14
Canadian Public Policy Does Full Day Kindergarten Help Kids? Author(s): William P. Warburton, Rebecca N. Warburton and Clyde Hertzman Source: Canadian Public Policy / Analyse de Politiques, Vol. 38, No. 4 (December/décembre 2012), pp. 591-603 Published by: University of Toronto Press on behalf of Canadian Public Policy Stable URL: http://www.jstor.org/stable/41756771 . Accessed: 16/06/2014 15:48 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . University of Toronto Press and Canadian Public Policy are collaborating with JSTOR to digitize, preserve and extend access to Canadian Public Policy / Analyse de Politiques. http://www.jstor.org This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PM All use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten Help Kids?

Embed Size (px)

Citation preview

Canadian Public Policy

Does Full Day Kindergarten Help Kids?Author(s): William P. Warburton, Rebecca N. Warburton and Clyde HertzmanSource: Canadian Public Policy / Analyse de Politiques, Vol. 38, No. 4 (December/décembre2012), pp. 591-603Published by: University of Toronto Press on behalf of Canadian Public PolicyStable URL: http://www.jstor.org/stable/41756771 .

Accessed: 16/06/2014 15:48

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

University of Toronto Press and Canadian Public Policy are collaborating with JSTOR to digitize, preserveand extend access to Canadian Public Policy / Analyse de Politiques.

http://www.jstor.org

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten

Help Kids?

William P. Warburton

Enterprise Economic Consulting Ltd.

Victoria, British Columbia

Rebecca N. Warburton School of Public Administration

University of Victoria, British Columbia

Clyde Hertzman Human Early Learning Partnership (HELP) University of British Columbia, Vancouver

La fréquentation de la maternelle à temps plein (MTP) est de plus en plus courante en Amérique du Nord, mais les impacts de cette réalité sont peu connus. Dans cet article, nous analysons l'impact d'un programme de MTP mis en place en Colombie-Britannique et visant les Autochtones et les enfants n'ayant pas l'anglais comme langue maternelle. Comme ce programme a été implanté par étapes, nous avons facilement pu neutraliser certaines variables, et ainsi faire, en évitant tout biais, deux estimations par les moindres carrés (variables instrumentales) des impacts du programme. Nos résultats montrent que le programme a eu des

impacts sociaux significatifs, puisque les enfants qui en ont profité avaient de meilleurs résultats par la suite en 4e année. Pour pouvoir concevoir et appliquer une politique de MTP dans l'ensemble de la population, il faudrait toutefois d'autres recherches impliquant un suivi sur une plus longue période et réalisées selon des méthodes rigoureuses.

Mots clés : maternelle à temps plein, variables instrumentales, évaluation d'un impact, Colombie-Britannique, Autochtones, anglais langue seconde, immigrants

Full day kindergarten (FDK) is expanding across North America, but program impacts remain poorly under- stood. Using administrative data, this paper reports impacts from a targeted program in British Columbia for Aboriginal and English as a Second Language (ESL) students. Staged implementation of FDK created natural controls, allowing unbiased two-stage least squares (instrumental variables) estimates of program impacts. We find that the targeted FDK program increased grade 4 educational attainment, producing sta-

tistically and socially significant impacts. Further research, longer follow-up, and rigorous methods are needed to guide FDK policy for the general student population.

Keywords: full day kindergarten, instrumental variables, impact assessment, British Columbia, Aboriginal, English as a second language, immigrant

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

592 William P ! Warburton , Rebecca N. Warburton, and Clyde Hertzman

Introduction

Full across day kindergarten

North America. (FDK)

In is the expanding

United rapidly States across North America. In the United States

the percentage of children in full day kindergarten increased from 10 percent in 1961 to 64 percent in 2003 (Lee et al. 2006; Child Trends 2003). Brit- ish Columbia, New Brunswick, Nova Scotia, and Quebec already offer full day kindergarten to five year olds (ECLA 2008), and Ontario is currently phasing in FDK.

Despite the high cost of FDK on the one hand and its potentially large benefits on the other, a consen- sus is forming that surprisingly little is known with confidence, even though a large body of literature seeks to quantify FDK's impact on children's long- term outcomes. For example, Cooper et al. (2010) conclude, "The evidence on the impact of FDK leaves much to be desired. The research designs are weak for making strong causal inferences." And Lash et al. (2008) report, "Studies of full-day kin- dergarten whose results have been reported within the last decade cannot provide strong evidence with which to judge the effects of a full-day program on student achievement. This is because the studies have used weak research designs."

The challenge in estimating the causal impact of full day kindergarten arises from the selection of schools that offer it. Full day kindergarten has long been viewed as a mechanism for reducing inequality (Karweit 1987; Plucker et al. 2004; DeCicca 2007), and for this reason FDK is gener- ally introduced in disadvantaged neighbourhoods first. Cannon, Jacknowitz, and Painter (2006) report that children in FDK in the United States were inter alia more likely to be poor, black, and from single parent families. The schools that they attended were less likely to be suburban, their teachers had less experience, and their class sizes were larger. With important observed differences between full and half day students and the schools they attend, there is a suspicion that there may be important unobserved differences as well.1 Research designs in studies to

date have often been classified as weak because they cannot control for unobserved differences.

This paper has a stronger research design. Rules regarding the supply of full day kindergarten in British Columbia allow us to control for unmeasured differences between full and half day kindergarten students and their schools. Between 1 996 and 20 1 0, the Province of British Columbia provided FDK funding for children who were English language learners (designated ESL by the BC Ministry of Education), who were of Aboriginal ancestry, or who had specified physical disabilities, but only if a full day program was available ? When demand for full day kindergarten increased (more eligible children) in a school's catchment area, the school would be more likely to offer a full day program or to add a FDK class. For this reason, the supply of spaces increased in discrete jumps. We use these jumps in an instrumental variables framework that includes a variable to capture year-to-year changes in neighbourhoods and dummy variables to control for fixed neighbourhood and year effects.

Data

For this project the BC Ministry of Education provided us with data from four sources. The first comes from a census of all students in the BC school system (public and private) on 30 September each year. These records contain information on eligibility for funding, including ESL status, self-declared Aboriginal status, and special needs codes, along with home postal code, grade, and school. The second data source is the province-wide Foundation Skills Assessments (FSA) in reading, numeracy, and writing, taken in grade 4. The third, provided to give context to changes on the FSA, records the K-12 education experience of individuals born in 1991 (most of whom completed the grade 4 FSA in 2000-01). The fourth, provided to give an indication of year-to-year changes in student characteristics across census tracts, consists of all 1 5 year olds in school and their subsequent graduation status.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten Help Kids? 593

We begin with the 330,434 non-special needs students in kindergarten in the school years 1996-97 to 2003-04 who are also in the school census four and five years later and so should have sat the FSA. Students classified as having a special need in kin- dergarten were dropped from the analysis because they constitute a distinct population (1.5 percent of students) that is not large enough for separate an- alysis. A further 616 records were dropped because they did not have a valid postal code, and 406 were dropped because their age in kindergarten was less than five or greater than six.

Columns 1 and 2 of Table 1 report selected characteristics of all students in full day and half day kindergarten between 1996-97 and 2003-04 and their grade 4 outcomes. Overall, 8.5 percent of students are in full day kindergarten. As expected, Aboriginal and ESL students, those eligible for funding for FDK, are overrepresented in FDK. FDK students are very slightly younger than half- day students (the mean age is 1 1 days less for FDK students), which would likely give the FDK group a slight additional disadvantage in terms of later outcomes (Bedard and Dhuey 2006). Consistent with the view of FDK as a remedial program, FDK students on average have worse outcomes in grade 4 than half day students. They are more likely to be classified as having a special need, are less likely to provide meaningful responses to FSAs, and among those providing responses, have lower average scores. The question that this paper seeks to answer is whether the outcomes of FDK students are better than they would have been in the absence of FDK.

We address this question using the opening of FDK classrooms in nearby schools. We locate stu- dents' homes, neighbourhoods, and schools using postal codes. We use Statistics Canada's census tracts as our neighbourhood and its Postal Code Conversion File (Statistics Canada 2005) to convert postal codes to latitude and longitude. We restrict our analysis to students eligible for FDK (ESL and Aboriginal) with postal codes that are different from school postal codes3 and are located in tracted areas

that had at least 1 5 students in kindergarten and at least five students in full day kindergarten over the entire study period. This left us with 19,935 eligible students in half day kindergarten and 19,904 eligible students in full day kindergarten; numbers in FDK increased as new classrooms were added. Columns 3 and 4 of Table 1 report selected characteristics of these students. While the FDK students still have worse outcomes than eligible half day students, simply restricting the table to students eligible for FDK eliminates about two-thirds of the differences and reduces the mean age difference to four days.

The Outcome Variables The province-wide Foundation Skills Assessments in reading, numeracy, and writing in grade 4 each year since 1999 form the backbone of our outcome variables. Scaled scores (mean ~ 0 and SD ~ 1) are available for reading and numeracy for about 90 percent of grade 4 students each year, and five-point scales are available for writing. The 10 percent of students who do not have a scaled score have either a blank score or "no meaningful response." In addi- tion to FSA scores, we have an indicator for gifted students and for special needs students.

Cunha, Heckman, and Schennach (2010) recom- mend producing a composite measure from disparate measures by weighting the disparate measures using coefficients from a regression in which the depend- ent variable is a later outcome of interest. While we report the impact of full day kindergarten on reading and numeracy FSA scores as well as on the likeli- hood of being classified as having a special need, our preferred outcome is the aggregated measure of educational attainment in grade 4 because the stu- dents who do not have scaled scores are not missing at random; dropping them from the analysis could introduce bias. (More detail on the weight calcula- tion, which uses grade 10 educational attainment as the dependent variable, is provided in appendix A.)

Factors Affecting Participation in FDK When there is choice in which school to attend, proximity is a well-established factor affecting

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

594 William P. Warburton, Rebecca N. Warburton , and Clyde Hertzman

Table 1 Descriptive Statistics for All Students and for Our Sample, by Full Day and Half Day Kindergarten

All Students Sample

HalfDay Full Day HalfDay Full Day

Kindergarten characteristics Self-declared Aboriginal 5.00% 23.60% 16.4% 7.1% Age 5.523 5.493 5.49 5.48 Standard deviation of age 0.29 0.297 0.29 0.29 French Immersion school 7.60% 0.60% 1.6% 0.2% ESL 9.30% 74.30% 84.2% 95.1% Self-declared home language not English 1 2.40% 68.40% 72.9% 86.8% Live in a tracted area 70.70% 79.70% 100.0% 100.0% Graduation rate in census tract NA NA 76.3% 74.3%

Grade 4 characteristics Special needs classification in grade 4 6.70% 9.20% 4.8% 4.7% Classified as gifted in grade 4 1.40% 0.30% 0.7% 0.4% N 301,527 27,885 19,935 19,904

Foundation Skills Assessments Grade 4 Reading FSA Mean 0.041 -0.358 -0.21 -0.28 Standard deviation 0.918 0.908 0.89 0.89 N 275,894 23,664 18,127 17,797 No record Reading 1.70% 2.90% 1.7% 1.8% No meaningful response Reading 6.80% 1 2.20% 7.4% 8.8%

Grade 4 Numeracy FSA Mean 0.061 -0.16 -0.06 -0.05 Standard deviation 0.943 0.997 0.97 0.98 N 274,685 23,527 18,023 17,669 No record Numeracy 1.70% 2.90% 1.8% 1.7% No meaningful response Numeracy 7.20% 12.70% 7.8% 9.5%

Grade 4 Writing FSA Does not meet expectations Writing 6.00% 7.70% 5.3% 5.8% Borderline does not meet 2.40% 2.50% 2.2% 2.0% Meets expectations 76.00% 69.30% 76.9% 75.8% Borderline exceeds 4.70% 3.10% 4.7% 3.6% Exceeds expectations 1.80% 1.30% 1.1% 1.5% No record Writing 2.00% 3.40% 2.0% 2.1% No meaningful response Writing 7.1 0% 1 2.80% 7.7% 9.2% N 301,527 27,885 19,935 19,904

Source: Authors' calculations using data from the BC Ministry of Education school census, Foundation Skills Assessments, and graduation records.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten Help Kids? 595

the decision (Card 1995; Alderman, Orazem, and Paterno 2001; Hastings, Kane, and Staiger 2007). Figure 1 reports the percentage of eligible students attending full day kindergarten by the distance to the nearest school with more than three students enrolled in full day kindergarten. Clearly, proxim- ity has a strong influence on choice of school for distances up to about a mile.

The popularity of distance as an instrument not- withstanding, we have reservations about accepting that distance to school is uncorrelated with personal characteristics. House prices are known to be higher near desirable schools because parents (particularly those with higher incomes, or those more interested in schooling) relocate to improve their children's access, making distance not fully exogenous (Gib- bons and Machin 2008).

For this reason, we construct a different measure of full day kindergarten availability - the percentage of all other kindergarten students in each census tract, in each year, who attend full day kindergarten. (This variable is calculated separately for each

student.) Because the placement of full day kinder- gartens is expected to be endogenous (the census tracts that need full day kindergarten are more likely to get them), we include dummy variables for each of the census tracts. In this way, the instrument cap- tures the within-census-tract, over-time variation in availability of full day kindergarten, without being directly tied to distance. It might be argued that par- ents will move to schools with greater availability of FDK, making availability not fully exogenous; however, the problem is much less than with the relocation decisions described above because of the difficulty of knowing at time of registration for kindergarten (February for September) which schools will offer FDK in a given year.

In addition, to capture year-to-year changes in the characteristics of students in a census tract, we include the graduation rate of 15 year olds in each census tract in each year. Age 15 was chosen because students must remain in school until age 16; the entire population of students is thus repre- sented, accurately reflecting any changes in student characteristics over time.

Figure 1 Percent of Eligible Students in FDK by Distance to Nearest FDK

Source: Authors' calculations using data from the BC Ministry of Education school census.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

596 William P. Warburton, Rebecca N. Warburton, and Clyde Hertzman

Table 2 reports the coefficients from a regression predicting participation in full day kindergarten. An F test for the exclusion of our instrument, attendance in FDK in that census tract and year, exceeds 10, Stock and Yogo's rule of thumb for assessing the strength of an instrument (2002). French Immersion reduces the likelihood of full day kindergarten - for the most part, parents may choose full day kindergarten for their children or French Immersion, but not both.

Results

Table 3 reports two-stage least squares estimates of the impact of full day kindergarten. We use 2SLS because we are interested in average marginal effects, which are reasonably estimated by the linear approximation (Moffit 1999), and because non-linear models such as logits or probits are inconsistent in the presence of heteroscedasticity. The instrument is the percentage of children in FDK in the census tract, in that year, excluding the reference child. Dummy variables are included for each census tract and for each year. Variables are included to control for demographic characteristics in all columns except column 1.

Column 2 reports our preferred results. It indicates a statistically significant impact of .16 standard deviations in aggregated grade 4 educa- tional performance (which is a weighted average of grade 4 FSA performance and special needs flags; weights are based on historical data, as explained in appendix A). Column 1 reports the results of the same equation without the demographic character- istics or census tract graduation rate - the estimate of impact changes from .156 to .152. (This small change suggests that our instrument is valid.) Col- umns 3 through 8 report the results for different subgroups. Impacts remain statistically significant when Aboriginal students are dropped, and impacts for females are significant at the 10 percent level, but, in general, small sample sizes resulting in wide confidence intervals make it difficult to draw conclusions regarding differences in impacts across subgroups.

Table 2 Coefficients and Standard Errors from Ordinary Least Squares Regression of Student Characteristics on Participation in Full Day Kindergarten

Instrument3 0.935*** [0.062] Female -0.002 [0.004] Age -0.018** [0.007] Age = 6 0.033 [0.034] French Immersion -0.152*** [0.027] Aboriginal 0.138*** [0.035] ESL 0.244*** [0.037] Graduation rate of 1 5 year olds in census tract in reference year -0.007 [0.042]

Home language indicator variables6 English -0.092*** [0.016] Punjabi 0.092*** [0.015] Korean -0.040* [0.022] Vietnamese 0.002 [0.012] Spanish -0.049*** [0.013] Tagalog (Philipino) -0.047*** [0.01 1 ] Hindi -0.082*** [0.017] Persian 0.019 [0.029] French -0.021 [0.055] Mandarin -0.007 [0.014] Portuguese -0.027 [0.034] Polish -0.050 [0.039] Japanese -0.048 [0.030] Russian -0.071** [0.029] Arabic 0.061* [0.032] German -0.132* [0.070] Urdu 0.028 [0.030] Italian -0.086 [0.062] Other non-English -0.018 [0.014]

Year indicator variables 1996 0.010 [0.010] 1997 -0.015 [0.009] 1998 -0.031*** [0.008] 1999 -0.020** [0.008] 2001 0.003 [0.008] 2002 0.016* [0.008] 2003 0.061*** [0.010]

Notes: Dependent variable is participation in full day kindergart- en. Robust standard errors clustered on census tract in brackets. N = 39,839. instrument is the percentage of all other kindergarten students in each census tract, in each year that attend full day kinder- garten. This variable is calculated separately for each student. bOmitted language is Chinese including Cantonese. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors' calculations using data from the BC Ministry of Education school census and graduation records.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten Help Kids? 597

¡S. 3 2 Oi -Q 3 CO 4- ' C a;

Q k_ v2 c § 03 "O c >N O 5 o tí s. E oj _£Z O wn £ £ 03 '^5 ,a ÜJ un

I- Sç in CM I- CM

S II gõ i I

I » jo •* I i o gõ i » •* o O

Ï .* I g,.á¡

.* Si £> o Ín ^ g S S

£> Ď * o-

gO g ™ ° ^ 5

"? 3 ä O ̂ .S¿ ^ «- cm O Oj ra S to no Ln S STÌ3 ra E # .£ >-

to £! sL 8 E cD # ld >- oo |>o sL

fig cD ld " oo °

O

-O E C ^ -O Qj E C (D O ■fa ^ O OI r ON PM O o, -£= <0 O un "<fr O S?1!

O o, s -£= <0

§> O ♦ g

un 2 "<fr

§ O

* S« ¡rü g - ° * 5 §

nj i - i -O F CÜ -O a» F ri ̂ Ü-& to 3 ^ S ^

2 IT! 9j ã S ^ ^

g °° S §05 S g S °° ®* c -c T : ^ o

rô* "T3 ^ - £5 a> ^ o TO Q

s|i = m$n o

-û mo g>0 = < -û CM mo ^ < < vo

ö

tn "O Qj o £,-âJ ° w M îîi a I s

< £, »

w S 2

g}<0 Zn m o O

LO "O a» o fa ̂ o o> vo S |1 2iO

s J.l|s m o 2iO m o ^ £ 0

& "S tí -O TO _3 01 CLI C Ü .£ "» .£ a»

^ s "S Ç3 ^ TO .TO Ç3 TO jo 'S "° S > jo = > vî - «2 £• o „ - 5 ïS'l?

o „ ? -ï c S S § 6 i" sis I S $ láoi

$ E 8 o v+- O a> cr> c fO OC ra c O c a» t: H3 s» a> -a c

Q 3 u_ O t3 S. E a» j= <-4- O co Ä to

jQ J ^ tf a £ tf CO -I ,< CO I - Ol

-C KJ ¿_ C on a» <¡¿ • S r- , -o it Il § ûc u H ^ CT* oj ^ 2 ^ no in -2 Oi

£S ■§ < ^

^ S- p 2 <

Is 3 00 ° ¡2 ft ■g on o ^ < 5 1 g ar 0 g to -g to e ■£ to aï to oc

11 il .2 m v-> G !=: to a»

¿o ^--S 2 o no ^ °9 2 .E c "S to ^ ¿o ̂3 - 2 o ^ no 2 .E c "S to -il <c jřS ^ I ° tí§

-§ p ^ ° 2 ^ C5 ? = ■§ a Qj I fO Qj fO ^ Q.

"5 l2 S jd

ë .s I !* Il „ il 1 8 g, 2

„ § o m II

- Tf -â; wo "2 Z; f ! a ^ g -5^p wo §E 1 ^ i -5^p S 0 a"S ö < I S1 -g II - £ ^ TO ^

_ o S S Cr, _ O = 5; ̂ ^ g- u m to dj c ;gai-o*rC g- u m to o

dj c 2 ^ o to -p o ^ o ^ ™ 3 w ? fN >■ I 8 m I P

^ ±i h C. ? =3 - ^ oj j: w fN >■ m P ±i h C. =3

U 0

j: 111 i i §

* 0

<1° 111 I #- ̂ flj o v <-> ■ A - •B ■ A

s § - 8 E 3 o> -"=

I -a J « •% M F ë 2 g .I S 5 è goo ™ O. S ci o» co - ; tí r y

1,5" -sl-g >s a? 2 2 fe >. ^ a? Si

2 5 5

t: 2

s E fe

f >. E ^ 5 t: s E . E LJ -•- • -i-» . ̂ LJ ro -•- ^ • -i-» ^ ^

I 3 Ü a» I <L» ^ ai 3 a» <L» ai £ 0 vS .E S 0 ->* <y S § g 2 ^ "S fc I S -a ß fc g--^ -a S 45 3 c - 7, "O •- c

c -p - w 5 E . •- 3 = Ä -p ¿ Ü2 to . =

m TO ̂ O 3 - r ° C m TO _3 t: _u - ( r v O S? -g -SQ^JP-I 1 i -S -S 1 S 3

-s -Q c 5 'ai o 3 -s s f

-Q i «

c 5 « 1 i

o -te 3

5 = > J2 S5<J o I

= £ g-

> J J2

-p ¿ S5<J ^ L- f

I „ £ gill

g- -p <=11

^ §1;

I latitili „

.-I S ä Já3é¿ i &ii 5

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

598 William P. Warburton, Rebecca N. Warburton, and Clyde Hertzman

The last five columns of Table 3 investigate the impact of FDK on different outcomes. The depend- ent variable in column 9 is aggregated educational attainment in grade 4 made using only data from the grade 4 FSA. The result is unchanged.4 For the remainder of outcomes examined, results are statistically significant only for reading and only at the 10 percent level, so no solid conclusions can be drawn. With that important caveat, the signs of the coefficients indicate that FDK has a larger impact on reading than on numeracy. Nonetheless, FDK is likely to improve numeracy, increase the likelihood of having a grade 4 FSA score, and de- crease the likelihood of being classified as having a special need.

Discussion

This study has estimated the average marginal effect of increasing attendance in full day kindergarten on aggregated grade 4 educational attainment for kin- dergarten students who self-declared as Aboriginal or who had an ESL designation. Because we have controlled for measured and unmeasured charac- teristics of kindergarten students and their schools, we believe that our study represents a significant advancement in the understanding of the impacts of full day kindergarten. Nonetheless, the applicability of these results to other segments of the population, or to other jurisdictions and programs, is unknown.

Does FDK help kids? We have shown that FDK improves outcomes for disadvantaged BC kids at grade 4. Whether these results will persist cannot be determined from these data. The results in this paper are consistent with 50 years of research on the impact of education (Card 1999): additional education has long-term beneficial impacts. They are also consistent with recent studies that find long-term impacts from kindergarten (Caseio 2009; Dhuey 2010; DeCicca and Smith 2011). The results are at odds with findings regarding the impact of full day kindergarten that are primarily based on

observational studies using ECLS-K data (Le et al. 2006; Aos, Miller, and Mayfield 2007). However, weaknesses in ECLS-K studies of the impact of FDK are discussed at length in the literature (Cooper et al. 2010; Lash et al. 2008), and the need for better studies to control for potential selection bias is ac- knowledged by the authors (Le et al. 2006; personal communication).

Currently, British Columbia is spending $130 million per year on FDK. At this stage it would be premature to conclude that FDK is a worthwhile public expenditure, but the potential is clearly there. Further study, longer follow-up, and use of rigorous methods will be required to substantiate or refute the supposition.

Conclusion

This study has examined the impacts of full day kindergarten as provided for ESL and Aboriginal students in British Columbia. In direct contrast to studies based on ECLS-K, this study finds that for these groups FDK has significant benefits through grade 4. However, many questions remain unanswered. Do other groups and jurisdictions experience similar benefits? What types of inter- ventions are most effective? What effect does full day kindergarten have on families and on children's health and well-being? Do the effects fade? And is there a net social benefit - that is, should govern- ment fund full day kindergarten?

Finding reliable answers to these questions is important for public policy. In this study we were fortunate to have access to data from a natural experiment (a staged implementation that created natural control groups), but policy-makers do not need to rely on good fortune. They can and should intervene to ensure that proper randomized trials are built into the implementation of FDK programs so that they can make decisions based on the highest level of evidence.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten Help Kids? 599

Appendix Aggregated Measure of Grade 4 Educational Attainment

As recommended by Cunha, Heckman, and Schennach (2010), we created an aggregate measure of grade 4 educational attainment (using grade 4 Foundation Skills Assessments and special needs classifications and a later outcome of interest, in this case, grade 10 provincial exam results). Much international work has been done on the attainments of students in grade 10, providing more context for grade 10 educational attainment than for grade 4 attainment, and making grade 10 outcomes an appropriate basis for assessing the importance of grade 4 attainments.

British Columbia replaced grade 10 FSAs with provincial exams in grade 10 mathematics, language arts, and science in 2003-04. The exams are set provincially, and marking is audited for consistency. Like the FSA, the grade 10 exams measure important skills in the provincial curriculum. Unlike the FSA, they are required for graduation, and for this reason, the educational attainment of those who do not complete the exams is less than those who do complete them.5 Like PISA, the exams cover the domains of reading, mathematical, and scientific literacy.

We assume that a latent measure of educational attainment in the domains of language arts, mathemat- ics, and science exists. For convenience, we assume that this latent measure is normally distributed with a mean of 0 and a standard deviation of 1. We further postulate that the rank order of performance on the standardized exams such as the grade 10 provincial exams, the grade 10 FSA, and PISA would follow the rank order on the latent measure of educational attainment. This seems reasonable based on the study by Cartwright et al. (2003) who examined the FSA and PISA scores of 2,800 grade 10 students who participated in both FSA and PISA in 2000.

To estimate the relationship between grade 4 performance and this latent measure of educational at- tainment, we divide grade 10 performance into 20 ranked groups based on their average mark on the three provincial exams. We then use ordered probit to parameterize this relationship. The sample consists of the 43,048 students born in 1991 who were in the BC school system each year from 1997 to 2005. The means of their grade 4 FSA scores, special needs classifications, and grade 10 exam results are presented in Table Al . We have scaled scores for all but about 7.5 percent of reading and numeracy FSAs. About 1 1 percent of stu- dents in grade 4 had a special needs classification; 7.8 percent of students did not have any grade 10 marks.

The coefficients are presented in Table A2. Grade 4 reading and numeracy results are strongly predictive of grade 10 performance. Missing values for reading and numeracy are associated with half a standard deviation decrease in grade 10 performance. Special needs classifications are associated with about one- fifth of a standard deviation decrease in grade 10 performance. Columns 2 and 3 show that the relationship between performance on grade 4 FSAs is not sensitive either to controlling for special needs and gifted status or excluding dropouts from the sample.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

600 William P. Warburton , Rebecca N. Warburton , and Clyde Hertzman

Table AI Descriptive Statistics 1991 Cohort

Grade 4 characteristics Special needs classification in grade 4 1 1 .5% Classified as gifted in grade 4 1 .7% N 43,048

Foundation Skills Assessments Grade 4 Reading FSA Mean -0.023 Standard deviation 0.936 N 39,847 No record Reading 4.5% No meaningful response Reading 3.0%

Grade 4 Numeracy FSA Mean -0.015 Standard deviation 0.938 N 39,847 No record Numeracy 4.5% No meaningful response Numeracy 3.0%

Grade 4 Writing FSA Does not meet expectations Writing 3.9% Borderline does not meet 4.2% Meets expectations 86.2% Borderline exceeds 4.9% Exceeds expectations 0.8% No record Writing 6.0% No meaningful response Writing 1 .7% N 43,048

GradelO Grade 10 mean 66.9 Standard deviation 14.1 No grade 10 marks 7.8%

Grade 10 exams mean 64.2 Standard deviation 14.3 No grade 1 0 exam marks 7.8

Source: Authors' calculations from BC Ministry of Education student level data, Foundation Skills Assessments, and provincial exam records.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten Help Kids? 601

Table A2 Coefficients and Standard Errors from Ordered Probit Regression of Student Characteristics on Vingtile of Grade 10 Performance

Variables (!) (2) (3) Base No Special Needs No Drop Outs0

Grade 4 Reading FSA 0.4259*** [0.007] 0.4312*** [0.007] 0.4369*** [0.008] Grade 4 Numeracy FSA 0.4039*** [0.007] 0.4083*** [0.007] 0.4091*** [0.007] No record Reading -0.6569*** [0.106] -0.7034*** [0.106] -0.5777*** [0.109] No meaningful response Reading -0.4678*** [0.041] -0.4944*** [0.041] -0.4673*** [0.041] No record Numeracy -0.5521*** [0.103] -0.5982*** [0.103] -0.6495*** [0.107] No meaningful response Numeracy -0.4481*** [0.038] -0.4683*** [0.038] -0.4513*** [0.039] Gifted 0.2897*** [0.040] 0.3246*** [0.040] Special need -0.2183*** [0.017] -0.2125*** [0.017] Does not meet expectations Writing -0.2727*** [0.026] -0.2914*** [0.026] -0.2810*** [0.027] Borderline does not meet -0.1873*** [0.025] -0.1961*** [0.025] -0.1797*** [0.025] Borderline exceeds 0.2952*** [0.024] 0.2975*** [0.024] 0.2902*** [0.024] Exceeds expectations 0.5019*** [0.057] 0.5030*** [0.057] 0.4910*** [0.058] No record Writing -0.0458 [0.052] -0.0713 [0.052] -0.0626 [0.052] No meaningful response Writing -0.2655*** [0.042] -0.2871*** [0.042] -0.2902*** [0.043] Observations 43,048 43,048 42,058

Notes: Dependent variable is vingtile of grade 1 0 performance. Standard errors in brackets. 'Excludes students without grade 10 marks who were not in the BC school system in 2006 and one subsequent year. *** p<0.01 , ** p<0.05, * p<0.1 . Source: Authors' calculations from BC Ministry of Education student level data, Foundation Skills Assessments, and provincial exam records.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

602 William P. Warburton, Rebecca N. Warburton , and Clyde Hertzman

Notes

Although this paper was undertaken at the request of BC's Representative for Children and Youth, the views expressed in this paper are solely those of the authors. This paper is not to be understood as a report under the Representative for Children and Youth Act, and no representation is made that it expresses the views of the Representative for Children and Youth, the Ministry of Education, or the Government of British Columbia. The authors thank the Michael Smith Foundation for Health Research for financial support and acknowledge the com- ments from two referees and from session participants at the Canadian Economics Association annual meeting. Any errors are the responsibility of the authors.

1 DeCicca (2007) argues that within-race differences be- tween full day and half day kindergarten students are small.

2 Half day programs typically do not have nap times, so sending children to two half day programs is considered too taxing for five year olds.

3 In some cases school officials enter the school's postal code when the student's postal code is unavailable.

4 The correlation between predicted educational attain- ment using special needs and not using special needs is greater than .99.

5 We check this contention by comparing another measure of educational attainment, the existence of any grade 12 mark, against the ordering of grade 10 outcomes. This measure also increases monotonically with grade 10 outcomes.

References

Alderman, H., RF. Orazem, and E.M. Paterno. 2001. "School Quality, School Cost, and the Public/Private School Choices of Low-Income Households in Pak- istan." Journal of Human Resources 36 (2): 304-26.

Aos, S., M. Miller, and J. Mayfield. 2007. Benefits and Costs ofK-12 Educational Policies : Evidence-Based Effects of Class Size Reductions and Full-Day Kinder- garten. Washington State Institute for Public Policy.

Bedard, K., and E. Dhuey. 2006. "The Persistence of Early Childhood Maturity: International Evidence of Long- Run Age Effects." Quarterly Journal of Economics 121 (4): 1437-72.

Cannon, J., A. Jacknowitz, and G. Painter. 2006. "Is Full Better Than Half? Examining the Longitudinal Effects of Full-Day Kindergarten Attendance." Journal of Policy Analysis and Management 25 (2): 299.

Card, D. 1995. "Using Geographic Variation in College Proximity to Estimate the Return to Schooling." In Aspects of Labour Market Behaviour , edited by L.N. Christofídes, E.K. Grant, and R. Swidinksy. Toronto: University of Toronto Press.

- . 1999. "The Causal Effect of Education on Earnings." In Handbook of Labor Economics , vol. 3, edited by O. Ashenfelter and D. Card, 1801-63. Amsterdam: Elsevier Science.

Cartwright, F., D. Lalancette, J. Mussio, and D. Xing. 2001. "Linking Provincial Student Assessments with National and International Assessments." Cat no. 81- 595-MIE - No. 005. Ottawa: Industry Canada.

Caseio, E.U. 2009. "Do Investments in Universal Early Education Pay Off? Long-Term Effects of Introducing Kindergartens into Public Schools." NBER Working Paper no. 14951.

Child Trends. 2003. Full Day Kindergarten. Accessed 28 May 2009. http://www.childtrendsdatabank.org/ pdf/1 02_PDF.pdf.

Cooper, H., A. Batts Allen, E.A. Patall, and A.L. Dent. 2010. "Effects of Full-Day Kindergarten on Academic Achievement and Social Development." Review of Educational Research 80 (1): 34-70.

Cunha, F., J.J. Heckman, and S.M. Schennach. 2010. "Estimating the Technology of Cognitive and Noncog- nitive Skill Formation." Econometrica 78 (3): 883-931.

DeCicca, P. 2007. "Does Full-Day Kindergarten Matter? Evidence from the First Two Years of Schooling." Economics of Education Review 26 (1): 67-82.

DeCicca, P., and J.D. Smith. 2011. "The Long-Run Im- pacts of Early Childhood Education: Evidence from a Failed Policy Experiment." NBER Working Paper no. 17085.

Dhuey, E. 2010. "Who Benefits from Kindergarten? Evi- dence from the Introduction of State Subsidization." Forthcoming in Educational Evaluation and Policy Analysis. Accessed 3 February 2011. http://homes. chass.utoronto.ca/~edhuey/datastore/files/docs/Who BenefitsfromKindergarten_Oct20 1 0.pdf.

ECLA. 2008 "Expanding Early Learning in British Columbia for Children Aged 3-5." Accessed 28 May 2009. http://www.bced.gov.bc.ca/early_learning/pdfs/ ecla_report.pdf.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions

Does Full Day Kindergarten Help Kids? 603

Gibbons, S., and S. Machin. 2008. "Valuing School Qual- ity, Better Transport, and Lower Crime: Evidence from House Prices." Oxford Review of Economic Policy 24 (1): 99-119.

Hastings, J.S., T.J. Kane, and D. Staiger. 2007. "Prefer- ences and Heterogeneous Treatment Effects in a Public School Choice Lottery." Accessed January 2011. http://cowles.econ.yale.edu/conferences/2007/75ann/ hastings.pdf.

Karweit, N.L. 1987. "Full or Half Day Kindergarten - Does It Matter?" Report no. 11. Johns Hopkins University, Center for Research on Elementary and Middle Schools.

Lash, A., S. Bae, V. Barrat, E. Burr, and T. Fong. 2008. "Full-Day Kindergarten and Student Achievement: A Literature Review." Published online by the Regional Educational Laboratory West. Accessed 14 January 201 1 . http://www.wested.org/online__pubs/REL_West_ FDK_Brief_ALL.pdf.

Le, V., S. Nataraj-Kirby, H. Barney, C. Messan-Setodji, and D. Gershwin. 2006. "Ready for School: Can Full- Day Kindergarten Level the Playing Field?" RAND.

Accessed 33 November 2012. http://www.rand.org/ pubs/research_briefs/RB9232.html.

Lee, V.E., D.T. Burkám, D.D. Ready, J. Honigman, and S.J. Meisels. 2006. "Full-Day versus Half-Day Kinder- garten: In Which Program Do Children Learn More?" American Journal of Education 112 (2): 163-208.

Moffit, R.A. 1999. "New Developments in Econometric Methods for Labor Market Analysis." Handbook of Labor Economics, vol. 3 A, edited by O. Ashenfelter and D. Card, 1368-94. Amsterdam: Elsevier Science.

Plucker, J.A., J.J. Eaton, K.E. Rapp, W. Lim, J. Nowak, J.A. Hansen, and A. Bartleson. 2004. "The Effects of Full Day versus Half Day Kindergarten: Review and Analysis of National and Indiana Data." Accessed 31 January 2011. http://ceep.indiana.edu/projects/project. php?id=76&category=3 .

Statistics Canada. 2005. "Postal Code Conversion File (PCCF), Reference Guide October 2005." Cat. no. 92F01 53GIE. Ottawa: Industry Canada.

Stock, J.H., and M. Yogo. 2002. "Testing for Weak In- struments in Linear IV Regression." NBER Technical Working Paper no. 284.

Canadian Public Policy - Analyse de politiques, vol. xxxviii, no. 4 2012

This content downloaded from 185.44.77.40 on Mon, 16 Jun 2014 15:48:24 PMAll use subject to JSTOR Terms and Conditions