Does Title I Increase Spending and Improve Performance? — IESP Working Paper 09-09

Summary

IESP Working Paper #09-09, August 2009, by Meryle G. Weinstein, Leanna Stiefel, Amy Ellen Schwartz and Luis Chalico, asks whether Title I increases spending and improves performance, using New York City public school data. Title I, part of the 1965 Elementary and Secondary Education Act, allocated $13.9 billion in FY 2008, reaching almost 17 million students, and the American Recovery and Reinvestment Act of 2009 allocated an additional $13 billion over two years. Using a regression discontinuity design, the authors find that Title I lets high schools spend more on direct services to students and improve pupil-teacher ratios, while elementary and middle schools increase spending less. They find that additional Title I spending does not improve achievement and may reduce average test scores in elementary and middle schools.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Issuer: Source documents
Document type: Does Title I Increase Spending and Improve Performance? — IESP Working Paper 09-09
Case: Does Title I Increase Spending and Improve Performance? — IESP Working Paper 09-09

Cite this document

“Does Title I Increase Spending and Improve Performance? — IESP Working Paper 09-09.” Original at files.eric.ed.gov. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/g-stiefel-l-schwartz-a/, SHA-256 73d10e2e8fd8157cd5378f56cb580f0a799506b56ee57759597b83994563c4e3.

Full text

IESP

Does Title I Increase Spending
and Improve Performance?
Evidence from New York City

Working Paper #09-09
August 2009

Meryle G. Weinstein, Ph.D.
Research Scientist and Assistant Director,
Institute for Education and Social Policy

Leanna Stiefel
Professor of Economics and of Education Policy,
Wagner and Steinhardt Schools

Amy Ellen Schwartz
Professor of Economics, Education and Public Policy
Director, Institute for Education and Social Policy
Wagner and Steinhardt Schools

Luis Chalico,
Ph.D. Candidate, Wagner School
Direct correspondence to Leanna Stiefel, Professor, Institute for Education and
Social Policy, 665 Broadway, Suite 805, New York, NY 10003. Email:
leanna.stiefel@nyu.edu. Any opinions expressed are those of the authors.

Editorial Board

The editorial board of the IESP working paper series is comprised of senior
research staff and affiliated faculty members from NYU's Steinhardt School of
Culture, Education, and Human Development and Robert F. Wagner School of
Public Service. The following individuals currently serve on our Editorial Board:

Amy Ellen Schwartz
Director, IESP
NYU Steinhardt and NYU Wagner

Sean Corcoran
NYU Steinhardt

Cynthia Miller-Idriss
NYU Steinhardt

Leslie Santee Siskin
NYU Steinhardt

Leanna Stiefel
Associate Director for Education Finance, IESP
NYU Wagner

Mitchell L. Stevens
NYU Steinhardt

Sharon L. Weinberg
NYU Steinhardt

Beth Weitzman
NYU Steinhardt
Abstract

Since its inception as part of the 1965 Elementary and Secondary Education Act
(ESEA), Title I has provided the largest amount of federal funding aimed at
improving the academic achievement of poor children. In this paper, we examine
the impact of Title I on school spending and school performance, using New
York City public school data. Based on a regression discontinuity design (RD)
with panel data, and including separate analyses for elementary/middle and high
schools, we estimate local average treatment effects of Title I.

Title I provides additional funding for schools serving high concentrations of
poor children, but there are few curricular or programmatic constraints.
Unfortunately it is possible that Title I funds supplant state or local funds,
resulting in muted net impacts on spending and student outcomes. At the same
time, the success of Title I in improving test scores offers important insight into
the effectiveness of school-based compensatory funding in general, such as
weighted student funding within districts or state compensatory aid across
districts.

Overall, the results indicate that Title I changes the mix of spending, enabling
high schools to significantly increase the amount of money they spend on direct
services to students and to improve their pupil-teacher ratios (while reducing
experienced teachers). Elementary and middle schools do not increase spending
as much, which is consistent with our finding that state compensatory education
funds may be supplanting some Title I funding in schools. Since schools just
below the Title I cutoff are similar to those just above the cutoff, this finding may
be an equitable, albeit unintended result.

Finally, additional Title I spending does not improve the achievement of students
and may even reduce school-wide average test scores in elementary and middle
schools. These effects for both spending and scores seem to increase with the
length of time schools are Title I eligible and to be stronger for ones that are
always Title I eligible compared to those that go in and out of eligibility.
1. Introduction

Since its inception as part of the 1965 Elementary and Secondary Education Act (ESEA),

Title I has provided the largest amount of federal funding aimed at improving the academic

achievement of poor children. As the fourth largest of all federal programs for low-income

children (behind Medicaid, food programs, and Temporary Assistance to Needy Families), Title

I allocated $13.9 billion in FY 2008, reaching over half the public schools in the country and

almost 17 million students. 1,2 The recent federal economic stimulus bill (the American

Recovery and Reinvestment Act of 2009 or ARRA) allocates an additional $13 billion to Title I,

over two years. Given the amount of money available, should we expect more spending in

schools eligible for Title I funding and better academic performance as a result? The answer is

not obvious. State or local funds may decline in schools receiving federal Title I funding,

resulting in little or no increase in total spending, and it is notoriously difficult to improve

student performance of poor children.

Although numerous previous studies estimate the effectiveness of Title I funding in

improving student achievement, many suffer from biases due to unobserved differences between

Title I and ineligible schools. Further, these studies evaluated the “old” Title I, in which Title I

distributed funds based on a mixture of student performance and poverty, rather than solely

based on poverty, as is done today. Finally, few studies have looked at the effects on high

schools and their students.

This paper improves on these earlier studies in several ways. First, we examine the

impact of Title I eligibility on school spending overall and on state compensatory education

1
http://www.ed.gov/programs/titleiparta/index.html
2
The numbers of participants has increased faster than the amount of funding or numbers of districts or schools
covered because school wide programs have replaced targeted assistance programs. In school wide programs, all
students are counted as Title I whereas in targeted assistance programs, only those receiving assistance are counted.
1 WP #09-09
funds separately, to gain insight into possible supplanting. Both spending and outcome analyses

use a regression discontinuity (RD) design with panel data. (Alternate specifications use borough

and school fixed effects.) Separate analyses for New York City (NYC) elementary/middle and

high schools provide different estimates by level. This design provides credible local average

treatment estimates of the effects of Title I on both spending and outcomes and controls for

additional unobservable characteristics with fixed effects. To our knowledge, only one other

paper combines an RD design with panel data and fixed effects (Cellini, Ferreira and Rothstein,

2009).

Since Title I is the largest federal compensatory aid program to education, its

effectiveness largely determines the success of federal spending in improving learning outcomes

for poor children. The 2001 No Child Left Behind (NCLB) legislation relies on Title I funding to

provide the federal resources for struggling schools. Further, the 2009 economic stimulus

package channels a large portion of new education funds through Title I, making it especially

important to understand the effectiveness of Title I in targeting funds and improving outcomes at

this time.

Additionally, the implications of our findings reach far beyond the federal program itself.

Most states also use compensatory aid programs to target additional resources to schools and

districts with low achieving and/or high poverty children. Like Title I, few detailed

programmatic or curricular specifications are attached. 3 Thus, analyses of Title I offer insights

that may be relevant to these programs as well. Further, most school budgets leave little

3
Title I is a funding mechanism rather than a specific program or policy aimed at educating poor students. In its
early years of implementation, school districts assigned staff and services rather than unrestricted dollars to schools
(Chambers et al., 1993; Goertz, 1988). The Improving America’s Schools Act (IASA) of 1994 gave principals
greater discretion over the use of their funds in return for greater accountability; NCLB further increased the
accountability measures and added sanctions for schools failing to meet these measures. The only current
requirement for how Title I funds are spent, in addition that they be spent for services to poor children, is that
schools must set aside ten percent of their Title I dollars for professional development and one percent for parental
involvement programs.

Does Title I Increase Spending and Improve Performance? 2
flexibility for principals in how money is spent. Funds from Title I and other compensatory

spending programs, however, are often unrestricted as to type of inputs, providing principals

with precious discretional dollars. If schools (and principals) know best how to educate their

students and how to use funds most effectively, these resources should be quite productive and

the dearth of evidence showing such productivity in previous studies is disappointing

Finally, understanding the effect of Title I funding on student performance will shed light

on the possible impacts of new district finance proposals such as “Weighted Student Funding

(WSF)”. 4 As with Title I, WSF allocates money to schools based on the characteristics of their

students, rather than allocating teacher positions, as historically is the case for intra-district

funding. In NYC, WSF aimed to give principals substantial control over budgets, with few

constraints on how money was used. Therefore, credible evidence on the effect of Title I on

student performance will prove useful to federal and state policymakers and provide evidence

relevant to changes in how schools are funded.

The findings presented in this paper are relevant to to other contexts, as its public schools

educate children that are in many ways similar to urban school children elsewhere. Indeed, while

some areas of NYC are quite unique (e.g., Times Square, Wall Street), there are large areas that

are representative of urban settings around the country (e.g., Queens, Staten Island). More

generally, a study of NYC schools and students provides valuable insight into the issues and

constraints faced by large urban schools in other states. NYC is not as different from most other

large, urban districts as is often thought. Recent research has shown striking similarities in intra-

district resource distribution (Rubenstein et al., 2007), efficiency and racial test score gaps for

example.

4
Other large cities have adopted this system (Edmonton, Canada, Seattle, Cincinnati, and Houston) and the
Fordham Foundation gathered a broad group of educators in support of such a school finance system in 2006
(Fordham Foundation, 2006).
3 WP #09-09
The rest of this paper is organized as follows. First, we discuss key features of the Title I

program, including how Title I funding is allocated, both nationally and locally, highlighting

features critical to modeling the effects of the program. We then review previous literature on

Title I funding and its impact on school spending and student achievement. Next, we describe

our methods, followed by data, empirical results, discussion and further exploration of results,

and, finally, conclusions.

2. Key Features of Title I

The Allocation of Title I Funds in General

The allocation of Title I funds from the federal government to local schools follows a

complicated process. First, each year Congress sets the total amount of funds available for

distribution by the U.S. Department of Education (USDOE) (Figure 1). USDOE then uses

several grant formulas based on the number of poor children to determine each local education

agency’s (LEA) share. Basic grants provide all LEAs with an allocation per poor student,

enabling almost all LEAs to be eligible for at least some funding. Other Title I formulas provide

additional funds based on increasing numbers and percentages of poor school-age children,

thereby giving large and high-poverty LEAs increasingly more money than small and low-

poverty ones. The total LEA allocation flows through each state’s Department of Education,

which then distributes the monies to the LEAs. Finally, states can keep one percent of the total

allocation for program administration to provide additional resources to schools in greatest need

of improvement.

LEAs allocate funds to their schools as follows. First, LEAs set a school “Title I

eligibility cutoff” for poverty, equal to the average child poverty rate in the district, below which

Does Title I Increase Spending and Improve Performance? 4
schools will receive no Title I funds. Then the LEA sets the Title I per-pupil amount of funds,

which is the total district allocation divided by the number of poor students in schools above the

Title I cutoff, including students from both public and non-public schools. Importantly, in

eligible schools all poor students are eligible to receive services. Conversely, no students in

ineligible schools are eligible to receive services, even if they are poor. 5

According to Title I regulations, all schools with at least 75% poverty must receive funds

and schools with less than 35% poverty cannot receive funds. 6 To reiterate, not all schools

receive Title I money, even if there are poor students at the school. 7

The Allocation of Title I Funds in New York City

Since 2000, NYC has spent over $12 billion dollars annually on public schools, with Title

I funding providing approximately five percent of the total (see Table 1). The amount of Title I

revenue NYC received increased between 1996-97 and 2003-04, with an especially large

increase (from $474 million to $638 million) between 2001-02 and 2002-03. The new federal

economic stimulus bill (ARRA) will send an additional $350 million to New York City in both

FY2010 and FY 2011 (NYCDOE, 2009), making this is a good time to assess the impact of Title

I on spending and performance.

5
If an eligible school has a high enough concentration of poor students to be a “school wide program,” and applies
for and receives school-wide status, then all students, whether they are poor or not, are eligible to receive services.
Since 1978, Title I schools have been able to adopt comprehensive schoolwide programs. In 1988, schools with a
poverty rate of at least 75% where able to use Title I funds for these programs while currently schools with 40% or
more poverty are able to have schoolwide programs. Nationally, the number of schoolwide programs has tripled
between 1994-95 and 2004-05 (National Assessment of Title I). In New York City, the number of Title I schools
approved for school-wide programs has also increased over time. Among elementary and middle schools, 40.5% in
1997 to 66.1% in 2003, among high schools 10.6% to 57.1%.
6
www.ed.gov/policy/elsec/leg/esea02/pg1.html
7
Some schools that do not have Title I status receive some funds for homeless students and those in foster care,
regardless of the poverty rate at their school. These are not usually large because incidence of children with these
characteristics is not large and most already attend schools that are Title I eligible. On average, Title I schools
received approximately $722,120 in 2003, with the top 10% of Title I schools receiving over $1.1 million. Non-
Title I schools, on average, received $7,870.

5 WP #09-09
An important feature of NYC, relevant to the modeling of Title I impacts, is that the city

allocates Title I funds to five counties (called boroughs in NYC 8 ), all of which are within its

borders. 9 This means that each borough receives a share of the city’s Title I money in proportion

to the poverty count in its own borough (as of the previous year). For most of the period of this

study (through 2003), the school Title I poverty cutoffs were the same across all boroughs,

resulting in a per-pupil Title I amount that differed by borough. (That is, the uniform cutoff

makes the eligible children a higher proportion of poor children in some boroughs compared to

others.) 10 For example, while the 2002-03 citywide per-pupil Title I amount was $948.17, it

varied from a low of $763 per child in the Bronx to a high of $2,194 per child in Staten Island

(NYCDOE, 2002). In terms of modeling impacts of Title I, this feature means that controls for

boroughs or schools may be needed.

NYC, like other school districts, also receives compensatory education funds other than

Title I. New York State provides Pupils with Compensatory Needs (PCEN) funds, which are

intended for programs for poorly performing students, particularly those students who are

English Language Learners (ELL), regardless of poverty status. While the total amount available

to NYC is less than half that of Title I, PCEN may still have an impact on the amount of total

compensatory funds available for schools (see Table 1 for comparisons between Title I and

8
Manhattan, Brooklyn, Bronx, Staten Island and Queens.
9
Prior to 1999, all Title I funds were allocated to counties rather than individual LEAs. The allocation process was
changed so that high-poverty LEAs in low-poverty counties received a more equitable amount of funding (Natriello
& Dill, 1999). NYC, however, differs from other LEAs in that it consists of five counties embedded within one
school district instead of a number of LEAs embedded within one county. Even though it is no longer required to do
so, NYC continues to allocate funds based on counties
10
Even after 2003, when the cutoffs by boroughs differ, the per pupil amounts also differ since the cutoffs were not
set to make them equal.

Does Title I Increase Spending and Improve Performance? 6
PCEN funding amounts) and thus it is necessary to look at this other compensatory aid in order

to understand patterns of total resources available to improve outcomes. 11

3. Previous Literature on the Effects of Title I on Spending and Outcomes

In this section we first review the literature on the impact of Title I funding on school

resources and then on student achievement.

Impact on School Resources

From the early authorizations, ESEA legislated that state and local funds must be

comparable between Title I funded and non-Title I funded schools and that Title I funds must

supplement not supplant state and local funds. 12 Enforcement is limited by the difficulty of

identifying the amount a school would have received in the absence of Title I funding.

Interestingly, although economic theory suggests that local school districts would fully substitute

intergovernmental funding for their own, there is substantial evidence of a “flypaper” effect,

whereby funds from a higher level of government “stick” where they are allocated (that is overall

spending increases). (See Hines and Thaler, 1995; Fisher and Papke, 2000; and Knight, 2002 for

a review of theory and empirical results on the flypaper effect.) Thus, understanding whether

(and how much) intergovernmental aid increases spending is an empirical question.

11
Anecdotal evidence suggests that PCEN funds in New York City have been used to “compensate schools with
insufficient numbers of low-income students to be eligible for Title I funds with additional PCEN funding.” See
correspondence between Noreen Connell, Executive Director of Educational Priorities Panel and Beverly Donohue,
Chief Financial Officer, New York City Board of Education, April 9, 1999,
www.edpriorities.org/Pubs/Opinion/Letters99_PCEN.4.9_Print.html, accessed October 12, 2006.
12
http://www.ed.gov/legislation/ESEA/sec1120a.html. SEC. 1120A. FISCAL REQUIREMENTS. (A) Except as
provided in subparagraph (B), a State or local educational agency shall use funds received under this part only to
supplement the amount of funds that would, in the absence of such Federal funds, be made available from non-
Federal sources for the education of pupils participating in programs assisted under this part, and not to supplant
such funds.

7 WP #09-09
Several researchers have employed quasi-experimental designs to disentangle the causal

impact of intergovernmental aid on districts and schools. Gordon (2004), using a national

sample of school districts, took advantage of a natural experiment arising from a switch in the

basis for eligibility for Title I funding from Census poverty data from 1980 to 1990. Updated

Census numbers provided an exogenous change in federal revenue for a large number of districts

in the United States. Gordon found that in the short run spending increased with almost no

offsetting reductions in state or local funds. Over the longer term, however, spending remained

flat due mostly to local offsetting responses. That said, Gordon did not examine the different

impacts on schools within large districts. Her results imply, however, that in the short run within

districts, Title I schools may receive additional funds with no reductions in funds for non-Title I

schools, but in the long run when district spending is flat, net school spending should not

increase.

Two studies focus on the relationship between Title I and total school spending -- that is

at the school rather than the district level -- using a regression discontinuity (RD) design. Van

der Klaauw (2008) used cross-sectional data from three years of NYC elementary and middle

schools and found no effect on school-level spending. Van der Klaauw, however, was limited to

three non-continuous years of data between 1993-94 and 2000-01 and did not use panel data

methods. Further, during one of those years (1993-94), non-Title I schools received money for

poorly performing Title I eligible students, therefore interfering with the regression discontinuity

design, which in its purest form, is based on a clear cutoff for eligibility that denies non-Title I

eligible schools any money.

Matsudaira et. al (2006) also used an RD design to study the impact of Title I funding in

a large northeastern city. They find that while federally funded revenues increased in Title I

Does Title I Increase Spending and Improve Performance? 8
compared to non-Title I schools, total direct expenditures did not differ. Their data were limited

to two years (2001 and 2002) and did not use longitudinal methods to address potential

unobserved differences between schools. In addition neither Van der Klaauw nor Matsudaira

included high schools.

Impact on Student Academic Performance

Evidence on the effectiveness of Title I funding in improving student academic

performance has come primarily from two large congressionally mandated studies, one meta-

analysis of seventeen studies, and the two regression discontinuity studies citied previously. The

first congressionally mandated study of Title I, “Sustaining Effects,” included five separate

analyses, one of which traced students longitudinally for three years between 1976 and 1979

(Carter, 1984). Overall students in Title I schools were found to achieve more than similar

students in non-Title I schools -- more in math than in reading and more in elementary than

middle grades. Further, very low achieving students in Title I schools achieved less and did not

advance enough to reach parity with “regular” students (Carter, pp 6-7).

The second Congressionally mandated study, “Prospects,” was conducted over four years

from 1991 to 1994 (Borman, D’Agostino, Wong, and Hedges, 1998). This analysis of

achievement growth of Title I students compared to non-participants found mixed evidence on

achievement. When more growth in achievement for Title I students was found, consistent with

the results from Sustaining Effects, it was for students who were initially in need of less

remediation.

Borman and D’Agostino’s (1996) meta-analysis of seventeen evaluation studies

commissioned or performed by the federal government (including Sustaining Effects and

9 WP #09-09
Prospects) found a “modest” impact for students in Title I schools, with an increasingly positive

impact as programs were in existence longer. That said, none of these studies claimed to have

found a causal effect of Title I. Although they constructed and reported on comparison groups,

the authors were clear that unobserved differences between the comparison students and the

students in Title I schools could not be ruled out. As discussed by Borman and D’Agostino

(1996), the impact of Title I varies by program implementation, time of year the test is taken,

grade level, and other unobservable differences. For example, since districts set different cutoffs

for Title I school eligibility, national studies compare students attending schools with varying

levels of poor students. Finally, the two previously discussed regression discontinuity studies

also found no school-level performance effect in either reading or math (Van der Klaauw, 2008;

Matsudaira et. al, 2006).

Thus, while the effects of Title I funding on student academic performance have been

studied intensively, shortcomings in methods or data have limited their ability to demonstrate

causal effects for schools and for more recently reauthorized versions of Title I, specifically

NCLB.

This study seeks to improve on these previous studies by using seven years of school-

level data and longitudinal quasi-experimental methods to estimate impacts. Thus we are better

able to estimate credible causal effects. Additionally, in contrast to the studies reviewed, we

examine the impact of Title I spending on high schools. Data on the impact of Title I at the high

school level is especially important in light of the next reauthorization of NCLB, which will

likely include accountability measures for high schools.

Does Title I Increase Spending and Improve Performance? 10
4. Methods

A key challenge in estimating the effects of Title I on school spending or outcomes is that

Title I eligible schools differ from non-Title I schools in numerous ways. Some of these

differences are known to determine spending and outcomes (for example, following Rubenstein

et al., 2006, size of school and percentages of special needs and poor students) and can be

included as covariates in models. Nonetheless, there are likely to remain significant unobserved

differences that could also determine spending and outcomes.

We use an RD design to obtain unbiased estimates. Our RD model exploits an important

feature of the Title I program – the use by school districts of a poverty cut-off rate to determine

whether schools are eligible for Title I funding. Specifically, school s is eligible for Title I

funding in year t if its poverty rate (pst), which is measured as the previous year’s school rate, is

greater than or equal to the district’s poverty cut-off rate for that year, pt. If pst < pt, the school is

not eligible. This discontinuity in the funding eligibility creates a quasi-experiment whereby

small differences in the school poverty rate affect eligibility but are unlikely to reflect significant

other observed or unobserved differences in the schools. Because a school’s poverty rate for

Title I is measured by its previous year’s actual rate, gaming to move that poverty rate is

unlikely.

There are several ways to implement an RD design. We use a parametric procedure that

employs data on all schools and years (see Hahn, Todd, and van der Klaauw, 2001, for

presentation of alternative methods). School (or borough) fixed effects control further for

unobserved but time invariant, school characteristics. This use of school effects also allows us to

estimate impacts on schools that change Title I eligible, becoming newly (in) eligible over time.

We estimate the following equation:

11 WP #09-09
Yst = β0 + β1TitleIst + g(pst) + β2Xst + β3Zst + θs + Φt + ust (1)

where s indexes schools and t indexes time. Yst is the outcome of interest, for example spending

on direct services or math achievement, TitleIst is a dummy variable that takes a value of 1 if

school s is Title I eligible in year t and 0 if it is not. Thus it identifies the discontinuity in Title I

funding. g(pst) is a polynomial function of the percent poverty (pst), where the percent is

centered on the Title I poverty eligibility cutoff for the year, and the polynomial is allowed to

differ on either side of the cutoff. This polynomial absorbs variation from schools far from the

eligibility cutoff. (See Imbens and Lemieux, 2008; Cellini et al, 2009, for explanations and use

of this techniques).

In this model, the coefficient on TitleIst (β1) captures the difference in outcomes at the

discontinuity and thus provides a direct estimate of the local average effect of Title I. Xst is a set

of school-level student characteristics such as the percentage of students who are eligible for

English language learning programs, Zst is a set of school characteristics, including size, θs is a

school or borough fixed effect, Φt is a set of year fixed effects, and ust is a random error term. 13

In the elementary/middle school models of academic outcomes, which are measured for grades 3

through 8, grade effects are also included.

In summary, this study is designed to provide unbiased estimates of the local average

treatment effects of Title I eligibility. It adds to the literature not only in the results but also by

applying the method to longitudinal data, which permit the addition of school fixed effects to

better control for unobserved difference. Further, one interpretation is that the school fixed

effects allow comparison between schools that change status while the borough effects illustrate

effects for schools that mostly remain either Title I or not.

Does Title I Increase Spending and Improve Performance? 12
Is the RD Appropriate?

There are a number of conditions that must be met before a causal effect of Title I status on

outcomes can be inferred from the estimates that are produced by the RD design. First, there

should be a truly discontinuous jump in the Title I status at the poverty cutoff. Second, the other

determinants of the outcome (X’s and Z’s) should be the same in a close band below and above

the cutoff, indicating that there are no other close discontinuities. Third, the outcome should be a

continuous function of the poverty at the Title I cutoff. Finally, the schools should not be able to

manipulate their Title I status, which means that the density of percent poverty above and below

the cutoff should be similar. These conditions are addressed below.

Figure 2 displays one year of elementary and middle schools receipt of Title I funds

graphed against the eligibility cutoff for funds. The cutoff produces a “sharp” discontinuity, with

no misclassifications. This holds for all years, save one in which three schools below the cutoff

were classified as eligible (less than .4% of the schools) and for all years of our high school

sample.

Table 2 presents descriptives for schools within five percentage points of poverty of the

Title I eligibility cutoffs in selected years for elementary and middle schools (panel A) and for

high schools (panel B). Statistically significant (at 5% or better) between Title I and non-Title I

schools are bolded. There are few differences and, when they occur, only in a few years,

indicating similarity in observed features of schools close to the cutoff. As expected the percent

poverty does differ, since the eligibility is based on a sharp cutoff in this variable.

Figure 3 shows the relationship between the percent poverty (and Title I cutoff) and direct

expenditures per pupil for one year, illustrating the continuous nature of this outcome at the

cutoff.

13 WP #09-09
Finally, as already pointed out, a school’s eligibility is based on its previous year’s poverty

level and the district poverty cutoff for eligibility is based on this year’s average poverty, which

makes it unlikely that schools will be able to manipulate their status.

5. Data

We use school-level data from 1996-97 through 2003-04 obtained from three NYCDOE.

sources: the Annual School Reports (ASRs), the School-Based Expenditure Reports (SBERs), and

the BOR1 Allocation Memorandum. ASRs are comprehensive reports containing school-level

student and teacher characteristics, test results, and four-year graduation and dropout rates.

SBERs provide school-level spending and are disaggregated in a variety of ways. 14 BOR1

provides data on Title I status of the school (i.e. whether NYC classifies the school as a Title I

school) and the percent poverty on which eligibility is based. Notice that status and eligibility

need not match since some schools are grandfathered. A total of 940 elementary and middle

schools (through grade 8) are in the database, with 776 of them having the crucial BOR1 data on

Title I status for all years in the analysis. The high school panel has complete data for 167

schools, the majority of regular high schools operating for more than four years as of 2001-02. 15

For elementary and middle schools, we use data for 1996-97 through 2002-03, during

which the only change in criteria for Title I eligibility in New York was in the citywide

14
These are organized by functional areas (for example, direct versus classroom instruction), program (for example
distinguishing special education versus general education programs), and location (for example, by borough).
15
For two reasons we exclude schools designated solely for full-time special education students (i.e., the school for
deaf children). First, many of these students were not required to take citywide or state standardized tests during
much of the study period. Second, many of these schools do not receive Title I funds since the delivery of
specialized student services are prescribed in each student’s Individualized Education Program and are paid for out
of other funding sources. We do, however, include special education students attending public schools with general
education students.

Does Title I Increase Spending and Improve Performance? 14
eligibility cutoff. 16 In particular, eligibility cutoffs in NYC’s five boroughs were the same. For

high schools, our analysis uses only three years of data, 2001-02 through 2003-04, because high

schools were often grandfathered into Title I status even after their poverty levels should have

made them ineligible before these years. Thus the discontinuity was not at all sharp. In 2001-02,

however, grandfathering stopped. Although a very small number of high schools were

misclassified as Title I (in)eligible, the number was too small to be substantively meaningful and

they were removed following sensitivity analyses. 17 We include 2003-04, even though Title I

cutoffs differed by borough beginning in that year, in order to have enough years for estimation,

and we modify the estimation equation appropriately as explained in the results section.

Table 3 shows the poverty eligibility cutoff and the distribution of Title I and non-Title I

schools by year. Between 1997 and 2003, the eligibility cutoff ranged from a low of 62% in

2002-03 to a high of 68.36% in 2000-01 and, in any given year, over 70% of the elementary and

middle schools (EMS) in New York City were eligible for Title I funding. The lowest percentage

of Title I eligible schools included in our EMS sample is 71.8% in 2001-02, when the eligibility

cutoff was the highest, and the highest percentage is 76.8% in 2002-03, when the eligibility

cutoff was 62%. Compared to the EMS sample, fewer high schools are Title I eligible. In 2003-

04, 54.6% of high schools were Title I eligible, but this varies considerably by borough. Staten

Island is not included in the analyses since only three of its schools were ever funded in the

period.

16
We do not include 2003-04 because cutoffs differed by borough and we do not need additional years to estimate
the effect of Title I eligibility well.
17
One of the misclassified schools was in the data twice; a few of the ineligibles that were classified in the BOR1 as
“Title I” were spending very low amounts of Title I funds (less that $80 per pupil, which is way below the average
of around $500 per pupil) and thus most likely “mistakes.” Models that included these misclassified schools yielded
the qualitatively similar results.
15 WP #09-09
Among the EMS sample, 12.4% of schools changed Title I status at any time during the

seven-year study period, while 19.3% of high schools did so, thus making estimation with school

fixed effects possible (Table 4). Of these changes, 53 (14) EMS (high schools) changed more

than once.

6. Results

In this section, we first present results for how Title I affects expenditures. We measure

expenditures as the per-pupil spending on direct services (money spent on classroom instruction,

support services, and school administration) and revenues in two categories: per-pupil Title I

funds and per-pupil PCEN funds (the state compensatory education program). Thus, we aim to

determine whether direct expenditures increase with Title I and then to explore whether state

compensatory funds substitute for Title I funds in schools below the cutoff. 18 Next, we present

performance analyses. Performance for elementary and middle schools is measured by test

scores in grade 3 through 8. These scores are reported as standardized grade by year “z” scores,

where to obtain grade/year scores, individual student scores for each grade are subtracted from

each grade/year mean score and divided by the system wide standard deviation. Thus, by

construction, system wide grade z scores in each year have a mean of zero and a standard

deviation of one. 19 Performance in high school is measured by the percent of students in a high

school cohort that begins together in 9th grade and that alternatively graduated in four years or

dropped out. These data are based on NYC DOE reports that track entering 9th graders over four

years.

18
We also estimated the same models with total expenditures and found qualitatively similar results. Direct
expenditures are most relevant because they are used in the schools themselves rather than for system wide
functions.
19
The student-level data, aggregated to the school-grade level, come from a dataset developed by two of the authors
for use in previous studies. See Schwartz, A.E., L. Stiefel, and D.Y. Kim (2004).

Does Title I Increase Spending and Improve Performance? 16
Complete regressions are shown for the first set of spending and outcome results for

elementary and middle schools, followed by coefficients on the Title I dummy alone for high

schools. Complete results for all models are available from authors.

Per-Pupil Spending and Revenues

Table 5 presents results examining the impact of Title I eligibility on per-pupil

expenditures and revenues in elementary and middle schools based on a cubic polynomial in

percent poverty. 20 Coefficients on all variables except middle school and year effects are shown.

The year effects indicate a continual increase in funding over the years, ceteris paribus on other

variables. Finally, model 1 (columns 1, 3, 5) includes borough fixed effects while model 2

(column 2, 4, 6) includes school fixed effects.

To begin, most of the coefficients on school characteristics are consistent with theory and

previous work. Larger school enrollment is related to lower spending per pupil, while larger

percentages of English language learners and of special education students result in higher

spending per pupil or have no effect. Increases in recent immigrants generally have a negative

effect or no effect on spending.

Turning to the effect of Title I on direct expenditures, model 2, which best controls for

unobservables, shows that there is an increase in spending of approximately $284 per pupil. This

is not a large effect especially since NYC receives about $500 per pupil in Title I funds (see

Table 1). In the alternate specification (model 1), which uses borough effects, we see no

statistically significant increase in expenditures as a result of Title I. Why don’t Title I schools

have higher direct expenditures?

20
Results of all degrees of polynomials are qualitatively similar (and available from authors), although the level of
significance of the effect at the discontinuity differs slightly between specifications. That is, in the quadratic and
linear (but not quartic) specifications, the effect of Title I eligibility on direct expenditures, with borough effects,
becomes statistically significant. The quartic and cubic specifications control better for the effect of poverty around
the cutoff and thus we prefer them.
17 WP #09-09
Perhaps the answer lies in the impact of Title I on Title I revenues and state PCEN

compensatory funds. On the one hand, with respect to Title I revenues themselves, as seen in

models 1 and 2 (columns 3 and 4), the difference between Title I and non-Title I schools is, on

average, between $478 and $508 per pupil. This amount is around the average per pupil Title I

funding shown in Table 1 over the period. Thus, Title I revenues appear to increase in Title I

schools. On the other hand, as shown in models 1 and 2 for PCEN funds (columns 5 and 6),

Title I decreases PCEN funding in Title I schools compared to ineligible ones by an average

between $119 and $127. This suggests that part of the reason that direct expenditures do not

increase more is that they are supplanted by PCEN funds.

Results for high schools are markedly different. As shown in Table 6, high schools that

are eligible for Title I funding spend approximately $600 to $650 more on direct expenditure

than do ineligible schools. As with EMS, they receive more in Title I funds (between $464 and

$488 on average), however they do not receive lower amounts of PCEN funding, 21 suggesting

the interesting finding that there is not supplanting at the high school level.

Summary of Findings on Spending and Revenues

The above analyses give us consistent estimates for elementary and middle schools that

show that Title I status contributes to increased per-pupil direct expenditures for those schools

that change Title I eligibility over time. We also find that while Title I funds reach EMS schools

targeted by the legislation, more PCEN funding goes to schools that are not eligible to receive

21
In the high school panel, in one set of estimations, the degree of the polynomial makes a difference to results – for
polynomials of cubic form, the direct expenditures do not increase in either with no effects or with school effects
regressions, at a statistically significant level of 10% or less.

Does Title I Increase Spending and Improve Performance? 18
Title I funds, implying supplanting. While the amount of supplanting is not large enough to

offset the total amounts available through Title I, it is not trivial either.

High schools are different, however. Direct expenditures and Title I funds increase in

eligible schools, but PCEN funds do not change. Thus, there is no evidence of supplanting of

state funds for federal funds and Title I increases direct expenditures considerably more in high

schools than in lower level schools.

Math and Reading Performance in Elementary and Middle Schools

Turning to the estimates of the impact of Title I status on student performance, we first

analyze school-level math and reading test scores for elementary and middle schools, averaged

across all grades that a school serves. We include controls similar to the ones in the spending

equations plus additional variables to reflect more student characteristics, consistent with the

education production function literature. Table 7 presents these results.

Again note that coefficients on control variables are consistent with most literature:

ceteris paribus, scores increase as the percentage of students that are Asian and female increases,

while scores decrease as the percentage of black, Hispanic, English language learners, and

special education students increases.

As to the effect of Title I, although none of the coefficients on the Title I eligibility

dummy is significant at the 5% level, a number are negative and significant at the 10% level,

indicating that there may be a slight negative local average treatment effect in the range of -.03 to

-.04 standard deviations. While these findings are disturbing, there may be several reasons why

we see these negative impacts.

19 WP #09-09
First, we found that Title I eligible elementary and middle schools do not receive a large

increase in funds, compared to non-Title I eligible schools. Therefore, the funds that these

schools do receive may not be enough to provide the services required to increase the test scores.

Second, we do not know how Title I schools are serving eligible students. Some schools

close to the cutoff may provide services only to eligible students (pull-out programs) while

others may provide services to all students in the school, regardless of eligibility. Lastly, some

schools may change Title I status over time and may receive different amounts of funds based on

the length of time they are Title I. We address this below.

High School Outcomes

Table 8 presents results for high school outcomes. The effects of Title I funding are

uniformly zero. At best, one could say that the funding and increase in direct expenditures that

follow Title I do not harm students, but they do not help them either.

7. Further Exploration of Results – The Timing of Impacts

The spending results for elementary and middle schools hint strongly at some reduction

of state and local funds (PCEN funds in particular) in schools that are Title I eligible, but with an

effect on direct expenditures not felt in schools that are newly Title I eligible. That is, in the

fixed effects regressions, Title I results in higher direct expenditures, perhaps indicating that

schools that change status (the newly eligible schools) receive more at first, but that over time

they see reductions. To explore this further, we created new dummy variables that define the

patterns of eligibility over the years in the panel and modified expenditure and test score

regressions. In Table 9, FirstTI indicates the first year that a school becomes Title I eligible;

Does Title I Increase Spending and Improve Performance? 20
TwoTI indicates that a school is in its second or higher years of Title I eligibility; FirstnotTI

indicates the first year a school exits from Title I eligibility; TwonotTI (omitted) indicates the

second or future years that a school is ineligible. Here we use only six years of data to be able to

determine cleanly the first year of (in) eligibility. Further we use only a linear variable in percent

poverty to preserve degrees of freedom. In effect, these newly created variables split the Title I

categories into two parts – the newly eligible (ineligible) and the longer eligible (ineligible). As

shown in Table 9, direct expenditures are higher in the second and future years than in the first,

although still significantly lower than the additional Title I funding of around $500. Title I

funding also is higher in the years after the first and declines a great deal the first year that a

school becomes ineligible. Finally, PCEN funding declines with Title I eligibility and more after

the first year. Thus, these regressions indicate a larger impact in years beyond the first rather

than a diminished effect as time goes on.

The Title I elementary and middle schools and to a greater extent the high schools spend

more on direct services than their non-eligible counterparts. What do these dollars buy? To

explore this question, we regressed three real resource variables on the Title I dummy. The three

resource variables are the pupil-teacher ratio, the percent of teachers with masters’ degrees and

the percent of teachers with more than five years of experience teaching. The pupil- teacher ratio

captures the teacher resources available in the school, although this ratio is generally smaller

than class size because teachers may tutor students or provide other help outside the classroom.

The percent of teachers with masters’ degrees measures additional education although it may not

correlate well with increased student performance. There is, however, some research that shows

that teachers with little experience do worse at increasing student performance than ones with

more (Rockoff, 2004; Rivkin, et al, 2005; Clotfelter, et al, 2007) and thus the variable indicating

21 WP #09-09
more than five years of experience is meaningful. Table 10 displays the coefficients for the Title

I eligibility dummy for elementary and middle schools (panel A) and for high schools (panel B).

There is little evidence of any effect on these resources for elementary and middle schools,

although the percent masters’ shows some minor negative effects for Title I schools at around the

10% significance level. For high schools, however, pupil-teacher ratios are lower by a little over

one student at just above the 5% significance level, but also teachers with over five years of

experience are considerably lower in Title I schools. These results makes some sense given that

the Title I high schools are estimated to spend around $600 per pupil more than the ineligible

ones, an amount considerably higher than the elementary and middle schools, for which they

obtain more teacher resources, but ones with less experience.

In light of somewhat higher spending, but not too much effect on the measured real

resources, we explore further the (slight) negative results for test scores in elementary and

middle schools. To do this, we re-estimate the score equations with the new Title I variables

created to indicate first and future years of (in) eligibility. Table 11 shows these results. Here

we see some indication (in the borough effects regressions only) that the negative effects show

up after the first year and again in the first year that schools are not Title I eligible. In yet another

specification shown in columns five and six, with the Title I eligibility dummy and dummies for

schools that are always or never (with sometimes omitted) Title I, we see that the ones that go in

and out of Title I do somewhat better (that is, always and never do worse).

These results indicate that the pattern of eligibility over time has an effect, but it is not

consistent with the idea that the latter years see a lessening of funds or of the negative scores, as

we hypothesized might be the case from the RD results. In addition, while Title I eligible high

Does Title I Increase Spending and Improve Performance? 22
schools show some improvement in their pupil-teacher ratios, they also have fewer experienced

teachers.

8. Conclusions and Discussion

This paper reports the results of a study of the impact of Title I on school spending and

student achievement in New York City. One of the primary differences between this study and

those done previously is that we use of panel data over a seven-year period that has consistent

program implementation instead of single years of data and years in which program requirements

change. We believe that the ability to use longitudinal data and school fixed effects improves

knowledge about expenditures based on Title I status. In addition, we are able to analyze results

for high schools, also over time.

Overall, the results indicate that Title I brings increased federal education funds to

schools and that these funds enable high schools close to the Title I eligibility cutoff to

significantly increase the amount of money spent on direct services to students and to improve

their pupil-teacher ratios (while having fewer experienced teachers). Elementary and middle

schools do not increase funds as much, which is consistent with our finding that New York City

may be using PCEN funds to supplant some of the Title I funding for elementary and middle

schools that do not have poverty rates higher than the eligibility cutoff. Since schools just below

the Title I cutoff are similar to those just above the cutoff, this finding may be an equitable, albeit

unintended, result.

Additional amounts of funds that Title I schools receive, compared to non-Title I schools,

seen not to narrow the achievement gaps between poor student and their more advantaged peers

and may even reduce school wide average test scores somewhat in elementary and middle

23 WP #09-09
schools. These effects for both spending and scores seem to increase the longer schools are Title

I and to be stronger for ones that are always Title I versus ones that go in and out of eligibility.

These findings are important and they also lead to several hypotheses that need to be

tested in order to understand why they occur. The results in the RD estimations indicate the local

average treatment effect, or what is happening right around the cutoff, but we do not know how

the schools near the cutoffs are serving the Title I students. There are informal reports that a

common use of Title I funds is to hire extra personnel to tutor students who are lagging. But this

tutoring could remove the students from their regular classroom causing them to miss material or

be stigmatized or both. Or, alternatively, since we are measuring school-wide scores, perhaps

the served Title I students in schools right above the cutoff are doing better than their unserved

classmates, but the school average is dominated by the unserved students. To explore such

hypotheses, student level, longitudinal analyses would be required.

Another hypothesis is that for schools near the cutoff, the administrative burden of

accounting for and using the extra Title I funding is large relative to the benefits. Strings may be

attached, such as making sure only Title I eligible students are served, that make the money

ineffective. In addition, in elementary and middle schools it is clear that PCEN funds go down

so that the net amount of additional funding is not great.

Perhaps when schools have higher percents of poverty (further from the cutoff) they

receive more funding or have more freedom (more likely to be school wide programs?), and it is

the ones near the cutoff that suffer most.

Finally, the performance effects over time need to be further explored. For example, the

evaluation literature almost always cautions against looking for effects in short time periods.

Some say it may take between five and 10 years to turn around a school (Darling-Hammond,

Does Title I Increase Spending and Improve Performance? 24
1988; Haynes, 1998; Hess, 1995; Levin, 1991).Thus the specific path and adjustment process

over a number of years may be important and differ from that found looking at one year and all

other years out as we did in Table 11. In addition, on the spending side, the federal government

may restrict how Title I funds are used, but these restrictions may contradict the plan that a

district has for funding its schools. For example, a district may wish to distribute funds

according to student characteristics without a sharp cutoff (making the reduction in PCEN funds

seem reasonable) or may even have a delivery plan that differs from the Title I restrictions that

require the school to serve only eligible students with the new funds.

What we have found is that around the Title I cutoff, there is significant substitution of

funds in schools that miss being Title I eligible, thus blunting the net amount received by the

eligible schools. In addition, at best, just above the cutoff there is no effect on performance (high

schools) and possibly even a negative effect for elementary and middle schools. These results

are consistent with other studies of districts that find reduction of Title I funds over the years and

are not inconsistent with all the Title I performance evaluations that rarely find any effect on

student performance. Still there are a number of hypotheses about why these results occur that

could be explored with longitudinal student data but are for future work.

25 WP #09-09
References

Borman, G. D., J. V. D’Agostino, K. K. Wong, et al (1998). “The Longitudinal Achievement of
Chapter 1 Students: Preliminary Evidence from the Prospects Study.” Journal of Education for
Students Placed At Risk, 3(4) 363-399.

Borman, G., & J. V. D'Agostino (1996). Title I and Student Achievement: A meta-analysis of
federal evaluation results. Educational Evaluation and Policy Analysis, 18(4), 309-326.

Brown, C. A. (2005). Federal Title I Funding Distribution: Are the poorest children receiving
their share of the funds? Paper presented at the American Educational Finance Association,
Louisville, KY.

Carter, L. (1984). “The Sustaining Effects Study of Compensatory and Elementary Education,”
Educational Researcher 13(7), 1984, 4-13.

Cellini, Stephanie R., Fernando Ferreira and Jesse Rothstein (forthcoming). “The Value of
School Facility Investments: Evidence from a Dynamic Regression Discontinuity Design,”
Quarterly Journal of Economics.

Chambers, J. C., T. Parrish, & M.E. Goertz (1993). Chapter 1 resources--Are they truly
supplemental? Paper presented at the annual meeting of the American Education Finance
Association.

Clotfelter, C. T., H. F. Ladd & J. L. Vigdor (2007). “How and Why Do Teacher Credentials
Matter for Student Achievement?” NBER Working Paper No. W12828

Darling-Hammond, L. (1988). Policy and professionalism. In A. Lieberman (Ed.),
Building a professional culture in schools (pp. 57–77). New York: Teachers College Press.

Fisher, R. C. & L. E. Papke (2000). “Local Government Responses to Education Grants,”
National Tax Journal, 53 (1):153-168.

Goertz, M. E. (1988). School districts' allocation of Chapter 1 resources. ETS RR-88-16.
Princeton, NJ: Educational Testing Service.

Gordon, N. (2004). “Do Federal Funds Boost School Spending? Evidence from Title I,” Journal
of Public Economics 88(9-10):1771-1792.

Hahn, J., P.Todd, and W. Van der Klaauw (2001). “Identification and estimation of treatment
effects with a regression-discontinuity Design,” Econometrica, 69(1):201-209.

Haynes, N. M. (1998). Lessons learned. Journal of Education for Students Placed at Risk, 3(1),
87–99.

Hess, Jr., A. G. (1995). Restructuring urban schools: A Chicago perspective. New York:
Teachers College Press.

Does Title I Increase Spending and Improve Performance? 26
Hines, J. R. and R. H. Thaler (1995). “Anomalies: The Flypaper Effect,” Journal of Economic
Perspectives, 9(4):217-226.

Imbens, G. and T. Lemieux (2008). “ Regression discontinuity designs: A guide to practice,”
Journal of Econometrics.

Knight, B. (2002). “Endogenous Federal Grants and Crowd-out of State Government Spending:
Theory and Evidence from the Federal Highway Aid Program,” The American Economics
Review 92(1):71-92.

Levin, H. M. (1991). Learning from Accelerated Schools. Philadelphia: Pew Charitable Trusts.

Matsudaira, J. D., A. Hosek, and E. Walsh (2006). “An Integrated Assessment of the Effects of
Title I on School Behavior, Resources, and Student Achievement.” Presented at the American
Education Finance Association Conference, March 24-26, Denver, Colorado.

Natriello, G. and E. L. McDill (1999). “Title I: From Funding Mechanism to Educational
Program,” in The Harvard Civil Rights Project, Hard Work for Good Schools: Facts Not Fads in
Title I Reform, Cambridge, MA: Harvard University.

National Assessment of Title I Interim Report: Executive Summary (2006). Washington DC:
U.S. Department of Education, Institute of Education Sciences.

New York City Department of Education, 2009 School Allocation Memorandum 5/20/2009.

New York City Department of Education. BOR Allocation Memo #1 FY 2001 Initial Allocations
Tax Levy and Reimbursable. New York City Department of Education. Available:
http://www.nycenet.edu/offices/d_chanc_oper/budget/DBOR/AllocationMemo/FY02-
03/NumberMemo/03BORno1/bor1.html [2004, December].

Nichols, Austin (2007). “Causal Inference with Observational Data,”
http://pped.org/stata/ciwod.pdf

Porter, J. R., (2003). “Estimation in the Regression Discontinuity Model,” Harvard University,
manuscript.

Rivkin, S. G., E. A. Hanushek, & J. F. Kain (2005). “Teachers, Schools, and Academic
Achievement,” Econometrica, 73(2): 417-458.

Rockoff, J. (2004). “The Impact of Individual Teachers on Student Achievement: Evidence from
Panel Data,” The American Economic Review, 94(2):247-252.

Rubenstein, R., L. Stiefel, A. E. Schwartz et al (2004). “Distinguishing Good Schools from Bad
in Principle and Practice.” In Developments in School Finance 2003, NCES, pp. 53-70.

27 WP #09-09
Rubenstein, R., A. E. Schwartz, L. Stiefel and H.B. H. Amor (2007). “From Districts to Schools:
the Distribution of Resources across Schools in Big City School Districts,” Economics of
Education Review, 26: 532-545.

Schwartz A. E., L. Stiefel, & D. Y. Kim, (2004). “The Impact of School Reform on Student
Performance: Evidence from the New York Network for School Renewal Project.” Journal of
Human Resources. 39(2): 500-522

Van der Klaauw, W. (2008). “Breaking the link between poverty and low student achievement:
An evaluation of Title I,” Journal of Econometrics. 142(2): 271-309

Does Title I Increase Spending and Improve Performance? 28
Figure 1: How Title I Funds are Allocated

Step 2
Step 1
States
U.S. Department of Education Individual LEA allocations are
Funds are allocated to local totaled and then allocated to
education agencies (LEAs) based each state Department of
on the numbers of poor Education. A state can keep up
children, whose count is to 1% of the total state allocation
obtained from the Census for administration of the Title I
Bureau. programs plus for schools in
need of improvement

Step 4

Schools
LEAs then allocate Title I
funds to individual schools.
All schools with at least
75% poverty must receive
money, but districts have
Step 3
the discretion of setting the
cutoff for Title I status as Local Education Agencies
low as 35%. The allocation to an LEA is based
on the same or more recent
counts of child poverty used in
Step 1, minus any state
administrative costs.

29 WP #09-09
Figure 2: Probability of Receiving Title I funds versus New York City Poverty Cutoff for Title I
(centered on zero), Elementary and Middle Schools, 2001-02

Title I Eligible

.4 .6

0
-60 -40 -20 0 20 40
Centered % Poverty

Notes:

Does Title I Increase Spending and Improve Performance? 30
Figure 3: Direct Expenditures per Pupil and Title I Poverty Cutoff (centered on zero),
Elementary and Middle Schools, 2001-02

25000

Direct Service Expenditures

10000 15000 20000

5000

-60 -40 -20 0 20 40
Centered % Poverty

Notes and draw line though points

31 WP #09-09
Table 1: Total Spending and Compensatory Education Revenues in New York City, 1996-97
through 2003-04

Total Spending Title I Funds PCEN Funds
Real $ Real $ % Per- Real $ % Per-
Per-
(in billions) (in millions) Total pupil ($) (in millions) Total pupil ($)
pupil $
Spent Spent

1996-97 12.01 11,232 543.45 4.5 506 369.34 3.1 343
1997-98 12.52 11,790 561.76 4.7 557 305.83 2.4 280
1998-99 13.01 11,951 501.17 4.1 492 291.91 2.3 274
1999-00 13.32 12,105 474.66 3.6 431 295.41 2.2 269
2000-01 13.67 12,365 467.21 3.4 422 286.80 2.1 260
2001-02 13.39 12,186 474.06 3.5 432 286.39 2.1 261
2002-03 13.86 12,691 637.94 4.6 584 292.18 2.1 267
2003-04 14.00 12,884 696.46 5.0 640 275.67 2.0 253

Sources: School-Based Expenditure Reports, System wide Summary, 1996-97 through 2003-04
(i) Monetary values expressed in 2004 Dollars

Does Title I Increase Spending and Improve Performance? 32
Table 2: Mean of School Characteristics, by Title I Eligibility: New York City Elementary,
Middle and High Schools, Selected Years, within 5 % Band of Percent Poverty on
Either Side of Eligibility Cutoff
Elementary and Middle Schools
1997-98 1999-00 2001-02
TI Not TI TI Not TI TI Not TI
% Asian 17.77 22.09 17.39 22.78 19.49 18.36
% Hispanic 29.74 20.69 28.04 26.97 28.60 26.01
% Black 35.18 26.41 30.98 24.77 32.68 30.46
% White 17.29 30.82 23.58 25.49 19.22 25.17
% Female 49.17 48.38 48.86 49.31 49.07 48.23
% ELL 15.06 15.51 14.06 15.53 13.09 11.16
% Recent Immigrant 11.27 11.59 10.02 11.56 10.12 9.18
% Full-Time Special Education 4.42 6.07 4.29 3.74 3.67 4.08
% Free Lunch 71.43 63.91 70.70 66.30 70.95 62.99
Total Registration 1097.18 894.42 983.54 868.4 961.12 944.7
N Schools 39 24 57 20 65 27
Bolded numbers statistically different

High Schools
2001-02 2003-04
TI Not TI TI Not TI
% Asian 5.13 1.05 7.58 9.30
% Hispanic 45.92 54.00 43.80 33.94
% Black 42.93 42.80 41.90 42.18
% White 5.98 2.15 6.70 14.55
% Female 57.26 49.20 45.12 46.83
% ELL 8.09 23.10 9.54 6.81
% Recent Immigrant 3.68 7.45 6.80 9.29
% Full-Time Special Education 3.80 7.74 5.52 3.91
% Free Lunch 70.76 67.25 63.84 52.81
Total Registration 1103.33 1320 1581 1300.16
N Schools 9 2 5 8
Bolded numbers statistically different
Note: Definition of variables is as follows:
% Asian: Percentage of Asian students in a school
% Hispanic: Percentage of Hispanic students in a school
% black: Percentage of black students in a school
% white: Percentage of white students in a school
% Female: Percentage of female students in a school
% ELL: Percentage of limited English proficiency students in a school
% Recent Immigrant: Percentage of recent immigrant students in a school
% Full-time Special Education: Percentage of full time special education students in a school
% Free Lunch: percentage of free lunch eligible students in a school
Total Registration: Total enrolment in a school
33 WP #09-09
Table 3: Distribution of Title I School Status by Year by Level

Poverty
Cutoff for
Title I Schools Schools Total
Eligibility Eligible for Ineligible for Schools
Title I Funds Title I Funds
% N % Change N % Change N
Ineligible Eligible to
to Eligible Ineligible
Elementary/Middle Schools
1996-97 65.99 571 73.58 -- 205 26.42 -- 776

1997-98 66.66 571 73.58 9 205 26.42 9 776

1998-99 66.30 579 74.61 15 197 25.39 7 776

1999-00 66.99 578 74.48 9 198 25.52 10 776

2000-01 68.36 572 73.71 10 204 26.29 16 776

2001-02 68.24 557 71.78 10 219 28.22 25 776

2002-03 62.00 593 76.42 43 183 23.58 7 776

High Schools
2001-02 68.24 87 52.10 -- 80 47.90 -- 167

2002-03 62.00 97 58.08 13 70 45.40 3 167

2003-04
Manhattan 60.00 42 75.00 5 14 25.00 0 56
Bronx 60.00 26 92.86 0 2 7.14 0 28
Brooklyn 60.00 29 56.86 3 22 43.14 0 51
Queens 57.86 8 25.00 0 24 75.00 0 32
Total -- 105 62.87 8 62 37.13 0 167
Note: i) Change columns are defined with respect to the previous year

Does Title I Increase Spending and Improve Performance? 34
Table 4: Change in Title I Eligibility over Time by School Level

Elementary/Middle Schools High Schools
1996-97 through 2002-03 2001-02 through 2003-04
N % N %

Always Title I Eligible 521 67.1 84 50.3

Never Title I Eligible 159 20.5 60 35.9

Change in Title I Status 96 12.4 23 13.8

(More than once) 53 14

Total 776 100.0 167 100.0
Notes:
i. Number of schools that changed at least once (i.e. some change more than once—see next row)
ii. Staten Island high schools not included because too few for estimations.

35 WP #09-09
Table 5: Per Pupil Expenditures and Funds, RD estimates, Cubic Polynomial Models, Elementary and Middle Schools,
1997-2003

Direct Expenditure Title I Funds PCEN Funds
(1) (2) (3) (4) (5) (6)
Title I 185.237 284.300*** 507.642*** 477.592*** -119.405*** -126.873***
(158.439) (91.745) (21.338) (23.492) (17.907) (17.489)
%Poverty -0.092 -37.951*** -2.133 -4.552* 3.316* 0.871
(16.660) (13.009) (2.065) (2.558) (1.950) (2.192)
%PovertySq -0.425 -1.682*** -0.141* -0.070 -0.008 -0.003
(0.635) (0.565) (0.076) (0.103) (0.072) (0.088)
%Poverty Cub -0.006 -0.019*** -0.002** -0.001 -0.000 -0.000
(0.007) (0.007) (0.001) (0.001) (0.001) (0.001)
TitleI*%Poverty 31.134 45.432** 12.061*** 7.188 -5.443* -6.524**
(28.807) (21.294) (3.998) (4.443) (3.003) (3.122)
TitleI*%PovertySq -1.730 0.485 0.128 0.167 0.155 0.329**
(1.605) (1.240) (0.230) (0.253) (0.160) (0.158)
TitleI*%PovertyCub 0.051* 0.040* 0.001 -0.000 -0.002 -0.006**
(0.027) (0.022) (0.004) (0.004) (0.003) (0.003)
Total Registration -1.907*** -4.377*** -0.118*** -0.360*** -0.060*** -0.126***
(0.733) (0.220) (0.010) (0.035) (0.006) (0.017)
%ELL 18.768*** -0.232 0.493 2.704** 6.369*** 2.797***
(2.568) (7.673) (0.416) (1.324) (0.258) (0.691)
%Recent Immigrant -18.930*** -3.151 -1.330 -1.067 0.319 -1.501
(4.848) (9.475) (0.829) (2.116) (0.545) (1.524)
%Special Ed 205.671*** 48.491*** 0.220 -1.247 -0.130 -2.282**
(4.882) (12.795) (0.923) (1.696) (0.412) (0.905)
Constant 12,281.01*** 13,841.03*** 174.256*** 592.462*** 381.235*** 461.292***
(429.647) (451.945) (68.860) (97.140) (64.027) (79.307)
Borough Effects Yes No Yes No Yes No
School Effects No Yes No Yes No Yes
N Schools 776 776 776 776 776 776
Observations 5432 5432 5432 5432 5431 5431
R-squared 0.67 0.90 0.75 0.85 0.32 0.65
(i) Robust standard errors in parentheses
(ii) * significant at 10%* * significant at 5%; ** significant at 1%
(iii) Dollars Adjusted by 2003 Consumer Price Index
(iv) Year dummies and middle school dummy included but not shown
(v) See definition of variables in note to Table 4

Does Title I Increase Spending and Improve Performance? 36
Table 6: Per Pupil Expenditures and Funds, RD estimates, Cubic Polynomial Models, High
Schools, 2002-2004

Direct Expenditure Title I Funds PCEN Funds
(1) (2) (3) (4) (5) (6)
Title I 589.14*** 391.70** 510.03*** 461.28*** -4.22 -14.61
(138.22) (139.60) (37.01) (70.52) (36.04) (26.46)

School No Yes No Yes No Yes
Effects
N Schools 167 167 167 167 167 167
Observations 473 473 473 473 473 473
R-squared 0.61 0.87 0.70 0.89 0.35 0.82
(i) Robust standard errors in parentheses
(ii)* significant at 10%; ** significant at 5%; *** significant at 1%
(iii) Controls include quartic in percent poverty differing above and below Title I cutoff, total
registration, %ELL, %Recent Immigrant, %Special Ed, and year effects.
(iv) Only cubic polynomial shows different significance levels than quartic, quadratic or linear.

37 WP #09-09
Table 7: Academic Outcomes, RD Estimates, Cubic Polynomial, Elementary and Middle
Schools 1997-2003
Math Reading
(1) (2) (3) (4)
Title I -0.035* -0.011 -0.037* -0.031*
(0.020) (0.017) (0.019) (0.016)
%Poverty -0.005** -0.002 -0.002 -0.000
(0.002) (0.003) (1.979) (0.002)
%PovertySq -0.000 -0.000 0.000* -0.000
(0.000) (0.000) (0.000) (0.000)
%Poverty Cub -1.37 -4.85 8.86*** -4.72
(9.74) (1.27) (9.31) (1.23)
TitleI*%Poverty -0.004 0.005 -0.012*** 0.000
(0.004) (0.004) (0.004) (0.003)
TitleI*%PovertySq 0.000 0.000 0.000 -0.000
(0.000) (0.000) (0.000) (0.000)
TitleI*%PovertyCub 1.56 5.76* -3.65 3.26
(3.31) (3.07) (3.12) (2.89)
Total Registration 0.000*** -4.29 0.000 -7.35
(7.48) (0.000) (7.27) (0.000)
%ELL -0.008*** -0.007*** -0.007*** -0.004***
(0.000) (0.001) (0.000) (0.001)
%Recent Immigrant -0.005*** -0.000 0.007*** -0.002
(0.000) (0.001) (0.001) (0.001)
%Special Ed -0.001*** 0.000 -0.001** 0.001
(0.001) (0.001) (0.000) (0.001)
% black -0.008*** -0.010*** -0.006*** -0.006***
(0.000) (0.001) (0.000) (0.001)
% Hispanic -0.006*** -0.010*** -0.004*** -0.008***
(0.000) (0.001) (0.000) (0.001)
% Asian 0.004*** -0.001 0.001*** 0.002***
(0.000) (0.001) (0.000) (0.001)
% female 0.018*** 0.003*** 0.019*** 0.004***
(0.001) (0.001) (0.001) (0.001)
Constant -0.237*** 0.639*** -0.434*** 0.270**
(0.094) (0.109) (0.089) (0.115)
Borough Effects Yes No Yes No
School Effects No Yes No Yes
N Schools 776 776 776 776
Observations 5432 5432 5432 5432
R-squared 0.66 0.83 0.63 0.81
(i) Robust standard errors in parentheses
(ii) * significant at 5%; ** significant at 1%
(iii)Year, middle school, and grade dummies are included but not shown.

Does Title I Increase Spending and Improve Performance? 38
Table 8: Academic Outcomes After Four Years, RD Estimates, Cubic Polynomials,
High Schools, 2002-04, Coefficient on Title I Eligible Indicator

Graduation Rate Dropout Rate
(1) (2) (3) (4)
Title I 8.61 3.59 -5.02 1.73
(6.44) (3.82) (2.97) (4.81)

School No Yes No Yes
Effects
N Schools 166 166 166 166
Observations 455 455 455 455
R-squared 0.31 0.92 0.32 0.90
Notes:
(i) Robust standard errors in parentheses
(ii) * significant at 10%; ** significant at 5%; *** significant at 1%
(iii) Controls include cubic in percent poverty differing above and below Title I cutoff,
total registration, %ELL, %Recent Immigrant, %Special Ed, and year effects
(iv) See text for definition of variables.

39 WP #09-09
Table 9: Per Pupil Expenditures and Funds, Elementary and Middle Schools, Exploration of
Time Path of Title I eligibility

Direct Expenditure Title I Funds PCEN Funds
(1) (2) (3) (4) (5) (6)

%Poverty 7.346*** -1.707 0.880*** -5.865*** 3.305*** 0.650
(2.064) (3.985) (0.249) (0.708) (0.215) (0.610)
TitleI*%Poverty -0.151 -3.865 8.011*** 10.963*** -2.482*** -1.075
(3.616) (5.436) (0.488) (1.015) (0.343) (0.718)
FirstTI 172.643 165.190* 388.839*** 464.611*** -109.725*** -112.485***
(160.748) (89.755) (20.569) (24.150) (16.635) (17.911)
Two TI 264.232** 236.166** 537.160*** 584.968*** -129.884*** -159.929***
(89.372) (87.195) (11.187) (22.243) (9.376) (19.768)
FirstnotTI 135.657 -5.152 60.474** 92.861*** 23.703 -1.343
(189.000) (84.713) (23.551) (25.098) (22.230) (20.614)
School Effects No Yes No Yes No Yes
N Schools 776 776 776 776 776 776
Observations 4656 4656 4656 4656 4656 4656
R-squared 0.64 0.90 0.75 0.83 0.32 0.66
Notes:
(i) Robust standard errors in parentheses
(ii)* significant at 10%; ** significant at 5%; *** significant at 1%
(iii) Controls include total registration, %ELL, %Recent Immigrant, %Special Ed, and year effects.

Does Title I Increase Spending and Improve Performance? 40
Table 10: Resources, RD Estimates, Cubic Polynomial Models

Panel A: Elementary and Middle Schools, 1997-03

Pupil Teacher Ratio Percent Teachers Masters Percent Teachers Greater
Degree than 5 Years Experience
(1) (2) (3) (4) (5) (6)
Title I 0.004 -0.567 -1.772* -1.292 -1.220 -0.591
(0.157) (0.722) (0.966) (0.809) (1.301) (1.022)
Borough Effects Yes No Yes No Yes No
School Effects No Yes No Yes No Yes
N Schools 634 634 776 776 776 776
Observations 4388 4388 5426 5426 5431 5431
R-squared 0.04 0.19 0.50 0.79 0.32 0.65
Notes:
(i) Robust standard errors in parentheses
(ii)* significant at 10%; ** significant at 5%; *** significant at 1%

Panel B: High Schools, 2002-04

Pupil Teacher Ratio Percent Teachers Masters Percent Teachers Greater
Degree than 5 Years Experience
(1) (2) (3) (4) (5) (6)
Title I -1.11 -1.09* 0.67 -4.80 -19.16** -9.92**
(0.77) (0.61) (4.31) (3.30) (7.05) (4.50)

School Effects No Yes No Yes No Yes
N Schools 164 164 164 164 164 164
Observations 452 452 434 434 434 452
R-squared 0.49 0.92 0.25 0.86 0.28 0.91
Notes:
(i) Robust standard errors in parentheses
(ii)* significant at 10%; ** significant at 5%; *** significant at 1%
(iii) Controls include cubic in percent poverty differing above and below Title I cutoff,
total registration, %ELL, %Recent Immigrant, %Special Ed, year effects and forums middle school dummy.

41 WP #09-09
Table 11: Academic Outcomes, Elementary and Middle Schools, 1998-2003,
Exploration of Time Path of Title I eligibility

Math Reading Math Reading
(1) (2) (3) (4) (5) (6)

%Poverty -0.006*** -0.002** -0.007*** -0.001 -0.006*** -0.008***
(0.000) (0.001) (0.000) (0.001) (0.000) (0.000)
TitleI*%Poverty 0.000 0.002*** -0.001 0.001 0.001* 0.000
(0.000) (0.001) (0.000) (0.001) (0.000) (0.000)
FirstTI 0.014 -0.001 0.005 -0.017
(0.020) (0.017) (0.023) (0.022)
Two TI -0.041*** -0.019 -0.041*** -0.020
(0.012) (0.016) (0.011) (0.018)
FirstnotTI -0.055*** -0.010 -0.040** 0.004
(0.020) (0.016) (0.021) (0.018)
TitleI -0.0146 -0.001
(0.0141) (0.014)
Always -0.031*** -0.049***
(0.010) (0.010)
Never -0.022* -0.013
(0.013) (0.013)
School Effects No Yes No Yes No No
N Schools 764 764 764 764
Observations 15,396 15,393 15,227 15,227 17,779
R-squared 0.66 0.84 0.63 0.81 0.63
Notes:
(i) Robust standard errors in parentheses
(ii)* significant at 10%; ** significant at 5%; *** significant at 1%
(iii) Controls include total registration, %ELL, %Recent Immigrant, %Special Ed, %black, %Hispanic, %Asian,
%female, grade effects, middle school dummy, and year effects.

Does Title I Increase Spending and Improve Performance? 42
The Institute for Education and Social Policy is a joint research center of NYU’s Wagner and Stein-
hardt Schools. Founded in 1995, IESP brings the talents of a diverse group of NYU faculty, graduate
students and research scientists to bear on questions of education and social policy. We are one of the
nation’s leading academic research centers addressing urban education issues and a substantial
amount of our work focuses on New York City public schools and students. More information is
available on our website at http://steinhardt.nyu.edu/iesp.

665 Broadway, Suite 805, New York, NY 10003 p 212.998.5880 f 212.995.4564

File and source

File: g-stiefel-l-schwartz-a.pdf
Size: 840,455 bytes
SHA-256: 73d10e2e8fd8157cd5378f56cb580f0a799506b56ee57759597b83994563c4e3
Our copy: g-stiefel-l-schwartz-a.pdf
Original: files.eric.ed.gov