -- administrative and that grows to 30% in 22. in terms of poverty rate, transient is largely unaffected by this and every year you have a downward shift between the cps and the link to data. showing theaph trend over time in retirement. we can see the beginningf the --ies that in the are only that the 2 are only a little bit apar rerkably, there has been no increase in retired inme received. , very quickly through this, we examine the retirement transition similar to wh sed has done and asked the questio

in terms of relative metric -- and we ask -- similar what pete has de and asked the question in terms of relative metrics, we pull many of the link data and examining similar individuals. the reason we do that is that is only way to measure the administrative record across time. these are just some dude tells of the analysis -- some details of the analysis. is it isee here separate for men and women. according to the cps, the decline in median income in blue and we look at the same thing for the administrativeecords prior to claim social security and the two lines are essentially on top of each other. you see this divergent, much less evident of the decf clng siaurit

we lfor w also lessdencf declinher divgee hapnsarlier cee it p to aisocial curity youk inbsute mees liover rnd we not muchvince oferty ng ieiererie f50%athe hst povey o30 of poy aningsoo sar f m ior ciming al secuty rerent.o wor--o foomen, appe elier. smarizecomefa much qukl tin t minirave rec. ere is n aabruptng r rerent anddo n e an edence povty

ra ris when there is a ri in to enter percent off poverty, the rise is -- in two large percentff verty,he rise is lower in of poverty,n 200% the rise is lower in records than the survey. everything we did tod examines incomes of the population of 65 and older through 2012. it will be a big change from db2 dc -- db to dc. there is evidence that debt levels are rising among older households. maybe we're overstating things somewhat. finally, we examined the retirement transition for an abrupt changes run social security claiming that cannot use ou analysis to ask the

question what happens 20 years after retirement? say thank youe to and re is our contact and pass it on. andrew biggs: thank you very much for our first discsant will be kevin moore othe federal reserve board. : thank you. a very interting session. there we go. started, i will say the analysis are mine and not those of the federal reserve or the staff or ybody else of the deral resee bank. i will give some bef comments on today's papers. off the to there are interestg things as a practitioner in somebody who in my day job runs of the survey for finances ande are terested i there is a proem with collecting data from people and especially josh and ad's paper ere could be problems with people reporting income.

and is somebody who uses the tax data and pete and his colleagues have really, i thi pushed the envelope on using an lot of the information in returns to bring good things to bear on tir analysis. i will start off here they go with adam and josh's papers from the headlines are like andrew was saying, itooks like retirement incomis higher than many people thought from all the things we hear the popular press d academic research looked at this for a lg time. some of it is due tthe data resources as not as comprehensive as we thought. it is because it is an administrative data which has been something that has come on the sins the last 15 or 20 years and has been -- has come on the scene in the la 15 or 20 years.

soi, whoith people at are superhigh quality it will thwart with them the last 30 years and he has been a great relationship. it looks like they a missing some of e moderate income. what does it mean? it is good news. it measures poverty as of the low end. it is like go new aunthe -- itcheportyt thlo en enhey lot thtpoind t minirave dat ver datae 30feree me. i threw in a number from the scf for 2012. our median is a littl lower so that is a concern for to defend our surveys, people got to the mean than the aggregate

amount matches up. there mabe underreporting at the median that we should focus on. this?s driving this is super interesting to me. , thishart from the paper exains the whole thing in the sense that in retirement income, when you compare the survey data, it looks pretty good. retirement income across t distribution of hoehold income from basically the fifth percentile through 95th is 15% to 30% higher risk thais whateg ing thi issings not secy. people will get security and they seem to know that and rert it well in theurvey. and db.

at was seen is, i guess, or any household survey or any other survey's retirement is complicated r people. the amount of knowledge that people have on ivaries. age. is some with those were younger may not be thinking about it. even iyou think you on the doorep on it, people are ill informed othe choices. we have a case where peopl supposedly in it a recving information and receiving income are still: four. .- ill informed a lot of it is about the margins of the retirement plans and not at people e not sure how much of that are getting because of they are calling it after taxes or amounts that only keep half of it and give their spouse of the other half or something along ose lines area it is not at that. they are not reporting that they have it.

-- along those line -- it is not that. db planso an issue for and that was one of the most surprising is you always think the d would operate like social security, you get a monthly paymt and you would think you would know. ,hea survey practitioner variability that makes it hard to report because these are things they may not consider every time that they think it is like taking money out of my savings accounts or my ira. hard in ourong and survey about how we can better caption that. one of the things we asked about withdrawals from iras after the assets. and how much and how much is the withdrawal. we do not wait to ask about

income. we also do that with other types dc plans and that helps to capture more. still not perfect. the silver lining is if people reported, they are get it fairly well. you have tget people to report it. on aggregate, your reporting with higher levs. it seems that is good for capturing the aggregate amount but does not help with the medi household. paper.te's like he spoke of, they use an that people use for research, really, really nice data and they are looking at individuals 55 to 61 through 10. this is a period around social securi. one of the things of that impressed me with this paper is

they went all out. i have probably sure that caused pain in terms of linking those together because i am sure that everything do not match up perfectly the rst time they ran the job. even tugh it is really good, is still going to those issues. is an impressive use of t endpoints to sort of innation in using administrative data beyond 1040. that is something they are doing a great work on. ande measures of income the concept of work-related income. one thing as a tax policy person i am happy to see in the paper is of the shift in taxation. you think about it should be painfully obvious ifeople are not working and move out of retirement and ince taxes go away. they do distribution analysis also.

one of the things you see is social security is doing its job. mosttake away is individuals ab to manage the level of income. -- it isays would be nice to see how that goes them over the income distribution and how they replace rates even across the top 1%. this is showing around that time they cla socia security and it looks like things are looking faly well in terms of replacing. peters:'-- io sen. have more to say about pete's papers. the issues is ts is a great results fopeople in those cohorts. both the authors are upfront and not claiming what they are

--ally in their papers finding in their papers are not predictions for the future. fed these cohorts have hh .ncidence of dc and db that is changing fast. we see it for our younger cohorts, it is declining and dc just going toot work and putting in the time in getting the payout for the formula, active saving has to happen. there are variations of how people are saving. people have vastly different amounts at the end of the time. some of these results should be that these may not be what the future looks like in terms of people younger today get the time. one of the last things i will touch on is because of the shift in db plans is s important, one

of the things we are focused on is wealth. assets thatrd, this people have will be mo important than the stream of plans.y from some db l a little takeaway we have ongoing is we're trying to make an expanded version and we ask a lot of questions about of wealth incling 401(k)s. there are things you missed, we do not ask people to tell us the net value of the social security benefits because nobody would know that answer. or even most economists would die voted to some long conversation about mortality rates and everything you should use. -- not evil most economists or it would be some long -- not even most economists or it woul be some long conversation about

mortality rate in everything you es and everything you should use. we wanted to know where and how usolds are matching resources. i will show you one chart and cohorts. h -- two a cohorts. the nonretiremt walled is everythingcohorts is not in your retirement and you see were the main wealth is. wealth, add in dc pretty decent effect on mean wealth. the real kick is this present value of social secuty wealth. you do if our older households and that is what you would expect, but the greater.

the point is we know we are not getting everything whenevewe ask people. our focus here is to build a concept of trying to measure from a wealth side of things along with the income si. one last plug for the scf is the result from the 2016 will be availablat the endf september. thanks. andrew biggs: thank you, kevin. all resected discusst is bruce meyer from chicago. -- our second discussant is bruce meyer from chicago. refer toer: pleas this, i was a late replacement. you can consider me the closer if you want. these pers are wonderful. the topics especially important given the magnitude o

expenditures on soal security given the modern government is pensionund that also ten -- that also happens to have an army. i like the papers also because ey hit othe fact that as a decline in poverty over the last 50 years has been one of the achievementserty , thee u.s. and the papers mitchell paper, in particular, indicates that the climate of poverty has been even greater than indicated by official statistics. becauseike the papers they support much of what i have been arguing for the las50 years, they fill in large gaps i have not been able to fill in.

i have been trying to get the social security administration to do a version of this be e-mitchell paper. i submit a proposal that it is rejected. hand, i don'ther in like the fact i am now in the age of the study group here. my twentysomething kids gleefully reported that not once but twice i gothe senior discount on a trip. summarize first mitchell paper. they fell by half recipients of income -- they say about half recipients of income do not report the pensionhey receive. they replace the self reports of

pension and reports of income from other transfer programs and it used tota remarkable findings. first, the poverty rates is much lower. second, the poverty rate falls more overtime, that is not qui how they sell it. figures, thet the -- thees between divergence between the survey and administrative numbers that the larger over much of the time and narrow at the end. once -- one should emphasize that the divergence pretty much constantly in a amount on a much smaller base over time. in percentage terms, the reduction is much greater

overtime. median income for those 65 and over are much higher than we thought when we bring in the administrative data and video incomes rise much more over time -- and mian incomes rise much more overtime. is small orincome noxistent and there are really interesting results. to put these results in context, i want to emphasize all surveys are seeing the deterioration in their quality. it is not just pension income that is underreported, half the staff is not reported in the cps in cash welfare. surveyk of the replaces report of rtain income components with administrative measures see similarly large effects. i have emphasized in my work

looking at consumption data if you look at consumption-based measures that look at what people are able to spend in terms of their housing, food, transportation and other goods muchervices and you find in it the way of similarities between the patterns and the consumption data and those a the corrected income data, corrected are the ones where the administrative data is substituted for the err- riddled survey results for we have jim sullivan who has argued that the looking at consumption data is conceptually a better way of looking at well-being than reported income. we conjectured and i will quote our brookings paper from012

"the two most plausible explanations for the difference between the changes in income and consumption poverty are measurement errornd saving to saving. there is conderable -- it is important for families with few resources and transfer income particularly relevant for these families, signifintly underreported in surveys and the extent of this underreporting hagrown over time." there's a lot of silarities with the findings in the papers here. the figure on the fourth page of the handout, there were reported consumption povertmeasures for men and years and the past 30 you will see that when you look circles which are or the open triangles

for women, those numbers start to diverge and show lower poverty rates and sometime around 1990 or a ltle later and the consumption measures higher decline in poverty over time r both men and women. bee andnt with wha mitchell are finding. a versionlide shows of this same issue applied not to be age but the overall population. here we are using data from the -mitchell used to we are looking at underreporting of snap, which is renamed food

stamps and housing. when you use the administrative values rather than survey purports, you find the poverty reduction for these three programs combined is 90% greater than ifouooatheury rertndheitifrsui aitors atreisd. wh y aou f t rafenrng o e usolsuey da,he fes tse pgrs reci per imu ear. wh a t picleon re e dey e chetr f anhe aea. its poanbeusitay afcthe wart fur an-perffts n, e ns buauilelse t per

rertoras yr at e meim tens llls -- ssewi rea aew ase ll t splenl veyeare ens llelse nemeurcaedhe ppmealovtyeare itoeso tngtht tnk e rll mtan aearef vey r e dey ats gh th t oiclatanon fo cldn,t mh we isorofhi tt is wfgleare tt lle leedlo wh effia meurtwwes om n,he rsit dss onisha eeiteliny

owhathportra o th eer imuorha thofcieareecsef endrertgf nsn co. en who tsorctn,e ouav kwnf vey s pealvetad r e dey. mo hf tic sndg foa mi ioungnd trsrtio % othe a oe reywnomana c, fdar aea tenar d , need t mata dta -inmes teed mntn e anrdfivg. mongn cng ince

itedt,eean d adl. a a colenty. omadntess s-vithr pernd e tcllap us loer hiorofasarngo isbltoet bte msu ofhathbali io i th bdyt l peacun foaxhi ive imrtt nc tesala l inetemt. lerurhath ofn es ltlbiwrin cungnvegean dis. wh w rllneedo t r nd aunishashref op dnoha eugto rve ooro wl

rerent inorely qstn mes miautwh sre eoe ve sstti fl tirtaarofivg? bdyt l.ap fd eier 1 opelehoav a llf lst 4 tee arutrothye bor rerentndenoro tcllhe ia lln inmeeli tlo-tm onme poro cimg , iia sstti dendonoyolo ait d w cheit y p onhaapneit0,0,% dendg h y msu

in? a ve hesrob tre ou bmoesrcoong atth src once r isro bau pbay i isoteay 1 ttava llf % me. therence src tt e t ptedn era e alat wlde ef tbrg thgsik sp,s wl d g ptu oling rcstceofhiro a i ulare omoongt nsptnhi ihi ia chetr wf ptinth lingtaarofhiro ginhe dabs ateoe rey ve e nel snserinhi

suey da iomis mieangndsi adniraveatby ielo a cbiti osuey da go aon w a svi t obmsitthsueyat d osatheotm ve reesrc tn hind rrt stutnsotpuic dratrehortn e edth iapar anewigha y vy chbre. amogizt oti a wt me re lvenoh me id o qstns i llum a me re lee ou te r qutis. osbu borfo ceaiy pe n igin ateavto ctnege da in aa inin tinde owowol tay e dng a tt nn tl

nesri wreeleil ben e fur wchs en e li qeson tt fa. t e meim a rcti tt ncathi goengendratial nsnsthgsav bn ecni. stee peio seiay, time io shldoulyisitth sa re eni ppl fr rirenf u d at meatofncasov te d u ulayhedeac ofetemt befs d plinwod aybo t sa. a sweso da okg ans irerentncess asedhrghhersat y heomidowhe retetohaes a a e-time enis er atam pio

y he fei f t trd t pt up o cas d h iom cng reti tpretemnt eain? haisitel rtnlsothg sechn ogssssi t loitin da rort i loinatner o arin rerentelivtohe ngtearng. thkf u bk t stt,ouaneehencen ofavg time iomw mu ler seon- dr bgsoueance fr pvaetemt? jouaitel: ive time, s. d mbeditthsoal cutyepcentat ud b y cldavso tanerou quti, sses vi stdas eroblmobl to mntn en lki a

eayerd omur pernd hes j iyolo - ifouoo a 10. th mhthe ela se thdirencs twn r peanthprio lirare isn enxpdire fl, eyerloinateoe th ionsshawe reri ihe9's d rhs eas tt vi stdas d an me rulyft riren, t veth 6 he meinof iveetemt co, wod pe tt vg anrd tbebltoe mataed : mef at a gog tsa iwh ihi i owndomofhaihoht kn. weood tasesrc

anits uenbl tt op ha bn ttg tt o i rerent a rgpa ixpsi o soalecit >>eoe veeegein beerf irerent no aar pt oth i wee d e paioof sury. se ppl s is f nlr ant bth peiopl ahas fe, t h aexnsn n e 0s pgr, ffen 'sesneto baenon otm a tt' hpea send, lofimeoe pu h my op g peioinmero cere wki. d e t tng tt bor ', t enhoas wki aan plerndovedy pl t ytngro i om o tsead0 ar yohato rirfm e pn g ytngroth pn al e tulyn e c,

thexntheatrnol, ncasuay e in psi iominhe70 an 's d tohe ely ocu o a '. d ke aot oth iue ft atot oy at op werdiore acal g setngrot esngie, tited tfiea a tn tueshaha ud ersndhe i pcuor t ne i a loedt dfentohtsnd th td sn crse heueioise' d rllba eno f my yes n, sm s stlin isnd t haa lg-rm tuov. tnkouermu. o ens w u to qstnsroheudnc se lkwi miopne iyo cldpe in t me peal bau he sp he d ve

ailti.ou heen aonig he t fnt >>rolypein ai t, t98 amdmtsomhaeded plemtat. er t pio scehe t or y'vtos athe bn gwtrelyn t rontofnce poed prat psis vio kis. am pio,he pretemt iomhagrn mkey ssqun s diriutn. mquti, d 'a stemt b is qutin iist t t ce iwh yr sryel thhrerentysm i

j ovy od exndg to rirenth iquit tt obrv iarng a oer co prerentsoha epci aaifuy d iakhaas te ay per y'r sady fthlla crsi dref iquit eriren >> d'thi t riren u--cueny s men,ouavroesve ci suryenit hele thk n lernce, roh naonf ct etorbefi o oft,utn nel,he iprabte nogog tcontacthgs at a hpeng bor isentani in siedo y g a

plemt teo oun' eecavth ir, >>ir, wa tthk bce lti mno thnaiswi a eecve haweg vey out sen aintriv da. ado i ulhae toheenrn dg f a bumyueio i o omy jo ito lk toheute cendstm anbo io tirias d hes y'rloinin st ,0,0 arfr n, ve t y'vobrv fm anti, n u n pte, atan uel m ist oi te % ilmo inmerothdmisate ta mclnt bed tsraitn?

>>th s mhte ees da srcweavon ts ghno owh wre e' tinto lk t roueshapepl wod he aiblanay tta tse socefopeplat time a st o is ojti o t s h a peleoi tdof eyav th p omoy snd on oththgs tt' s,ouc ouit owthe'so oises o inrn ctr f ppl ft,s u t t c. pnsndheha t magehi rhtndo w hltagth gen shks gennyth kd shksha hpe rht fr ou oeoe eyou bbeer inal tireaht tmro se thsees althreurs ateoe ulha a rirenwi th st eand ase wee yi tget d en

anooathis keow cldtrchh o ro le snssi tin keorli resndhis keha qstn.cu 'sraht wh loof sutisbo at gng hpe heigstyokn, ecofhaoue ki i s intt d. unrtaty crsehe ceaiy ha cmes.ke f hi fstfllcomis da atalin is a tli y wt haen. 'r ls cutet focain buthe'aewhis. smso apion atveonetedita ldat a psi. heatr a tys enonha cta ris. din beft naavag p auay ve loofisinha t befi a triy ck ad. y nnlha two f a mpy r lo te g tlee e mpy oso rm

rerentge bau aotf omal.t d t iahaevyo, rtullyownce ppl ona olotf soon w rir fm diniot e of l-iompeonno tt wt eijo i th'r pbay lstide co, g? w d'tnowh'soi t haen the'ceaiy sk o de therehis ate mht inarbeerbot, rtullyorow iom tt dhexnt won w tdo re lelbe he wkfcehayo bud isve te,t n asacoad oerhi y c lk ishas inonnt tse thgsani in--o rtnl treayavbe so wd llfrm fid netstheoe r enitth we ner tulyund,uts ars negog tohepls, u' lki acoinit emoyonibio, plee gngon 1 opa to.cla.

ann'thk 'ver d atoi io din pn. wdotn wt e tu isoi t b i'nogog t-- d'tha cst bl. ers rtnl uerin. t e aumioth wre fm pfe wldo peecor, eye o lsft rl, ers rk,ndt'erunrtn at gngo hpe y y okt t da, lo athe atre rhtt time,e' h ahi fr wt pefla w sftroho ma pplha i t ceiocaedomprle foyoge ppl b n time, %f usol hahiinmewoin eeehds a erha otr reah rtnlwh he ty y meurth a y dsea if ieah om.b t c.hi an'seng mh a dli ora ith uerintt' i n'te y wt'gog haen i20 yrs

sha tson putoicunn dngig ath cysr baou eidaobst -wh ihohtasoi toe pt meobtued t bmo tn rtim s thheedal orrnnt orsht brd er ro d e eson i whheth wlde mir a ryer blo. ro fin siifannclroem athipot. aso delo soceth wanutnt puto rooveife re inin tba tmutwee noabo. wod y a n ilid ssehai in the'a anou fli o ndsd atueo co tstctal cng ma t iansuaiblnd

oniclyrongveth tu. soheho awe inohe wi bnoaiut focod lko ngss eye cledo d i tt.uiolw- t ialofhe ces ecedebanhe's serede a whhrl ic t prt uncud btncdehaof eenonun. ic eofueo nstuonwhh ss t scicayc at ttheenal litn btsheos seorndossere in cyserth ce atou he5%f eew mpyto wt t heecedbthi btot cmm dn aep f ctonhe doar nsvaveanfr mke s i rech stncyppedo e am

uw.fhe mqutis w hi t sendsomne prenngheveig brd ataslledhe veme ay00of t psi h c y se th at >>'lbeer qck th i o-ticndoue intoosmo o t kw thgh w ouprtic eoa hnstud et drat cng t t p sye treilbe0% nenonenit alg wh rgcu i pein wlean ofthhis. he'soby t ian d skelds t isnd bitenons,e tinsbe icrits o eotakg gnict saifes ueio, rrit the. i jt aitenfhe ntesd. nodisusaiten

riren oth fit watoomimt e neonherentio tt we me icer unrsndleo a lpeon niob i veueioabt eor nh m b srt ueray core thageax a etr netshama aibl d etr u n lk aut eecofhamitapn, popotil pepeivon t kd fiin atouavrit owhaigfft >>reoualinbo ifng01)srenon aot bis hi trereti chge a i gssod o autoolywee he wt ans heax

pocyreei pseedy thadnitrio i tki abt thottienit a e aibl tpele diertevs ncmend ay i wchhe xema b ali tpele dfent lelofnce. jh,et dyohany outsn tt? is erbrd pian m t into lueehas into pp. ugstnto meini n'wa tapn r re twl y ha aooon tatrtmt enonana t opross tt pped t ptndowhe afcted t treme. d iig rerouoha oni veomsht bog y d'wa tre 2 pa bk,heesibso o si ghpot y toha b ion wto apn,hamit rtullyft'sothg dotan

>>osisoon eer he. a'. tei ve aueiofoyo the'aotn he,ot pe tt e ry iestg d alng aotf e sptns l ousav m lki ath gph f iomquti a inho y'r sinth n miaishe reacenra n gngow buit auay gngp. d iuive, w w unrsanha wal sw at t inea ipralymo the w ctueo rkft clmi siaseri. f tsehoreoonr woin is slht iree, ig'sots socomis ve dge w eanghe tleane rmtoesrod em da,net t in ttee tbe ourng

atontial ocoinng or tsin towt inle9 co d'tav a iar rucon rngswheathhier pas t iom diriti, enho w cotieo rky s a eaeainhr anednarng sohait sges mis heotm o a ve l iommabesi iss wa splen mae eyer luity nsaid. 's bstoesrc a th'rsi itoncas thrnce,ay ty ul efe eylaed ciecit b wdot la.sh anin ion wt rd o ch to tlmbs, t ndncen,ia tt okli tsethe wk ectiueo hang dfent tn os at top ifhecoin torth'r ki pttmu wt e reakg boran ain

adtial iom naib qstn] edn s. naib qstn] wl, l o tmav otr ors. i25onhe tt, 2 a oy ryi o soalecit avtonoth t soalecitsyemas veroesveenit stcte. aowarr,ou a fuay plinve10 ur ara letearng inatndjte n we nftibu ers very high replacement rates at the bottom, they're getting higher replacement rates for social security, but there are some that have nsions and have other income. question] >> well, the replacement rates are inflation adjusted.

about inflation adjusted income in d. plus one before retirement. the periods, it wouldn't make much of a difference. >> what ias talking about is when people talk the bottom of the formula is 90% of wage adjusted lifetimearnin, turns out at least historically, we hope it happens in the future, wages hav grown inflation so as a percenwha what they actlly earned in real terms, you're replacing a higher percentage. to add one more thing, the tes are in there, too. tax is going to be huge and social security is not be taxed so their net of the tax bill goes down and that helps also. >> i'm glad you read it more recently than have because i forgot that. >> if you had somebody in the 90% replacemt factor at the very low end, but they're losing

security and medicare they're nx and paying txes in retirement, you're getting close to 100%. goldman, senior penon fellow at e american acemy of actuaries so there is an room.y inhe my question gets back to this convertingblem of savings into lifetime income and could it be one explanation, peter, for t. plus 3 is that people draw down enough assets same standardeir of living, but if you were to continue your study and look at t.20, we may see marked people's ability to maintain standard ofiving? it's good news if people saved last ree years but do we know anything beyond that? >> so from my paper, i can't say anything beyond three years, but i can tell you work that we have done with the data. all, although we use

t. plus 3 in t paper, pa of trade-off. the further we look out, the fewer people we can look at. we basicallsad what happens if we look two years, three years out. those.ooked at there isn't a huge differce and actually, 've done sensitivity analysis and looked further back. those don't seem to change it. the other thing we did do is we ope each cohort in '99 and percentage that had income afterwards. at other age ranges. 'll look at people who were 70 '99, 60 to 70, and so forth. the problem with that is that he 55 to 61se years old who are worki, the tax da is representative of sample, nonfilers a not an issue. i don't know that for the other group. those who were filing in '99 of those age groups and we

look out, the perctage that have pension ince is not off.ng if we follow people for 10 years and all these different cohorts up as a synthetic cohort, it goes up and it down.t really go now, i'm not sure how representative those other groups are becse i don't have non-filers and i think non-filing goes up as you age. but noing i can see in the it's going away, but i can't just from the analysis i've done rule it out certainly. >> you don't know how much wealth they have accumulated, right? thing we can look at forma wealth from the 5498, but we can't look at d.c. wealth, we don know d.b. and other things. we don't know any other taxable wealth. flows. have the income >> josh, do you have any thoughts on that? >> we also have the same problem the data ending in 2012 so

we can't do our synthetic cohort the future, but we do look in the cross-section, one of the main results i presented, 85 plus group did have substantial boosts in income rative to the survey the fact that poverty rates higher.y sowhat obviously, that's mixing cohorts and ages, but that evidence there is consistent with our toatively high ability mainin some absolute living andard. >> i think the questions you're you might want to look at work done on data, something like mike heard or from rand who use the hrs c survey. some work with scf. at older retirees where you think these are the folks money,e running out of you would expect they would cut back substantially on expenditus,y gifts to people, charitable

contributions, things like that. it, and illy see think the work, which follows individual household over time find thathrs doesn't many people are ruing out of money. questions? oh, sure. >> i'm with the economic pic ant toe and i congratulate a very interesting panel, i've really appreciated it. those papers are interesting. my question -- well, first of josh, bick question for you mentioned at the end of your paper that you are looking at 2014 resign of cps ambigguous -- and i us whoust for those of are sort ostuck with using the corrections to make for the cps data, very much appreciate it if you would repeat the exercise with

administrative data, but the y for both youis and kevin moore, this is very speculative, but the most muctheinghing is how d.b. income is underreported. we knew was underreprted, but kevin moore was surprising and if you have any idea why it is that it's so underreported given a steady source of income. >> so we've looked into this in checking if these d.b. withdrawals are one time or really doe and they seem to be steady payments as you say so it's a b of a puzzle. to explain it in terms of demograpc groups and so on and we don't find much there. i think the standard explations for survey may apply here, such as fatigue when the questions getsked and the of questions, people may

interview. end the also, some notions about what is asome, people may view this i earned this during my working not viewing it the way we would as income. issue.another potential and then lastly, i know that from some other surveys there's th finding that people tend to report their income better when they actually turn around and spend that income. just getxample, if you a distribution and it goes into and thisngs account could be d.b. or it could be d.c. and from one account to another, you might forget to report that as income for the didn't spende you period.year or that >> so one of the things i knew coveredto this is we d.b. income penon well, but d.c. i've never been satisfied with that explanation and someone who

does the survey can explain it better to me, but i've never been surveyed, but if you just ad the survey questions, the literally says, do you receive any payments of anything other than social security and part of the things it includes is profit sharing plans. now, many people may not know th, but 96% of profit sring are 401(k) plans. the regular payment item was ther brought up until second question, which is what is the source of this, and then option.was an so we'veone a lot of studies, comparisosve done of the cross-sectional data of cps, josh's of the paper came out, maybe some of the things won't be a shock to find many of the same things and, you know, it be the d.c. that's the undreporting, it's just seems to begh, it regular payments and back in

1990, when there weren't a lot d.c.s, we were underreporting this. we are measuring this badly and i don't know if we really know measuring it badly and let me just put a little caveat intoying to separate this d.b. and d.c. i would like to sarate this into d.b. and d.c. hard to do with the tax data. however, since there are many things i knew, i don't want want knowap to a new thing i that really isn't true. i think some humility is needed. exactly really know where this is coming from and we can try different ways to get at it, but until it's important for tax authority to track whether it's d.b. or d.c., as would wantmakers it, it's going to be hard to figure this out and i like going plans and we can try to narrow it down and identify it. just a little caution on what's d.b. and what's d.c., i don't think we really know and even know what's a non-qualified annuity payment.

we're going to have to call time now and i want comingk all of you for today and i want to thank all of speakers. i'm hoping our speakers and be willing toll stick around afterwards and didn'tany questions we get to during the event. thank you for coming. [applause]

>> andow, more le coverage on c-span. joining the global taiwan institute for a discussion on china-taiwan relations. scholars,earing from journalists and a former cia analyst about the