Cost of Preschool in NJ 2007

8/9/2019 Cost of Preschool in NJ 2007

1/54

Clive BeleldQueens College, CUNYHeather SchwartzTeachers Columbia, Columbia University

THE COST OF HIGH-QUALITY PRE-SCHOOLEDUCATION IN NEW JERSEY

December 2007


2/54


3/54

CONTENTS

Summary 5

1. Introduction 11

2. Preschool Programs in New Jersey 13 2.1 PreschoolintheAbbottdistricts 13 2.2 AbbottpreschoolexpendituresinNewJersey 15 2.3 PreschooloutsidetheAbbottdistricts 18 2.4 EvaluatingtheAbbottprogramrequirements 19

3. Costing Out Preschool: Methods and Evidence 22

3.1 Methodology 22 3.2 Priorevidence 25

4. Estimating the Costs of Pre-School 27 4.1 Costfunctionmethod 27 4.2 Pre-schoolprogrammodel 33

5. Total Cost Requirements in New Jersey 36 5.1 PreschoolcostsforAbbottdistricts 36 5.2 PreschoolcostsacrossNewJersey 38 5.3 Totalprogramcosts 39

6. Conclusions 40

References 41

Appendix Tables 47


4/54


5/54

THE COST OF HIGH-QUALITY PRE-SCHOOL EDUCATION IN NEW JERSEY | 5

SUMMARY

Tis report calculates the ull cost o providing well-planned, high quality pre-schoolor children in New Jersey, as required underAbbott vs. Burke(153 NJ 480 1998).

The benets of pre-school

Te case or investments in pre-school is strong. High-quality pre-school helps chil-drens cognitive growth and the benets are stronger or children rom less advantagedbackgrounds. New evidence on the benets or children rom participating in theAbbott district pre-school programs shows that children who attend these programsobtain higher scores in language, literacy, and mathematics. Tese gains persisted intokindergarten such that children who attended pre-school or two years had closed overhal o the gap against national averages or vocabulary. Importantly, investments inpre-school and investments in high quality K12 schooling are complementary. Onedoes not substitute or the other and the gains rom pre-school need to be reinorced

through eective elementary and secondary school education.

State investments should reect this consensus, such that well-planned and high-quality programs are made available, particularly or disadvantaged children or thosein low-income amilies. However, current state expenditures may not ully equate tothe amount needed or high quality pre-school.

Pre-school in the Abbott districts

Across the New Jersey Abbott districts, the main component o the pre-school programis six hours o provision or 180 days; the program requirements are high, including

small classes and qualied teachers. In addition, the program includes extended daychild care and summer provision or 10 hours per day.

Overall, the Abbott district pre-school requirements reect the research consensus onthe components o high-quality pre-school. From a resource perspective, the mostimportant regulations pertain to stang and class size: certied teachers and masterteachers (who mentor and train classroom teachers) are expensive, as is the need orclass sizes o 15. Tere are also signicant supplemental services, as well as managerialand accountability resource requirements. Te last item is important or ensuring thatprograms are implemented aithully. Te program is delivered through the publiceducation system. It is a public program, unded and regulated by the Departmento Education. Abbott pre-school programs may be oered by private providers, who

contract with the school district, as well as through the public school system. In total,there are approximately 600 separate centers or providers. Head Start programs arealso available or pre-school children in New Jersey Abbott districts. Tese Head Startproviders may oer Enhanced Head Start which must meet Abbott pre-school stan-dards so as to be eligible or state unding.


6/54

6 | THE COST OF HIGH-QUALITY PRE-SCHOOL EDUCATION IN NEW JERSEY

Across the 31 districts, 38,900 children are enrolled in pre-school (approximately 75%o the cohort). Over hal o all children are enrolled at private centers, with one-thirdin public school programs, and the remainder in Enhanced Head Start programs andindependent special education centers.

Current ependitures on pre-school

Funding or Abbott pre-school comes rom two agencies: the Department o Educa-tion (DOE) and the Department o Human Services (DHS). Te DOE unds the6-hour educational component o the day and the DHS unds the wrap-around ex-tended day and summer care.

Te DOEs estimate o its total expenditures in 200607 is $501 million with DHSspending an additional $222 million. DOE unds are allocated to districts. Teseunds are then allocated to in-district programs, private providers or Head Start cen-ters. Districts do hold back unds to provide district-wide services and administrationor the regular pre-school program.

In 2006, per-child state aid or the main component o the program was: $12,585or children in private centers; $10,900 or children in public programs; and $7,241or children in Enhanced Head Start. Providers ace dierent costs or teacher hiring,location, and or acilities; private providers may ace higher administrative costs be-cause they cannot share acilities with public K12 schools; and Head Start providersreceive ederal subsidies to supplement this state aid.

Tese amounts are high relative to spending in other states. However, the prices oeducational services in New Jersey are almost 25% higher than the national average.Te adjustment makes a big dierence. For example, absolute pay or new teachers inNew Jersey is the sixth highest in the nation; when we adjust or relative prices it isthe seventh lowest.

Importantly, these expenditures only tell us what is spent on pre-school, not what itcosts.

Enhancing the quality of pre-school

Tere are a number o reasons why pre-school unding may be below what is needed.

Compensation or pre-school teachers may need to be increased. Tis upward pres-sure reects several actors. First, the program requires a specic credential and mustinclude proessional development. Second, Abbott pre-school teachers in private set-tings receive less generous benets than teachers in the public schools. Tird, pre-

school teachers have less experience than the average teacher because the program isnew; as these teachers gain experience the salary structure is such that their pay willincrease. Finally, as the program expands, higher wages must be oered to attract moreworkers.

A second consideration is acilities. Te growth o the pre-school program has putpressure on capital, such that current acilities are probably insucient. In the pub-lic system, many schools are very old and were not constructed or the purpose o


7/54


educating very young children. In the private sector, capital unding may be limitedand many centers lease rather than own their space. Also, support or public-privatepartnerships to build acilities has been limited. Hence, there is a need or additionalunding to build new acilities and upgrade older schools.

Tere are additional pressures on pre-school budgets.

Systems o accountability must also be included in the costs o pre-school programs.Te Abbott requirements include a rigorous set o state inspections (every three years)and sel-assessments and validations by the providers (annually). But these are otennot explicitly counted in budget statements or are insuciently costed.

Pre-school provision or children with special educational needs must also be appropri-ately unded. DOE data shows that the majority o children with special educationalneeds are in sel-contained settings, which may increase costs unnecessarily. However,the absolute proportion o children in special education in pre-school is low (less than5%), suggesting its incidence is under-diagnosed. More accurate diagnosis, along withscreening services such as Recognition and Response, would raise costs.

Te scale o the Abbott pre-school program may also put pressure on costs. Te Ab-bott program is genuinely universal and must be ully unded by the state. Universalprograms may be more expensive because they have to accommodate students withheterogeneous backgrounds and must serve all students in the locality, including thosewho would not have otherwise attended and those who require more supports orlearning.

Eective pre-school should also be aligned with K12 schooling. Alignment includesactors such as the physical movement and transportation o children and inorma-tion exchange across preschool and kindergarten teachers. Te costs or alignment are

typically not explicit in budgets.

Ensuring efciency

It is important to ensure that pre-school programs are eciently provided. Te chargeo ineciency is typically made because public enterprises do not have clear goals, arenot directly accountable to customers or revenues, and do not have competitors toorce them to meet the needs o clients and keep costs low.

However, it is important not to over-emphasize this argument. Eciency gains shouldbe sought (regardless o the amount o money allocated). But we have to identiyineciency and nd a remedy and this is not easy. A center is not inecient simplybecause it spends more than another or seems to spend a lot. Te prices o inputsmay dier, as may the quality o provision; and public and private providers may servechildren with dierent characteristics. We have already noted above that prices oeducational services in New Jersey are some o the highest in the nation.

Also, the evidence on public schools ineciency is surprisingly weak and the mostcommon argument or ensuring eciency is competition through market orces.Tere is evidence that competition does increase eciency. But the Abbott pre-school


8/54


system is already competitive: there are many small private centers competing with in-district programs to supply a well-dened, uniorm service. It is also closely regulated.Tis is the classical economic denition o an atomistic, competitive market.

Calculating the full costs of pre-school: Methods

We use several methods to predict the ull amount o resource needed to provide awell-planned, high quality pre-school program.

Our primary estimates o the costs o pre-school are based on cost unction methodsusing data on Abbott district pre-schools. We estimate the determinants o the averagecost o pre-school in private centers and public program separately.

We use data on the costs and quality rom inspections o classrooms in 210 privatecenters and 535 public programs. We correlate the average cost with the observedquality to see whether improvements in quality are associated with higher costs. Ourcorrelations adjust or the prices o teachers in each district and demographic charac-teristics. Tis cost unction equation tells us how much more it costs to provide high

quality pre-school.

o check our results we review all extant cost templates or pre-school. Tese templateshave been produced by researchers across the U.S. or dierent pre-school programs.Tey are researchers best estimates o what pre-school costs. Our search produced 11dierent templates which estimated the cost o pre-school. We adjusted these tem-plates to meet the Abbott district requirements and to account or dierences in pricesin New Jersey.

Calculating the full costs of pre-school: Results

Our cost unction estimates show that higher quality programs do cost more and that

hiring more experienced pre-school teachers will cost more. We do not nd strongevidence that larger centers have lower unit costs.

Our cost unction estimates indicate that current unding should be increased by5.8% or public programs and 6.8% or private centers. Tis increase in costs wouldensure high quality pre-school programs across the districts. Te increase reects onlytwo actors: the cost o improving quality and the cost o paying teachers more to bet-ter reect the necessary amount o compensation.

For the academic year 200708, average per-child expenditures or pre-school in theAbbott districts should be $11,993 or public programs and $13,978 or private cen-ters. Tese amounts account or the necessary increases to ensure quality and year-

on-year ination. Using current unding ratios, we derive unit costs o $8,189 orEnhanced Head Start provision. Tis cost is calculated as current Enhanced HeadStart spending plus the absolute amount extra that we predict or public centers.


9/54


Based on the template estimates, we nd that the cost o pre-school across all providertypes in the Abbott districts is at least $12,300 per child. Tis is a lower bound becausemost cost templates only consider operating costs.

Per-Child Costs of Pre-school in Abbott districts

(adjusted to 200708 prices)

AveragePublic

programsPrivate

providers

Currentspending $11,982 $11,333 $13,088

Costtoensurewell-planned,highqualitypre-schoolusingcostfunctionmethod

$12,736 $11,993 $13,978

Lowerboundestimateusingcosttemplates

$12,276 $11,611 $13,410

Conservative nature of results

Tese estimates o unit costs or the Abbott districts are almost certainly conservative.Tey do not take account o possible under-unding o: acilities; programs to alignpre-school with K12 education; accountability systems; and special education.

Other actors will also raise costs. For example, i more three-year olds are enrolled,our estimates o unit costs understate the true resource requirement. Expanding theprogram to serve more children is likely to urther increase costs. Wages would have togo up; and we nd little evidence o savings i centers are larger such that economies

o scale can be realized.

o indicate the extent to which these gures are understated, we investigated twoactors. One is the cost o quality, i.e. what it costs to raise quality. Our estimates arethat an additional 2% in unding would be needed or this aspect alone. Alternative,plausible estimates are closer to 5% or 6%. Tus, unding might be increased by anadditional 3 or 4 percentage points. Te second domain where our results are conser-vative is or acilities. Te above calculations assume no additional acilities costs. I weassume that currently the underinvestment in acilities is the same as that or teachers,we should add another 1 or 1.5 percentage points to the required unding.

Implications for total funding

otal unding may be calculated as the unit cost times the numbers enrolling. I en-rollment is expanded by 4% to cover 40,456 children, annual unding or high qualitypre-school in the Abbott districts would total $528 million in 200708.


10/54


Statewide epansion

Pre-school outside o the Abbott districts is publicly unded under the Early Child-hood Program Aid (ECPA) program. Currently, ECPA helps und pre-schoolprovision or 7,270 children in 101 districts across the state. Per child state aid is ap-proximately $4,400 and total state aid is $30 million.

In addition, the Early Launch to Learning Initiative (ELLI) is a small-scale programoering public pre-school to children outside the Abbott districts. Participating schooldistricts are required to blend state aid with other sources o unding, including localrevenue, IDEA, itle I, and parental contributions. In 2006, approximately 1,500children across 28 districts participated. State aid unding was $2.19 million.

Potentially, many more children across New Jersey might benet rom high qualitypre-school. Price indices show that the unit costs o a given quality o pre-school arevery similar across Abbott districts, across non-Abbott ECPA-unded districts, andacross the state as a whole. Tereore, our cost estimates or the Abbott districts mayapply to other districts. However, additional state-level administrative costs and set-up

costs will be incurred i the program is expanded to other districts.

Conclusion

Given the benets o pre-school, it is important or the state to invest in high qual-ity programs. Our cost analysis shows that current spending is somewhatbut notgrosslybelow what is needed. In conclusion, we note two important aspects or theuture o the program.

First, the additional increment in investment is not signicantly above the rate o in-ation. Tereore, making sure that the rate o change o prices is calculated accuratelyis important to ensure that the program is adequately unded.

Second, it is important to ensure that the workorce is eciently structured and thatuture wage increases are anticipated. Te Abbott pre-school program has grown rap-idly over a short time period. Many pre-school sta (now highly qualied) are new totheir positions. Retaining them will be important to ensure program quality but it willalso put pressure on costs: many o these teachers may have oers rom other schoolsand districts as the labor market changes; and as these teachers gain more experiencetheir salaries will increase correspondingly. I current trends continue, upward pres-sures on wages will necessitate additional unding.


11/54


1. INTRODUCTION

Tis report calculates the cost o providing adequate pre-school or children in NewJersey, as required underAbbott vs. Burke(153 NJ 480 1998). Pre-schooling is an im-portant component o the states Constitutional requirement to provide a thorough

and ecient system o education. We derive reliable, evidence-based estimates o thecosts o well-planned and high-quality pre-school. We ocus primarily on provisionor children residing in the Abbott districts, but we also extrapolate our estimates orthe purpose o statewide expansion.

Te case or investments in pre-school is strong. High-quality pre-school helps chil-drens cognitive growth.1 Tis conclusion is conrmed by evidence rom small-scaleexperiments, welare reorms, national data, and evaluations o state preschool pro-grams and Head Start.2 Yet, not all children have access to preschool. Afuent ami-lies are more likely to enroll, as are parents with more education. I we include HeadStart, Arican-American children enroll at rates higher than white children; nationally,Hispanic children enroll at relatively low rates in all programs.3

Te benets o preschool are stronger or children rom less advantaged backgroundsand or Hispanic children.4 New evidence on the benets or children rom partici-pating in the Abbott district pre-school programs is presented by Frede et al. (2007).Children who attend these programs obtain higher scores in language, literacy, andmathematics. Tese gains persisted into kindergarten such that children who attendedpre-school or two years had closed over hal o the gap against national averages orvocabulary.

However, the quality o preschool is highly varied across the U.S. Children inhigh-income amilies tend to experience the highest quality preschool. Childrenin low-income amilies do typically enroll in better quality programs than those in

middle-income amilies, but this is because they are eligible or means-tested publicsubsidies such as Head Start. Importantly, investments in pre-school and investmentsin high quality K12 schooling are complementary. One does not substitute or theother and the gains rom pre-school need to be reinorced through eective elemen-tary and secondary school education.5

1 Te evidence is mixed on the eects o early education on behavior. Using national data, Loeb et al. (2005) andMagnuson et al. (2005) nd that center-based care negatively eects childrens socio-behavior. Te reasons, though,are not clear. Colwell et al. (2001) hypothesize the cause is the cumulative amount o time spent in child care; Loebet al. (2007) nd earlier entry age is associated with more negative eects on behavior.

2 On experiments, see Beleld et al. 2006; Reynolds et al. 2002; Campbell and Ramey 1994. On welare reorms,

see Zaslow et al. 2002. National data are in Magnuson et al. 2007. For state results, see Henry et al. 2003; Gormleyand Gayer 2005; Barnett et al. 2005. On Head Start, see Love et al. 2002.3 On socioeconomic status dierences, see: OBrien-Strain et al. 2003; Hoerth et al. 1995; Fuller et al. 2004; Bain-

bridge et al. 2005; West et al. 1995; Magnuson et al. 2005. On participation dierences, see: Liang et al. 2000. Inpart, Hispanic children may enroll at lower rates because they live in regions where state programs are smaller.

4 On benets by race, see: Peisner-Feinberg and Buchinal 1997; Loeb et al. 2004; Magnuson et al. 2007; Gormley etal. 2005. On income eects, see: Belsky et al. 1996; Phillips et al. 1994.

5 Research by Magnuson et al. (2007) nds that how pre-schoolers progress into schools makes a dierence to theduration o academic gains. Other research has ound pre-school impacts to ade out, in part because o low qualityK12 schooling. See also the model generated by Heckman and Cunha (2007).


12/54


Trough linking dierences in the observable characteristics o pre-school programsto childrens outcomes, it is possible to identiy programs that genuinely are higherquality. Evidence reveals, or example, that teachers with more education and bettertraining tend to provide higher quality preschool as shown by improved childrensacademic perormance. Likewise, programs with lower child to sta ratios are better,as are programs with more intensive learning and longer duration.6

Tis evidence establishes that high-quality pre-school is academically benecial (espe-cially or disadvantaged children) and that eective programs can be specied. Stateinvestments should reect this consensus, such that well-planned and high-qualityprograms are made available, particularly or those children in low-income amilies.However, current state expenditures may not ully equate to the amount needed orhigh quality pre-school.

Based on the above evidence, and direct analysis o budgetary data in New Jersey, wecalculate the costs o high-quality pre-school or the state. Te report is structured asollows. First, we describe current pre-school provision in New Jersey. We report onenrollments and current program standards and document current unding levels; wealso consider additional requirements that might be included in the Abbott programrequirements. Second, we describe the possible methods or estimating costs and howthese might apply to the Abbott districts. We also review the existing evidence. In thethird section, we calculate our baseline estimates o the costs o pre-school or the Ab-bott districts. We use two methods: one is the cost unction approach; the other is avariant o the evidence-based reorm design applied to pre-school rather than K12schooling. Using multiple approaches, we can place bounds on the unit costs. Fourth,we report our estimates o the per child cost o pre-school aggregated up to the districtand state level. We ocus on provision in the Abbott districts, but also consider thecosts o expanding similar services to other districts. Finally, we consider importanteconomic issues related to pre-school provision and expansion across New Jersey.

Our analysis is based on current patterns o enrollment and expenditures. Tese g-ures involve thousands o students and many separate expenditure items. Some mea-surement error is possible, although such error is as likely to overstate as understateour ndings. Also, as with all education systems, pre-school in New Jersey is evolvingand student mobilityparticularly among disadvantaged groupsis high. As such,a xed picture o pre-school is not easily depicted. Our analysis is based on the bestavailable data at the current time and our calculations are expressed as average costso provision or marginal costs o expansion; these costs are not heavily dependent onthe baseline numbers. Finally, we include sensitivity tests to see how costs vary underalternative assumptions.

6 On quality, see: Peisner-Feinberg and Buchinal 1997. On teacher quality, see: Loeb et al. 2004; NICHD 2000;Arnett 1989; Fuller et al. 2006; Howes 1997; Blau 2003; Gormley and Philips 2007. On child-sta ratios, see:Clarke-Stewart and Allhusen 2005; Ruopp et al. 1979; Howes et al. 1998. On intensive learning, see: Robin et al.2006; Fuller et al. 2004. However, programs with stricter regulations are not necessarily better, see: Kisker et al.1991; Helburn 1995; Gormley 2007.


13/54


We do not consider broader policy issues in relation to the organization o schoolingin the Abbott districts. Tese considerations are beyond the scope o this analysis.

Troughout this report money values are expressed in 2007 dollars (unless explicitlystated otherwise). Costs and prices are weighted according to a district-specic orstate-specic price index (CWI) generated by aylor and Fowler (2006). We note here

that, as prior indices have shown, New Jersey has one o the highest cost-o-educationvalues o any state: prices o educational inputs are 24.4% higher than the average orthe nation. In act, the CWI is likely to understate the dierences in costs when ap-plied to pre-school education and or the most disadvantaged districts. First, it doesnot capture any variations in cost attributable to working conditions in specic dis-tricts (aylor and Fowler 2007, p.4). Second, the CWI does not include workerswithout degrees or those who work less than hal-time because they are not directlycomparable with teachers (p.7). Tese understatements mean that our cost estimatesare probably conservative.

2. PRE-SCHOOL PROGRAMS IN NEW JERSEY

2.1 Pre-School in the Abbott Districts

Te Abbott districts provide pre-school or 3 and 4 year olds ollowing the New JerseySupreme Courts ruling and codied in the state Department o Education programguidelines. Tese stipulations are given in Box 1. Te specic inputs that correspondto these requirements are given in Appendix able A1.

Overall, these requirements reect the research consensus on the components o high-quality pre-school. From a resource perspective, the most important regulations per-tain to stang and class size: certied teachers and master teachers (who mentor andtrain classroom teachers) are expensive, as is the need or class sizes o 15. Tere arealso signicant supplemental services, as well as managerial and accountability re-source requirements. Te last item is important or ensuring that programs are imple-mented aithully.

As part o the comprehensive ull-day, year-round pre-school provision in Abbottdistricts, wraparound childcare services are oered beyond the six hours o educa-tion programming.7 Tese wraparound services consist o 4 hours per day during theschool year and 10 hours per day during the 65-day summer session. Te Abbott pro-grams thereore have three components: a regular pre-school program; extended-daywraparound provision; and summer wraparound provision.

7 Until July 2007 all amilies were eligible or extended day coverage at no charge. Ater July 2007 amilies with in-comes greater than 300% o the ederal poverty line must pay the ull cost o these wraparound services.


14/54


Bo 1Pre-school program standards: Abbott districts

DurationFor180+days(thelengthofthedistrictsK12schoolyear),6hourseducationalcomponent+4hourswraparoundservicesForadditional65days,10hoursofwraparoundservices

Child eligibility

All3and4yearoldslivingintheAbbottdistrict

Districtsmustplantoenrollatleast90%eligiblestudentsindistrict.

Eligible providersDistrictcenters;district-contractedcommunityprivatecenters;andHeadStartprogramsthatmeetAbbottstandards

Class size Maximumat15

Stafng per classroom1Certiedteacher(BA+P3certication)1assistantteacher

Other staff

1+masterteacherper20classrooms1on-sitedirector1+socialworkerper250300preschoolstudents1preschoolinterventionteamper750children

Supplemental services Socialservices,health,transportation

Students with special needs ProgramsmustserveESLanddisabledstudents

Mandated curriculum Developmentallyappropriatecurriculumthatmeetscontentstandards

Assessment NJEarlyLearningAssessmentSystem(mandatory)

Monitoring andaccountability

Validationvisitbystateonceevery3years.Annualinternalassessmentbyproviders.

FacilitiesNewfacilitiesmustbe950sqft,nohigherthanthe2ndoor,withoutdoorspaces,naturallight,andstoragespace

Professional development AsperK12schools.

Source: NJAdministrativeCodeTitle6A,Chapter10A.www.state.nj.us/njded/code/current/title6a/chap10a.pdf

Te Abbott pre-school program is delivered through the public school system. It is apublic program unded and regulated by the Department o Education. Abbott pre-school programs may be oered by private providers, who contract with the schooldistrict, as well as through the public school system. In total, there are approximately600 separate centers or providers. Head Start programs are also available or pre-school children in New Jersey Abbott districts. Tese Head Start providers may oerEnhanced Head Start which must meet Abbott pre-school standards so as to beeligible or state unding.

able 1 shows the enrollments in pre-school in the Abbott districts. Across the 31districts, approximately 75% o eligible children are being served. O these 38,900children, 55% are in private centers and 35% are in-district, i.e. in public schools.O the remainder, 8% are in Enhanced Head Start programs and 3% in independentsel-contained pre-school disabled class rooms.


15/54


Table 1Demographic Information on Pre-school in the Abbott districts (20062007)

Abbott school districts

Numberofdistricts 31

Percentofeligiblechildreninpublicly-fundedpre-school ~75%

Actualenrollmentinpublicly-fundedpre-school 38,900

Privateproviders 55%

In-districtproviders 35%

HeadStart(enhanced) 8%

Specialeducation(self-contained) 3%

Three-yearolds 45%

Four-yearolds 55%

Sources: NJDOEOfceofEarlyChildhoodEducation.www.state.nj.us/education/data/.

Notes: DataretrievedAugust52007,priortonalauthorizationbythestate.

Across the 31 Abbott districts there is variation in what proportion o children livingin the district are served, which settings they are in, and in how many three-year oldsparticipate. Full details or the 2006 cohort are given in Appendix ables A2 and A3.Districts vary rom only two-thirds participation to practically complete enrollment(population numbers are predictions and potentially mis-state participation rates). Somedistricts elect not to contract out pre-school services through community-based privateproviders, other districts have no Enhanced Head Start places, and one district has nopublic provision.8 Although three-year old children are participating at high rates, thereare still some districts where they are signicantly under-represented (e.g., Camden).

2.2 Abbott pre-school ependitures in New Jersey

Funding or Abbott pre-school comes rom two agencies: the Department o Educa-tion (DOE) and the Department o Human Services (DHS). Te DOE unds the6-hour educational component o the day (the regular program). Te DHS unds thewrap-around extended day and summer care.9

Te DOEs estimate o its total expenditures in 200607 is $501 million, and basedon the DOEs enrollment gures, DHS expended an additional $199 million.10 DOEFunds are allocated to districts. Tese unds are then allocated to in-district pro-grams, private providers or Head Start centers. Districts do hold back unds to providedistrict-wide services and administration or the regular pre-school program. Centers

8 Hoboken has pre-school in public schools that are operated by private providers.9 Te ocus here is on the amounts spent, not the sources o unding. Ritter and Lauver (2003) analyze the sources

o unding and disparate burdens across local communities in New Jersey. We recognize that unding is blendedrom a range o sources (e.g., Abbott pre-school expansion aid and ECPA aid to Abbott districts).

10 Tese gures are approximate since neither DOE nor DHS has nal reconciled numbers or the 20062007 year.DHS calculates actual expenditures rom retroactive monthly attendance counts provided to DHS by pre-schoolproviders. Te DOE initially tabulates enrollment rom projected enrollment gures provided by 31 Abbott dis-tricts, with nal reconciliation at the end o the year. Te two methodologies or enrollment yield disparate enroll-ment gures. For this report, the authors estimated DHS expenditures by multiplying the DOE enrollment counto 38,900 by the DHS per pupil allocation or wrap around services.


16/54


dier in the budgets per child: on average, in-district centers receive allocations o$10,987; private providers, $12,585; and Head Start providers $7,241. (Tese guresinclude provision or children with special educational needs. Te Head Start resourc-es do not count ederal transers).

For wraparound services, DHS provides unds to all the providers at a per-child

monthly unit rate. Te calendar year 2007 rate or accredited pre-schools is $394 permonth or the hal-day services (4-hours per day or 10 months) and $587 per monthor ull-day services during the two summer months. Assuming a child attends wrap-around services or the ull year, the annual unding or these services is $5,110.

able 2 shows the trend in regular expenditures in the Abbott districts since 2002. By2006 enrollments have stabilized at 38,900 (although there is still some local unmetdemand with children on waiting lists). Similarly, per-pupil DOE expenditures arestable at approximately $11,500 in real dollars. able 3 reports expenditures per-childby center type in each o the Abbott districts. Expenditures in private centers arehigher than in public settings, by $1,687 or 15%. Tere may be many reasons or thisdierence, such as the need to nd acilities, the types o students that are served byprivate providers, or the locations o centers. Public centers may be able to piggybackon school resources such that some o their costs are hidden. Notably, private provid-ers serve a large proportion o children; without them, the number o available placeswould be much lower.

Te range o spending across private centers is also signicant, rom $10,149 to$15,341; unding dierences across public programs are also substantial. Te bulk othis disparity in unding levels may be attributable to variation in teacher qualica-tions: the DOE salary schedules remunerate centers at higher rates or teachers withmore years o experience, academic credentials, or state certication. So a pre-schoolthat hires more experienced sta will be reimbursed more.

Note, however, that these expenditures only tell us what is spent on pre-school, notwhat it costs.

Table 2Ependiture information for the Abbott districts 20022006

Academic year

Total annualependitures($ millions) Average spending per child

Nominaldollars

2007dollars

2002 $379.92 $9,645 $11,092

2003 $406.33 $9,637 $10,6972004 $451.82 $10,519 $11,466

2005 $463.93 $10,754 $11,292

2006 $460.00 $11,521 $11,752

Sources: NJDOE.www.state.nj.us/education/data/.Totalannualspendinginnominaldollars.Spending

doesnotincludeDHSfunding.Enrollmentin2006isactual,notapproved.


17/54


Table 3State Ependitures

In-district Private provider Head Start

AsburyPark $8,113 $15,341 $7,544

Bridgeton $10,528 $10,780 $4,028

BurlingtonCity $9,269 $10,518

Camden $8,698 $11,161 $3,666

EastOrange $9,902 $11,196 $7,550

Elizabeth $13,364 $15,305

Gareld $10,884 $10,724

Gloucester $11,098

Harrison $11,712 $13,407

Hoboken $11,906 $10,076

Irvington $11,911 $13,043 $6,945

JerseyCity $11,262 $12,837 $5,413

Keansburg $10,315

LongBranch $9,014

Millville $9,817 $10,904 $4,654

Neptune $10,120 $13,273

NewBrunswick $8,346 $12,719 $3,354

Newark $9,194 $13,785 $8,026

Orange $8,931 $12,318 $6,784

Passaic $12,294 $13,703 $6,157

Paterson $7,793 $11,989 $7,035

Pemberton $9,109 $11,766

PerthAmboy $12,898 $12,299

Phillipsburg $10,516 $12,586 $8,457

Plaineld $7,532 $12,102 $6,309

Pleasantville $10,547 $12,287 $5,309

Salem $10,524 $7,407

Trenton $10,666 $13,054

UnionCity $9,992 $10,149 $7,499

Vineland $10,712 $12,768 $6,403

WestNewYork $10,684 $12,881 $7,130

WEIGHTEDAVERAGE $10,900 $12,585 $7,241

Source: NewJerseyDepartmentofEducation(OPRAsubmission).

Notes:Missingcellsdenotenoprovisionorinformationnotavailable.Costsincludedpro-ratedsharesof

district-wideadministrativecostsforpre-school.SpendingdoesnotincludeDHSfundingforwraparound

services.Weightedaverageaccountsfordistrictenrollmentpatterns.


18/54


2.3 Pre-school outside the Abbott districts

Pre-school outside o the Abbott districts is publicly unded under the Early Child-hood Program Aid (ECPA) program. Full details o the program are in Box 2. Stateunds are available or districts i 20%40% o the students are rom low-incomeamilies. Currently, non-Abbott ECPA helps und pre-school provision or 7,270 chil-

dren in 101 districts across the state. Per child state aid is approximately $4,400 and isallocated based on K12 enrollment. otal state aid is thereore $30 million. However,state unding may be supplemented rom other local and ederal sources.

In addition, the Early Launch to Learning Initiative (ELLI) is a small-scale programoering public pre-school to children outside the Abbott districts (but including otherECPA districts). Participating school districts are required to blend state aid withother sources o unding, including local revenue, IDEA, itle I, and parental con-tributions. In 2006, approximately 1,500 children (600 o whom were low-income)across 28 districts participated. State aid unding was $2.19 million.11

Bo 2

ECPA pre-school program standards: ECPA non-Abbott districts

Duration Minimumof2.5hours/dayforacademicyear.Half-dayprogramcomprises75%enrollees,andfull-dayprogramfortheremaining25%ofenrollees

Child eligibility4and5yearolds.(3yearoldsiffundspermit)Highpovertydistricts(Atleast2040%ofchildrenenrolledindistrictqualifyforthefederally-subsidizedFreeMealsorwhosefamiliesearn130%ofthepovertylevel)

Eligible providersPrimarilydistrict-provided(89%enrollees),althoughcommunity-basedcontractingallowed

Class size ratioof1:25

Stafng per classroom TeachersmusthaveBAandbecertiedinearlychildhood

Support staff Norequirements

Supplemental servicesRecommendedtoincludehealth,nutrition,andotherservices.Districtsshouldaddresstransportationbarriers.

Students with special needs DistrictshouldtrytomainstreamIEPstudents

Mandated curriculum 5recommendedcurricula

Assessment Nomandatedassessment

Professional developmentEachdistrictmustsubmitaprofessionaldevelopmentplanintheir1-yearoperationalplan.

Inspection

Districtmustcertifythatprogramsmeetqualitystandard.7-yearmonitoringcycleby

state.

11 Te regulations or ECPA and ELLI provision are much less stringent and detailed than those or Abbott (e.g.the latter has smaller classes and higher sta-child ratios). Tus, processes across these three program types are noteasily compared.


19/54


Finally, many amilies across New Jersey will enroll independently in private pro-grams. Nationally, one-third o all our-year-old children are enrolled in private pro-grams (Blau and Currie 2006).

2.4 Evaluating the Abbott program requirements

Relative to many other states programs, the Abbott program requirements are high.Absolute public spending in New Jersey on early education per child Abbott pre-schoolis also higher than almost all other states (unadjusted or CWI). In part this reectsthe unusually extensive set o Abbott pre-school services when compared to otherstates. It does not imply that the Abbott program is suciently well-resourced or thatthe standards are suciently high (given the research on the benets o pre-school).So, although the Abbott program requirements broadly meet quality standards, thereremain a number o issues relating to the quality o the programs currently provided.

We consider a range o actors that may cause a divergence between current expendituresand what it actually costs to oer well-planned, high quality pre-school in New Jersey.Te goal is to see whether pre-school programs are under-unded or over-unded.

Teacher Quality

Te rst issue or high-quality preschools is to ensure that all teachers are qualied(Early et al. 2007). National Institute or Early Education Research (NIEER) qualityratings emphasize teacher quality, such that: the lead teacher must have at least a BA,have specialized training, and perorm annual in-service training o at least 15 hours;and the assistant teacher must have a Child Development Associate (CDA).12 As o200607, over 99% o Abbott pre-school teachers have at least a BA degree and 43%have a P3 certication.

However, in the early years o the Abbott program, nding adequately credentialedteachers was a challenge.13 Many pre-school lead teachers have recently obtained cer-tication.14 But, the preschool labor market and career routes are not as sharply estab-lished as or K12 teachers. Pre-school certication programs may be less valuable ormore risky i they do not transer across labor markets (e.g., to other states where pre-school teachers do not require degrees). Also, the most recent NIEERYearbook(2006)shows the Abbott district standards only miss one o the ten quality benchmarks: theexpectation that assistant teachers have a CDA credential.

A second concern is that pre-school teachers are relatively inexperienced. Abbott dis-trict pre-school teachers have on average 5.6 years o experience, compared to a national

12 Similarly, NACCRRA (2007) quality standards expect teachers to have a CDA credential or Associates degree inearly education; annual training, including health and saety; and center directors must have at least a BA. We notehere that neither o these rameworks considers second language competency or pre-school teachers.

13 Early survey data showed that New Jersey teachers tended to be academically qualiedthe proportion o teacherswith only high school or an Associates degree as their highest qualication was relatively low in New Jersey (13%in the Abbott districts but 27% nationally)but the proportion o teachers with Child Development Associatecredentials was only 15% in the Abbott districts compared to 23% nationally (Gilliam and Marchessault 2005).

14 Certication has been very recent: as o 2004, over 80% o teachers in the Abbott districts were working toward aP3 credential (Ryan and Ackerman 2004).


20/54


norm o 8.2 years. wo-ths o teachers have less than ve years experience (Ryan andAckerman 2004). Inexperienced teachers are paid less, but experience is also positivelycorrelated with educational outcomes such that program quality is sacriced.

A third concern is proessional development (Odden et al. 2002). Te Abbott require-ments mandate that districts provide at least one master teacher per 20 Abbott preschool

classrooms in the district who may perorm on-the-job training. But Abbott pre-schoolteachers report less time than their peers or individual and team planning: per week, thehours on these tasks are 3.3 and 1.5 in the Abbott districts, but 4.1 and 2.2 nationally.Moreover, across New Jersey pre-school teachers may have ollowed a less traditionalcareer path; they may be older than K12 teachers and less likely to be in ull-time em-ployment (Ryan and Ackerman 2004). Tereore, their training needs may be greater.

Each o these actors is likely either to put upward pressure on salaries or to meanthat pre-schools hire rom a less-qualied or inexperienced labor market. Abbott pre-school teachers in private settings are paid on the same scale as public school teachers,but their health and retirement benets are not as generous (because they are more ex-pensive to purchase). So there are two eects: one is the relative position o pre-schoolteachers to school teachers in the Abbott districts; the other is the relative position oteaching in the Abbott districts to other occupations or teaching in other districts.

Unortunately, wage comparisons must account or the many dierent circumstancesacross labor markets (Stoddard 2005). In absolute money amounts, pre-school teach-ers in New Jersey receive compensation that is comparable toand oten higherthannational norms. In New Jersey and across the U.S., hourly wages are $24$30and annual salaries average $40,000; and the majority o teachers receive health andretirement benets (see Appendix able A4). Critically, price index adjustments makea big dierence: aylor and Fowler (2006, p.15) note that, although the unadjustedsalaries or beginning teachers are relatively high in New Jersey, Adjusted or geo-

graphic variations in hiring costs, beginning teacher salaries in New Jersey were notthe sixth highest in the nation, but instead the seventh lowest. As pre-school salariesare set in correspondence to K12 teachers, this adjustment also applies to pre-school.Regardless o national comparability, however, requiring more credentials o teachersand nding more experienced teachers will necessitate higher pay.

Facilities

Abbott pre-school program enrollments have grown substantially over a very short pe-riod o time (see able 2). Plus, the Abbott requirements o 950 square eet per class-room exceed the minimum space in pre-school designs (o 35 sqt per child, Sussman2005; Butin 2000). Tis requirement puts pressure both on existing providers andon new centers. In their study or Pennsylvania, Gill et al. (2002) report that start-up

costs are a substantial constraint on new centers.

Across the Abbott districts, resources or acilities and capital construction are prob-ably insucient. In the public system, many schools are very old and were not con-structed or the purpose o educating very young children. In the private sector, capitalunding may be limited (Fuller and Wright 2007, p21) and over two-thirds o centerslease rather than own their space (Rice and Ponessa 2004). Leased spaces may not


21/54


have been designed or the purpose o providing pre-school. Also, support or public-private partnerships to build acilities has been limited. Hence, there is a need oradditional unding to build new acilities and upgrade older schools.

Additional Cost Considerations

We identiy a number o additional cost considerations. Most o these are likely toraise the pressure on pre-school resources.

Systems o accountability must also be included in the costs o pre-school programs.Quality rating scales emphasize these systems.15 But they are oten not explicitlycounted in budget statements or are insuciently costed (as shown in the literaturereview below). Te Abbott requirements include a rigorous set o state inspections(every three years) and sel-assessments and validations by the providers (annually).Across all district-wide services, it is unclear whether these internal and external ac-countability systems are being ully allocated to pre-schools (Applewhite and Hirsch2003).

Pre-school provision or children with special educational needs must also be appro-priately unded. Special education programs are on average 1.9 times more costly thanregular education programs (Chambers et al. 2003). DOE data shows that the major-ity o children with special educational needs are in sel-contained settings (contraryto the requency in K12 education). Tese settings are normally more expensive thaninclusive settings. Tereore, i there is over-classication into sel-contained settings,then the state is over-paying or special educational services in pre-school.16 However,the absolute proportion o children in special education in pre-school is low (less than5%), suggesting that the incidence o special education is under-diagnosed. Moreaccurate diagnosis, along with screening services such as Recognition and Response,would raise costs. In addition, very ew children with special educational needs are inprivate centers. In part this may reect expertise and training within the public sec-

tor.17

But it also reects the weaker incentive or private providers (and these cannottypically cross-subsidize such provision).

Te scale o the Abbott pre-school program may also put pressure on costs. Unlikeprograms in other states, the Abbott program is genuinely universal (not targeted) andmust be ully unded by the state. Universal programs may be more expensive in thatthey have to accommodate students with heterogeneous backgrounds and serve allstudents in the locality. Universal programs must also cater to students who would nothave otherwise attended pre-school and who may require more supports or learning.(Tese costs may be oset by reduced administrative costs or determining eligibility).ypically, state unding or pre-school is not the only source o unding (with ederaland local governments oten contributing).

15 NIEER ratings account or whether regulatory agencies make site visits to ensure compliance with requirements.NACCRRA ratings expect: centers to be licensed and to be inspected at least 4 times per year; licensing sta to haveat least a BA and a center caseload o ewer than 50 programs; and inspection reports to be publicly available.

16 In 2007 75% o preschool children with disabilities statewide and 60% in the Abbott districts are in segregatedsettings (www.nj.gov/education/specialed/ino/spp/partb.doc).

17 Communication, Ellen Frede.


22/54


Eective pre-school should also be aligned with K12 schooling. Alignment includesactors such as: the physical movement and transportation o children; parental prepa-ration or school; inormation exchange across preschool and kindergarten teachers;and sta collaboration. Using national survey data, Magnuson et al. (2007) nd thatchildren progress aster when the transition to public elementary school is easier. Yet,in a recent study o the Abbott districts, Rice (2007) nds that alignment in each o

these actors is not adequate. However, absent inormation on the costs o improvedalignment, we do not include this actor in our cost calculations.

Finally, the cost o pre-school might be below current expenditures i there is sub-stantial ineciency in provision. Te claim o ineciency is typically made becausepublic enterprises do not have clear goals, are not directly accountable to customersor revenues, and do not have competitors to orce them to meet the needs o clientsand keep costs low. However, it is important not to over-emphasize this argument.

Certainly, eciency gains should still be sought, regardless o the amount o undingallocated. But we have to identiy ineciency and nd a remedy. Tis is not easy. Acenter is not inecient simply because it spends more than another. Te prices oinputs may dier, as may the quality o provision; public and private providers mayserve children with dierent characteristics; and in less densely populated areas ewerchildren may enroll or transport costs may be higher. For the last example, a remedyis not easible (unless amilies are orced to move closer together).

In act, the evidence on public schools ineciency is surprisingly weak and the mostcommon argument or ensuring eciency is competition through market orces. Teevidence on competitive pressures raising quality and eciency o education is reason-ably strong; there is even evidence or pre-school (Henry and Gordon 2006). But theAbbott pre-school system is already competitive: there are many small private centerscompeting with in-district programs to supply a well-dened, uniorm service. Tis is

the classical economic denition o an atomistic, competitive market.

Expenditures and Costs

Overall, these actors suggest that current budgets are below what is necessary to oerwell-planned and high quality pre-school in the Abbott districts. Below we calculatethe ull (predicted) costs to meet the program requirements and satisy the New JerseySupreme Court mandated standards.

3. COSTING OUT PRE-SCHOOL: METHODS AND EVIDENCE

3.1 Methodology

Te purpose o this costing out study is to calculate the cost (or minimum amount oresources) needed to achieve a specic level o child development through pre-school(Levin and Schwartz 2007). In this case, the specic pre-school program is one thatmeets the Abbott requirements.

It is important to clearly dene the pre-school program, particularly in relation toother costing out exercises. For the Abbott districts, we distinguish between the costs


23/54


o: (a) the regular pre-school program (6 hours over 180 days); (b) the regular programwith wraparound services at the end o each day (item (a) plus 4 hours o childcareover 180 days); and (c) the total costs o the ull program (item (b) plus 10 hours ochildcare over 65 days in the summer). ypically, child care programs are less costlythan pre-school programs: the ormer typically hire sta at lower pay rates, place lessemphasis on early childhood instruction, and are less likely to be sited in a public

school (Blau and Currie 2006). Overall, these childcare programs have less stringentstandards. Tereore, we calculate childcare costs and pre-school costs separately.

Tere are our possible approaches to costing out pre-school: the proessional judg-ment (resource cost) model; the successul district model; the evidence-based reormdesign; and the cost unction estimation (Downes and Stieel 2007). In the proes-sional judgment model, a group o experts convenes to establish a prototype educa-tion, listing the elements needed such as sta, instructional time, curriculum content,administrative overheads to achieve that education. Tese elements are then costedout. In the successul district model, the approach is to identiy schools or districtsthat are already providing the prototype education and then calculate the inputs re-quired to achieve the same results or other students. Te evidence-based reorm de-sign starts with a specic reorm (e.g., Perry preschool model) and then identies thecosts associated with the inputs required or that reorm. Finally, the cost unctionestimation method uses perormance measures (e.g., preschool quality ratings) andcurrent cost data to determine how much a given school would need to spend to ob-tain a given rating, taking into account the characteristics o the school and students.Here, we apply the last two approaches to calculate a baseline amount o unding perchild in pre-school. We then calibrate these baseline gures to account or variationsin district characteristics and demography.

We orego the proessional judgment model (PJM), which is the most popular method.18PJMs conducted or similar settings have yielded large variations in costs (Downes and

Stieel 2007). o some degree, the PJM model might overlap with the evidence-basedschool reorm design because the inputs required or successul pre-school are well-speci-ed in several pre-school interventions. Indeed, i the experts involved in the PJM panelsare provided with the research evidence, then such evidence might be determinative orpre-school, i.e. the research dictates which inputs are needed. More importantly, theinputs required or pre-school in this case are set down in the Abbott rulings.19

Essentially, all these costing out methods assume that: (a) current expenditures aresomewhat reective o need; (b) providers are in some sense maximizing educationaloutcomes given the resources available to them; and (c) a ailure to satisy quality stan-dards is primarily a consequence o inadequate unding (Baker 2005). Tese assump-tions seemprima acieplausible and they are implicit in our costing out exercises.

18 PJMs have been perormed primarily in K12 settings. Examples include: Augenblick and Myers 2002; Augen-blick and Silverstein 2003; and Odden et al. 2006. All o these examples ocus on K12 education.

19 Alternatively, without this research evidence, the experts may ail to reach a consensus: because the Abbott pre-school programs are relatively recent and there is not a well-developed history o student-level assessments theproessionals may not agree on how to relate pre-school inputs to outcomes.


24/54


Nevertheless, these methods have been criticized. Hanushek (2006) claims that:(a) there is no evidence that more resources will lead to better outcomes; (b) systemicreorms should be applied beore committing urther public unding; and (c) inputsare purposeully chosen with the goal o increasing expenditures. Each o these con-tentions has signicantly less merit when applied to investments in pre-school or theollowing reasons.

First, there is a good understanding o the best inputs to use, as well as the purposesand outcomes rom pre-school (see the ootnotes in the Introduction above). Tecorrelation between cost and quality is shown in Currie and Neidells (2007) study oHead Start. But leaving aside the well-documented achievement gains or children,there is research showing how: screening reduces subsequent health problems; howparental engagement improves school-amily relations; and how recognition and re-sponse programs ameliorate placement in special education. Without resources orscreening, parental engagement, and recognition and response programs, these ben-ets cannot be realized.

Second, all public service providers may have an incentive to over-identiy need,so that more unding is made available. However, in the case o pre-school where asizeable portion o services is provided through the private sector, this distortion maybe less signicant. In New Jersey, as in other states, there are private providers o pre-school, such that competition between providers will discourage them rom inatingneeds (see Henry and Gordon 2006). Also, there are many instruments to evaluateand monitor pre-school programs, such that genuine need can be determined.

Tird, this costing out model is based on the Abbott requirements, which predeter-mine the inputs rather than allowing providers to allocate inputs in ways that pur-poseully increase expenditures. Tis costing out exercise involves pricing out theseprescriptions under the assumption that they are implemented aithully.

Certainly, the disadvantages o the cost unction approach are that it treats the deliveryo education as a black box, i.e. it does not consider the process o education suchas modes o instruction or curriculum content. It assumes that centers are minimiz-ing costs and use all types o input. It also assumes that all inputs can be varied andthat any dierences in inputs are reected in their prices (see Blau and Mocan 2002;Gronberg et al. 2004).

But a key advantage to the cost unction approach is that we can use inormationon the many individual pre-school providers; we do not solely rely on district-levelrelationships which cannot identiy cost consequences attributable to within-districtheterogeneity. Moreover, in applying more than one method and relying on many

independent sources, we can provide a more objective estimate o costs. Tese separateapproaches amount to a series o sensitivity tests that should make the costs estimatesmore reliable (as advocated by Duncombe 2006).


25/54


3.2 Prior evidence

Tus ar, the our costing out methods have not been extensively applied to calculatean adequacy standard or pre-school. A state-level adequacy study that addresses thecosts o pre-school in Kentucky was undertaken by Lawrence O. Picus & Associates(2003a) but this study ocuses primarily on K12 schooling. It does not include a ull

costing out model or pre-school. In their study or Arkansas, Lawrence O. Picus &Associates (2003b) recommend expenditures o $100 million to expand the existingABC program by 22,300 places. However, this program is only 36 weeks per year,with class size o 20.20 Notwithstanding, there is a substantial body o related researchusing econometric methods and school reorm costing templates.

Several econometric studies derive cost unctions or childcare centers (not pre-schools). Using data on 266 childcare centers rom the Cost, Quality and Outcomesstudy, Blau and Mocan (2002) estimate a cost unction; parallel research was per-ormed by Mukherjee and Witte (1993) and Mocan (1997). Powell and Cosgrove(1992) estimate a cost unction or 206 children care centers across the U.S. Helburnet al. (2002) use data rom 116 pre-school age day care centers in Caliornia, North

Carolina, and exas. Zellman and Gates (2002) do so using data rom 59 childcarecenters across the U.S. military services.21 Many o the ndings rom these studies,discussed below, are quite plausible and may apply to pre-school.

otal costs are aected by the prices o inputs (Blau and Mocan 2002). However, therelationship diers depending on whether the center is non-prot (NP) or or-prot(FP). As wages go up by 10%, total cost increases by 3% (NP) or 8% (FP). As theprices o materials go up by 10%, total cost increases by 2% (NP and FP). As the priceo capital goes up by 10%, total cost increases by 0% (NP) or 2% (FP). Tis last nd-ing is unsurprising: non-prots oten rely on donated or cross-subsidized space, suchthat when the price increases their costs are not signicantly aected.

otal costs are also aected by the scale o the center. Larger centers are associatedwith lower unit costs, i.e. there appear to be economies o scale. But the eect is notdramatic. As the total hours o pre-school oered by a center goes up 10%, total costsgo up by 89% (Mocan 1997; Powell and Cosgrove 1992). Evidence collected usingother methods also suggests that economies o scale are not large.22

20 For Connecticut, Augenblick et al. (2005) used a proessional judgement panel approach to determine spendingor a program which would add on average $650 per student (applied to all students enrolled in a district regardlesso student age) or small districts, $813 per student in moderate districts, $1,144/student in large K12 districts,and $746 per child in K6/8 school districts.

21 However, the cost unction specication is linear and the control variables are o limited pertinence or NewJersey.

22 In other states the pre-school market is not highly concentrated. Ater over a decade o ree entry into the pre-school market in Georgia the median provider oers pre-school programs at only two sites (Levin and Schwartz,2007). In other states, no large enterprise has taken over the market. Tis revealed behavior suggests that disec-onomies o scale do occur; possibly, these are managerial diseconomies in operating multiple sites. Golin et al.(2003) estimate that unit costs will be virtually unchanged as a program expands rom 60% participation to 80%participation (unit costs might dier by 0.2%).


26/54


For-prot centers report lower costs per child (unadjusted or quality). Powell and Cos-grove (1992) estimate that private centers have costs that are 9% lower than publicones.23 Tis dierence is reected in wage payments: non-prots pay 918% more thanor-prots; overall, they spend 78% o their budget on sta, compared to 62% by theor-prots (Helburn 1995). Centers that are either Head Start or enroll children withspecial educational needs have variable costs that are 25% higher (Blau and Mocan

2002).

Pre-school programs or three-year olds are more expensive than or our-year olds(Powell and Cosgrove 1992). Younger children typically need more supervision, whichmeans smaller class sizes. Zellman and Gates (2002) report that the costs or toddlersare 32% higher than or pre-school aged children and 90% higher than or school-agechildren.24

Te important relationship or public policy is the one between unding amounts andthe quality o the program (Hanushek 2006). Blau and Mocan (2002) calculate thatin order to score one-unit higher on the Early Childhood Environment Rating ScaleRevised (ECERS-R), total costs must be 5.6% higher. Te other studies use qualitymeasures that are less ideal than the ECERS-R ratings, but they too nd a positiveassociation between costs and outcomes.25

Our analysis or the evidence-based school reorm approach draws on two separatelines o research.

First, there are programs that have been identied as eective using high quality re-search methods. Tese programs include the Perry High/Scope Program and the Chi-cago Child-Parent Center Program. Te ingredients in these programs are known andthey can be priced out or the New Jersey Abbott districts.

Second, there are a number o costing templates or pre-school.26

Tese templates ol-low a similar protocol o: determining enrollment; identiying quality measures suchas class size, teacher qualications, and curriculum content; calculating baseline costs;calculating upgrade costs; and modeling program expansion. However, these templatesare either specic to a state or only set out the most basic cost structures. In the lattercase, their actual dollar values are illustrative, rather than calibrated to actual budgetsand adjusting or district characteristics and demography. For comparison purposesthese templates are included in our evidence-based school reorm design analysis.

23 Zellman and Gates (2002) nd that the Department o Deense acilities were more cost-ecient than those run

by private providers. Settings matter even or model programs: the Abecedarian program is one-sixth more expen-sive in a public school center than a child care development center (Masse and Barnett, 2002).

24 Mocan (1997) nds modest evidence o economies o scope, such that it should be lower cost to oer programsor 3-year olds and 4-year olds at the same site.

25 For example, Helburn et al. (2002) nd a positive but quite modest association between ees and the Family DayCare Rating Scale. Also, costs are higher i class sizes (a proxy measure o quality) are smaller: an increase in thesta-child ratio o 10% raises costs by 35% (Blau and Mocan 2002).

26 See Barnett and Kelley 2002; Barnett and Robin 2006; Golin et al. 2003; and Stebbins and Hanson Longord2006; Yonce et al. 2006.


27/54


4. ESTIMATING THE COSTS OF PRE-SCHOOL

4.1 Cost function method

Data and model specications

We estimate cost unctions or pre-school private centers and public programs sepa-rately. As applied to K12 school districts, the generic specication or cost unctionsis cost per child as a unction o enrollments, input prices, attributes o providers, localdemographics, and provider quality (Duncombe 2002).

We merge three datasets to estimate these cost unctions. Budgetary data are takenrom Department o Education budget statements; we used data or 200607 orthe private centers but or 2003 through 2007 or the public programs. For the mostrecent years these are budget statements rather than actual expenditures, but the twoare likely to be highly correlated (given the high level o accountability to ensure thatproviders do spend their allocations). Tese data include the unit costs or each indi-vidual private provider but only district-aggregated unit costs or the public providers.

Inormation on private center and district-level enrollments are also available. o thesedata we merge inormation on the quality o the providers. Tese data are rom theEarly Learning Improvement Consortium (ELIC) evaluations o classrooms (Frede etal. 2007; Esposito Lamy 2005). Te evaluations, which are part o the state account-ability system or the Abbott pre-schools, use the ECERS-R scale, which is widely rec-ognized as a valid measure o pre-school quality (Blau and Mocan 2002; Burchinal etal. 2000). Finally, we augment these data with district-specic demographic variablesrom the Census.27

Our specications are close to the estimation by Blau and Mocan (2002). We estimatethe unction using ordinary least squares, which has the advantage o being robust.28Te dependent variable is the log o each centers average per pupil costs or the 6-hour

educational portion o the pre-school day lasting an academic year. Te independentvariables are teacher wages (input prices), provider/district attributes, and local demo-graphic variables. We estimate a log-linear unction to account or the distributiono the dependent variable (which is censored at zero). Coecients on independentvariables may thereore be considered as percentage changes in costs. We also includedemographic variables as controls or unobserved local heterogeneity. Based on dataavailability, the specications or public and private centers dier in terms o provider/district attributes. o ensure a sucient sample size we pool data across the availableyears: complete data is available on classrooms in 210 private centers and 535 public

27 We apply several alternative approaches to estimating a cost unction. Each suggests that centers would benetrom additional unding to improve quality. Based on state-level data and quality benchmarks rom the NIEER2006 yearbook, we nd that spending is higher or states meeting 8 o the 10 quality benchmarks (Levin andSchwartz 2006). Simple dierences are given in Appendix able A1. (For both items where there is no correlation,the nding is plausible: Early et al. (2006) nd that a BA alone is not a strong indicator o program quality).

28 It does not control or the endogenous determination o quality with costs, as modeled by Duncombe (2006)and others. However, Blau and Mocan (2002) report that instrumental variables estimation, a method they usedto circumvent the potentially circular relationship between quality and cost, did not aect their results (as doesImazeki 2006).


28/54


programs.29 Frequencies or the variables are reported in Appendix able A4 or theprivate centers and Appendix able A5 or the public programs.

In act, there are a number o reasons why this estimated relationship between qualityand cost is likely to be an understatement. First, i relatively inecient districts havelower perormance and higher spending, then the coecient on the quality measure

will be biased downward (Imazeki 2006). Second, coecients will be biased down-ward i there is measurement error or multicollinearity. Tird, ECERS-R ratings maynot capture all dimensions o the outcomes desired rom the program; the inspectionsmight also not precisely identiy quality.30 Finally, the Abbott pre-schools are improv-ing in quality such that it is hard to identiy dierences in quality.31 ECERS-R scoresinclude some process measures (e.g. in regard to acilities) such that a centers score isunlikely to go down year-on-year; low quality provision is thereore not ully reectedin the scores.

Cost function estimates

For private centers the cost unction is reported in able 4. Tis specication explains

52% o the variance in costs. We note several key ndings; these indicate that ourresults correspond well with the existing evidence. First, higher quality pre-schoolingcosts more: a center would need to spend 2% more to raise its ECERS-R score byone unit, holding all other things constant. Tis is lower than the 6% ound by Blauand Mocan (2002) but our sample is rom a more homogeneous set o providers.Second, pre-schools in labor markets where teachers wages are higher have higher ex-penditures. Tis relationship is particularly strong or assistant teacher wages: or leadteachers, increasing wages by $1000 would raise costs by 2%; or assistant teachers,the same increase would raise costs by 14%. Tird, the link between costs and scaleis weak, suggesting that pre-schools with ewer Abbott classrooms do not necessarilycost more. Te coecient on Abbott-eligible enrollment is not signicant, as with thecoecient on whether the center operates multiple sites. However, costs are lower i

the private center enrolls any children who are not eligible or Abbott subventions.32

Tus, larger centers probably do have lower costs, but the savings are not large. Finally,or-prot centers have higher costs.

For public pre-school providers the cost unction is reported in able 5. Again, the de-pendent variable is the log o the average cost per child in a given academic year or the6-hour educational portion o the pre-school day or the length o the academic year.Tese results are broadly similar to those or the private centers reported in able 4.

29 Te cost unction or the public programs is less well-determined than or private providers. Costs are only avail-able at the district level, whereas ECERS-R quality measures are per center (or per classroom). Furthermore, there

are only 31 Abbott districts. Tis motivates our decision to pool the sample across 20032007. Across some years,it is necessary to interpolate missing observations. We use averages rom the beore and ater year values.30 Although this eect is not strong: the year-on-year correlation o ECERS-R scores or centers that are inspected

twice in two years is 0.4 (p-value


29/54


Table 4Cost function: Abbott district private pre-school centers

Log average cost (per child)

Coeff. (SE)

Programquality:ECERS-Rscore 0.0196 (0.0090)*

Input prices: Leadteacherwage(000s) 0.0215 (0.0074)**

Assistantteacherwage(000s) 0.1367 (0.0159)**

Center-level characteristics:

Abbott-eligibleenrollment 0.0001 (0.0001)

Anynon-Abbotteligibleenrollees 0.0348 (0.0132)**

Multi-sitecenter 0.0304 (0.0227)

For-protcenter 0.0384 (0.0144)**

Leasefacilities 0.0100 (0.0132)

Demographic controls:

Percentmarried 0.1137 (0.1126)

Percentcollegeeducated 0.1161 (0.2445)

Averageincome(000s) 0.0303 (0.0390)

Population(00000s) 0.0386 (0.0125)**

R-squared 0.52

N (classes in centers) 210

Notes: ECERS-Rscorestakenfrominspectionsin2005and2006.Thedependentvariableisapre-school

providerslogaverageperchildannualcostfortheDOE-funded6-houreducationalcomponentofthe

schooldayfor180daysperyear.Demographiccontrolsaremeasuredatthedistrictlevel.Constanttermis

included.*signicantat5%;**signicantat1%.

An improvement in the quality o the pre-school program (a one unit increase in theECERS-R score) is associated with a 2% increase in average costs. An increase in thewages o assistant teachers o $1,000 is associated with a 1% increase in costs; butthere is no statistically signicant eect on costs i lead teachers wages are increased.33Te district-level characteristics show again that there is no signicant impact on costso district size; scale eects are not strong. Tere is no correlation between unit costsand the percentage o three year-olds enrolled or the percentage in private centers. Tepercent o Head Start enrollees is associated with higher public program costs.

o check or robustness we estimated the public program cost unction separately oreach year.34 able 6 reports key results rom separate cost unctions or the samples

rom 2003 to 2007 (the specications are identical to those in able 5 except that

33 Tis result (higher pay does not cost more) is counter-intuitive, but it may be explained. First, there is a very highcorrelation between the two wage values across districts (this multicollinearity reduces our ability to identiy cor-relations). Second, it may reect a very high degree o substitutability across stang types. Tird, centers may usesta whose wages are not included in these equations because o missing data.

34 For private centers we included scores rom 2006 to improve the sample size (but only or centers which werenot inspected in 2007). A dummy variable to distinguish these scores is not statistically signicant. We thereoreinterpret this to mean that year on year uctuations in resources do not substantially inuence quality.


30/54


there are now no year dummies). Te eect o program quality is strongest in 2003,with a 1 unit increase in ECERS-R rating associated with costs 4.6% higher. Tis isplausible: as centers and programs develop over time and the rating standards becomeknown, pre-school provision will more closely track this rating. able 6 also showsthat the relationship between proportion o 3-year olds and costs is typically positive,conorming to extant evidence. Te null result in the pooled regression is driven byone year (2006), and this is a year or which our measure o the proportion o 3-yearolds is interpolated. Hence, we might anticipate that serving more 3-year olds willentail higher per-child expenditures.

Predicted pre-school costs

Based on the above cost unctions, we calculate the costs o well-planned and high-quality pre-school or private centers and public programs. Tese predicted costs as-sume that the current expenditures should be increased in several ways.35

35 We cannot use the model to generate predicted values o costs or each district. Tese cost unctions are only asample o all centers and programs (as well as being spread over multiple years).

Table 5Cost function: Abbott district public pre-school programs

Log average cost (per child)

Coeff. (SE)

Programquality:ECERS-Rscore 0.0159 (0.0078)*

Input prices: Leadteacherwage(000s) 0.0021 (0.0012)

Assistantteacherwage(000s) 0.0094 (0.0015)**

District-level characteristics:

Pre-schoolenrollment(000s) 0.0001 (0.0000)

Pre-schoolenrollmentsqd(000s) 0.0000 (0.0000)*

Percent3-yearolds 0.0930 (0.1517)

Percentprivatecenterenrollees 0.0477 (0.0325)

PercentHeadStartenrollees 0.8900 (0.1486)**

Demographic controls:

Percentmarried 0.6926 (0.1491)**

Percentcollegeeducated 1.3531 (0.3097)**

Averageincome(000s) 0.0124 (0.0046)**

Population(0000s) 0.0011 (0.0006)

R-squared 0.36

N (classes in centers) 535

Notes: ECERS-Rscorestakenfrominspectionsin20032007.Thedependentvariableisapre-school

providerslogaverageperchildannualcostfortheDOE-funded6-houreducationalcomponentofthe

schooldayfor180daysperyear.Demographiccontrolsaremeasuredatthedistrictlevel.Fouryear

dummies(20042007)andaconstanttermareincluded.*signicantat5%;**signicantat1%.


31/54


First, program quality must be raised directly. We assume that the average ECERS

quality o each center would need to be raised to six; this is the mid-point o the scorehigh-quality programs receive. Raising the average ECERS-R score to six equates to anincrease o 1.07 units per private center and 1.3 units per public program. Costs willbe higher by 2.10% and 2.07% respectively.

Second, we also assume that salary costs would need to increase to ensure the Abbottpre-schools hire well-qualied, experienced teachers. o estimate the dierences in sala-ries across teachers within New Jersey we use individual-level administrative data rom200607 (NJ DOE). We compare Abbott pre-school teachers with kindergarten teach-ers rom across the state. Regressing wages on experience, race, and education, we ndthat there is no dierence between the two types o teacher in how these characteristicsinuence pay. For each job a year o experience is associated with an increase in pay

o 2.4% (see Appendix able A7, columns 1 and 2). Tis is reassuring because theseteachers operate on the same pay scale.36 As noted above, these NJ DOE data showone signicant dierence between the Abbott pre-school teachers and kindergartenteachers: the ormer have on average 9.5 years o experience compared to 12.1 or thelatter group. Tis in part reects the newness o the Abbott pre-school program as newhires typically have less experience than the average worker. As the Abbott pre-schoolteachers tenure tends toward the mean, their wages will increase accordingly. Te mostplausible assumption is that their tenure will equal that o pre-school teachers in otherdistricts across the state, which is 11.3 years. Tereore we estimate that the steady stateexperience levels o Abbott pre-school teachers are 1.8 years below those o other pre-school teachers. Tis translates into salaries that would be 4.32% higher than currently.Increasing lead teacher pay by 4.32% raises unit costs in private centers by 0.41% but

does not increase unit costs in public centers. I assistant teacher pay goes up by 4.32%unit costs will rise by 1.3% in private centers and by 0.72% in public centers.

36 Te raw average wages across the jobs are signicantly dierent: Abbott pre-school teachers earn $55,504, com-pared to $57,204 or kindergarten teachers. When we estimate a pooled wage equation or both jobs, there is nosignicant dierence in wages or each job (see Appendix able A7, column 3). However, this equation does notadjust or working conditions properly and it controls or dierences in teacher experience.

Table 6Cost functions: Abbott district public pre-school programs by year

Log average cost

2003 2004 2005 2006 2007

Programquality:ECERS-Rscore 0.0460(0.0194)*

0.0098(0.0078)

0.0218(0.0114)

0.0123(0.0189)

0.0009(0.0075)

Percent3yearolds 0.7123(0.2734)*

0.0446(0.2192)

1.4773(0.4319)**

1.8116(0.7870)*

0.8010(0.3305)*

N 103 109 114 119 90

Notes:EstimationincludesallvariablesasperTable5(exceptyeardummies).

*signicantat5%;**signicantat1%.


32/54


Tird, the pre-school program is intended to be universal. Expanding enrollments tocover a higher proportion o children might be expected to inuence costs (as mighta policy change to promote larger centers or programs). wo eects need to be ac-counted or. One is economies o scale (i.e. how average costs change as rms growlarger). However, neither the extant evidence nor our cost unction estimates showsignicant economies o scale. Indeed, almost hal o all providers in the Abbott dis-

tricts oer one or two classrooms; only one-in-seven oers more than six classrooms(Fuller and Wright 2007). Moreover, the Abbott districts may have more capacity:they are budgeted to serve more students than they actually do enroll (Applewhite andHirsch 2003). But the second eect o expansion is that more enrollments will neces-sitate more teachers and other personnel and higher wages will be needed to attractthese workers. Based on estimates o labor supply elasticity, to expand the program by4% would require an increase in wages o 5% (Blundell and MacCurdy 1999). Tisexpansion would increase costs by approximately 3%.

Tereore, the unit cost o well-planned, high-quality pre-school in the Abbott dis-tricts is anticipated to be above what is currently spent by 6.8% or private centersand 5.8% or public centers. Tereore, using the gures in able 2, average unit costsacross the Abbott districts should have been $13,441 or private centers and $11,529or public programs in 2006. For uture academic years, an ination actor o 4%must be applied.

For many reasons, we believe this is a conservative gure. Te estimation procedurelikely understates the coecients on quality and input price eects. Notably, our datado not yield evidence on other pressures on resources such as: capital costs (but seeBlau and Mocan 2002); K12 alignment; accountability systems; and special educa-tion. On the last item, we note that pre-school special education enrollments are verylow and that the biggest inuence on the costs o special education is placement ininclusive or sel-contained settings. Finally, i additional enrollments are younger than

average (aged 3 instead o 4), this might raise costs. Our estimations give modest sup-port to this assumption, but the extant evidence gives strong support. o maintain theconservative nature o our estimates, we do not include a cost premium or serving ahigher proportion o three year-olds.37

o illustrate the size o the understatement, we consider two domains where our re-sults may be conservative. One is the cost o quality, i.e. what it costs to raise theaverage ECERS-R score. Our estimates are that an additional 2% in unding wouldbe needed. An alternative, plausible estimate is that rom the 2003 equation (the rstyear o available data and so closer to the underlying relationship between costs andquality) or that rom Blau and Mocan (2002). Tese estimates would mean undingshould be increased by an additional 3 or 4 percentage points. Te second domain

where our results are conservative is or acilities. In our calculations we assume no ad-

37 We do not investigate amily commitments. Low-income amilies are less able to contribute in-kind resources(books, materials, or time) even as these are oten an important contribution to making pre-school eective. Forexample, Muenchow and Scott (2005) estimate in-kind resources at 30% as large as state contributions. In addi-tion, we do not include any transport costs. Tese are conventionally estimated at over 7% o per pupil expendi-tures (Verstegen, 2006).


33/54


ditional resources are required. Instead, we might assume that expenditures on capitalacilities are sub-optimal to the same extent as expenditures on teachers (although inmost public education systems capital spending is usually more decient than spend-ing on teachers). I we assume capital acilities should be upgraded to the same extentas wage payments that would add another 1 or 1.5 percentage points to the requiredunding.

4.2 Pre-school program model

Te second approach to costing out pre-school is the evidence-based school reormdesign. Tis model is based on evidence-based school reorm costing templates, as ap-plied to K12 schooling (Odden et al. 2006). Te ad

Cost of Preschool in NJ 2007

Documents