Joseph Mullins on Valuing Parental Time and Children’s Development in the Design of Cash Transfer Programs

Dave Chancellor [00:00:04] Hello and thanks for joining us for the Poverty Research and Policy podcast from the Institute for Research on Poverty at the University of Wisconsin–Madison. I’m Dave Chancellor. It’s well-documented that both money and parental time matter for children’s development and ultimately for the collection of skills and personal resources that they’re able to collect in life. We often call that human capital for short. However, when it comes to cash transfer programs like welfare for single parents and especially mothers, most of the evaluation and economic modeling efforts have focused on how those programs impact the amount of paid work single parents do. But much less attention has been paid to the value of parental time. For this podcast episode, I spoke with economist Joseph Mullins of the University of Minnesota. Months developed an economic model for U.S. Cash transfer programs that attempts to place an accurate value on parents’ time when it comes to assessing cash transfers. And he says his model suggests a very different structure for cash transfer programs. If you want to best balance children’s need for money, resources, and parental time for the healthy development. Let’s turn to the interview.

Dave Chancellor [00:01:16] Professor Mullins, thank you so much for being here. I appreciate your time today.

Joseph Mullins [00:01:21] Thanks for having me.

Dave Chancellor [00:01:23] We are talking today about a paper you wrote. It’s currently a human capital and economic opportunity working group working paper called Designing Cash Transfers in the Presence of Children’s Human Capital Formation. In this paper, I mean, you start with the introduction with this with this one sentence argument. It says “A proper accounting of the costs and benefits of cash assistance programs for households with children must include the long run impacts on children’s human capital.” And so that I mean, to me, that doesn’t seem particularly controversial. So, what’s the what’s the story here?

Joseph Mullins [00:01:58] Great question. I’m going to borrow a thought that I can’t attribute to myself. So let me refer to a working paper by Hilary Hoynes and Anna Aizer and Adriana Lleras-Muney. Yet as many you know, around the time of the child tax credit is being discussed in public policy discussions, they released this working paper talking about it. They make the point that the Congressional Budget Office, when they’re evaluating how much it will cost to, for example, expand the child tax credit by a certain amount, they take into account short run stuff, which is more the classic public economics programs. A dollar spent on the it doesn’t cost the same as a dollar spent on TANF because of labor supply responses. So, it’s accounting for those behavioral responses is really important. And that seems to be that’s a paradigm that’s existed in public economics since forever, since at least the fifties. But this other piece of like thinking about the long run, thinking about human capital formation, that’s new. So, for whatever reason, I don’t think it’s found its way into mainstream policy analysis, let’s say. So that’s the missing piece. And I think in addition to a lot of the other stuff, that that’s a big picture thing. So, I’m not the only person to be contributing, I think, along those lines.

Dave Chancellor [00:03:13] You know, I do want to talk about this model that you’ve put together for this paper. But I’m also kind of hoping that you can talk to me about sort of economic modeling more generally. It feels it feels to me like models sometimes get a bad rap. But, you know, as a non-technical reader, I sometimes I sometimes skip over these parts of the papers, you know, and I think other times I see people being skeptical of macroeconomic models, you know, others saying maybe they’re not that useful or maybe they don’t reflect reality or perhaps that they don’t really incorporate lessons from other disciplines. Well. So and what do you what do you say to that?

Joseph Mullins [00:03:52] Oh, we could talk about this all day. Right. Yeah. It’s juicy topic. So I teach in my Ph.D. class to my second year students. We talk about models a lot. And, you know what purpose the models have. I think I would start by saying, good, we should be skeptical of models. The key challenge is always that. Once you add layer upon layer, you’re losing transparency. So, you know, I’m going to come to you with a policy recommendation and it’s going to be built on layers of assumptions. And it’s difficult to do things transparently when there’s many complicated layers. So, I think I totally respect that critique. I would say I think it’s worth noting, almost tautological that anybody that’s giving you a number has a model in mind. Now, it might be a simple model. It might just be a number that I made up based on my reading of the literature, which you’ll be surprised to find happens pretty often. So I think like the first layer is just to say, well, a bar we should set for ourselves is if I’m going to give you a number, I’m going to write down every assumption I needed to get to that number. Right now, if I have a really simple model, I maybe don’t need that many assumptions, but that really simple model isn’t going to necessarily life is the world is complicated. Right. And so it’s not clear how well these models do it. Extrapolating or thinking, building in all the mechanisms that we think are relevant when we do policy analysis. So, you know, I think that’s how you end up getting led down the dig out and posits adding in more layers in order to hold your model accountable to more aspects of the data. Now that explains how we end up where we end up or people like me end up where we end up where you have know all of a sudden complicated. So I guess a question you have to answer a lot is why make the model so complicated now? One reason will be just the more data we can hold the model accountable to, the more we should believe in it. And I think there is this other layer. This is a whole other kettle of fish how we don’t have time to get into it. But economic models specifically provide us with notions that are useful for policymaking, notions of efficiency and welfare based on revealed preference. I find that to be personally, I’m not a sophisticated thinker when it comes to politics, but I personally find that a powerful rhetorical tool to say, Look, I’m going to assume that people know what’s good for them. Did they make decisions according to that? And here’s a policy that’s going to make everybody better off. Now, there is political content baked into that statement, of course, but that that’s something that appeals to me personally. So that that’s obviously you may have a set of philosophies that doesn’t jive with that way of looking at the world. But again, that’s like I’m drawn to that sort of really core feature of economic reasoning that gives us notions of efficiency, in particular this idea that like, are we looking at a situation where we can make everybody better off if we are, let’s do that. Let’s make the pie bigger and redistribute.

Dave Chancellor [00:07:00] So you said just a bit ago that sort of a basic requirement that we should hope for of people that are making models is that you list out the assumptions that you do. And so can you talk about sort of the pieces of the framework that you’ve developed here and how they can help us think about how cash transfers really do matter for children’s human capital development?

Joseph Mullins [00:07:20] Totally. I think we can divide the model conceptually into two big blocks that matter a lot. Now, the first big block is about quote unquote preferences or the behavioral model. So if I change the incentives of a particular set of transfers, for example, if I introduce time limits on welfare participation, or if I increase the earned income tax credit by some percentage, what is the response going to be in labor supply? Okay. So that that’s a classical piece that’s sort of as we said before, it’s at the heart of all public economics to think about how individuals respond to incentives. And that’s an empirically well founded sort of piece that I need to get right in the context of the data that I’m using. So that’s one piece. The second piece is the is human capital formation or the effects of two key resources, household income and maternal time at home. Now, I say maternal because, you know, I have this since a broad brush stroke type of statement that I guess I should be upfront about that it’s really a model of single parents. Single parents participate in welfare overwhelmingly single mothers. Forgive me for all future uses of maternal time as the shorthand for single parent time. These are two resources that, like prior literature, tell us both matter. And so that’s really the key, is to say, actually, all this huge literature we build around thinking about how best to design taxes and transfers. They ignored the fact that every dollar matters for kids’ skills. And every hour matters every hour at home. It matters for kids’ skills. And so we have to quantify that, too. So those are the two big pieces. The first is really where all the modeling comes in, because that’s one of the classical assumptions pieces of an economic model, preferences and technology and so forth, wages. That’s sort of, let’s say, more standard. The second piece we’re trying to let the data tell us as credibly as possible what we think the marginal effects are of additional income for kids’ skills and mother’s time at home. So those are the two big pieces that.

Dave Chancellor [00:09:40] With that model, with that framework, you do two exercises in this paper. And in the first of these is you look at what an optimal cash transfer structure would be if policies really did account for children’s human capital development. And we sort of set out in that initial argument. So can you.

Joseph Mullins [00:09:59] Tell me about that? What I do is I study a problem or a class problem in the public economics literature. That’s, you know, we refer to it as optimal non-linear taxes. Sorry, it’s a bit of a mouthful, but we’re just saying, look, if the government was free to draw any sort of line, you know, imagine a graph where you have on the X axis an individual’s earnings, anything between zero wins the most that we can imagine, somebody in the population earning on the Y axis, we’re going to draw taxes or transfer. So tax transfers will be in the positive region of the Y axis and taxes will be in the negative region. We’re imagining a government that can draw any kind of squiggly line they like basically. Right. And if we give ourselves that freedom, what would the optimal schedule look like? What do we mean by optimal? Well, we have to take a stand on that’s our standard version, is just to say, I have some notion, some concept of welfare that the models are delivering to me. So, I’m going to maximize an average of that subject to the marginal value of the public dollar. So here we’re imagining that we want to give to we want to give transfers and potentially tax the population of single mothers. They will have kids. And there’s a marginal value of dollar that we could spend on this program. Or we could send it off to somewhere else, send it to the military, send it to the education sector, wherever it goes. Sorry. That tells us what the costs are of the program and the benefits are built there into this welfare function. Okay. So this is a classically started problem and the classic answer is, well, it all comes down to it turns out the labor supply responsiveness of the individuals that we’re giving money to. And so that’s going to tell us basically what kind of bang we can get for our buck if people are really responsive. What I’d love to do is really concentrate my taxes at the bottom, but it’s going to be hard to do that if everybody’s really responsive because then they’ll all you know, they’ll all crowd to the bottom and you know, so disincentivizing labor supply, if that is true. Now, of course, the extent to which that’s true, that’s an empirical question. But these are the classic sort of. Pieces of the puzzle in this literature. And we’ve dedicated a lot of effort to learning about, quote unquote, labor supply elasticities. So now all I’m doing is just adding a little piece of the accounting that says, well, there’s the cost today of doing this. But there’s also the effect on future resources through human capital formation. So now I’m looking at labor supply decisions entirely differently because we might have mothers going to work. The net effect on child human capital is going to depend on how important that additional income is relative to the reduce time at home. Now, I’m going to do this in a sort of highfalutin way where we do the optimal nonlinear thing. So you get lots of, you know, do lots of fancy calculus, basically. But I think the messages are pretty simple. We think of boiling down this human capital formation process into two pieces. The first piece is what is the effect of a marginal dollar of household income on cognitive and behavioral skills? And what is the net effect of a single mother going to work? So she’s going to arrange some sort of care for the child and there’s going to be some impact on their skills, positive or negative. If we find out that the net effect. Is. Is costly. That is, if time is really important, we might see negative net effects of maternal employment. That really is going to change the planners accounting to say, well, actually before I was worried about providing disincentives to labor supply because it makes redistribution more costly. Well, there’s is now there’s this benefit, even if mothers are more productive at raising their kids than they are in the labor market. That pays off in the future. Okay. So that’s one piece. The second piece is to say overall. If money still matters, I have a goal for what I think net income should be for every household. The more important money is, the more I’m looking at boosting net income. So I think of this as like the general rubric is size and shape. What matters for size? It’s the overall importance of money that tells me how much I’m pumping in to this population. And shape is like how important this time because that’s going to tell me how much I want to use labor supply incentives over just focusing redistribution at the bottom. That’s sort of the rough heuristic that’s hidden behind first order conditions, as we call them.

Dave Chancellor [00:14:56] I’m hoping that you can kind of give me a sense of that sort of size and shape as we think about it in terms of business or the income distribution.

Joseph Mullins [00:15:04] Here’s what I’ll say. I will define size here, quote unquote, to be when everything is said and done. I set my schedule. People make their decisions. What is the average net income for every household in the population now? The formula that I derive says. If we define that to be size, it’s directly proportional to. A parameter in the model that I’ll just defined as the overall importance of money for skill formation. And that’s perfectly intuitive, right? If money is more important, I should have a goal for getting average net income to some amount. So that’s what we mean by size. Shape is more complicated because again, there’s infinitely many versions of a squiggly line I can draw. Right. And all of them are different in terms of their quote unquote shape. I’d say that the rule of thumb here is like, how much does your squiggly line, how much does it implicitly tax the decision to work in terms of if I didn’t work, what would I be getting for the government versus if I do work? How much is that payment reduced relative to my to the additional income I get from earnings? That gives us sort of what we call the participation tax rate. And often like when we design these policies, if they’re highly redistributive, that participation tax rate is really high like it was in the early 1990 for participants in welfare. So it provides in some sense a strong disincentive to work because if I decide to work, maybe my, my, my entitlements, let’s say my welfare entitlement, if I’m participating in welfare that reduces it’s only just replaced by my income. And so maybe my participation tax rate is close to 100% in the sense that my net household income doesn’t increase very much when I decide to work. That’s the that’s what we mean, I guess, by shape is one way to think about shape is like what is the average tax rate on participant work participation? And again, the punch line is to say, well, if time doesn’t matter so much, then we want to we want to encourage work by subsidizing employment, really reducing those participation tax rates. If time matters a lot, which I find for the data I’m looking at it, it appears to it’s not optimal to do that. We actually have a—the optimal system looks much closer to the way it used to look in the 1990s with AFDC and that you have some payment, some entitlement for people not working that we reduce at a almost constant marginal tax rate with earnings.

Dave Chancellor [00:17:49] I just want to dwell on that from when I when I read this. I had to reread it because I want to make sure that I saw it right. Because you suggest that an optimal level is fairly high with little to no earnings, but then it drops off very quickly for low amounts of earnings. So it’s saying like it’s not worth it. You know, like your time is more important. I mean, I think that’s kind of provocative.

Joseph Mullins [00:18:11] Yes, quite provocative. I mean, I. If I’m if I’m being honest, maybe I shouldn’t be dishonest. But it gave me pause. That result is not. I mean, it is dramatic findings. And I think one way to think about it is honestly to own up to what’s incomplete about the study or what. I think that we might have a chance for me later to confess or the other issues with what I’m doing. But the big one, in my opinion, is childcare. So I’ve been saying I’ve been using this phrase a lot. What is the effect of mothers going to work reducing time out? Now, obviously, that depends crucially on the quality of care it’s available to them when they go to work. So that’s, first of all, not going to be the same for every individual. And perhaps more crucially, it’s not policy invariant, which is to say, you know, the quality of candidates available to them is amenable to policy. Now, if I a sure way of answering your question is to say part of the reason for the dramatic result, I find, is that the quality of care that’s available to women in the in the theoretical study I do is health fixed that I think is contributes to the sort of dramatic finding that we get which is like you know that the data is telling us that the net effect of that time there’s some positive effect on mothers staying home, that they’re more productive, raising their kids and they are on the labor market. And that number really adds up. And it ends up being much bigger than what can be contributed to the economy. You know, you have to remember also, these are high school, largely high school educated women in the 1990s who are occupying sort of the bottom end of the earnings distributions.

Dave Chancellor [00:20:08] So that was the first of the two exercises of the paper and the second and we’ve already kind of been to that, right. And so you sort of use this model to evaluate the really significant safety net changes that took place in the mid to late 1990s. And I mean, I think that we have a sense of where, you know, this is going to head. But tell me about that.

Joseph Mullins [00:20:29] As you say, I think the first quantitative exercise combined with sort of theoretical framing of it, is telling us, well, it’s time is really important. What do we know about what do we think we know about the effects of welfare reform? Well, we think that it was successful in shifting women off of AFDC, off of welfare rolls and into the labor force. And, you know, they are largely working, you know, what we might call low wage jobs. So now there’s a lot of heterogeneity there. I think that’s the main takeaway from this exercise and. I sort of think of this as being important. If you just look in the cross section, what you’re going to see is many more people working. About the same amount is being spent on transfers. It’s just that we are spending a lot less on what we call colloquially welfare. We’re spending a lot less on TANF and a lot more on the earned income tax credit. So, you know, we’re shifting away from shifting towards these more conditional cash transfers, if you like. You know, using the magical panel data, I get to see the same individuals through the 1990s becomes pretty clear that you’re actually redistributing away from some individuals to others. And so the net effect is, you know, there’s a negative impact on child human capital for kids whose mothers have a high propensity to participate in welfare if they eligible for it. Those women are getting less from the government now than they did at the beginning of the 1990. And then you have people that have lower costs of work, who are less have a lower propensity to participate in welfare. If they’re eligible for it, they’re getting more and their kids are doing better. So it’s more of a the deep aggregate effect of this is an overall redistribution. And again, it’s you need to you need the panel data. You need to be able to track the same individuals to see that showing up. The aggregate effect on kids is pretty small. And it’s such that if you were running, you know, if you were trying to leverage, say, state, I don’t know if this is I’m allowed to use this jargon without introduced priming everybody that if you were to leverage state variation to like sort of estimate the average effect on kids, you might not see very much because it’s there’s really like it’s a highly redistributive reform.

Dave Chancellor [00:22:56] I want to talk about some of the limitations of this model, this framework that are that you think are important for understanding kind of what it does do and what it doesn’t do. And I guess related that, you know, I mean, your paper fairly clearly call out you know it said that more research is needed here and so what do you think are some of the priorities for more research that.

Joseph Mullins [00:23:18] Yeah. Well we’ve already talked about I’m very eager to talk about this because I think it’s always the first thing that when you put your work out into the world, that’s front of your mind is everything that’s wrong with it. So it’s good to exercise those statements. Now, the one, the big one I think we talked about it is that you can think of there being child care in the model, but you have to treat that as fix the quality of it. It’s being fixed and there’s two reasons that we should care about that. The first is in terms of getting a more accurate, accurate picture of the world. We would want to think about heterogeneity or differences, both in terms of differences in the relative quality of maternal care to whatever care is available to mothers when they go to work. Now, that’s one as we said, one piece is just getting the right picture of the data. The second is maybe more immediate, which is, you know, policy stuff. So like, how drastically does the optimal I’m talking about pulling one level, which is transfers, but we have these other levers that are being in policy discussions are sort of mentioned in the same sentence, like expanding Headstart or providing some sort of version of universal pre-K or even increasing access to child care subsidies. These are things that could, in theory, reduce the developmental cost of women going to work. And so that the evidence exists, I think, to help discipline such a policy exercise is not available in the data set that I use in my opinion, but the evidence does exist. And so I think, you know, if I ever finish this paper and that for me is the next thing I’d like to do, I’d like to work on is think about using these policies in tandem because that’s really you know, they’re clearly complementary, I think, now. I should also say this what I’m calling the effect of maternal employment on skills. There are different estimates of that in the literature. I find in this paper that it’s that there is a negative maternal employment effect on kids using this dataset. I have my own work and others have worked. Actually, I believe your director Katherine Magnuson, has work on this showing using experimental data, slightly different population, but there doesn’t seem to be a lot of evidence of negative maternal effects in those data. So clearly the numbers are we need more data, more evidence to discipline some of this stuff, to summarize better data, but broader range of evidence, and then bringing evidence on childcare into the mix and thinking about what we can do with these policies in tandem. Those are the things that I’d love to sink my teeth into if I ever get the chance.

Dave Chancellor [00:26:14] So I want to think about all that we’ve been hearing here in light of the current policy environment, because you’ve been working on this paper for four years, but I think we’re also sort of in this unique moment. So at the end of 2021, we saw that the temporarily expanded child tax credit, it was a very different approach to the safety net than what we’ve had in the last few decades, really. And at the same time, I think because of the pandemic, or at least related to that, it seems like there’s a good argument to be made that a lot of families are sort of rethinking the way that they balance time versus money. And so there are sort of all these issues going on. And how do you think about all that? And, you know, what do you see going forward here?

Joseph Mullins [00:26:58] One thing you mentioned, which I can barely wrap my mind around, it does cause issues. I think you’re suggesting this notion that, like when the universe intervenes and force us, forces us to work less. Does that make us reconsider the value of working class? That in the model you’d be thinking.

Joseph Mullins [00:27:23] A preference as they have a formation in preferences or something. You know, is it like we don’t allow God to interpret data if you allow people’s true rankings over outcomes to be in flux?

Dave Chancellor [00:27:36] And to be clear, I don’t have I’m not presenting evidence that that’s just sort of my observation.

Joseph Mullins [00:27:42] Your data point and my data point and probably in agreement there if I’m being honest. But again, we said that one of the core principles of economics. Right, is to think about this notion of welfare and use it to make definitions of efficiency. And so it’s important to get preferences right. I think that that is first of all, it is to ask, well, is it possible that preferences for work, let’s call them, are malleable to know larger social equilibria? That’s big picture and that would like changes. I think in the way we understand work decisions would alter our understanding of a lot of questions in public economics, not just my little corner. Well, okay, so does that child tax credit more generally? I think if I had time, it’s the third exercise I would have done here is to evaluate, you know, in theory, the model is an apparatus that says, okay, does this we’ve written down, taken a stand on what’s socially desirable and what isn’t, and we’ve done the accounting. So like, let’s simulate the expansion, the permanent expansion in the child tax credit. What are the returns? And is it is it a good idea? You know, everything is there in the model to do it, you know, in terms of how people respond to the incentives of it and also the returns to the additional income. My guess is that if I were to plug those numbers in, the estimated model will tell us would tell us it’s a good idea. Well, no one cares what I think. But I did try to advocate for it. It was limited to my limited capacity. But, you know, I personally I felt that it was a missed opportunity to let that to not permanently expand the child tax credit.

Dave Chancellor [00:29:24] Thanks again to Professor Joseph Mullins for taking the time to talk. You can find him on Twitter at @JosephMullins. And if you’d like to read the paper we’ve been talking about, we’ll have a link in the show notes. The production of this podcast was supported in part by funding from the US Department of Health and Human Services Office of the Assistant Secretary for Planning and Evaluation. But its contents don’t necessarily represent the opinions or policies of that office and other agency of the Federal Government or the Institute for Research on Poverty. Music for the episode is by Martin DeBoer. Thanks for listening.