 Thank you so much, Sue, and it is such a pleasure to be here back on campus. Although when I was a student at the Ford School, we definitely did not have an auditorium like this. We were over on the fourth floor of Lurch. So to all of the current students, your lot has improved greatly compared to previous graduates. But nonetheless, it was one of the best experiences of my life, so I'm very pleased to be here. And as Sue mentioned, definitely interrupt with questions. I usually present to audiences of economists that aren't shy. And the sort of biggest insult is to not ask any questions at all. So do not hold back. And as Sue mentioned, today I'm going to be looking at federal student aid, in particular the Pell Grant program and college pricing, and looking at how institutions respond to their students' Pell grants. Great. So, you know, we all are in the field of higher education, so I probably don't need to spend much time motivating why we should care about affordability and access to college. There's an extensive literature and labor economics suggesting there are large private returns to earning a college degree. And there's suggestive evidence of positive spillovers from having a highly educated workforce. If some students face credit constraints and can't borrow against their future income to finance college education, then the overall level of education and the population may be inefficiently low. So this is one of the main rationale that the federal government and state governments use to justify their provision of need-based student aid. The largest of these programs, as Sue mentioned, is the federal Pell Grant program. And I should say by need-based student aid, what I mean is that at least one of the criteria for eligibility is depends on family income. So you have to be sufficiently needy, hence the name. But back to the Pell Grant program. So the federal Pell Grant program is the largest source of need-based aid in the U.S. In 2011, 9.5 million students received $35 billion in aid through this program. It is a large proportion of the Department of Education's discretionary budget. And what I'm going to be looking at today is essentially for every, say, $100 of Pell Grant aid a given student gets, how much does this actually reduce their cost of attendance? How does this actually reduce the price that they pay? And the answer to this depends on whether this aid sticks where it hits. So from an economist's perspective, we can think of this in a tax instance framework. The students are the statutory recipients or the sort of party that is entitled to receive Pell Grant aid, but they may not receive the full economic benefit if schools also respond to the fact by raising tuition or lowering institutional aid. So this... Yeah, great. So we also think of it in the play paper framework, right, where it's one level of government, the federal government sending funds down to another level. Right. To what extent does that actually change total spending on it? Yes, yes. Although in this case, the intended recipient from the federal government's perspective is the student, not the school, but there is definitely... Yes, yes, exactly. You can add tuition prices, but most of the students are going to public institutions. Yep, yep. So you're subsidizing public schools. Exactly. I mean, there is definitely a close tie between sort of thinking about subsidy incidents and the play paper literature. Great. Yes. Don't hold back with questions, please. Okay. So this paper is also motivated by a shift in the organization of higher education with the emergence and growth of the for-profit sector over the past, say, 20 years. So these are schools like University of Phoenix, ITT Tech, many smaller single-standing institutions that disproportionately serve low-income students such as Pell Grant recipients. And over the past five to 10 years, there's been a growing concern of these schools. There's been allegations of deceptive marketing practices and that these schools are taking advantage of federal aid programs. So in this paper, I'm going to specifically look at the for-profit sector and see whether their response to Pell Grant aid and the fact that their students, say, get $100 Pell Grant, how... If that differs from the responses of other schools. Okay, great. So I'm going to just broad sweeps, talk about what I'll talk about for the rest of this time. I'll briefly go through sort of what do we know about federal aid proud out in higher education. Spend most of the time talking about my paper. And then since this is a policy seminar, spend probably not enough time talking about what are the policy implications of my findings. Great. So my paper is not the first to look at this question. Do schools respond to the fact that their students are getting sort of federal grants or federal loans? This behavior actually has a name. It's called the Bennett Hypothesis, named after a former secretary of education that essentially asserted if we increase federal grant aid by $100, colleges will just respond by increasing tuition by $100. Sort of the pass-through from students to schools will be 100% with students not benefiting at all. Right. So there's been sort of a number of papers looking at this question in terms of tuition increases. And the identification or sort of how these papers are going to disentangle the impact of changes in need-based aid on changes in tuition is time series variation in the maximum Pell Grant award. So that's what this figure is showing. So this green line is the maximum Pell Grant award students could receive in a given year, since the program's inception in 1974, in nominal terms. And the blue line is the maximum Pell Grant award in inflation-adjusted terms. So just looking at this picture, there may be concern if we're sort of looking at how tuition changes as the maximum Pell Grant award changes and inferring a causal sort of channel. For instance, if you focus on the late 1990s, these were good economic times. And these are the years when the real value of the maximum Pell Grant award increased by the largest amount. And I could tell similar stories sort of about this variation being correlated with other things going down in the economy that may affect schools' decisions or state's decisions over tuition. And so not surprisingly, this research sort of finds inconclusive evidence with some papers finding Pell Grant's increased tuition, some papers finding Pell Grant's decreased tuition. And furthermore, one thing I'd like to emphasize is really this channel of increasing tuition is maybe not the way schools would choose to go after Pell Grant aid. So for most schools, Pell Grant aid really is a drop in the bucket of total resources. Okay, great. So inferring a causal relationship rather than a correlation. So by identification, what I am saying is we want to say, the question they're asking is, does Pell Grant aid cause tuition to increase? Or do increases in Pell Grant aid cause tuition to increase? My argument is that these papers can show that increases in Pell Grant aid are correlated with increases in tuition in some cases and not so in others, but we can't look at this correlation for a causal relationship because changes in the maximum Pell Grant award are correlated with lots of other things going on that may also affect tuition. Thank you. Yes. So you're going to be answering the question of incidents, which is great. But I'm also, as a public economist, interested in the quantity response, right? So it's going to be an increase in supply, which I guess is coming from these four profits. So I'll talk a little bit. Are there also, are you going to be able to expand on this later or to be able to get the quantity responses to see how much, where it's coming from and then we get all the different elasticity to supply and demand of education? It would be awesome. So if you all invite me back in, say, years time, I would be happy to give that paper. That's not something I'm going to talk about today, but that is something that is important. I have, you know, pay lip service to this at the end. So what I'm going to show you today is sort of the very short run response of schools to Pell Grant aid not allowing for entry, right? But in the long run, exactly. If in fact I find schools are capturing Pell Grant aid, then this may induce other schools to enter the market. So if you're not allowing for entry, then you're holding supply fixed and theoretically in a competitive market, then the hypothesis will be right. So this market is far from being perfectly competitive, and I'll talk about that in a little bit. Yep, yep, so the setup is coming. Any other questions? Okay. So I mean, most, 80% of it is public. So right there, it's not a profit maximizing, you know, 80% of these students are attending public institutions and then another, what 10 are going to nonprofits and then 10% are going to for profits. That, so yeah. So, on average, right? We still might think for profits would respond and there is some evidence that they do, but in the short run, in the very short run in a given year, they're sort of not going to be able to respond. That's why, and Sue is exactly right. So the majority of Pell Grant recipients are attending public institutions. Although there is this growing proportion going to for profit schools. Okay, so really what I'm going to argue is that for the vast majority of schools, which are by and large public schools serving Pell Grant recipients, tuition isn't the margin through which schools want to respond. And the one exception is the for profit sector. So these schools, some of them serves up to 90% of the students in their population are receiving Pell Grant. So this may be sort of the one sector that we might think would respond on the tuition margin. And just to say, like, why is the for profit sector important? Well, actually they received 25% of all Pell Grant awards in 2011. This number has been growing. A few years ago it was 80%. Or less than, you know, around 10%. And now it's 25%. So just a little diversion about the for profit sector that I think is rarely focused on. This figure is graphing the for profit share of all Pell Grant aid. And when people talk about sort of the dramatic increase in Pell Grant aid going to for profit schools, focusing on the data points that are sort of after this line that I'm drawing. So if you start in 1994 you see there's a large growth. But if we were to sort of extend this time series going back, we actually see that this point that we start at is the bottom of a trough. And in the late 80s there was actually another spike in the proportion of Pell Grant dollars going to the for profit sector. Sort of what led to this decrease? Well there was tightening of federal student loan regulation. There was sort of many for profits that exited the market. This is not something I'm going to talk about, but I think it's an interesting fact that doesn't get mentioned a lot when thinking about the for profit sector. So why are we more worried in more recent years? Well part of the reason is that overall Pell Grant aid, which is this green line, has increased dramatically. So even though if we look in 2009 and then the percentage of students attending for profit schools is around 25 percent, which is similar to the percentage back in the late 80s, the percentage of Pell Grant dollars. Are you going to have those 10 years on the share of Pell Grant aid? No, but it is... That's the 10 percent I was referring to. So it's at a high right now about 10 percent, where 10 years ago it was maybe like 5 percent. So it's growing as well. So this is a share of Pell Grant aid. But let's get into it. Let's get into it. So my speculation is that it's related to regulation with respect to student loan default rates and eligibility, which affects eligibility for federal student aid. My point here is that this pattern that we've seen over the last 10 years is not unusual if we're thinking about the share of Pell Grant aid. What is unusual is the total amount of money on the table. So 25 percent of 5 billion dollars is a lot less than 25 percent of 35 billion dollars. So sort of thinking about the sheer magnitude of dollars. So even though this cycle is something we've seen before, it's perhaps more relevant now, yes. I've read his paper on Pell Grant. Okay, great. The other thing that changed was the definition of who's eligible for Pell. Yes. So in the past you had to have received a high school diploma or GED to receive a Pell Grant award. And then in the 90s this requirement was loosened and schools just had to show that you had the ability to benefit from higher education. Also in this sort of period, the regulation regarding online education was loosened. So schools before had to serve a certain amount of students in person and that regulation was loosened, which sort of hastened the entry of schools that survey mainly online audience. I'm not going to tell you the exact date this regulation changed because I'm not going to say something wrong on videotape or in general, but it was around that period. Okay, so great. So the for-profit sector, sort of maybe we've seen this before, but maybe not because of the sheer amount of money on the table and they're sort of the only sector where Pell Grant aid could constitute a significant share of total revenue. I just want to pay lip service to this recent paper by Stephanie Cellini and Claudia Golden, who are sort of looking at this tuition margin and they find that for-profit schools that are primarily offering sort of lower than associate degree programs do increase tuition relative to schools that can't disperse federal aid, suggesting that the benefit hypothesis may be appropriate for the for-profit sector. But in other sectors, I'm going to argue that sort of the real margin on which schools should respond will not be the tuition margin. So what's the sort of setup or the framework I have in mind and how does the financial aid process work? So students that wish to qualify for federal student aid, this is loans and Pell Grant aid, have to submit a FAFSA, a free application for federal student aid. If you've gone to college recently or have kids who've gone to college recently, you know how this works. It's fairly complicated. You input a lot of information about family income assets structure and it gets put through a complicated formula and what comes out is the federal government's measure of need, your expected family contribution. Then this information gets provided to the schools the student is interested in attending. Most students who are sort of new college entrants only list one school on their FAFSA, so they're only considering one. And at this point, schools observe this measure of need. They expect a family contribution. They observe everything that went into the FAFSA, so family income assets, family structure. And then they calculate the student's eligibility for outside aid. So they calculate how much of a federal Pell Grant are eligible for, how much state grants are eligible for. And at that point, they decide how much to give you in terms of an institutional grant or a discount from the list price of tuition. So the world I have in mind is one where schools set an overall level of tuition, but then prices are going to vary sometimes substantially across students through these individual discounts or through the provision of institutional aid. And this is exactly the margin that I am interested in examining. So in economics, we refer to this as price discrimination and colleges and universities are in a somewhat ideal situation to practice price discrimination since they observe sort of a great deal of information about students, including a measure of their willingness or ability to pay. So specifically, I'm going to be looking at how this sort of more flexible component of price, the individual discount, varies as a student's Pell Grant award, varies. I'm using data from the National Post-Secondary Student Aid Study. It's a nationally representative cross-section of college students attending schools in all different sectors of higher education, so public schools, nonprofit schools, for-profit schools. And just to be very clear here, the specific questions I'm going to be looking at are A, what is the economic incidence of Pell Grant aid? Or specifically, how much of every Pell Grant dollar is passed through from the intended recipients from students to schools? And then second, I'm going to be looking at whether this behavior varies across different sectors of higher education, where I'm defining the sector by the control of the school, so public, private, or nonprofit, and by selectivity. And selectivity is going to be sort of different than the way we generally think about things as faculty or students at University of Michigan, University of Maryland. It's going to be a binary indicator for whether a school is open admission, except sort of every student that comes through its doors, or whether it has some, it rejects some number of applicants. On average, these are not going to be highly selective institutions. They are not going to be University of Michigan, Harvard, Columbia. These are going to be schools that are on average accepting around 60% of their applicants. Yes? Can you help me think about price discrimination for nonprofit schools? So you say how much of every dollar is passed through schools. When I think about price discrimination, I think of some firm owner who's then going to put the money in his pocket. If more money comes to Michigan, then the students somehow you would think get that back. Right. So the school is still going to want to maximize its total pot of money to use for whatever its objectives are. So it may not be to maximize profits. It may be to offer scholarships to really poor students or to build new dorms. But given that sort of schools observe how much students are able to pay, then they're going to use this information to figure out what exact price students would be willing to pay to come. So this is complicated and I'll get to it. That's a good question, but stay tuned. I have to have some suspense. So does behavior vary across sectors? So it's sort of preview. Yes, it does. And then can we say something about differences in what schools care about or in schools' objectives from how they respond to Pell Grant aid? Yes, we can. Okay. So I made this argument earlier that this variation that other people have used in the maximum Pell Grant award, this time series variation, is maybe problematic because there's other things going on that also vary with the maximum Pell Grant award. What I'm going to take advantage of is the fact that most students who receive Pell Grants are most eligible students receive less than the maximum award. And how much Pell Grant aid I specifically receive is going to be both a function of the maximum Pell Grant award and this measure of need that the federal government calculated my expected family contribution. And I'll talk exactly how this formula works in the next slide, and I've already talked about how the expected family contribution works. Okay. So this figure is graphing how much Pell Grant aid students received in my sample of NPSAS students according to their distance from the Pell Grant eligibility threshold. So just to be clear, all of these students over here are receiving no Pell Grant aid and they have an EFC, an expected family contribution that makes them ineligible for Pell Grant aid. And then students to your left of the eligibility threshold are eligible for the Pell Grant program and they are receiving positive amounts of Pell Grant aid. Each one of these circles represents the average amount of Pell Grant aid received by students who are a particular distance from the eligibility threshold. Okay. Great. So it's driven by the formula. So this formula specifies that it's your expected family contribution standardized to represent the distance from the threshold. So every student gets this expected family contribution EFC and in a given year there is some maximum value of EFC above which no one is going to receive a Pell Grant. Below which students receive this minimum Pell Grant award which is $400 in the years I'm looking at. And then as your need increases for every dollar, your statutory Pell Grant award also increases by a dollar. Okay. And this is all just based on the federal government's formula. And this is going to lead to two sources of variation that I can use to get at this causal question. So not just how is Pell Grant aid correlated with institutional aid but does Pell Grant aid affect institutional aid? Does Pell Grant aid cause institutional aid to fall or rise? Thank you. Just to clarify. So you figure out a distance of say minus 2,000 and we're seeing people getting about 1,300 around then. So what's it kind of for that $700 difference? So imperfect take up? Yes. Non-compliance and if you do not stay enrolled for the full year, you don't get your full Pell Grant, right? So if everyone was taking up and everyone was staying enrolled, it would be almost a diagonal for $400. There's also an adjustment for whether you're full-time, part-time or quarter-time. Roughly like what part of the term say if you're talking about someone coming in like fall semester of the academic year, like is it like middle of the term? Is it like first few weeks of the term? Well basically... When are they getting the award? Like when... Maybe I don't quite understand the question why I want to answer but like when... Is there a point in the term where basically the institutions get to keep the funds? No. No, no, no. No. Either you students qualify or they don't, right? Ah. Colleen, do you know the answer? Yeah, thanks, Steve. If a student drops out at a certain percentage of the way this semester then the institution essentially keeps the funds. I see. But this is not actually the institution... So but this isn't the institution keeping it, it's that the student still has to pay tuition. Yeah, it's essentially... So the student gets to keep it, right? Because they're still on the hook for tuition. Yes, but the... So they essentially then they end up owing money to the school on the amount of the Pell Grand Award. So... They owe the school tuition. Yeah. Okay. Right, so if the student drops out too soon they're not going to get their Pell Grand Award but they still owe the school is what you're saying. Correct, but after a certain point in time during the semester then the school keeps essentially all of the aid and the student might still have some funds to pay the school. If they haven't paid their bill for whatever reason, for example. That time period would usually have to drop out. Right. Right. But that's not what this is about. This distance from the maximum is about part-time versus full-time. If you look at this separately, I mean I looked at this particular NPSAS, this particular graph. It's basically if you look at the full-time people it looks right and then this is people who are going part-time. Got many experts in the audience. The best kind of audience. Okay, great. So there's two sources of variation in Pell Grand Award that I'm going to use to disentangle this causal question. So the first is this discontinuous increase in students' Pell Grand Award as you say move from being a dollar above the threshold to a dollar below the threshold. And this is driven by this minimum Pell Grand Award. So no one receives less than $400 in Pell Grand Aid if you qualify for a Pell Grand Award. And for the people who like econometrics in the room I can use this discontinuous change in the level in a regression discontinuity design. The second source of variation I'm going to take advantage of is a discontinuous change in the relationship between need and Pell Grand Aid. So specifically what I mean by this is for ineligible students we can think about sort of increasing their need by a dollar, right? So we take a student out here for a Pell Grand Award for a dollar, but there's zero dollars change in their Pell Grand Award, right? They're all ineligible, this is a flat line, the slope is zero. Conversely, after we cross the eligibility threshold for every dollar increase in students' need there's a statutory increase in their Pell Grand Award of a dollar if they were full-time full-year and empirically we see about a 70% increase in their expected Pell Grand Award. So this is a discontinuous change caused by the kink in the program schedule which I can use in the aptly named regression kink design. Okay, great. So for those of you who like econometrics here's a brief segue for the rest of you you can tune out for a second because hopefully all will be clear. Regression kink design, if you haven't heard of it is very similar to the regression discontinuity design. We're taking advantage of changes in the level or the slope of Pell Grand Aid right at this eligibility threshold and the key assumption I need to make to use this to say something causal is that basically students who are on either side of this threshold have imperfect control over their expected family contribution and they're essentially going to look the same in terms of their observable and un-observable characteristics. So we can think of students on either side as being as good as randomly assigned to receive say a zero dollar Pell Grand Award or a four hundred dollar Pell Grand Award. Essentially this is assuming individuals cannot fully manipulate their expected family contribution to move from say this point to this point, right? So this is a point where I usually get a lot of questions and I'll try to clear things up in advance. The argument I am not making is that students and families take no action to increase their need, right? So we all know there's online calculators you can use to estimate your expected family contribution now. There's lots of guides about what actions you can take to maximize your need, maximize your financial aid. That's okay. All I need is for this person right here to not realize ahead of time, oh, I'm not going to get a Pell Grand. I need to do this X, Y and Z to get a four hundred dollar Pell Grand. And so first I'm going to give you sort of institutional details why this won't be happening and then I'm going to show you things in the data that suggest this won't be happening. So first of all, many of the inputs for the FAFSA depend on information that you're reporting in your tax returns such as AGI. But in many years where this threshold actually falls is not reported until after the end of the tax year, right? So sort of everything that is going into your 1040 into your tax return has already been fixed. The tax year is over. You could still lie on your FAFSA, right? You could put in a different AGI, but actually around one third of all FAFSAs are audited through the Department of Education verification process. This is much, much higher than the IRS audit rate. And in recent years there's becoming more and more linkages between the data report to IRS and data report to Department of Education making it easier to audit these FAFSAs. So you can lie, but chances are you will be found out. The other thing that I'm going to show you, yes. How often do they catch people? How often do they catch people? So they report this data online. I do not know it out of my head. And so the question is, are they catching people who are cheating or are they catching people who are confused by this massive, complex FAFSA form and not understanding what they need to report? Things are hard to disentangle. Okay, so I made this argument. Students are not sort of manipulating this measure of need. There's also things we should see in the data. So for instance, if we plot the number of students with a particular EFC, and this is just a standardized measure of the number of students, we shouldn't see a jump in the number of students right at the hell-grant eligibility threshold. If we saw a jump, this would suggest that sort of students were realizing they were right over here and doing something to become eligible for Pell Granted. And we see no jump. We can also look at the distribution of observable characteristics, which I'm happy to talk about afterwards. I'm not going to show you now for the interests of time. But sort of all evidence in the data suggests sort of what we might expect, that students have imperfect control over this measure of need, and they're not able to perfectly choose their expected family contribution to take advantage of the eligibility threshold. Great. So really quickly, I'm going to go through what we would expect to see in the data if, say, schools were capturing 100% of Pell Granted, and if on the other hand schools were capturing no Pell Granted. So I've drawn this stylized depiction of the Pell Grant Program schedule where sort of everyone as they comply or everyone attends full time, we've got a change in the level of $400 due to the Minimum Pell Grant Award and a change in the slope or this relationship of negative one. Your need increases by a dollar or you get a dollar more Pell Granted. Okay, what would happen if there was no crowd out? So what this blue line is representing is a hypothetical relationship between institutional grant aid and need. And what we're really looking for is what's happening right around where the Pell Grant changes discontinuously. And here in this hypothetical scenario there's no change in the slope and there's no change in the level, right? So students on either side who are as good as randomly assigned to say get $400 of Pell Granted or get $0 get the same amount of institutional aid. Okay, so what would we expect to see if there was full crowd out? If schools captured students Pell Grants dollar for dollar. Well, what we expect to see as students are responding through this institutional aid mechanism is both a discontinuous decrease in how much institutional grant aid you're getting, right? You get $400 more Pell Granted. The school reduces your grant by $400 and a change in the relationship because as your need increases you get more Pell Granted but then your institutional grant aid sort of follows dollar for dollar, right? So these are all hypotheticals. I just want to make clear what we're looking at in the data and what we're looking for. Okay, so now I'm going to show you what's actually in the data. And this figure is graphing A, the relationship between Pell Granted and need, similar to what I showed you a few slides ago and the relationship between institutional grant aid and need around the eligibility threshold. And each of these markers is going to be demeaned to take into account the fact that I'm looking at four different years of data. So I'm including your fixed effects and I'm including school fixed effects to take into account that students are attending different schools. So I'm just taking out the average, say, institutional grant aid for students attending Washnaw Community College versus University of Michigan, although there's very few Pell Granted recipients, to take into account their sort of unobservable differences. So what is going on here compared to those pictures I just showed you? If we look at the level of Pell Granted right at the threshold, what do we see? We see a discontinuous increase in the level of Pell Granted. So students get $400, say, more of institutional grant aid and they also get more Pell Granted. This would suggest that schools respond to Pell Granted by increasing institutional aid or that Pell Granted crowds in institutional aid and I can put this into an econometric framework and estimate that if we just look at the change in the level that schools respond to a dollar Pell Granted by increasing institutional grant aid by 50 cents. So that's one story, right? What happens if we look at the change in the slope here? So as for students that are eligible as their Pell Grant are using similar formulas to the Pell Granted eligibility threshold, at least some places are like if we deem that we are eligible for Pell Granted we also deem you worthy. Is that evidence of that? So that's a possibility. There's a lot of schools and there are a lot of states. Okay. But the EFC, so where our students are able to qualify for the minimum Pell Granted award is changing over time. So it would have to be not just sort of a relationship between this measure of need and the provision of state grant aid but also the fact that on one side of this particular value of EFC in a given year students are getting Pell Granted. Many states have those rules. We see the state grant program. We see the current EITC and the state EITC and the climate. Right. So you're saying that what is the... So I guess what I'm saying is that there's a possibility of that member that's a possibility. Okay. I would plot state grants. Are you going to? No. There are definitely state programs and these programs are only for Pell Granted recipients. And so there's a and I believe there are also I don't know if Michigan has such a thing. No. Michigan's names that way. But there are definitely states in which it's structured such that here's a grant for Pell Grant people. So what that would matter here in terms of the interpretation is it would just make the coefficient smaller. Going on out there. That would actually increase this further. The discontinuity would be even later because you're also getting state money. Exactly. Are you going to talk about COA? I will talk about it now. Okay. So when I talked about the Pell Grant program formula I said students Pell Grant primarily depends on their expected family contribution and the maximum Pell Grant. There's another component that may matter for students attending very low cost schools or students attending sort of very part time. And that is the cost of attendance. So cost of attendance equals tuition and fees books and supplies and living expenses which depend on where you live. So room and board at dorms a smaller expense if you're living with your parents an expense if you're not living in the dorms and not living with your parents. And so if this cost of attendance is less than the Pell Grant award you qualify for then you will only get as much as the cost of attendance. Now what's going on here is that I'm thinking about what's happening to students right? And so the variation in the Pell Grant award is from say zero to a thousand dollars, right? So we would have to have students attending schools that had very low tuition or were attending sufficiently part time that this cost of attendance requirement would be binding. In practice around the threshold this is really not a concern. If the minimum Pell Grant award was much larger if it was ten thousand dollars then we would worry we wouldn't worry about this but this would be something I would talk about more. So this is a constraint but in practice for the students that I'm looking at it's not going to matter. It's part of the slope though too in the sense that you know you're going to go on further you're going to have students who are running their COA and their institutional grant is getting reduced so I'm going to look at remaining need and see if this is a mechanical relationship I'm not going to show it here but it's not. So students that are even receiving the maximum Pell Grant award have on average even those attending non-selected public schools which are very cheap have on average I'm not going to get the numbers right. It's in the paper they have a sufficiently large amount of unmet need that once you take into account Pell Grant aid and state grant aid that this couldn't be a mechanical relationship. But that's a fair point. So I think the point that even if the state aid isn't using exactly the same threshold as the Pell which probably isn't or maybe it is I don't know but even if that's fair you still could have a problem particularly with the RK because that's estimating the slope difference on each of these specialties and so even if the threshold was like $100 to the left that's not going to change your estimate of the slope because so much of the data is helping to estimate the slope. So I mean I can take into account the fact that things might be going on out here by including controls for flexible functions of expected family contribution I can also look at neural ranges around the threshold and things stay relatively similar although they get noisy because I'm losing data. So to the extent that state grant aid is piggybacking on top of Pell I just need to scale scale down my estimates by sort of however much actual total grant aid is increasing here right and I've done that and things look pretty much the same but I'm not going to show it here you'll have to take my word on it if you really care I'll show you some other time but I want to get to the punchline are there any other questions first? No. Okay, great. So if we just look at the change in the level of Pell grant aid we would come to the conclusion that Pell grant aid crowds in institutional grant aid so schools respond to a dollar Pell grant aid by increasing institutional grant aid by about 50 cents however what happens if we look at the change in the slope of the relationship between institutional grant aid remember as your need increases you get a dollar more Pell grant aid a statutory dollar more empirically about 70 cents however for institutional grant aid as your need increases your institutional grant aid actually falls so if we're just looking at sort of how the slope of the relationship between institutional grants and need changes at the threshold we would actually come to the opposite conclusion that Pell grant aid crowds out institutional grant aid by about 20 cents on the dollar okay so before I get into interpreting this result I'm going to show you some heterogeneity across sectors and the first is looking at public institutions now note that these two scales these two axes have different scales and again I'm plotting the relationship between institutional grant aid and need this sort of black short dashed line here is just escalating from this relationship between institutional grant aid and need for ineligible students what this is basically saying is if we think sort of this represents how colleges are giving out aid depending on need this line will represent the counterfactual of how much grant aid Pell grant recipients would have gotten if the Pell grant program didn't exist but if we focus right around the eligibility threshold we can really see that public institutions are leading the charge here in terms of this crowd in and I'll quantify sort of how much is going on here if we look at non-selective private institutions so this is grouping together for-profits and open admissions non-profits and I'll show you in a few slides that these schools are behaving similarly to Pell grant aid we don't see any change in the level at the threshold and we do see a change in the slope suggesting that the only thing going on here is crowd out and then finally we can look at more selective non-profit institutions and I always want to make clear here that these are not going to be the Columbia's and Harvard's these are just non-profit schools that have some criteria for admission they're still admitting most of their students and in this case we see the largest degree of crowd out so note again that we have different scales on these axes suggesting that for every dollar Pell grant aid received by students attending these schools their institutional aid falls by a substantial amount okay so what I'm showing you is that there's a lot of heterogeneity across sectors so how schools respond to Pell grant aid is going to vary depending on whether they're a public a non-profit or a less selective non-profit and in the paper I go through a model suggesting that this is not these patterns that we observe are not consistent with a world where schools maximize profit so they respond to Pell grant aid to maximize profit so this is reasonable no one I'm not going to stand here and argue that we all thought public schools were profit maximizers that's not realistic but we can give an alternative framework where schools have preferences over the characteristics of students they serve right so they may care about serving students that are from different economic backgrounds, geographic backgrounds and they may care about whether or not you receive a Pell grant right and so in this world receiving a Pell grant affects how the school treats you or sort of their willingness to pay for you to come so why might schools care about Pell grant recipients well schools might simply have preferences for diversity right they may care about having lower income students in their population the US News and World Report recently started reporting the percentage of students in each school that are receiving Pell grants so this may be a reputation issue talking to fundraisers one of the statistics that is most effective in getting people to donate is we are 30% of our students are needy they're receiving Pell grants we serve a needy population why don't you contribute and then finally thinking about state grant aid programs so the framework I have in mind here is not necessarily the one that has been mentioned but there are some states that give schools a lump sum amount of grant aid depending on in part the neediness of their student body so you get more Pell grant recipients today you can get more grant aid from the state in the future period right so this might be another reason why states would care about serving Pell grant recipients yeah so I would love to do that right I run out of data when I start cutting the data that finally so I there's a reason why I show you these graphs for students way out away from the eligibility threshold it's because I don't have enough data to sort of get reliable estimates and when I start cutting into smaller and smaller categories it just makes the problem more something I would love to look at with different data so anytime the department of education wants to provide me with similar information I could definitely look at that but it's not a question I can answer yes Kevin how does merit aid fit into this this wouldn't be a concern with the RD but with the R Kink and that EFC potentially is correlated with achievement or something so if you're talking about the students way on the left-hand side they might have sort of lower achievement scores and they get less merit aid from the institutions so I'm going to go back to this argument that my estimates are thinking about students are coming off of students right around this eligibility threshold and so you may not believe that as I flexibly control for expected family contribution that's eating up sort of all of the relationship between schools willingness to give merit aid and need I can look at sort of narrower windows and the results get noisy but they're consistent so but not sort of necessarily a large away from the threshold in the same way the regression discontinuity requires observations I guess in the limit you need just one on the other side with the regression Kink you need two but okay great so sort of if any of these reasons cause schools to care more about not care more but care about whether or not you're a Pell Grant recipient then there's going to be two things going on as students move from being Pell Grant and eligible to eligible right there's going to be a change in your outside aid you're getting Pell Grant aid right and the school would like to capture this regardless of whether they're profit maximizing or care about other things so they have sort of the maximum amount of resources to redistribute sort of in the way they'd like but then students also receive a label of being a Pell Grant recipient and this may affect some schools willingness to pay for them and this is going to work in the opposite direction if sort of any of these channels are working right you get a Pell Grant you are more valuable in the school's eyes and they're willing to provide more grant aid for you so I do some fancy econometrics which I'm happy to talk about afterwards but I can basically disentangle these two separate parameters using these two in the kink and using the discontinuity and estimate on one hand how much of each dollar of Pell Grant aid is captured or is crowded out by reduction in institutional grant aid and then how much does this labeling affect change schools willingness to pay for Pell Grant recipients and overall I estimate that 20 cents of every Pell Grant dollar is crowded out by reduction in institutional grant aid so I get $100 of Pell Grant aid which results in me getting $20 less of institutional grant aid so my total grant aid package increases by $80 however for students that are right around this eligibility threshold for students that are receiving Pell Grant this labeling effect leads schools to provide an additional 330 or an 18% increase in institutional aid because of this labeling effect around this eligibility threshold this willingness to pay is going to dominate if we're just looking at these students lead us to conclude that Pell Grant aid increases institutional aid when overall this is not the case but first I'm going to look at different sectors so here I've broken out schools by their control and their selectivity so these top two are public institutions that are open admissions or more selective this first parameter is what I'm calling pass through or the capture so you can think of every dollar of Pell Grant aid among open admissions public schools leads to an 11 cent decrease in institutional aid among more selected public schools that leads to a 13 cent decrease but due to this willingness to pay open admissions public schools provide $220 additional dollars to Pell Grant recipients which is a 92% increase at the mean and selective schools provide an additional $860 to Pell Grant recipients which is a 120% increase at the mean but focusing on the remainder of these schools so these are the third row is the open admissions nonprofit schools there's no sort of preferences for Pell Grant recipients no willingness to pay 18 cents of every dollar Pell Grant aid is crowded out in the for profit sector I estimate that crowd out is around 10 cents on the dollar or 10% which is actually the lowest of them all and finally if you remember that third graph I showed you with the sharpest kink as that would suggest students in attending more selective nonprofit schools experience a great crowd out so every dollar Pell Grant aid leads to a 72 cent reduction in institutional grant aid and there may be some willingness to pay here but it's very noisily estimated so I'm not going to hang my hat on it so bottom line here is we can say something about differences in objectives public schools clearly seem to place value on the fact that they're serving Pell Grant recipients crowd out is largest among nonprofit schools and open admissions nonprofits and for profit schools have the on average the smallest degree of crowd out so they're putting this together with Chewini and Golden would have the dependent variable in the net price yes and that would suggest a very different answer well I don't know it might be only for the for profits basically you're manipulating something but you could do this I'm looking within I'm looking within schools I'm looking yep okay mm-hmm right mm-hmm mm-hmm mm-hmm mm-hmm mm-hmm mm-hmm mm-hmm no this is an important caveat right if these schools find the easiest to raise tuition because all of their students are receiving Pell Grants and this is not capturing the whole story exactly and I will raise that again before the end before we conclude sort of for profits are capturing the least but in terms of this particular identification strategy right now I'm looking within schools within years right and so I could look within schools across years I have to think about that it would be cool if I was able to say something about the the net price mechanism okay so Sue asked can we put this together and say something about the overall degree of crowd-out so on average how much of every hundred dollars of Pell Grante is passed through to all schools or to schools in a particular sector so we're back to our very nice hypothetical diagram here where this green line again is this hypothetical depiction of the Pell Grant award we've got the discontinuity in the kink and this blue line is sort of an exaggerated version of what I observe and the relationship between institutional grant aid and need where there's a discontinuous increase and the change in the slope right so this is what we see in the data okay so what assumption am I going to make the assumption I'm going to make which I talked about a little bit earlier is that if we look at this relationship between institutional grant aid and need for students who are just barely not eligible for Pell Grant aid we can use that to extrapolate and predict how much these students who are receiving Pell Grants would have received an institutional grant aid if there was no Pell Grant program right and so that's what this dashed line is that's the counterfactual or the hypothetical relationship between institutional grants and needs for students that are receiving Pell Grants if the world was one where there was no Pell Grant program and was imposing some assumptions on the data that I can say that I can look at these people here and say something about these people there that are going to be very different okay so these results you should look at with a little bit more skepticism than the ones I was showing you in the past which was saying what's going on right around this threshold okay so if we look at this area under the green dashed line this is the total amount of Pell Grant aid that the federal government is giving to students right we add up the Pell Grant awards and we add up how many students are receiving each one and that's that's that 35 billion dollars um this area here underneath this dashed hypothetical line is how much granted from the school these students would have received in the absence of the Pell Grant program and conversely this area under the actual institutional aid line is how much students actually receive right and so we put these two areas together and they give us um Triangles A and B which are representing the winners and the losers so people in Triangle B are the winners right they're actually getting more institutional grant aid so they're getting this amount than they would have in a world with no Pell Grant aid Pell Grant program due to this willingness to pay um and conversely the people in Triangle A are the losers so they're getting less Pell Grant aid sorry less institutional aid than they would have gotten in a world with no Pell Grant program right so we have to sort of account for the winners and losers um we subtract out how much people in Triangle B are gaining from how much people in Triangle A are losing and that gives us and divide it by the total amount of Pell Grant aid going out and that gives us an estimate of the average percentage of Pell Grant aid that is crowded out um by schools responding by decreasing institutional grant aid how much of every dollar Pell Grant is passed through from students to schools so on average um around 15% of all Pell Grant aid is passed through to students to schools from students although if we look sort of at a the confidence interval so sort of if we allow there to be uncertainty in this estimate we um can bound this amount by falling between 9% and 20% okay um if we look in public institutions on average crowd out is around 4% so it's very very small what this is telling us is that these two Triangles so the total amount being um lost by the losers and the total amount being gained by the winners is very close to being equal right so in the public sector sort of we know where these captured Pell Grant funds are going they're going to supplement the Pell Grant of students who are just barely eligible right so the funds captured from people in area A are being transferred to students in area B um on average these open admissions private schools so these are nonprofits and for-profits are getting around 18 cents on the dollar and these more selective nonprofit schools are capturing close to 70 cents on the dollar okay okay great yes so there's two reasons why you could see like a zero in this one is because an institute a set of institutions just doesn't have institutional needs so there's nothing to crowd out so what we'd ignore on this would be as a proportion of the aid that they give out so much to crowd up and the other is that they give a lot of aid but it's unrelated to the Pell Grant schedule yes yes or that these two areas cancel each other out so I mean in this case it's the areas canceling each other out it's neither right so they're giving out aid they're just um it's mm-hmm yep but what this is saying is on average the Pell stays the same at those places there can be a place that doesn't give out much you could have two places the same tuition price one of them doesn't give out much the other one does the Pell is the same at each right so we're getting the weighted average of those two right so we're including schools that sort of can't respond because they don't have any institutional aid to give out and schools that can respond right and so right so yeah yep no no that's a good point um so public schools the open admissions public schools give out the least amount of aid right but we're still seeing this crowd out so if I if I mm-hmm but this is masking heterogeneity right so these students are gaining and these students are losing and they're approximately equal um so it's a zero overall but it's representing a lot of transfers between the most needy Pell Grant recipients to the least needy Pell Grant recipients mm-hmm yes this is in percentage terms right so it just has to be there's some aid to crowd out what I would want to do is eliminate schools that absolutely can't not eliminate them but that's that's sort of part of this picture right if you're a school that doesn't respond because you have no aid we want to take that into account um and that's going to drag these results towards zero but it's part of the big picture right yes Sheldon I do think you want to look at that so essentially the difference between the community yes mm-hmm don't have mm-hmm aid and the example of yes presumably if there were no Pell Harvard would still fill that in so it's a hundred percent of Harvard right and it's zero if Harvard has this policy that they're going to right provide aid to everybody below 50,000 then all of these people are getting full aid right it's a hundred percent of Harvard and it's zero in a community college without it would be interesting to see what it's like without some of those so first to answer the Harvard part there's very few Pell Grant recipients going to schools that have these policies by and large these are going these students are going to schools where they are still sort of at risk for remaining costs so they're not schools that have a blanket policy of covering full tuition for students that are poor um it's just the nature of where Pell Grant recipients go but I agree with the second part that it is worthwhile to look at schools who have some margin to respond but if we want to think about sort of the overall response taking aside the Harvard's which are sort of not in this set these are not Harvard's these are schools that admit 60 percent of their applicants so they're just private schools that are nonprofits that have some criteria for admission by and large so assuming that there's not this mechanical proud act I think we still want to take into account schools that don't respond because they have no institutionally because that's part of this overall big picture I don't think anyone's arguing about the 50.148 when you break it down by sector yep then turns into there's differences across sector and it's because of perhaps preferences and perhaps it's okay budget constraint so it could be one of the two your point is well taken great okay so I'm going to argue it's differences and preferences stay tuned if I find something different I'll change the punchline here I show that public schools appear to value the fact that a student is a Pell Grant recipient so there's this labeling effect which increases schools willingness to pay for these students for-profit schools along this margin differently than non-selective non-profit I'm not looking at tuition though and the big caveat here is that many of these schools may find it more effective to respond by increasing tuition rather than altering institutional aid I find the greatest degree of pass through among students attending selective non-profit schools so this is not a mechanical effect these students on average have over $10,000 of remaining need so it's not the fact that they have no need left to fill at the eligibility threshold and I argue that this may be due to differences in market power so these schools have fewer competitors and so they can extract larger from these students and so the bottom line is under these stronger assumptions that I can trace out the counterfactual relationship between institutional aid and need if there was in the case of no Pell Grant program I estimate that on average 35% for every dollar increase in Pell Grant aid so if we take this to the data in 2011 like I said at the very beginning students received 35 billion in Pell Grant subsidies and this suggests that 5 to 6 billion of this amount was passed through from students to schools via price discrimination now these are big numbers but what does this mean for policy so does Pell Grant aid have any credibility this is a resounding yes so students are receiving the majority of their Pell Grant aid if we look at how much information schools have if this was a market where schools were truly profit maximizers we would not see students getting 85 cents for every dollar this is actually a large proportion given the setting that we're in so I want to make that clear what sort of can we think about doing in order to aid in reaching the students and reducing their price so one way that is possibly the most uncontroversial I don't know people can argue with me is to increase students information about what schools are doing if I know I can go to this school and I'll only get 30 cents of my Pell Grant versus going to school B and I get 90% of my Pell Grant this may affect my decision so students the most clearly observable measure of price is this listed price of tuition but I'm arguing that there's a large variation in the actual prices students pay the federal government has started trying to make this information easier to obtain so they have a net price calculator you can put in some information about your family circumstances and it will tell you for a given institution sort of on average what students are paying it looks like you and the recent college scorecard is also intended to provide more information to students that may affect their decisions but neither of these is sort of dealing with the particular problem I am posing today which is this crowd out or this institutional aid response to Pell Grant aid and so one idea would be to use the FAFSA as a tool to provide additional information so for students that apply through the FAFSA there could be a way to provide these students in turn with more information than just their expected family contribution so in the FAFSA students list what program what school they're interested in attending and the department of education actually has information on what students paid the net price listed price of tuition minus institutional aid that students paid for that particular program if they were able to work with the IRS and so for students that are applying to a particular program they could actually there could be a way to help them observe not just the average cost and average institutional aid but what students with their sort of very specific EFC were getting from this school there might also be a way to use the FAFSA to provide information about other schools in the local area so you only list one school but there's five other schools that are within 20 miles of you and we could also send you information about what sort of institutional aid these schools are giving to their students who will look like you and then the final thing is that really we can't just look at sort of overall levels of tuition when thinking about college affordability or even overall net price so this is graphing the net price that students paid in particular sectors according to whether students were receiving the maximum Pell Grant so these are the most needy any Pell Grant or whether they were in eligible and this is sort of at least letting you judge how much institutional grant aid schools are giving to students in particular sectors so for instance the difference in net price between the neediest student and needy students who are not receiving the maximum Pell Grant award is sort of the largest perhaps among four-year non-profits and doctoral granting institutions and I'm not arguing that this figure is the end of the story this is looking across institutions students that receive Pell Grants are going to different institutions but something like this would be a useful source of information to say look at schools in your surrounding area if you were not willing to provide if it was impossible to provide individualized information so at least you could say the poorest students are paying less than poor students and students who are ineligible for Pell Grant aid okay, yes yes I'm slowly learning this literature great money into aid programs like Pell Grant like student loans if the federal government gives the money here's the money if they don't provide like administrative support then that might be another policy information I'll put the expert answer this question well I mean no so these are typically these are just kind of mandates where staff has to be provided in order for schools to meet their kind of minimum functions but the government doesn't provide any kind of subsidies to financial aid offices to staff additional people in their class of policy that's getting required extra hands to process those dollars so say like if like I don't know like maybe like grants in general to like university researchers like funds kind of set aside for like you know administrative support could that could that be like another policy implication for how to get more of the aid to the students I think that institutions would have a hard time going along with the policy that requires them to hire a certain number of staff when they don't know the institution's needs or kind of like the institution's capability in processing aid so I think that would be kind of an infringement on institutions themselves in terms of like their payroll and benefits and things like that because likely the grant would cover that amount but also the government can't really doesn't specify a number per program that you should have employed in your financial aid office either so I think it would be putting a lot on the institution so there's an administrative allowance that goes with and it's hooked is this the right phrase that they use that FSA uses? administrative allowance that travels with and I don't think it's like person or per dollar it's tiny but basically it's the hidden cost of the programs that they administer it at most it's like 2% or something of 2 cents and a dollar it's small and I don't know if it's per dollar or per head I don't know off the top of my head either because you think if they do essentially need to recapture the cost of buying it that would be an interesting number to know but basically this is how they get their overhead yes exactly so that would be a nice thing to add to this yeah something that's probably knowable okay but just to be clear these policies that I'm recommending are all sort of demand side policies so none of them are imposing additional regulation on schools they're aiming to increase the amount of information students have to maybe help them choose schools where they get the most bang for their buck okay great so okay so with using net price I think that part of the problem with using kind of a net price framework is the fact that institutions don't necessarily have a uniform awarding methodology so as that varies across institutions that can be kind of an issue in assessing net price but there's also kind of the added issue of small schools might have very few students in a certain kind of income set that the net price calculator can skew your year greatly based on so you would want to throw out small cells or sort of not base policy decisions on small cells mm-hmm do you have a sense of if there's a wide distribution of institutional aid around like EFC price points so if someone's expected family contribution is let's say a thousand dollars under the Pell threshold is there a pretty wide is there a lot of variation if you're looking across all different sectors, yeah so could there be a role in that given the Pell grant students are now more willing to accept like enrollment from the lower end of like this distribution um is that contributing to the widening of the right so all of these, the estimates I'm showing you are looking within schools so it's not being driven by the fact that students, I don't have the picture students over here attending different schools and students over here it's all coming from within schools right so I guess I'm thinking like without the Pell if maybe you know the only people who accept you know a financial aid package of people who receive two thousand dollars or more in aid but now that you get a Pell you're willing to accept fifteen hundred dollars oh is it changing the types of school students go to yeah so this is a really interesting question that also I'd love to address in future research I can look at that here and I don't find a lot of evidence of sorting across different sectors right but this is probably going to depend on what a student's specific situation looks like if there's only one school within fifty miles of me and by and large Pell Grant recipients go to schools that are very local I'm not probably going it's not going to matter right and so what would be really interesting is to look at students who have choices and whether receiving a Pell Grant affects the choice of what school to go to so I can say in my noisy data I don't see that happening but that's not going to rule it out right but it's going to be sort of a different if student so I can look at measures of school quality okay and to see if students are upgrading in terms of things like institutional resources so spending on instruction for full-time equivalent spending on institutional grants for full-time equivalent and I don't find any evidence of that going on but that's not to say it's not right that's a good point okay so I'm about to be cut off okay