survey of income and program participation

please stand by we're about to begin. good day, and welcome to the self-sufficiencyresearch clearinghouse�s survey of income andprogram participation: overview, reengineering, and self-sufficiency research webinar. today�scall is being recorded. at this time i�d like to turnthe call over to zakia redd, research scientist at child trends. please goahead. thank you, and good afternoon, everyone. beforewe delve into today�s webinar on the survey of income and program participation,or the sipp, i would like to provide a brief overview of the selfsufficiencyresearch clearinghouse, or the ssrc, which

is sponsored by the office of planning, research,and evaluation, in the administration for childrenand families, u.s. department of health and human services.the ssrc is a virtual portal of research and other resources related to self-sufficiency.it functions as an online community for researchers, practitioners,and other stakeholders interested in self-sufficiency, employment, and family and child well-being.the ssrc�s purpose is to disseminate quality research. thereare currently over 4,400 items in the library, and new resources are added on a regular basis.the library�s materials are organized into 12 topical areasthat are shown in the drop-down screenshot

on this slide. everyitem included in the library is reviewed for relevancy, and users may search by key word,or use filters like topic area, target population, geographiclocation, or research methodology to browse the collection. everytopic under the browse topics tab includes an our librarian recommends box that highlightsresearch and resources recommended by the ssrc libraryteam. each topical area page also includes relevant federal lawsand regulations. under the stay connected tab, you�ll findentities involved in self-sufficiency research, a robustevents calendar, and a list of organizations

that have partnered with the ssrc to hostwebinars and other events. please use this tab to access previouswebinar recordings and additional materials. on the screen is asnapshot of the new students� corner and professors� place. through these tabs, professorsmay access selfsufficiency related course material or reading, and studentsmay browse through the ssrc library resources to supplement their coursework, as well asto explore learning opportunities within the field of self-sufficiency. finally, the ssrc houses a number of datasetsand sources that can be used to support program management, or to explore research questionsrelated to low-income self-sufficiency, such

as the sipp, whichwe�ll be hearing about today. on the right side of your screen, you�ll find some quicklinks to the ssrc. select the title, and then click the �browse to�button for those links to open in a new window. dr. jason fields, who directs the sipp atthe u.s. census bureau, was unable to be with us today.however, we�re excited to have with us in his place dr. matthew marlay, the assistantsurvey director for content and evaluation. and, after matthew,we will also hear from ashley edwards, a survey statistician inthe poverty statistics branch of the u.s. census bureau.you may submit questions to our speakers at

any time during this webinar using the questionand answer feature at the bottom right side ofthe webinar platform. questions will be answered after bothpresentations. if we do not get to all of the questions during the webinar, responseswill be posted to the ssrc following the event, along with otherwebinar materials. and, finally, we encourage you to join today�sconversation on twitter using the ssrc webinar hashtag displayed on the screen. tweets usingthis hashtag will be displayed on the left side of the webinarplatform. and, with that, i will now turn it over to matthew marlay.hi. good afternoon, everybody. we here at

the census bureau are really excited to beparticipating in this webinar. i think one of our core missions right now is to sortof broaden the user base and get the word out there about what sipp is,what it does, and what it can offer you. so, i am really glad to behere. i thank zakia and the other people for setting this up, and i hope you find it very,very useful. so, just a brief overview of what i am goingto be talking about today. i will go over a little bit ofwhat sipp is, the kind of content that is in the sipp, some of the recent changes � andby that, i mean we are in the process of reengineering the instrument,and i will walk through what that means for

the content, aswell as for the survey administration itself. i will talk a little bit about where you canget sipp data, how to access it, and then some things you can dowith it, as well as some considerations to keep in mind. and theni will turn it over to ashley, and ashley will talk about actually using the sipp data,and give some examples of research that we have done using the data.so, sipp is a nationally representative, longitudinal, multi-stage stratified sample. it is a surveythat has continuous data that are in panels roughly three to four years long, startingin the mid-1980s and continuing through the present day. we justrecently wrapped up interviewing for wave

2 of a panel thatstarted in 2014, and will run through 2017. the sample of the sipp, it covers civilian,non-institutionalized households in the united states.so, it does not include people who are on active duty in the military, or people whoare in institutional group quarters, such as prisons or nursing homes.the mission of the survey, we want to provide a nationally representative sample that helpspeople evaluate annual and sub-annual dynamics of income, and that is particularly monthlydata as income and other measures. we want to look at movementsin and out of government transfer programs, and that isthings like wic (women, infants, and children)

or tanf (temporary assistance to needy families),food stamps. we want to look at family and socialcontext of individuals and households, as well as families, andinteractions between these items, because none of this happens in a vacuum. it�s allconnected, and the sipp has the data that allows you to see thoseinteractions. i will say that several of the slides andthe content and you are going to see are courtesy of lukeshaefer, who is a professor at the university of michigan. he is in the school of socialwork there. he uses the sipp extensively in his research, and he hasdone some similar webinars and workshops,

and he was kindenough to give us some of his materials for that.so, a little bit of background about the sipp. the sipp was originally designed to compensatefor some limitations that were inherent to thecurrent population survey, or the cps. the march supplement ofthe cps, which is now known, which is now called the asec, the annual social economicsupplement, it uses a fairly long recall period and it has traditionalquestion design. and the downside of that is that it is not goodat measuring the irregular or odd sources of income. so, if you have a regular 8:00to 5:00 office job, the cps

can cover that pretty well. if you work, ifyou have spotty work, or you work and then you don�t work, and thenyou work again, it�s less good at capturing that. and so, there is a need to get thatdata. asec also suffers from high levels of under-reporting of programparticipation, and it doesn�t always capture changes in familystructure. and so, there was a real need for a survey that addressed those problems, andsipp was the result of that. sipp is designed to fill that need.it is designed to have a shorter recall period, and that enables peopleto collect, or enables the census bureau to collect monthly dynamics of its respondents.so, it is meant to provide better estimates

of income and public program participation.as a result, the survey offers the most detailed incomeand comprehensive program participation variables of any of themajor nationally representative surveys from the census bureau or elsewhere.between 1984 and 2008, the sipp, the content was collected in separate sections. so, everywave, we asked a set of core questions, and thosewere things like demographics, as well as some labor forcequestions. and then, at the end of the core sections, waves would have different setsof topical modules. and so, topical modules could be anything fromchildcare to commuting and work schedule,

to residents� history.it covered a whole gamut of topics, but it wasn�t asked every time. sometimes it wasasked once per panel. sometimes it was asked once per year, so severaltimes per panel, but it wasn�t asked consistently. the contentwithin the topical modules was not asked consistently. with the 2014 redesign, which i�ll givemore detail about a little bit later, the survey has beenchanged such that we are now collecting everything together. there is no longer this distinctionbetween core and topical modules. it is a single interviewthat collects the full year�s-worth of data at once.so, you can see from this slide that we have

run a number of different panels, startingin 1984. and, originally, you can see that there isa 1984, a 1985, a 1986 panel. those panels were run overlapping. so,the �84 panel was still ongoing when the �85 panel started. the �85 panel was stillongoing when the �86 panel started, etcetera. more recently, it has changedto a design where it is no longer overlapping panels. so, the�96 panel began in april of �96 and ended in 2000. then, the 2001 panel started, etcetera.and, you can see that in february of 2014, we went into thefield with the first wave of the 2014 panel, and that panel is stillongoing as we speak.

so, as far as what�s included in sipp, andthis especially in this section, i am more talking aboutthe old design of sipp, but i am also including, to a large extent, the 2014 panel. but, thecore files for the 2008 and earlier sipp include demographics � sothat is race and ethnicity, age, sex, etcetera. dr. marlay: income measures, and that�sincome of the person, families, and households, as well as the totalincome, property income, other income, as well as information about poverty and whetheror not people receive severance pay.and then, there are a series of questions about employment, and the employment questionsinclude whether the person was employed or

unemployed, or not in the labor force. andthey collected data for up to two jobs per months, as well asthe industry and occupation of those jobs, what kind of job it was,whether it was a public government job or a private sector job, how big the firm was,how much the person made, and whether the person had health insurance.it also included program participation measures, and you can see that we ask about a largevariety of programs, from unemployment insurance tofood stamps, to receiving and paying child support, whetherpeople have health insurance, whether their children are eligible and receive the freeor reduced price lunch

at their school, whether they get veteranspayments or have pell grants, and other programs, other federalprograms as well. and, the advantage of this is that most ofthe program variables include both a dichotomous, didyou-receive-itvariable, and then if you say you did receive it, it subsequently asks how much you got.so, it is a good measure of total household income,because you can see exactly how much in every month theyreceive from every program. but i will say this is not a comprehensivelist. i have gone through a lot, but it is a very rich survey,and if you want to see the link here, this

goes to our user�s guide, which containsa full, a full set of all the variables that we have in the survey.that was the core content and, as i mentioned earlier, there is a differentiation betweencore, which is asked every time we interviewed youin the sipp, versus topical modules, which are asked once perpanel or multiple times per panel, but they�re not asked every wave. and so, these topicalmodules include questions that are added to the core, andthey include topics such as fertility history, migration history, foodsecurity, assets and liabilities, healthcare utilization, what kinds of medical expensesyou might have paid, as

well as your work schedule.as before, this is not a comprehensive list. the link on the slide there points you toour website, which has a complete list of the topical modules,as well as what wave they are asked in and how frequentlythey�re asked. so, all of this information covered the sippthrough the 2008 panel, but at that time, sipp was indanger of going away. it is an expensive survey. there is a heavy respondent burden. and so,the census bureau had a mandate to, if they wanted tokeep sipp, they needed to make some changes. and they wantedto do things like reduce cost, reduce respondent

burden, improve the data processing system,modernize the instrument � that is a significant one becausethe sipp used to be administered on a dos-based instrumentand we needed to move it to a windows-based instrument � as well as expanding the useof administrative records. so, with the redesign, the sipp hasnow moved to an annual interview. so, we talk to you once a yearinstead of three times per year, but there is an event history calendar. and i will showan actual visual example of that in a minute. but the event historycalendar literally is a calendar, and it lets you see throughout theyear, month by month, what changed within

your family situation or your household.so, this table is just a quick comparison of sipp from 2014, and what we now lovinglyrefer to as sipp classic, which is 2008 and before. so,you can see that it went from a dos-based instrument to aninstrument that is coded in blaise, which is a survey software that the census bureauas well as other large survey organizations use. it�s still donemostly by personal visit with some telephone follow-up. we nowinterview annually instead of three times a year, and that leads to a reference periodthat is now a full previous year instead of the previous four months.the panel length is about the same. it varied

from two-and-a-half to five years. now we�replanning on about a four-year panel, but the instrument is flexible enough that the panelcan be longer or shorter as needs dictate. sample size is slightlylarger than it�s been in the past, but it covers the same universe,the civilian, non-institutionalized population of the united states.the content is comprehensive in both. we did not lose a lot of content in the redesign,but one of the big advantages for data users, andi hope for a lot of the people on this call, is that the file structure ismuch more simplified. it is a person-month data for the full calendar year in a singlefile, so it eliminates a lot

of the merging files and joining them togetherthat used to, it used to be an issue for a lot of people whenthey were using sipp data. so, just a few of the key design changes thathappened as part of this reengineering. i will sayagain that the scope is pretty similar. it�s broader than what sipp core used to be, andit includes a lot of the key topical module content. and the advantageis that it integrates the concepts a lot better. we now havean event history calendar, which allows you to integrate reporting across domains. andby that i mean if, for example, you are telling us about your healthinsurance, and you say, �well, i stopped

getting health insurancebecause i lost my job,� well, then the field representative can very easily go up to thejobs topic and say, �okay. well, you told us that you lost your healthinsurance when you lost your job, and that was in march of thisyear. let�s talk about that,� and then they can talk about health insurance. andthen they can say, �well, because of that, i also started receivingfood stamps.� and then, the interviewer can go down to the foodstamps section and immediately start entering data about the food stamps. so, it allowsa much more flexible and conversational interaction between thefield representative and the interviewee.

dr. marlay: another advantage is that thetopics that used to be add-on topical modules are now integrated,and the instrument is designed to facilitate hooks that allow supplements if we need toask about additional content. so, overall, it greatly enhancedthe flexibility of administration, as well as the people actuallyadministering the survey and receiving the survey.so, this slide shows the event history calendar that i was discussing a second ago. you cansee that along the top are the different monthsin both the interview year as well as the reference year. and then,along the side are the different topics. so,

we had them start talking about residences,and then they can go down through marital history, education, wecollect information up to seven jobs, and then we get intogovernment programs like ssi (supplemental security income) and food stamps. and, finally,we ask about a number of types of health insurance. but,as i said before, they don�t have to march through it line by line.they can go through it in a way that is natural to the respondent, and hopefully that theresult of that is that we get better data, because it facilitatesrespondent recall. the reengineering has increased a lot of benefitsfor us. it has increased our efficiency, both

inprocessing and producing data products. it has certainly given us flexibility in administration.we can now interview any time between january and juneof the interview year. it used to be the case that, the way theinstrument was set up, if you did not interview in the month that you were set to interview,nothing would work. and so, you would just have to not takethat interview until you talked to the person in the next wave.now, it is a lot more flexible. there is also a dynamic reference period. we have it setat a year. you can see on the previous slide that there is this barhere to indicate, to set apart the interview

year from the reference year,but we can adjust that as necessary. we could actually extend the reference period to bethe full 18 months if need be.another advantage is that, because it is a longitudinal survey, there are a lot of thingsthat we know that we learn in the first wave, thatwe don�t need to ask about in the subsequent wave. and so, weincorporated what we called dependent data, or feedback data, into the instrument, sothat things you tell us in a wave that aren�t going to change, likeyour date of birth or your name, or your sex, and a few other things,as well as the fact that you have health insurance,

or that you have a certain job, or whatever,that can come back into the instrument, and we don�t haveto go through all the questions again. we can take what therespondents have already told us and streamline the subsequent interviews for them.so, all in all, the reengineering, which i have talked about pretty quickly, but covereda period of several years. we did a number of field testsand a number of iterations of the instrument. but all in all, wehave concluded that it was a very successful reengineering. it produced estimates, thecurrent version of the survey produces estimates that are not substantiallydifferent from that in the old survey. we

have done anumber of comparisons with administrative records, and the new survey lines up wellwith the administrative data, at least as well, often better, thanthe old sipp did. and, by administrative data, i mean things like wehave some records from the social security administration telling us who received ssi,for example. and so, we can compare people that tell us they getssi with what social security has on file. and so, those estimatesare better, generally better in the new sipp. we do still have seams. that is sort of justa function of being a panel survey. the transitions falldisproportionately on the seams. now the seams

have been moved to december to january, insteadof being staggered throughout the year. so, if youlook at the data, you can see that there are sort of spikes at thebeginning of the year, because people say, �oh, yes. i have that, and i started atthe beginning of the year.� now, all of that said, i don�t want to takeanything away from the 2008 panel. there is still a lot ofgood stuff there. there is a lot of time to use it, but actually, it started in 2008,but was administered through the end of 2013. so, it is a pretty recentsurvey, especially in its later waves. in 2014, the content that the survey coversis listed here, and if you have questions

about specificcontent, we can cover those in the q&a, but here is just kind of an overview of the differentkinds of content. so, when you first get into the interview,it asks a few questions about the housing unit. it then gets into theroster and asks demographic questions about everybody in the household, including peoplewho lived there during the year but don�t live there currently.and, we refer to those as type 2 people. so, we collect someinformation about those people as well. and then, for each person in the household, weproceed through the interview.so, we then proceed to the event history calendar

(ehc) and we collect this information abouteach person in the household. and then, at the end of the ehc, there is a set of questionsabout things like health insurance, asset ownership, householdexpenses, healthcare utilization and healthcare expenses, aswell as disability and fertility history, and some information about parents� nativityand mortality. and, after that, then we close out of theinterview. the respondent identification policy is we getthe respondent�s explicit consent to feed their data back. so, i talked before aboutdependent data and how that can help us shorten the interview insubsequent ways, but to do that we need the

respondent�s explicitconsent, so we collect that. and then we collect contact information for each person in thehousehold, so that if the household splits up, we don�t haveto only ask one person. we have a cell phone number or some kindof contact information about each person in the household.the current status of the survey, we completed fieldwork for wave 1 in early june of 2014.the sampling frame was drawn from the 2010 decennialcensus sampling frame. we were the first census bureausurvey to use that sampling frame. we interviewed just under 30,000 households, for a responserate of just

over 70 percent, and that covered calendaryear 2013. and, at the end of may of this year, we completedfieldwork for wave 2. so, we returned to all ofthe households, all of the 30,000 households that we had interviewed in wave 1, as wellas following movers. so, if people had left that household, ifthe household, somebody had gotten divorced or if a child had goneoff to college or whatever, we followed those movers to their new residences and interviewedthem there, as well as additional people in that, in thenew household. we used the dependent data from wave 1 to helpwith the recall. our response rate, overall

response rate was just under 75 percent, andthat covered calendar year 2014.as far as data goes, we are planning to release a research file in december of this year thatcovers a significant portion of the content fromwave 1 of the 2014 panel. so, it is going to include residents� history,all the demographic data, all the labor force and participation data, as well as the programreceipt data. and then, after that, we will release the fullversion of the file early next year, so spring of next year.i talked a little bit before about how the instrument is flexible and it allows you toadd

supplements if need be. and, in one case here,the social security administration realized that, for what theyneed to do, they use sipp to improve a lot of their modeling and their forecasts of whowill be using their programs in the future. but the reengineeredsurvey no longer included some of the things they needed. andso, they hired us to field a supplement for them, and that supplement included, it wasa telephone survey of all of the interviewed wave 1 sipp respondents.so, all of the people in the 30,000 households that weinterviewed we called back in september and october of last year to ask them these questionsfor the social

security administration. and, that includeda detailed marital history, which was a full history of all themarriages that the person had had, whether they received retirement pension and how muchthey received, and then a much more detailed disability questionfor both adults and children, and then, for adults, if theywere unable to work because of the disability. i hope after all that, you�re very excitedand you�re thinking, �hey, that�s great. how do i get in andactually access the sipp data?� well, there are a number of ways to do that.the easiest way is through the sipp website, and the url is there. we have pointers toa number

of the other resources that i�m going toname here. that�s a good place to start, just in general. you can getthe data. you can get documentation. you can get information about how to contact us.if you know what data you want, you can go to the official census ftp (file transferprotocol) site and download the full wave files there.additionally, the national bureau of economic research, although they are not affiliatedwith the census, they are a great resource. they havea lot of sipp files, and they have a lot of good documentation.and, the url is there. they have a lot of stuff that is in stata format. we primarilyhave stuff that�s in sas

format, but they have stata code, and theyhave stata files that you can download and use, as well as spssfiles. there�s some other online tools that youcan use as well. if you only want to look at a couple ofvariables, or you know what variables you want and you just want to create a quick tableof that, you can use the census bureau�s dataferrett tool. dataferretthas all the public use data. you can go through it. it will giveyou a list of the variables. you can go through and select those variables, and here are justa few examples of some variables you can get. and then, it givesyou person-month estimates of those, of those

variables. that�snot necessarily the best option if you want a lot of variables, or if you want a fullfile. you�re better off using the ftp site. but for quick tables�you mighthave used american factfinder for acs (american communitysurvey) or decennial data. dataferrett is more like that, where it will give you tablesand quick results. also coming soon, we have contracted withorlin research to produce a data tool that helps youwork with sipp data. so, it gives you the ability to do hierarchical analysis. it letsyou create recodes on the fly. you can analyze spells, and if you don�twant to analyze months, you know that somebody

had healthinsurance for a six-month period, you can analyze it based on spells instead of foreach month individually. and it has also integrated documentation rightthere with the data variables. so, we are hoping to release thatpretty soon. we have been working hand-in-hand with orlin to get that out.just a couple of quick notes about public use files. i mentioned the ftp site. the datathere are in ascii format, with sas input statements. butthe national bureau of economic research, they have ascii datawith sas, stata, and the spss input. so, depending on the statistical software package that you�remore

comfortable with, you might want to use oneor the other of those sources. there�s also another site here which isnot, again, not a census bureau site, but it gives you somestata commands if you want to convert sas datasets into stata datasets. the link thereis from going to sas to stata, but there is a parallel commandto go in the opposite direction if you want to save stata datasets insas format. we have also produced a fair amount of technicaldocumentation for the sipp. there is the user�s guide, which is the comprehensive source ofall information about the sipp. that has been updated multipletimes through the 2008 panel, and we are in

the process of rewriting that from the groundup for the 2014 panel.the same with the data dictionary, the ftp site has data dictionaries for each of thedata files that it has available. and, again, as we releasethe 2014 datasets, we will release data dictionaries for those as well.content of most variables stays the same across the panels. the names change occasionally,but the idea is to keep it as consistent as possible,both to make it more user friendly, and just for our own internalmaintenance, we don�t want to change anything we don�t have to change.just as a visual idea of if you are going

to use the sipp classic, and one of the reasonswe did the reengineering. so, you can see that each wave,which is wave 1, wave 2, wave 3, wave 4, is divided intorotation groups. and so, we would interview one rotation group a month for four months.and so, at the end of four months, we would have completed thewave, and then we would start over. so, if you want to look atdata for a given month, you have to join the files from four different rotation groupstogether. so, if you want to look at, say, here is march of �96, youneed wave 1, reference group 4, reference month 4, rotation group2, it is wave 1, reference month 3, etcetera.

one of our goals in reengineering was to simplifythis. and so, you can see at the top, the top partof that graph is what i was just showing you, where you have the staggered files and theyare overlapping. the bottom portion of the graph, the reengineering,everything, if you want the whole year, it�s all right therein one file, and it�s so you don�t have to do all this joining. you don�t have toconsider what month each of the files covers. you know that everything isright there in a single file. so, our goal for that was to greatly simplifythat for data users. just some additional things to keep in mindas you are using the sipp data.

if you want to identify unique respondents,you need to � again, this is for the classic sipp, less soin the 2014 � but you need to make sure that, because there are four observationsper person per wave, you are using the right person identifier. andso, the slide there lists the correct variables that you need, as well asgives you some sample stata syntax to quickly create that identifier. and so, the code thereconcatenates the variables into a single variable, and thenyou can just refer to that instead of the combination of all threevariables. and, keep in mind that you need the form to match across waves. so, if it�s101 versus 0101, you

need to make those consistent so that theymatch and they can merge them over time. because the files are fairly large, this canlead to memory limitations on your machine, and maybe worth investing in extra hardware to boost the ram (random access memory). you probablywant about two gigabytes of storage space, if you arecreating a fairly sizable sipp dataset. and when you load in a dataset,we recommend that you keep only the observations and variables that you need to cut down onexcess storage.just a suggested practice, we think that you should keep a complete sipp file in its originalstate,

and then as you�re doing analyses, createsubsets of that in a separate area that you can easily create and thenget rid of when you are done with them. that way, you�ve only got the variables thatyou want. you�ve only got the observations that you want, and ifyou do something to the file, you haven�t lost your original file.and, once you have created the program, you can always go back and create the datasetagain if need be. what is sipp good for? well, sipp is goodfor a whole lot of things. here�s a few examples of that.you can use it to�if you want to do a series of cross-sectional analyses, you�ve gota number of panels, a

number of ways you can do it that way. youcan use it to generate monthly estimates or annualized estimates.results do appear to be most accurate in the fourth month, the last reference month, i.e.the month that we interviewed you in. and, as i mentioned before,that is commonly referred to as seam bias. you do need toadjust the estimates for the sample design, and files include the waves that allow youto do that. however, sipp is most powerful when you�re usingit longitudinally. it is designed as a longitudinal survey. it can giveyou the point in time estimates, but it is really powerful when you�re doing longitudinalestimates.

there are a few things that it is not idealfor. it is not ideal for the top of the income distributionfor the most part. it�s partially because we top code income to protect respondent confidentiality,and also just because the sample is such that therearen�t a lot of people at the very top of the income distribution.although it is a longitudinal survey, it�s not ideal for long longitudinal analyses.in other words, we don�t follow somebody over the entirecourse of their lives. we follow them for a four-to-five-year periodwithin that life, within that life course. so, keep in mind that we don�t, dependingon what you�re interested

in, it may not be a long enough period foryou to see the effects that you�re looking for.however, sipp used to not be ideal for annual estimates every year. in the new design becauseit is an annual survey it is much better forthat. and there are some complexities in household and familycomposition. people�s lives are messy. stuff happens. people get married. people get divorced.people have kids, and all of that is reflected in thesipp data. so, the data can look very messy, but on the flip side it is veryrich. it gives you these transitions that you might not otherwise see.one final word of caution, for smaller states,

be careful about using sipp to produce state-levelestimates. sipp is not ideal for smaller states, just due to the small sample size. the estimatescan have a lot of variability, but in larger states that�sless of an issue. so, overall, use the sipp when you want thecomplexity, when you want to deal with messy questions about people�s lives and changesin people�s lives. if you want estimates of poverty or the incomeof the poor, sipp is a good choice for that. or if you want to look at program participationrates, sipp is again very good for that because that it part ofthe survey�s focus. and, if you want to do longitudinal analyses overa fairly short period, up to about four years,

sipp is an excellent dataset for that.that is a lot of what sipp has been and what sipp is currently doing. a little bit aboutwhat we are continuing to do and what we are planningto do, as we release more data from the new survey, the 2014panel, we are of course going to release additional documentation. and one of the pieces of documentationthat is most important for the data users is a series of crosswalks from the old surveyto the new survey, because a lot of people have been using thedata for years. they know, they have their variables. they knowwhat it looks like, and they want to make sure that this concept is also available inthe new survey. well, with

the crosswalks it is going to have�okay,your variable was named this in the 2008 panel. here is what it isnamed in the 2014 panel, and it will also go the reverse. so, if you see a variablein 2014 and you want to see if it�s available in the older panels, youcan do that as well. we�re doing workshops like this. we�vealso done a number of multi-day workshops at universityof michigan and at duke. we did a workshop here onsite at census in may. we�re goingto do more of those, so i would encourage you to sign up. usuallywe make announcements on the sipp website and on the sipplistserv, and you can see that contact info

for those things down at the bottom of theslide. one of the things we�d like to do that wehave never done in the past is create code repositories,and a lot of people do the same things, right? you�re creating similar recodes, or you�resub setting the file in a certain way, and there�s no need for eachdata user to have to reinvent the wheel. so, we would like to sortof put up a series of short pieces of code that helps everybody, and can really speedup your analyses. we�re also creating more data utilities.i mentioned the orlin system, we want to do more thingslike that, again, with the goal of making

the data easier to use. and we want to startup a users� group. we have the listserv, but we�d also like amore vibrant users� group, where people can ask questions about sippor discuss sipp, get input from us, get input from other data users, and so on.so, that�s a brief overview of all that sipp has to offer. i�m going to turn itover now to ashley edwards, who is an analyst in the povertybranch, and she�s going to give some examples about how to usethe sipp data and some findings that we�ve had. and i will say, before i turn it overto ashley, if you have any questions about what i�m working on as wellas � or what i was presenting on as well

as what ashley presentson, please enter it, please submit it as a q&a, and we will try to get to it at the end.and now i�ll turn it over to ashley. good afternoon. so, i�m going to be buildingon what matthew discussed by talking more about how census analysts, as well asoutside researchers, have used sipp data across various researchdisciplines, specifically focusing on issues of economic self-sufficiency. and the reportsthat i am going to be referencing are all based on data from the2008 or earlier sipp panels. so, in addition to reviewing and preparingpublic use data files, subject matter analysts

withinthe census bureau also release official sipp reports on various timelines. these officialsipp reports are referred to as the p70 series, and as you can see,they cover a range of subject matters, and these reports are oftenupdated within or across panels to allow for comparisons across time periods. again, theseand all official sipp reports can be found on the sipp website,in addition to detailed table packages, technical working papers,and conference research. this is also just a brief review of recentexternal releases that have used sipp data. you see herethere�s a number of journal articles, some

federal reserve bank discussion papers, andsome work by policy research organizations.my presentation today is really designed to serve two purposes. one is to provide somebackground on topics related to self-sufficiency, by summarizing recent data releases basedon the sipp. and, two, to provide some inspiration on how youmight use the sipp moving forward in your own research.i am primarily going to be focusing on issues related to the measurement of poverty, asthat�s my particular subject matter focus, but iwill also be bringing in some other subject areas that touch on thesethemes of self-sufficiency, specifically research

related to the receipt of government programs,family support structures, as well as asset measurement.so, i will start with my own subject area research, which has really focused on thedynamics of poverty over time. by using the sipp longitudinally,i can observe individuals over the course of 2009 to 2011,and document how many times they entered poverty, how long their poverty spells lasted, andhow poverty entry and exit rates vary across the demographicgroups and by calendar years. i can then compare those2009 to 2011 estimates, which cover the period during and immediately following the mostrecent recession,

to estimates from the earlier 2004 sipp panel,which covered calendar years 2005 to 2007, the periodimmediately before the recession. so, i really like this figure from the mostrecent report, because i feel it does a good job ofshowcasing the unique ways that sipp can be used to measure poverty. the bars shown inblack in the middle give you the annual poverty rate from thesipp. however, because the sipp collects monthly data, this is theonly survey where we can also get the percent of individuals who were in poverty for atleast two months of the year, which is the higher episodic povertyrate. alternatively, we can also calculate

the percent ofindividuals who are in poverty for every single month of the calendar year, which is the lower,chronic poverty rate.you can also see from this graph that the reference period used makes a difference.in 2011, 23.6 percent of individuals had at least asingle spell of poverty lasting two or more months. however, whenyou look over a longer reference period, in this case from 2009 to 2011, you see thatalmost a third of the population was in poverty for at least twoor more months over the longer three-year period. the same is truefor chronic poverty, but in the reverse direction.

more people are in poverty every month of2011 than are in poverty every month from 2009 to 2011. andso, by using the sipp monthly data to look at these alternatemeasures, such as episodic and chronic poverty rates, you are able to get a better senseof how individuals experience poverty than when only lookingat annual poverty rates. using sipp data to look at the duration ofpoverty spells, we find that generally poverty spellsare temporary, with the majority of spells lasting for less than one year. however, 15.2percent of poverty spells observed from 2009 to 2011 did lastfor more than two years.

when looking at how individuals change povertystatus across years, you see considerable movement both into and out of poverty. however,approximately half of individuals who exited poverty, from2009 to 2011, continued to have annual incomes below 150 percent of their poverty threshold.external researchers have also used the sipp to investigate measures of extreme poverty.this paper doesn�t follow individuals longitudinally,but it uses sipp data to create custom poverty thresholds andincome definitions in order to evaluate the depth of poverty, as well as the impact ofsocial safety net programs. by looking across multiple sipppanels, the authors of this study were able

to evaluate the impactof the 1996 welfare reform act on poverty alleviation.here, the authors are looking at families, particularly those with children, and theyare focusing on those who are living on virtually no income,in this case defined as less than $2 per person per day. theauthors use sipp data to see how the size of this population has changed over time,calculating estimates across the �96, 2001, 2004, and 2008 sipppanels. the authors find that extreme poverty has increased steadilyover the past 15 years, as shown by the dark black line, which is using the official censusbureau definition of

family income, and the author�s definitionof extreme poverty. by adding and subtracting the value of varioussocial safety net programs, the authors can thenevaluate how the extreme poverty rate would vary in the absence of those programs. so,you can see by the progressively divergent bottom dashed andgreen lines, that when you add the cash value of snap(supplemental nutrition assistance program) benefits, tax credits, and housing subsidies,the families� incomes, the decline in the extreme povertyrate is getting progressively larger in later years, indicating thatthese programs are playing a greater role

in keeping families out of extreme povertythan in 1996. however, you see that when the authors subtractthe cash assistance provided through tanf, which is included in the official povertyincome definition, that while in early 1996 that subtraction led to largeincreases in the extreme poverty rate, the impact of those cash transfer programs inkeeping families out of extreme poverty has declined sharply in lateryears. additional research on the receipt of means-testedassistance programs has also used the sipp longitudinally, allowing researchers to understandboth how many people are participating in a program in agiven month, as well as how long individuals

are recipients of that assistance program.here, the top graph is using monthly data from the sipp cross-sectionally, to createan average monthly participation rate over the courseof a sipp panel. the authors found that around one in fiveindividuals participated in one or more major public assistance programs in a given monthin 2012. as you can see, participation in medicaid is highestamong the major means-tested programs, followed by snap.on the bottom graph, we are looking longitudinally at individuals who received means-testedassistance for at least one month from 2009 to 2012, and looking at the total accumulatedtime they spent

participating in that program over the period.of major program recipients, 43 percent were receivingprogram assistance the majority of the four-year period, from 2009 to 2012, although 31 percentreceived assistance for a year or less.when you look at participation by program, you see that although average monthlyssrc webinar transcript participation in housing assistance is relativelylow, that program has more long-term recipients, with almosthalf of recipients receiving support for nearly the entire reference period. alternatively,tanf recipients receive more short-term benefits, with themajority of recipients receiving support for

a year or less.here, the authors report median monthly benefit amounts for people receiving means-testedtransfers, and you see that ssi recipients receive the largest median monthly cash transfers.the authors also find that, although monthly participationrates are much higher for families with a female household or nohusband present, those families actually receive a smaller median benefit amount than otherfamily types. but so far the research i have presented hasfocused on the core content of the sipp that isquestions that were asked in each wave. however, as matthew discussed earlier, each sipp wavealso has a

number of what we refer to as topical modules.generally, each wave will have around three topical modulesthat include a detailed battery of questions on select topics. this research, done by censusanalysts, uses the core sipp longitudinally to look at changesover time in children�s households, and many use data collectedin the wave 10 child well-being topical module to see how those earlier transitions may haveimpacted children�s achievement and social engagement.so, this is the kind of detailed information that you can get from a sipp topical module,and in this case focusing on child well-being. thesequestions about extracurricular activities

were not asked in eachwave of the 2008 sipp panel, but were asked in repeat topical modules in waves 2 and 10.and so, as you can see, children who are in families with income-to-povertyratios greater than 200 percent had higher extracurricular participation than those livingbelow the 200 percent of poverty. but the authors then use the sipp longitudinallyto look at major family transitions over the period from 2008 to 2011. here, they are lookingat the percent of children who are in families where a parentlost or gained a job, had the spouse or cohabitating partner of their reference parent enter orleave the household, or changed addresses at any pointover the period. over half of children under

the age of 17experienced at least one of these transitions between 2008 and 2011, the most common transitionover the period being changes in parental employmentstatus. this figure is designed to show how thesechanges in families� economic situations oftenoverlap with changes in family structure and residential location. children in the middleof the diagram who experienced all three transitions were morelikely to be in poverty, while children in families at or above 300 percent of povertyhad the lowest rates of transition. by combining this data with the child well-being topicalmodule, the authors are then able to investigate

to what extent this instability might haveon child well-being. as the final example, external researchershave also used sipp data for multiple topical modulesthat referred to in a similar topic. in this case, the 2008 panel, waves 10 and 11 topicalmodules. by using the consecutive assets and liabilities, and retirementand pension plan coverage topical modules, researchers arebetter able to understand the vehicles by which individuals are saving for retirement,as well as the extent to which their assets should be expected to covertheir retirement expenses. sipp topical module data, as matthew mentioned, offer a comprehensivelist of respondents�

financial assets, including the value of theirbank accounts, bonds, securities, stocks and mutual funds, lifeinsurance policies, ira accounts, defined contribution accounts, real estate holdings,as well as home and business equity. using this rich sipp data,the authors find that the average net worth of near-retirementhouseholds is around $183,000 per single person households, and around $500,000 for married-couplehouseholds. this table shows the total liquid assets ofthe near-retirement population in 2011 to 2012, andthe figures represent financial assets that can be easily liquidated. so, here you areexcluding the value of any

homes. the authors find that 30 percent ofu.s. households at or near retirement age have less than $10,000in liquid assets, with the next 24 percent also having too little saved, less than $100,000,meaning that 54 percent of near-retirement households willrely almost exclusively on social security and any defined benefitpensions they may be eligible for. so to conclude, even though this presentationshas covered a variety of topics, there is still somuch that i�m not able to summarize here. what i really hope is that this presentationhas provided you with more information about what sort of contentis available on the sipp, as well as giving

you some ideas on howyou might be able to use that sipp data to investigate the subject matters that you workon. again, i appreciate your time, and we welcome any questions thatyou might have. great. thank you, guys, so much. thank youso much, ashley and matthew, for those very informative presentations. it�s really excitingto hear about how the sipp is being reengineered to make iteven more useful, and it was also great to hear some specific examples about the typesof research that can be done using the sipp.i�m now going to open it up for questions. again, as a reminder to those listening in,you can submit questions to our speakers using

the question and answer feature at the bottomright-hand side of the webinar platform, and responses to the questions thatwe don�t get to today will be posted on the ssrc following theevent, along with other webinar materials. so, go ahead and ask those questions thatyou have. we only have a few minutes, so i don�t know that we�llget to all of them, but we�ll definitely post responses afterwards forthose questions we don�t get to. i�ll start with the first question: whatwould the process be if a federal agency was interested inadding questions to the survey? is it only possible to do so through a new supplement,or can it be done

through topical modules?that�s a very good question. we don�t generally add questions in the middle of apanel, just because we want to keep the questionnairepretty much the same from wave to wave. we are certainly opento adding questions when we start a new panel. i would encourage the person who asked thatquestion to send me an e-mail. this is matthew. my contactinformation is up there, and you can also send it to jasonfields, who is the survey director of sipp. i think we�re always looking for input fromfederal agencies or other data users about things that would make thesurvey better for them, or stuff that we�re

just missing on thesurvey. great. great. second question: how successfulare you in obtaining social security numbers tomatch to ssa administrative data? so, we do not ask for social security numbers.we used to, i think in the �96 panel, but maybe2001. but since then, we have not asked for social security numbers. we ask enough, wehave information from the respondents, and we use that to matchto social security records and other data based on thingslike name and date of birth. and i believe, with the 2014 data, we were successfully ableto match about 92

percent, something like that. we do give respondentsthe ability to opt out if they do not want their datamatched, but the opt-out rate was low. it was under 1 percent.great. no, that�s very helpful. another question: how is the sipp for looking at ruralresidents? is it good for research on rural residents?so, because it is a nationally representative survey, it definitely includes a representativeportion of rural respondents. as i mentioned in themain part of the presentation, for states that are predominantlyrural, like wyoming or montana, i wouldn�t use the sipp to generate, to generate estimatesin that way, just

because the sample size is too small. butif you want to use it to generate national estimates of the ruralpopulation, you are able to do that through sipp.okay. and, it is a minute before 3:00. matthew and ashley, do you have a couple of minutesjust to go through the remaining couple of questions,or do you have a hard stop? no, i can go. go ahead.oh, okay. great. what population cutoff constitutes a small state?i believe it is the 20 most populous states, and that has varied from panel to panel basedon the sample size. i don�t have the list in frontof me, but it is generally the 20 most populous.

ssrc webinar transcript survey of income andprogram participation (sipp) okay. great. and, can you use the sipp togenerate state-level estimates? i know you referencedthat briefly, but do you have any further guidance about that?for states like california and new york, definitely. for states like montana and wyoming, andsome of the smaller new england states, probably not. i mean, you can always generate them.they are just not all that reliable.okay. okay. and, i see a response also from the sipp director that there is not a specificsize cutoff. variances just increase significantly, i guess,with those smaller states.

right. and if you want the list of states,we are happy to get that to you. we have that in ourdocumentation. okay. great. and, another question. will programparticipation items include questions on reasons why families who might be eligible for programsaren�t accessing or applying for assistance? yes. so, if you say that you don�t get theprogram, we don�t ask you, �why not?� but if you say thatyou do get it and that it ended, we ask you both why you started getting it and why youstopped getting it. okay. great. i had another question that wassubmitted separately right before the webinar. it saysit appears that looking at foreign-born respondents,

a majority of those from latin american countriesoutside of mexico and cuba are not counted as hispanic/latinoon the origin question. how much is the sipp hispanicgroup slanted towards mexicans and/or cubans? yes. i mean, i think certainly when you justlook at the population of the united states, themajority of respondents in the 2010 census that reported hispanic origin were respondentsof mexican origin, or 63 percent. and then, puerto ricans were9 percent. so, that�s just reflecting the demographics of thecountry. but when you�re talking about people who have origins from nations that would suggesta

latino/hispanic origin and they�re not self-identifyingas such in the origin question � and you see this in a lotof various surveys both from the census bureau and other research agencies � that can befor any number of personal reasons. the census bureau, the eorigin, hispanic origin question, is based on self-identification. so,we�re not assigning that based on any sort of ethnic, ancestry, race, or country of originbasis. it�s just based on how people self-identify. and, like yousay, you do see that that identification varies based on respondents�personal characteristics. pew has used some of the american community survey data, andhas found that

while 99 percent of immigrants from mexicowere to refer to themselves as hispanic, only 67 percent ofrespondents from panama. so, there is some variation on how people identify and how theyrespond to that question. and the census bureau is constantlyinvestigating changes. obviously, our country is becomingmore diverse, and we have to have questions that resonate with people�s self-perceptionsof both their race and ethnicity.okay. great. well, thanks so much to you both for your flexibility, going a little long.it looks like a lot of our listeners stayed on as well. so,they seemed to be interested in hearing what

you had to say. so, ijust want to thank you all from the u.s. census bureau for your time, matthew marlay and ashleyedwards, and i want to thank you all for joining us.please have a great afternoon. thank you for having us.

Just got my check for $500 

Sometimes people don't believe me when I tell them about how much you can make taking paid surveys online...

 So I took a video of myself actually getting paid $500 for paid surveys to finally set the record straight.