Thinking again about earlier-age versus later-age interventions in skills development

The recent paper on the Kalamazoo Promise, by me and my colleagues Brad Hershbein and Marta Lachowska, found that this program, which provides up to 100% free college tuition for graduates of Kalamazoo Public Schools, increases college completion sufficiently to have very high benefits. The program is estimated to have an annual rate of return of over 11%, and to offer a benefit-cost ratio of over 4 to 1. These benefits are estimated based on the likely higher earnings due to the additional college credentials induced by the Promise’s tuition subsidies.  (See previous blog post for a summary of this study and its results.)

These returns are quite high. In my recent book, From Preschool to Prosperity, I estimate that high-quality preschool has a benefit cost ratio of a little over 5 to 1. I also estimate that high-quality child care/pre-K from birth to age 5 for disadvantaged families has a benefit-cost ratio of 1.5 to 1. A similar benefit-cost ratio of 1.5 to 1 is found for the Nurse Family Partnership, which provides pre-natal and parenting assistance to first-time disadvantaged moms from the pre-natal period until age 2.  So the Kalamazoo Promise has a similar benefit-cost ratio to high quality preschool, and a higher benefit cost ratio than some high-quality earlier-age interventions.

In addition, research by Nobel-prize-winning economist James Heckman and his colleagues finds that the Perry Preschool program has a social rate of return in the range from 7 to 10%. For a variety of reasons, these figures are calculated so differently from my figures that they are not exactly comparable. Still, the suggestion is that the Promise intervention that occurs at ages 18 -28 (students have 10 years to use the Promise) may have benefits that are comparable to very high-quality early childhood programs

This raises again the issue of earlier-age versus later-age interventions. Some in the policy community have sometimes argued as if there is some strong regularity, that earlier interventions almost always have higher rates of return.

For example, New York Times columnist Nick Kristof has argued the following:

“When we hear “early childhood education” we mostly think of pre-K. In fact, the earlier the intervention, the better. Helping pregnant moms avoid substance abuse is highly cost-effective, and then helping them through home visitation programs like Nurse Family Partnership in the first couple of years of life is crucial as well. By the time you get to age 4, it’s a little late, and children are so far behind that they never catch up.”

Mr. Kristof may be in part relying on a widely-circulated figure, which shows a rate of return to human capital investments at different ages at which the investment is made. The figure shows the rate of return as starting very high for investments made near birth, and then steadily declining as the age at which the investment is made increases.

This figure can be found at a variety of places around the internet; for example, see it in various materials at the website  , such as at page 6 of this brochure.

This figure appears to go back to some research by Nobel-prize-winning economist James Heckman. He argued in a 2008 paper that

“As currently configured, public job training programs, adult literacy services, prisoner rehabilitation programs, and education programs for disadvantaged adults produce low economic returns. Moreover, for studies in which later intervention showed some benefits, the performance of disadvantaged children was still behind the performance of children who experienced earlier interventions in the preschool years…Remedial interventions for disadvantaged adolescents who do not receive a strong initial foundation of skills face an equity-efficiency tradeoff. They are difficult to justify on the grounds of economic efficiency and generally have low rates of return.”

I think this research by Heckman has been misinterpreted by some to imply that there is some iron law of human capital investments that earlier investments are always better and always have higher benefit-cost ratios. This is not true. Although later investments may have some limitations on whom they can reach, and how high a percentage effect they can have, well-designed later human capital investments often can have high rates of return and high benefit-cost ratios.

I am not the first researcher to point out that the relationship between social rates of return, and the age at which human capital investments are undertaken, is more complex than might be implied by some interpretations of the age versus rate of return figure.  For example, in 2010, noted early childhood researcher Arthur Reynolds and his colleagues reviewed the literature on the benefits and costs of human capital investments at various ages. They concluded the following:

“Figure 8.5 [in this chapter] shows the returns per dollar invested for several types of programs with available cost-benefit analyses over the first 9 years of life by the age of entry into intervention. These include family-centered home-visiting programs, preschool and prekindergarten programs, full-day kindergarten, and class-size reduction programs…Although programs at all ages show evidence of positive economic returns exceeding $1 per dollar invested, preschool programs for 3- and 4-year-olds generally show the highest returns.” (Reynolds et al., p. 181)

Another recent paper, by Susan Dynarski, Joshua Human, and Diane Whitmore Schanzenbach, compared the cost-effectiveness of various policies, including policies at various ages, in increasing college enrollment. This analysis does not of course capture all benefits of these policies, but it does consistently analyze one benefit. They concluded the following:

“…The amount spent by Head Start to induce a single child into college is therefore $133,333….For Abecedarian [high-quality child care and pre-K birth to age 5 for disadvantaged families], the figure is $410,000….The amount spent in Project Star [the Tennessee experiment to reduce class size in grades K-3] to induce a single child into college is $400,000. If the program could be focused on students in the poorest third of schools…, then the cost would drop to $171,000 per student induced into college.

Upward Bound [which provided at-risk high school students with increased instruction tutoring, and counseling], if [it] could be targeted to students with low educational aspirations, [would have an] implied cost of inducing a single student into college [of] $93,667. .. The Social Security Student Benefit Program, which paid college scholarships to the dependents of deceased, disabled, and retired Social Security beneficiaries, [had a] cost per student induced into college [of] $21,000…The FASFA experiment, [which] randomly assigned families to a low-cost treatment that consisted of helping them to complete the FASFA, the lengthy and complicated form required to obtain financial aid for college, [had] an implied cost per student induced into college of $1,100. “

In a recent report on Michigan school finance by Kevin Hollenbeck and several Upjohn Institute colleagues, including me, we noted that a number of interventions from birth to age 18 have benefit-cost ratios of greater than one, if we focused on the increased present value of career earnings due to the intervention, compared to costs. (See previous blog post for a copy of the relevant table). Preschool and other early childhood interventions have high benefit-cost ratios; for example, high quality preschool has a ratio of increased future earnings to cost of 5.3, that is increased future earnings whose value is $5.30 for each dollar of investment. But later interventions also work.  High-quality summer school for children who are academically behind has an increased future earnings to cost ratio of 9.0. High school career academies, that provide a more career-oriented education for students who are so inclined, have an earnings benefits to cost ratio of 8.7. Finally, a program of math tutoring for disadvantaged 9th graders, combined with cognitive behavioral therapy, has an estimated earnings benefits to cost ratio of 10.8.

What is one to make of all this? It is true that human brains are more malleable at earlier ages. Therefore, if we invest earlier, we can make a larger percentage difference to later outcomes. We can also make a larger difference to a greater variety of people, as options will not have been foreclosed by early damage to human development.  As Heckman argued in his 2008 paper:

“Skills beget skills and capabilities foster future capabilities. All capabilities are built on a foundation of capacities that are developed earlier. Early mastery of a range of cognitive, social, and motional competencies makes learning at later ages more efficient and therefore easier and more likely to continue.”

But this also implies that these earlier investments have the highest returns when they are followed by later cost-effective investments. Heckman also argues in this 2008 paper that

“The advantages gained from effective early interventions are best sustained when they are followed by continued high quality learning experiences….Due to dynamic complementarity, or synergy, early investments must be followed by later investments if maximum value is to be realized.”

In other words, later investments also matter. These later investments may also have high rates of return.

What is true about later investments is that in order to be cost-effective, they have to be more targeted, in two senses. First, they have to be targeted at particular groups of people who at these latter ages are capable of greatly increasing their future prospects. Second, these later interventions have to be targeted at the particular barriers that are impeding progress, and that are not overcome by the normal operations of our society.

So, for example, the Kalamazoo Promise’s high estimate returns in our paper are due to our estimates that the program induces 12% of eligible students to get a post-secondary credential who otherwise would not do so. The other 88% of eligible students do not have their credential attainment affected in our estimates – either they would have received a post-secondary credential without the Promise, or the Promise was not enough to induce them to receive a credential, at least within the 6-year window post-high school considered in our paper.  These 12% of “induced” students are affected by the Promise’s combination of a generous scholarship plus whatever cultural shifts in expectations and support accompany that scholarship offer. For at least this group of students, more money and changed attitudes help overcome a barrier to success. Furthermore, the Promise’s high returns occur for an urban school district, with many disadvantaged students, although the benefits seem to occur for both advantaged and disadvantaged students.

Some type of targeting is also helpful for other later interventions. Class-size reduction works best for low-income schools, as noted by Dynarski et al. above. Upward Bound works best when targeted at students with lower educational aspirations. Social Security benefits for college tended to target low-income students with high financial barriers to college. Help in completing financial aid forms will only help the future economic prospects of high school seniors who have the skills needed to effectively use the increased financial aid that results.  Summer school works for students who are academically behind but motivated to improve. Career academies work for some students interested in career-oriented education, who have the capabilities in high school to respond to this alternative approach.  Math tutoring and cognitive behavioral therapy is targeted at disadvantaged ninth graders, and is specifically  oriented towards overcoming some of the hard skill and soft skill issues for this group.

What about earlier-age interventions? Although brains are more malleable at earlier ages, and therefore the potential benefits of earlier interventions may be greater in percentage terms, costs may sometimes be greater as well.  Particularly in infancy, many interventions have to be one-on-one or have very small class sizes in order to be high-quality, which raises the cost per child. This may be part of the reason for the finding, in both my work and the work of Arthur Reynolds and his colleagues, that high-quality preschool tends to have greater benefit-cost ratios than some earlier-age investments. As I said in my 2014 book, From Preschool to Prosperity,

“Pre-K services at ages three and four target an age range that is a “sweet spot”: the child’s brain is still malleable enough for modest interventions to have large long-run effects, but the child is old enough that the child is ready to learn in larger groups that are cost-effective to run.”  (p. 50)

In sum:

  • Investments at a wide variety of ages can have very high benefit-cost ratios.
  • Later investments may need to be more targeted in what groups are served, or what services are provided, in order to obtain high benefit-cost ratios.
  • Early investments may yield the largest percentage effects, but sometimes can be more costly, which can hold down their benefit-cost ratios.
  • The best strategy includes well-designed investments at a variety of ages, as such investments complement each other.
Posted in Distribution of benefits, Early childhood program design issues | Leave a comment

Kalamazoo Promise boosts college completion by one-third

In a paper released on June 25, 2015, the Kalamazoo Promise college scholarship program is estimated to increase college completion by one-third.  The college completion effects of the Promise would be expected to significantly increase future earnings. Based on predicted future earnings effects, the annual rate of return to the Promise’s tuition subsidies is over 11%.

This paper is the first to examine the effects of the Kalamazoo Promise on post-secondary outcomes. The paper was written by Brad Hershbein, Marta Lachowska, and me, all economists at the Upjohn Institute.

The Kalamazoo Promise is a simple, generous, and near-universal college scholarship program announced in November 2005. Under the program, graduates of Kalamazoo Public Schools are eligible for scholarships that pay up to 100% of college tuition and fees at any Michigan public university or community college.  (Starting with the graduating class of 2015, some Michigan private colleges are also included.)  The “Promise” has relatively few conditions: students must graduate from KPS, must have attended KPS since at least ninth grade, and must live in the district. There is no high school GPA requirement: students must simply graduate from high school and be admitted to a college or university. The program is a “first-dollar” scholarship program, and so is not reduced by other scholarships.  Scholarships are generous: scholarships are 65% of tuition and fees for students attending KPS since 9th grade, and then go up to 70% for students attending since 8th grade, and so on, with students attending since kindergarten eligible for a 100% scholarship.

The Kalamazoo Promise is funded by anonymous private donors. Its stated purpose is to promote Kalamazoo’s economic development, by attracting parents and businesses to the area in the short-run, and, in the long-run, by increasing the local supply of college-educated labor by increasing the educational attainment of KPS graduates, some of whom will stay in or return to the Kalamazoo area.

Although the Kalamazoo Promise has many unusual features, its effects are highly relevant to ongoing debates about how to increase educational attainment and promote greater economic opportunity and economic equity. The Kalamazoo Promise was the first of many “place-based” scholarship programs. Since 2005, over 30 communities around the U.S. have adopted similar programs, in some cases using public funding or imposing additional restrictions on scholarship eligibility. Can such programs work? The effects of the Kalamazoo Promise are obviously relevant to this growing place-based scholarship movement.

More broadly, there is the issue of how much college costs and college scholarship design matter to educational attainment. Can money make a big difference to college success, at least if money is handed out in a relatively simple and straightforward manner, with few requirements? The Promise’s effects are at least suggestive of whether reduced college costs and more college scholarships can make a difference, although obviously the devil may be in the details of any particular scholarship program.

How did we estimate the effects of the Kalamazoo Promise? Because the Kalamazoo Promise is a near-universal program, it was not possible to do any random assignment experiment. However, the details of the Promise’s design provided a good “natural experiment”. Students are eligible for the Kalamazoo Promise if they started in at least 9th grade, and ineligible if they started after 9th grade.  Changes over time for these two groups help reveal the effects of the Promise.  The paper compared the change in post-secondary success, before and after the Promise, of the “eligible student group” (including pre-Promise graduates who would have been eligible if the Promise existed) with the “ineligible student group”. What we found was an abrupt increase in post-secondary success for the eligible group that began in 2006, the first graduating class that could use the Promise, but no such increase for the ineligible group.  The most plausible explanation for the changing relative success for these two groups is that the Promise’s tuition subsidies helped increase post-secondary success.

Among the estimated effects of the Promise are the following:

  • The Promise is estimated to increase enrollment in a 4-year college by about one-third, from 40% for the comparable group in the pre-Promise period to 53% for the eligible group in the post-Promise period.
  • The Promise increase college credits attempted at 2 years, 3 years, or 4 years after high school graduation by close to 15%. As of 4 years or 8 semesters after high school graduation, these effects correspond to students attempting an additional 2 to 3 college classes.
  • As of 6 years after high school graduation, the Promise is estimated to increase the receipt of any post-secondary credential (certificate, associate degree, bachelor’s degree) from 36% of comparable KPS graduates in the pre-Promise period, to 48% for Promise-eligible KPS graduates in the post-Promise period, an increase of about one-third (12%/36%).
  • As of 6 years after high school graduation, the Promise is estimated to increase the percentage of KPS graduates getting a bachelor’s degree from 30 percent of comparable KPS graduates in the pre-Promise period, to almost 40% for Promise-eligible KPS graduates in the post-Promise period, an increase of about one-third in the number of BA/BS graduates.

The Promise effects are not restricted to more advantaged groups. Although estimates for different sub-groups of KPS graduates are more imprecise, Promise effects appear to be similar for KPS graduates from low-income families compared to KPS graduates from middle-income families. Because baseline college success for students from low-income families on average is smaller, the relative effects of the Promise on college success are higher for low-income students. For example, the Promise is estimated to boost the number of low-income students attending a 4-year college by over 50%, over twice the percentage effect observed for middle-income students.  The Promise had at least as great and sometimes greater effects for non-white students compared to white students. Across gender, the point estimates found statistically significant effects on bachelor’s degree attainment for female students, but bachelor’s attainment effects for males were not statistically significant.

More coverage of this study’s findings can be found at the Upjohn Institute website, by Julie Mack in the Kalamazoo Gazette, and by Ron French in Bridge Magazine.

Posted in Distribution of benefits, Economic development | 1 Comment

Increasing educational performance and reducing educational disparities is more feasible if pursued through high-productivity interventions, including but not limited to early childhood education

On May 27, 2015, the Upjohn Institute released a report on Michigan’s school finance system and how to reform it to improve student performance in Michigan, and lessen disparities among children in various income groups.  The lead author of the report is my colleague Kevin Hollenbeck, and I am one of the co-authors, along with Randy Eberts, Brad Hershbein, and Michelle Miller-Adams.

Although the report is obviously focused on Michigan, the report’s lessons are more generally applicable to school finance reform and school reform throughout the U.S. Our report argues that money does matter to educational performance. More resources will improve student achievement, and can reduce test score gaps between disadvantaged students and more advantaged students. But solving educational problems through increasing spending across the board is expensive enough that in many cases it may not be politically feasible. A more feasible route to educational improvement is to increase resources in a targeted manner: increase spending on educational interventions known to have a high benefit-cost ratio. Such interventions include early childhood educational interventions. For example, high-quality child care and high-quality pre-K programs have been shown by rigorous evaluation to have a large effect on improving future prospects per dollar spent. But other educational interventions also have been shown to increase student achievement by a large amount per dollar spent. These include reduced class size in early elementary school, high-quality summer school for elementary students who are behind, longer school year for high-poverty schools, small group tutoring for high school students who are behind, and high school career academies for students interested in a more career-focused high school education.

It is sometimes argued that money doesn’t matter to educational performance. Michigan’s experience shows that this argument is wrong. Michigan dramatically changed its school finance system with passage of Proposal A in 1994. Proposal A essentially shifted most schools from locally-controlled spending per pupil to a more uniform system of state-controlled funding per pupil. The result was a natural experiment: some low-spending districts, mainly in rural areas, ended up with higher funding per student, whereas other school districts did not. As shown in studies by Leslie Papke and Joydeep Roy, and as confirmed in our report, the districts that experienced higher funding per student showed  improvements in student achievement relative to school districts that did not (Figure 2-1 on page 8 of our report).

However, the influence of money on student achievement is of a magnitude such that reaching educational goals through across-the-board spending increases will often be politically difficult. For example, these estimates of money’s influence suggest that for Michigan to match a leading state such as Massachusetts in student achievement would take an extra $10,000 in annual spending per pupil. To eliminate the achievement gap between low-income students and other students would require spending an extra $19,000 per pupil on low-income students. (Similar results would occur for achievement gaps across income groups in other states.) While one could make a case that such spending boosts would have benefits greater than costs, obviously the costs are large enough that the political feasibility of such funding boosts is doubtful.

However, there are educational policies that improve student achievement and adult outcomes by far larger amounts per dollar spent than across-the-board spending increases. Early-age interventions frequently have higher benefit cost ratios, presumably because younger brains are more malleable, and early learning tends to build on itself and create future learning. Such early interventions include high-quality child care and pre-K, and lower class size in early elementary school. Later-age interventions can also be cost-effective if they are highly targeted on the particular learning needs of students. For example, targeted tutoring or extended school years can work for students who are academically behind, and career-oriented education can help students attracted by that approach to education.

If one asks what boost to the present value of future earnings is brought about by an educational policy, more cost-effective educational policies frequently have benefit/cost ratios in the range of 2 to 1 up to 13 to 1. Such interventions can be over ten times as effective in improving student outcomes as is true for across-the-board spending increases.

The below table is taken from our Michigan report (see Table 5-1, page 44; the version below is slightly rearranged). It shows the increase in present value of future earnings per child for a given educational intervention, compared with the cost of that educational intervention per child. Our report provides references to the research behind these numbers.

Relative Costs and Future Economic Benefits of Various Educational Policies

Policy Effects on Present value of Future Earnings per Child Program costs per child Economic benefit to cost ratio Target group
General school funding effects  $        7,000  $     11,000 0.6 All students
Reduced class-size K-3  $      22,000  $     11,000 2.0 All students
Full-time full-year child care from birth to age 5 (Educare) for disadvantaged families  $     134,000  $     87,000 1.5 Disadvantaged students
Universal full-day pre-K  $      53,000  $     10,000 5.3 All students
Mandatory elementary summer school for one year for children who are behind  $      18,000  $      2,000 9.0 Students who are behind
High school career academies  $      26,000  $      3,000 8.7 Students interested in CTE
5 best school practices (longer school year & school day yielding at least 25% more time, small group tutoring, frequent feedback to teachers, more use of testing to guide instruction, high expectations) $26,000 per year $2,000 per year 13.0 High-poverty schools
1 hour per day math tutoring plus cognitive behavioral therapy for disadvantaged 9th graders  $      54,000  $      5,000 10.8 Students who are behind

As shown in the table, simply increasing educational spending per pupil by $11,000 would be estimated to increase the present value of future earnings per student by a little more than half that spending increase. Education of course has other benefits than increased earnings, such as lower crime, higher civic involvement, etc. With these other benefits, it is quite plausible that dramatic across-the-board increases in educational spending could be justified as having benefits exceeding costs.  Still, across-the-board increases in spending are not the most cost-effective educational policy for boosting future prospects, as shown in the table. Other policies have benefit cost ratios of up to 13 to 1.

In this blog and my recent writings, I have extensively argued for early childhood programs as a way to promote the economic development of the U.S. economy, or for particular states to promote their own economic development. This is not because early childhood programs have the absolutely highest benefit-cost ratios, but rather because these policies combine very high benefit-cost ratios with other advantages.

The policies with the highest benefit-cost ratios are actually later policies that are targeted in some way, either at students who are academically behind, at high-poverty schools, or at students with particular career interests. What is particularly attractive about universal pre-K as an educational policy is that it is perhaps the most cost-effective intervention that is relevant for all students, not just targeted groups of students. This means that universal pre-K can be more easily scaled up to have truly large effects on the overall quality of the labor force of a state or of the nation, and thereby to have truly large effects on overall economic development. (Lower class size in early elementary school also helps improve achievement for all students, but is not nearly as cost-effective as universal pre-K.)

High-quality child care from birth to age 5, similar to the Educare program, is attractive because among all these interventions, it has perhaps the highest estimated gross benefits per individual student. If we want to dramatically affect the life prospects of a student from a high-poverty background, this intervention does the most. Its benefit-cost ratio is lower because its costs per child are so high. However, educational policy is not solely concerned with benefit-cost ratios, but with also achieving large effects per child.

Based on these findings, our report ends up recommending that Michigan increase school funding, but do so in a way that encourages more resources to be devoted to more cost-effective educational policies.  Given the difficulty of increasing taxes to finance public spending increases, a similar policy course may also make sense for other states.

Posted in Distribution of benefits, Early childhood program design issues, Economic development

The adequacy of state pre-K funding for quality and access

The latest edition of “The State of Preschool-2014” was released on May 10, 2015 by the National Institute for Early Education Research. This annual release of data and analysis has become essential to understanding what is going on with publicly-funded pre-K in the United States.

Using this latest edition, in this post I analyze funding for age-4 pre-K, and compare current funding to what would be needed to ensure adequate quality for pre-K for all age-4 participants, and what would be needed to ensure both adequate quality and adequate access to pre-K for all age-4 children in the United States.

Based on the NIEER data, I estimate that in aggregate, total funding for state-sponsored pre-K for 4-year-olds was about $5.2 billion in the 2013-14 school year.  (All dollar figures in this blog post are stated in year 2013 dollars.) This total includes not only state funding for state-sponsored pre-K programs, but also what is known about local and federal funding for such pre-K programs. (An additional note for data wonks: To get this estimate, I used NIEER estimates of total pre-K funding per child by state, and multiplied it by the number of 4-year-olds in pre-K in each state, and then summed this product over all states. Because the NIEER estimates of per-child funding included both 3-year-olds and 4-year-olds, the implicit assumption is that this funding per child did not differ much between 3-year-olds and 4-year-olds, or that such differences canceled out when summed over states. )

NIEER also estimates what it would take for the state to have adequate resources per child to consistently deliver a high-quality program. The estimates of the per-child funding are derived from estimates from the Institute for Women’s Policy Research, in their 2008 report, “Meaningful Investments in Pre-K: Estimating the Per-Child Costs of Quality Programs”. For each state, these estimates of per-child funding reflect the mix of half-day vs. full-day and other length of program design features of the state’s pre-K programs – that is, the required per-child funding for quality would be much greater if a state had all full-day slots vs. all half-day slots. The estimates are then adjusted to 2013 national dollars, and also adjusted for cost differences across states, as described in NIEER (2015, p. 18), using estimates from Lori Taylor.

If we add how much it would take to consistently deliver quality programs for the CURRENT number of 4-year-old children served in each state, we come up with a figure of $7.4 billion. Therefore, current funding of $5.2 billion would be needed to be increased by $2.2 billion, or over 40%, simply to consistently deliver a quality program to all 4-year-olds being served by state-sponsored pre-K, in the mix of half-day versus full-day programs that is currently being delivered.

(Another data note: This calculation simply increases all states that are below the needed quality spending level per child to the needed spending level per child. States for which funding is currently above the estimated required funding level are not reduced. If we simply adjusted all states up or down to the estimated required spending level for quality, total required funding would be somewhat lower, at $7.1 billion.)

What would it take to offer universal access to quality pre-K to all 4-year olds in the United States? This depends upon what one assumes about how many families with 4-year-olds would participate in a program that offers universal access, but obviously does not require participation, as pre-K is a voluntary program. It also depends upon what one assumes about the length of the program in terms of full-day versus half-day.

Suppose we assume that the program is a full-day (six hour) academic year program, with a class-size ratio of 15 students to 2 teachers, and with the lead teacher paid wages and benefits similar to what is paid to public school kindergarten teachers.  Suppose further that we assume that in a voluntary program that offers universal access, the percentage of 4-year olds in state-sponsored pre-K would be similar to what it is today in Oklahoma, at 76.4% of all 4-year-olds. The remainder of families would have their 4-year-olds in Head Start, in private preschool or child care, or at home. As was done with NIEER, and as I have done in previous work, I rely on the IWPR cost figures.

The resulting estimate is that full-day universal-access high-quality pre-K would cost $31.1 billion per year. This is an increase of $25.9 billion over current state-sponsored pre-K funding. Although $26 billion is a great deal of money, this amounts to less than 2% of overall state and local government tax revenue. (See p. 70 of my book, From Preschool to Prosperity.)

However, the cost of universal access to pre-K could be considerably less if we assumed the program would not be full-day for all. For example, suppose we assume that the program would have the mix of full-day versus half-day that is true of current state-sponsored pre-K programs. Then the estimated total cost of universal access to pre-K would be $19.8 billion per year. This amount would be sufficient to enroll 76.4% of all 4-year-olds in a program whose funding allowed high-quality, but with the program mix skewed mostly to programs offering half-day services.

In the real world, I think that to achieve universal access, and elicit adequate participation from all income groups, it might be necessary to expand on current full-day options. Therefore, the best estimate is that a universal program offering some mix of half-day versus full-day services would probably cost between $20 billion and $30 billion per year.

How do these figures compare with where states were before the aftermath of the Great Recession, and the expiration of the federal stimulus, led to state budget cutbacks? In NIEER’s data, it appears that state pre-K funding was significantly cut after the 2009-10 school year. So, let’s look at funding in the 2009-10 school year, using estimates from the 2010 NIEER Yearbook.

In 2013 dollars, total funding in 2009-10 for state pre-K programs for 4-year olds was $5.6 billion. So funding in real terms in 2013-14 had dropped by $0.4 billion, from $5.6 billion to $5.2 billion.

In 2009-10, the total funding estimated that was needed to consistently deliver quality services was $6.1 billion. This is only $0.5 billion above the actual funding. In contrast, the “quality funding” gap for 2013-14 was $2.2 billion ($7.4 billion versus $5.2 billion). The increased quality funding gap, going from 2009-10 to 2013-14, is largely because the mix of pre-K services has shifted to either more hours or days of service, which increases the needed funding level per child.  A more minor factor is the number of 4-year-olds served in state pre-K funding increased slightly, from 1.11 million in 2009-10 to 1.16 million in 2013-14.   States have not increased real funding to match the expanded hours of service per child and the increased number of children served.

What would have been needed in 2009-10 to finance universal pre-K? Using assumptions similar to what was used for 2013-14, the estimated cost, in 2013 dollars, of universal full-day pre-K would have been $32.5 billion. This is slightly higher than the estimated cost of $31.1 billion in 2013-14, largely because the number of 4-year-olds declined slightly from 2009-10 to 2013-14.  However, the costs of universal high-quality pre-K, if we assumed the mix of part-day versus full-day services actually provided in 2009-10, would have been $18.1 billion, again in 2013 dollars. Overall, it seems likely that a real world high-quality universal pre-K program in 2009-10 would have cost somewhere in the range from $20 billion to $30 billion, similar to the estimated cost in 2013-14.

What do these figures indicate? First, we are still well-short of the resources needed to have full access in all states to high-quality universal pre-K. We would need somewhere between $15 billion and $25 billion in additional resources to achieve universal access to high-quality programs.

Second, there is a growing gap between current funding levels and what is needed to fund the current numbers of children and the current mix of services in a high-quality manner. The main concern is that under-funding of pre-K will reduce the effectiveness of pre-K programs in changing a child’s future life course. A secondary concern is that under-funding of pre-K may undermine public support for expanded pre-K.

Both access and quality matter greatly to the aggregate economic impact of pre-K. As we continue to recover from the recession, all levels of government need to devote additional resources to address both access and quality issues in current pre-K funding.

Posted in Early childhood program design issues

Forum discusses universal versus targeted preschool, age 4 versus earlier age programs, and public school versus private provision of preschool

On May 5, I spoke at in a forum in Minnesota on issues in designing preschool programs. The forum had six speakers with diverse perspectives on how to design early childhood programs.

The context for the discussion was a debate about the future course of Minnesota preschool policy. The policy choices facing Minnesota are similar to the choices facing other states, and therefore the Minnesota debate is relevant to all state and local areas considering expansions of their early childhood education efforts.

Minnesota currently has quite modest funding for preschool. The state’s current approach to preschool is somewhat unusual. The state funds scholarships for low-income children to go to both private and public preschools at ages 3 and 4.  At current funding levels, these scholarships only go to perhaps 10% of the eligible low-income children, and have a maximum size of $5,000, which limits what proportion of costs are covered. The state also has a “Parent Aware” rating system to help inform parents of the importance of preschool quality, and the quality features of different preschools.  The current Minnesota model is to create an informed market for quality preschool among low-income parents, with some funding to help low-income parents to pay for at least part of the costs of quality preschool.

Minnesota Governor Mark Dayton has proposed a significant expansion of Minnesota’s early childhood education efforts. The main increase in funding would go to help support universal pre-K for 4-year-olds. This funding would go directly to local public schools, although the public schools in some cases might choose to contract out to private preschools to provide preschool services.  The state would also double funding for early learning scholarships for low-income children, but extend the age eligibility for such scholarships to birth to age 5.

The coalition that has backed Minnesota’s early learning scholarships would prefer different uses of any expanded early childhood funding.  These groups, which include some business interests, some private charities, and some private preschools, is advocating for much larger increases in funding for the existing system of early learning scholarships for low-income students. In addition, this coalition would want larger funding increases for earlier age programs for low-income students, such as home visiting programs to improve parenting.

At the forum, some of the perspectives of the “Minnesota model” coalition were expressed by members of the speakers’ panel, including Art Rolnick, former director of research for the Minneapolis Federal Reserve Bank, and currently co-director of the Human Capital Research Collaborative at the Humphrey School of Public Affairs of the University of Minnesota. Some perspectives of the coalition were also expressed in audience members’ questions and comments.

Some of the concerns expressed at the forum about Governor Dayton’s program included the following:

  • There are concerns about prioritizing universal access to preschool over more income-targeted approaches. The argument made by Art Rolnick at the forum is that the highest return to early childhood investments comes from investments in low-income children, not middle-class children. The focus should be on closing the achievement gap between low-income children and other children.
  • There are concerns that putting most of the funding into age 4 preschool doesn’t put a sufficient priority on earlier-age (birth to age 3) early childhood programs. Art Rolnick argued that age 4 preschool was too late for many low-income children, and that based on the brain science, the child’s brain is more malleable at earlier ages, and therefore we would get the greatest effects on closing the achievement gap by investing earlier in low-income children.
  • There are concerns that utilizing public schools to deliver preschool will leave out private preschools. Although public schools can contract out, they may choose not to do so. Leaving out private preschools was argued by Rolnick to decrease parent choice. Some private preschools at the forum were concerned that greater reliance on public schools for age 4 preschool would cause financial problems for private preschools, which were argued to use age 4 preschool revenue to cross-subsidize more expensive care at earlier ages.

I was asked by the forum organizers to focus on why I favor universal over income-targeted preschool.  A previous blog post presented my detailed prepared remarks on this topic.  The essence of my argument was two points:

  • The research evidence suggests that preschool for middle-class kids has similar benefits to preschool for low-income children. Preschool for both income groups has very high benefit-cost ratios. Therefore, any state wanting to maximize its future should invest more in preschool for both low-income and middle-class children.
  • Universal access to preschool has some practical advantages. Most notably, universal pre-K is likely to have sustained broad political support for maintaining an adequate number of preschool slots and adequate preschool quality.

Art Rolnick brought up a counter-argument at the forum. He argued that the empirical evidence I presented only showed that the direct earnings benefits of preschool for middle class children and low-income children was similar. But he argued that it was reasonable to believe that some social benefits, such as lower crime and reduced special education costs, might be greater for preschool for low-income children.

My response to this counter-argument is as follows. There is actually no empirical evidence on whether preschool’s benefits in reducing crime and special education costs are greater for low-income children. It is a reasonably plausible assumption, but it has not been shown to be the case in empirical studies.

In addition, there are other possible social benefits of preschool that might imply greater social benefits from preschool for middle-income children. For example, there is empirical evidence that there are substantial economic spillover benefits for the entire economy from increasing the percentage of workers with college degrees, benefiting all workers, not just those who get a college degree. But the empirical evidence does not support such economic spillovers from more high school degrees. If preschool for the middle class has more of its earnings effects through helping children who would have been high school grads to become college grads, while preschool for low-income children has more of its earnings effects through helping children who otherwise would become high school dropouts to become high school graduates, but not necessarily college graduates, then preschool for the middle-class might have greater “skill spillover” benefits for the entire economy. There is no direct empirical evidence for these greater skill spillover benefits for middle-class preschool, but it seems as reasonable an assumption as the assumption that the anti-crime benefits of preschool are greater for low-income preschool.

On the issue of age 4 preschool vs. programs at earlier ages, I argued in discussion at the forum that although earlier-age programs can achieve higher effects in increasing a child’s future earnings than age 4 preschool, possibly because the child’s brain is more malleable at earlier ages, the benefit-cost ratio is higher for age 4 preschool than for earlier-age programs. I have argued this position before, and similar findings have been reported by Reynolds et al.  Why is this the case? Moving to earlier-age programs, we do get higher effects on earnings, but costs per child go up even faster. Child care programs at ages 1 and 2 must have lower child to staff ratios to be high-quality than is true of age 4 or age 3 preschool. Parenting programs such as the Nurse Family Partnership must use a lot of 1 on 1 staff time with parents to be effective. Preschool at age 4 or age 3 may be a “sweet spot”, where the child’s brain is still malleable enough to have significant effects, while children are old enough to have larger class sizes, which lowers costs per child.

On the private versus public preschool issue, I argued in discussion at the forum that either a public school approach or a private preschool approach can work. For example, both Oklahoma and Georgia run preschool programs with some evidence of success, yet Oklahoma runs preschool primarily through the public school system, with some contracting out, while Georgia uses a charter school approach. Furthermore, in the broader K-12 area, the evidence suggests that market competition via charter schools has not resulted in clear dramatic improvement in average school quality. Some charter schools are better than the average public school, and some are worse.

Therefore, Minnesota (or other states) can use any mix of public versus private approaches to delivering preschool, and still get results. Public preschool might have some advantages in tying preschools more directly to the K-12 system, which will make transitions to K-12 somewhat easier. Preschool in the public schools might also strengthen the public schools as an institution. Public schools might tend to be more universal and less selective, while some private preschools might be more selective in which students they try to serve, although this selection can of course be regulated. On the other hand, private preschools might tend to be more innovative, and might tend to offer more diversity of curriculum.  But public preschools can be encouraged to be more innovative and diverse in curriculum choices with appropriate state policies.

One point I made at the forum, and more extensively in private conversations in Minnesota, is that the choices facing Minnesota are broader than the contrast between Governor Dayton’s original proposal, and the expansions favored by advocates for the scholarship model. For example, there are preschool expansions that would be more “universal” than Minnesota’s current scholarship system, yet not fully universal. We could imagine funding preschool fully for students up to 300% or 400% of the poverty line, and charging fees above that family income level. Or we could imagine funding universal preschool for school districts exceeding some minimum percent of students eligible for free or reduced price lunch, but providing more limited funding for school districts with more affluent student bodies, and allowing those more affluent school districts to decide whether they wanted to provide the local funding needed to make preschool universal, or instead wanted to charge fees.

As another example, we could imagine the state providing some funding for public school provision of preschool, but also significantly expanding funding for private preschools at age 4. The state could directly allocate some share of the funds to private preschools, or require local school districts to allocate some share of funding to private preschools.

A final comment is on Art Rolnick’s argument that the main focus should be on closing the achievement gap between disadvantaged students and other students. In general, I think it is more appropriate for public policy to be based on multiple goals rather than exclusive focus on one goal. Lowering the achievement gap to help the poor is one worthy goal. Increasing the skills and earnings of all workers, including the middle-class, as part of an effort to improve the overall economy, is another worthy goal. Public policy towards early childhood programs, and other public services, is more likely to be effective and gain greater sustained political support if it tries to balance a variety of worthy goals rather than allowing one worthy goal to trump all other worthy goals.

It is tempting for advocates for the poor to adhere to the doctrine that all that matters for public policy is what public policy does for the poor. But a democratic government that seeks to raise taxes from a broad range of people , and promises to provide useful public services to a broad range of people, must have a broader vision. The interests of the poor should perhaps receive a greater weight in public policy, but that weight should not be 100%. Furthermore, if the interests of the poor are prioritized within a universal system that provides some benefits to all, public support will be greater. I suspect that Social Security, with its broad public support due to its near-universal benefits, but with a formula that favors low-wage workers, has done more to reduce poverty in the U.S. than income-targeted welfare programs, which are highly unpopular.

Putting all the weight on the poor might appeal to some conservatives and some business leaders as a way to hold down taxes, because enhanced spending is only required for a relatively small percentage of the population. But if broad public services such as universal preschool provide economic benefits greater than costs, then this is too narrow a view. Universal preschool that includes the middle-class will expand the U.S. economy more than income-targeted preschool. The initial costs of universal preschool will be higher, but in the long-run, the economic pie will be larger, which will increase business sales and profits, and provide a broader tax base that will allow improved public services combined with lower tax rates. The long-run interests of the business community, and of those favoring low taxes, are better served by the universal preschool approach, because it has larger effects on long-run economic growth.

States face a wide variety of choices in designing their early childhood education efforts. A pragmatic political strategy will forge appropriate compromises in these design choices between different policy goals. There is not one “ideal system” for early childhood policy.

Posted in Distribution of benefits, Early childhood program design issues

Why preschool should be universal

On Tuesday, May 5, I was asked to speak at a forum in Minnesota, on why I think that preschool should be universal rather than income-targeted. Below are my prepared remarks:

My main reason for arguing that publicly supported preschool should be universal is that the research evidence suggests that preschool’s benefits are universal. Preschool benefits for middle-class children are almost as large as benefits for lower-income children.

For example, the research evidence in a study I co-authored of Tulsa’s universal preschool program suggests that test score effects of pre-K for middle-income children are 90% as large as those for low-income children. Based on available studies of the relationship between early test scores and later earnings, we would also expect the dollar effects of Tulsa pre-K on future earnings for middle-class kids to be 90% as great.

These estimated earnings effects, for both middle class and low-income kids, are that for each dollar invested in pre-K, the present value of future earnings for those kids increases by about $5. This is a very favorable benefit-cost ratio for investing in both low-income and middle-class kids.

Even though the benefit-cost ratio is about the same for preschool for both middle-class and low-income kids, a universal preschool program would still significantly redistribute income. Because low-income children on average have lower baseline levels of future career earnings, the same dollar boost to earnings will cause a larger percentage boost to future earnings.

For example, in my work for Tulsa, I’ve calculated that a full-day preschool program costing $10,000 per year would boost the present value of future career earnings by about the same $50,000 for both low-income and middle-income children. But because expected future income is lower for low-income kids, this same $50,000 boost is about 10% of baseline career earnings for low=income kids, twice the 5% effect on career earnings for middle-class kids.

This evidence from Tulsa is backed up by similar evidence from Boston’s universal pre-K programs, and from random assignment experiments in both Utah and Rhode Island. The Boston and Rhode Island results for universal preschool programs suggest that preschool’s benefits for middle class children are 70% as great as for low-income children.

In contrast, the available evidence for earlier age interventions, such as parenting programs and child care programs for ages 0 to 2, suggests that they only pay off for low-income groups. For example, a study of one early child care program for ages 0-2, co-authored by Aaron Sojourner at the business school here, suggests that this program only helps children from families below 180% of the poverty line.

Why is there this pattern of benefits, in which preschool benefits all income classes, but earlier interventions in child care and parenting only benefit the poor? I think the most plausible explanation is the middle-class families are generally able on their own to provide adequate parenting or child care services. But high-quality preschool is hard for both middle-class parents and low-income parents to provide on their own. Preschool provides various skills, in particular social skills, that are best provided in a group setting. And quality preschool costs a lot: one school year of full-day high-quality preschool costs around $10,000, which is hard for many middle class families to afford on their own.

This interpretation of the research evidence is not my own unique view. A November 2014 “Consensus Letter” signed by over 500 early childhood researchers attempted to define the research consensus on early childhood education. Among other things, the letter stated that for programs such as preschool, and I quote, “benefits outweigh costs for children from middle-income… families.”

The economic research argument for universal preschool is backed up by some practical considerations. Universal programs are easier and cheaper to administer, without dealing with the complexities of income determination and recertification. Universal programs are less stigmatizing to the poor. Universal programs find it easier to encourage income mixing of children in preschool, which provides positive peer effects for low-income children. As argued in a recent report for The Century Foundation, economic diversity in a preschool program should be considered an important element of preschool quality – no-one thinks that running preschool in an income-segregated manner, as we do with Head Start, is ideal.  Finally, universal programs tend to have more persistent strong political support for maintaining access and program quality over time.

So in sum, I think that the research supports providing universal preschool at age 4, and income-targeted parenting and child care services for low-income children from the prenatal period to age 2. This combination of universal preschool services and targeted early age services will provide valuable benefits to all income classes, and boost the entire economy, while also providing an extra percentage boost to the poor and reducing income inequality.

Posted in Distribution of benefits, Early childhood program design issues | 1 Comment

The importance of neighborhoods for child development

On Monday, May 4, the New York Times gave prominent coverage to two recent papers that provide strong evidence that better neighborhoods or local areas for young children make a large difference in increasing future adult earnings and income for these children. This research was conducted by Harvard economists Raj Chetty, Nathaniel Hendren, and (on 1 of the 2 papers) Larry Katz.

One paper (by Chetty, Hendren, and Katz) concludes that providing housing assistance to low-income families, combined with a requirement that the family move to a low-poverty neighborhood, increases future earnings of a young child in these families by around 30%. This result only holds for children who are between the ages of 4 and 12, with an average age of 8.  In contrast, for children 13 and over, forcing a move to a better neighborhood appears to be counter-productive, possibly because the disruptive effects of the move outweigh any benefits of the better neighborhood.  For children under the age of 4 at the time of the move to a better neighborhood, the authors do not have data.

The other paper (by Chetty and Hendren) suggests that children from low-income families do much better in adult income and earnings due to growing up in some local areas (counties or local labor markets) compared to other local areas, and that this is a true causal effect of the local area. Because of data limitations, the study can only look at where the child grew up back to age 9. But the study finds that the more years during childhood the child spends in a “good” local area, compared to a “bad” local area, the higher are the child’s future adult earnings. At least from age 9 on, each extra year in a “good” local area yields the same predicted improvement in the child’s future percentile rank in the adult income distribution.

My initial reactions to these two papers include several points:

First, these estimated effects of either the child’s “neighborhood” or “local area” are quite large. For example, the 30% boost to the child’s future earnings from housing subsidies tied to neighborhood quality exceeds the 26% boost found in studies of the effects of the Abecedarian program, which provided full-time child care and preschool to low-income children from birth to age 5. As another example, a quality full-day preschool program for one-year might reasonably be predicted, based on current research, to increase future earnings for children from low-income families by 10%.

Second, the gross fiscal costs of realizing these gains from better neighborhoods might be quite high for some families. As Chetty, Hendren, and Katz point out, for families who would have received low-income housing subsidies anyway, the fiscal costs of tying housing subsidies to moving to better neighborhoods are not great, perhaps about a one-time cost of $4,000 in extra counseling costs of facilitating the move to a better neighborhood.  Therefore, there is a very strong argument for trying to use existing housing subsidies for low-income families to facilitate family moves to better neighborhoods.

But most low-income households do not currently receive large housing subsidies.  If we wanted to help all low-income families move to better neighborhoods, this would be quite expensive. Typical low-income housing vouchers in the U.S. cost around $8,000 per year per recipient family. The exact cost per child depends on how many children are in each family, and the relative ages of the children. But paying for better neighborhoods for all low-income children from birth to age 18 could be quite expensive. For a one-child family, this would be a per child cost of over $140,000 over these 18 years, although the average cost per low-income child would be lower, sometimes considerably lower, in families with more children.

These large costs might well be worth it, solely in terms of benefits for children. For example, Chetty, Hendren, and Katz estimate that the 30% increase in future earnings would amount to an increase in the present value of earnings for the typical child in their sample of $99,000, discounted back to age 8. As the typical child in their study moves to a better neighborhood at age 8, maintaining that better neighborhood would require spending $8,000 per year for 10 years, from age 8 to 17. The present value of this would be around $70,000. Even if the family only has one child, the earnings benefits exceed the costs.

However, the point is that facilitating better neighborhoods for the poor through housing subsidies for better neighborhoods would not be a cheap program if pursued as a comprehensive program.  For example, if we provided $8,000 per year in housing subsidies to all families below twice the poverty line, which is around 23 million families, the annual costs would be over $160 billion. In contrast, interventions such as universal pre-K education for all 4-year olds would cost around $30 billion annually. Benefits for low-income children from a comprehensive housing subsidy program might well exceed costs, but it is unclear whether the benefit-cost ratio for such a comprehensive program would be greater than the 5 to 1 or greater ratio that is sometimes achieved for preschool programs.

(Note to policy wonks: this ignores entirely the value to parents of the housing subsidies, which could be considered an offset to the costs of housing subsidies.  The value to parents of early childhood programs is also sometimes ignored in benefit-cost analyses. Furthermore, determining this value is not straightforward. Finally, it often seems that the political system in the U.S. places little value on what income transfers can do for low-income adults, but does place some value on what income transfer programs can do for low-income children.)

Third, if better neighborhood quality is pursued solely through moving low-income households to more income-integrated neighborhoods, this requires an enormous social rearrangement. This necessarily implies as well that middle-income and upper-income households would usually be living in neighborhoods with more lower-income households. These middle-income and upper-income households would likely fear that this would lead to a loss of neighborhood quality that might damage their prospects and their children’s prospects.  In the paper by Chetty and Hendren, they argue that current evidence suggests that upper-income households are less sensitive than low-income households to features of the local area that are related to the income segregation  of the area. The argument is then that the children of the poor benefit from greater neighborhood income integration, but the children of the middle-class and the upper-class do not suffer from greater neighborhood income integration. But whether this finding is robust to more detailed analysis of the effects of specific neighborhoods, rather than overall area traits, is unclear.

Fourth, the limitations of a housing subsidy approach to improving neighborhood quality increases the importance of improving neighborhood quality through a variety of mechanisms, and not limiting the neighborhood improvement mechanisms to housing subsidies. In both these papers, what exactly constitutes the “quality” of a neighborhood or larger “local area”, from the perspective of improving a child’s future prospects, is still to a large extent a black box. Some correlations are presented that relate different local area characteristics to the child development effects of the local area, but the relative importance of these different neighborhood or local area characteristics is unclear. We are a long way from defining with any great assurance what aspects of the neighborhood and area child environment are most crucial in shaping child development and the child’s adult outcomes.

Fifth, it is reasonable to infer from these results that some early childhood interventions will help directly improve local neighborhood quality. Income transfers to families with very young children will raise average income in low-income neighborhoods, which we would predict would lead to higher overall neighborhood quality. Preschool and other early childhood education programs will raise school test scores, which is suggested in Chetty and Hendren’s paper to be one of the area characteristics associated with better adult prospects.

These spillover benefits from better neighborhood and area quality due to early childhood interventions may add to the overall benefits of early childhood interventions. One implication of this work by Chetty and Hendren is that many social interventions may have important neighborhood spillover benefits. If educational and social interventions are pursued on a large scale, they may improve neighborhood quality enough to add considerable economic and social benefits.

Posted in Distribution of benefits, Early childhood program design issues, Economic development