Children and Youth Services Review 40 (2014) 13–19

Contents lists available at ScienceDirect

Children and Youth Services Review

j ourna l homepage: www.e lsev ie r .com/ locate /ch i ldyouth

Does reform in Kazakhstan improve access to childcare? Evidence fromnationally-representative surveys

Nazim HabibovSchool of Social Work, University of Windsor, 401 Sunset Avenue, Windsor, Ontario, N9B3P4, Canada

E-mail address: nnh@uwindsor.ca.

http://dx.doi.org/10.1016/j.childyouth.2014.02.0130190-7409/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 8 December 2013Received in revised form 20 February 2014Accepted 22 February 2014Available online 1 March 2014

Keywords:Early childhood education and careDaycareCentral AsiaKazakhstan

During Soviet era, Kazakhstan enjoyed universal free-of-charge access to childcare. After the commencementof economic and political transition, and the achievement of independence in 1991, attendance in childcare pro-grams dropped significantly. This reduction in attendancewas accompanied by a growing gap in access caused bywealth, language,mother's education, and regional disparities. Responding to the reduction in attendance aswellas growing inequalities with respect to attendance, the government of Kazakhstan initiated a bold program ofreforms aimed at improving access to childcare. This paper represents an initial assessment of the success ofthese reforms using a unique set of nationally-representative surveys. We found that, in general, the reformwas successful in increasing childcare attendance. However, the results of the reform fell short of their target.However, the reform significantly reduced the role of household wealth as a barrier to attendance. Nevertheless,the wealth of a household remains an important determinant of attendance. Additionally, although the reformsuccessfully mitigated gaps in attendance which were based on language spoken and education of mother,regional disparities remain significant.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The purpose of this paper is to assess the reforms whose aim hasbeen to improve access to childcare in Kazakhstan. Like other formerSoviet Union republics, Kazakhstan enjoyed universal free-of-chargeaccess to childcare (Habibov, 2012a, 2012b; Lokshin, 2004). After thecommencement of economic and political transition and independence,which was achieved in 1991, the inherited childcare system sufferedfrom severe under funding, leading to a considerable drop in childcareattendance (Anderson & Heyneman, 2005; Habibov, 2010a, 2010b;Stewart & Huerta, 2009; UNESCO, 2004). Furthermore, during the initialperiod of transition, the drop in attendance did not occur evenly. Signif-icant gaps in attendance surfaced based on family income and resi-dence, as well as language spoken in the family and the educationlevel of the mother (Anderson, Pomfret, & Usseinova, 2004; Giddings,Meurs, & Temesgen, 2007).

Responding to the reduction and growing inequalities in attendance,the government of Kazakhstan initiated a bold program of reform(UNESCO, 2004, 2005). The program set up a specific numeric targetsfor increased childcare attendance. It predicted that 70% of theKazakhstan children would attend childcare by 2010. It also envisagedsignificant reductions in the existing inequalities in childcare atten-dance. To achieve the above-mentioned goals, the program initiated acomprehensive set of measures, including an improved accountability

on the part of the regional administration, for developing andmaintain-ing childcare facilities, a significant increase in the number of newlybuild facilities, various tax breaks, the expansion of childcare in theKazakh language, and financial support for low-income families.

The information about the reforms is still very limited. Hence, thispaper represents an initial assessment of the success of the reformsusing a unique series of high-quality nationally-representative surveys.The first survey was conducted in 2005 at the inception of the reforms.This survey provides us with a baseline, or a snapshot of the situation asthe reforms were beginning to take place. The second survey wasconducted in 2011, by the time the targets of the reforms should havebeen achieved. Consequently, the specific research questions whichare addressed by this paper are:

(1) How did childcare attendance change after the reforms?(2) How did the determinants of childcare attendance change after

the reforms?

Answering the above-mentioned questions has both theoretical andpractical significance. From a practical standpoint, the findings of thispaper will help policy-makers, social administrators, and internationaldonors understand and evaluate the successes and pitfalls of thereforms. The finding of this paper will also help with impendingdecision-making about the future course of the reforms. It must alsobe highlighted that in addition to the reforms aimed at improving accessto childcarewhich have already been put in place in Kazakhstan, similarreforms are also currently being planned in the neighbouring countries

14 N. Habibov / Children and Youth Services Review 40 (2014) 13–19

of Kyrgyzstan, Tajikistan, and Uzbekistan (Government of Kyrgyzstan,2007; Government of Tajikistan, 2010; Government of Uzbekistan,2008). Therefore, the findings of this paper will allow neighbouringcountries to learn from Kazakhstan's experience in reforming childcare.From a theoretical standpoint, the findings of this paper contribute to abetter understanding of the over-time variation in childcare attendanceand its determinants in transitional countries.

2. Context

2.1. Soviet era

During the Soviet Union era, the development of childcare inKazakhstan followed a unified all-union model. This model had threeexplicit objectives (Habibov, 2012a). The first objective was to provideuniversal services for children and their families in order to reduceexisting income, educational, and regional inequalities. The secondobjective was to facilitate the increase in womens' participation in thelabour market for economic reasons, and to improve the status ofwomen in family and society. The third objective was to enhance thepsychosocial and cognitive development of children through earlychildhood education. These objectives were achieved through a two-leg childcare system, where the first leg consisted of state-run childcarefacilities which were fully funded by the state budget, and the secondleg consisted of enterprise-run facilities funded by large enterprisesand collective farms.

In the last years of the Soviet Union, access to childcare in CentralAsia was as high as in the countries of Eastern Europe (Giddings et al.,2007). Moreover, by 1991, Kazakhstan enjoyed the highest level ofpreschool access among all the countries of Soviet Central Asia (Penn,2004). This emphasis on childcare access was hardly surprising givenKazakhstanis view on children (UNESCO, 2004). In Kazakhstan a childis considered to be a gift, and families with children are regarded aslucky. Utmost importance and respect are given for the opportunity ofraising a child, and this is particularly true with respect to teachingchildren in their earliest years. The great Kazakh humanist andphilosopher Abai Kunanbaev stressed that children have two needsto be satisfied: to eat and sleep, and to learn.

Indeed, the Soviet era childcare system stressed mutuality andcitizenship through the high-quality universal delivery of service(Penn, 2004). It must be noted however that as it followed the all-union model, the childcare system in Kazakhstan during the Soviet erawas highly-centralized, rigid, and teacher-oriented. Despite this, itachieved significant success in mitigating existing social disparitiessuch as income, education, regional variations, and it allowed a sig-nificant number of women to join and excel within labour market(Lokshin, 2004; Stewart & Huerta, 2009).

2.2. Early transition

The dissolution of the Soviet Union, accompanied by the transitionfrom an administrative-command economy to a market economyprolonged economic recession, and political instability, and thus nega-tively affected the childcare system. Both legs of childcare systeminherited from the Soviet era were badly shaken (Habibov, 2012b).The state-run enterprises and collective farms were transformed to pri-vate businesses and farms, which then halted their support to childcarefacilities. These profit-focused enterprises did not consider it appropri-ate to spend precious financial resources on supporting a social infra-structure which they saw as being the responsibility of the state.Furthermore, newly-emerged enterprises and farmers often did nothave sufficient funds to fund childcare facilities. At the same timeduringtransition, the drastic reduction in state revenue severely diminishedthe ability of the state to support a childcare system. As a result,childcare attendance plummeted. At the time of the dissolution of theSoviet Union in 1991, about 50% of children attended childcare in

Kazakhstan (Penn, 2004). In 1998, the share of children attendingchildcare dropped to just 11%.

Furthermore, the plummeting attendance in childcarewas not even-ly distributed among the different groups in society. In fact, the collapseof the former childcare system in the period of early transition led to adramatic increase in inequalities in access to childcare. Evidence avail-able from Kazakhstan and neighbouring transitional countries suggestsseveral main reasons for the growing disparities in access to childcare.

During the early days of transition, the disparity in childcare accessbetween wealthier and poorer households grew considerably. The lackof public and business funding for childcare during transition led tothe skyrocketing of both official and unofficial out-of-pocket fees foraccess. As a result, household wealth emerged as a strong determinantof attendance inasmuch as children from the wealthier householdsbecame the most likely to attend childcare (Habibov, 2012a, 2012b).

A considerable disparity in access to childcare existed betweenchildren from Russian-speaking and non-Russian-speaking households(UNESCO, 2005). During the Soviet era, Kazakhstan's elite tried to getRussian-speaking education for their children since Russian was theofficial language of the Soviet Union. Without mastery of the Russianlanguage, one could not receive a high-ranking position in Kazakhstan.Thus, most of the childcare facilities of the time had Russian as thelanguage of instruction. Russian speaking children from native Russianfamilies and Russians-speaking native-Kazakh elite were at an advan-tage because they also communicated Russian at home. On the contrary,Kazakh-speaking and other native language speaking families didnot send their kids to childcare because childcare was in Russianand not their own languages. Therefore, children from Russian-speaking households had a higher likelihood of childcare attendance.

The disparity in attendance was also be explained by education(Giddings et al., 2007;Mertaugh, 2004).Motherswith higher educationallevels could be more aware of the benefits of early childhood education.They could also have a better understanding of how deregulatedchildcare works in terms of rules, regulation, fees, and availability.As a result, this may have been another reason that children withmothers with higher educational levels had a higher likelihood ofattending childcare.

Regional disparity became noticeable during the early phases oftransition (UNESCO, 2004). The former centralized system of childcarehad been abolished since independence. However, no mechanismthrough which to match regional needs with financial resources forchildcare was developed in its stead. Regional governments holddifferent attitudes towards childcare. A disparity has been notedbetween the amounts of resources different regional governmentsprovide to childcare. Furthermore, there is noticeable inequalitywith respect to the resources available to regional governments.Typically, the Astana region, which includes the capital Astana, andthe Almaty region, which includes the former capital and the largestcity of the country, have access to greater resources for childcare. Atthe same time, the demand for childcare is much higher in the Astanaand Almaty regions. As a result, attendance greatly fluctuates betweencapital cities and regions, and between the regions.

2.3. Reforms

Against this background, the government of Kazakhstan initiatedreforms aimed at addressing unequal access to childcare. The re-forms commenced in 2004 and consisted of several steps (UNESCO,2004). In early 2004, the government instructed heads of regionalgovernments that they must personally take the responsibility for re-storing the network of childcare facilities across their regions. Wherepossible, mostly in urban areas, new childcare facilities were to bebuilt. In rural areas with relatively small densities of population, andwhere the construction of new facilities was unfeasible, childcareclasses were to be organized on existing school premises.

15N. Habibov / Children and Youth Services Review 40 (2014) 13–19

The next step was to be the implementation of the ambitious “StateProgramme on Education Development for 2005 to 2010.” The programfocused specifically on the issue of access to childcare for the period of2005–2010 (UNESCO, 2005). It proclaimed that an equal access todevelopment, health, and education should be ensured for all childrenin the country. One of the most important steps to ensuring childcareaccess stipulated by the program was compulsory preschool educationfor all children in Kazakhstan at the age of 3 by 2010. More specifically,it was planned that at least 75% of children would have access tochildcare by 2010.

To achieve this goal, the government planned to construct 164preschool facilities which could provide 23,000 places and organize800 community preschool centres which could be attended by 20,000children (UNESCO, 2004). Childcare groups were to be established insecondary schools and preschool organisations. Support was to be pro-vided for the children from poor families in order to facilitate their ac-cess to childcare. Childcare in government-run facilities was to be fullyfunded by the state budget, and privately-run facilities were also to re-ceive considerable support in form of a wide range of tax benefits suchas exemption from Value Added Tax and land and property taxes wereprovided to private-run facilities to provide themwith more affordabil-ity for low-income families. Childcare facilities were allowed, and evenencouraged to raise funds and seek sponsorship for their activities.

Standardized regulations for the management of childcare serviceswere approved by the Ministry of Education (UNESCO, 2004). Thesestandards became mandatory for both public and private providers ofchildcare. The standards included a basic curriculum, basic healthcareservices, and meals for children. For instance, even as of the beginningof the reforms in 2005, between 100 and 92% of children receivedmeals while attending childcare.

To address the language gap, the government planned to increasethe number of childcare facilities with a native Kazakh language of in-struction. The program also stipulated the right of other ethnic groupsto access childcare in their native languages on the territories wherethey constituted the majority. In addition, the program specificallystressed the necessity for all children to master both the Kazakh andRussian languages, and to broaden their understanding of the cultures,traditions, and customs of different ethnic groups living in the country.

Finally, the Ministry of Labour and Social Protection was assignedwith developing effective social assistance schemes to provide benefitsto low-income families. It was assumed that such assistance wouldrender childcare more affordable for low-income families and henceeliminate the persistent gap between the wealthier and the poorerwith respect to access to childcare.

To summarize, the main goal of the reform was twofold:

(1) to increase childcare attendance to at least 75% by 2010;(2) to reduce the growing attendance gap caused by wealth,

language, mother's education, and regional determinants.

3. Method and data

3.1. Data

To capture change in childcare attendance and its determinantssince the beginning of the reforms, we use two rounds of the MultipleIndicator Cluster Surveys (MICS). The first round of the MICS wasconducted in 2005 just as the reforms began to be implemented. Thesecond round of the MICS was conducted in 2011 when the first stageof the reform had already been realized. Both rounds of the MICS usedthe same questionnaire and same sampling methodology. Hence, thedirect comparison of the data from both rounds allows us to uncoverand assess the results of the reforms.

The MICS is a high-quality cross-sectional nationally-representativesurvey designed and conducted by the United Nations Children's Fund(UNICEF) in co-operation with the national statistical authorities—in

this case the Kazakhstan's Agency of Statistics. TheMICSwas developedto provide analytical evidence with the intent of using it for thepurposes of advocacy for children, as stipulated by child-related goalsand commitments of the Millennium Declaration and the World Fit forChildren Declaration and Plan of Action (UNICEF, 2004). Consequently,theMICS collects information about child development, including infor-mation about attendance in childcare programs. The MICS also collectsinformation about the child (gender and age), the mother of the child(date of birth and education level), the household (wealth and mainlanguage spoken), and the place of residence (rural vs. urban andsocio-economic regions).

TheMICS has several advantages whichmake it especially pertinentto our study. Importantly, the MICS is a nationally-representativesurvey. Consequently, the results of estimations based on the MICScan be generalized for the whole country and so are useful in policyanalysis. In addition, theMICS collects information about childcare atten-dance in two ways. One way is through a binomial variable – attendedversus non-attended. Another way is with a continuous variable –

hours of attendance for those childrenwho attended childcare. Assessingboth indicators allows us access to a rich picture of changes in childcareattendance which resulted from the reforms. Finally, the questionnaireused in the survey allows us to link childcare attendance with the deter-minants of child, mother, household, and place of residence. It furtherallows us to fit a multivariate regression model to estimate the effectsof these determinants with respect to the probability of childcare atten-dance and hours of attendance. Due to these advantages, the MICS wasactively used for comparative evaluation of childcare attendance andits determinants (Habibov, 2012a, 2012b).

3.2. Outcome variables

The main outcome variable of interest in this study is childcareattendance. The MICS asked the caregiver of every child aged 36–59 months about the child's attendance at childcare. The actualquestion is constructed as follows: “Does (NAME OF THE CHILD) attendany organized learning or early childhood education programme,including in a private or government facility, including kindergartenor community child care?” We use “mother” for describing the care-giver throughout the following text inasmuch as the caregivers,in 99% of all cases are mothers, although a caregiver could also be afather or grandparents. The answer to this question is binomial—attended versus non-attended childcare.

The second outcome variable of interest in this study is hours ofattendance. For those children who participated in childcare, the MICSasked about the child's hours of attendance in childcare. The questionwas constructed as follows: “Within the last seven days, about howmany hours did (NAME OF THE CHILD) attend?” The answer to this questionis continuous—hours of childcare attendance.

3.3. Analytical strategy

To reveal the effects of the reformon childcare attendance and hoursof attendance, we commence with a descriptive and bivariate analysis.Keeping in mind that the main declared purpose of the reforms was toincrease the attendance to at least 70% in 2010, we compare statisticsfor childcare attendance and hours of attendance before the reformsin 2005 and after the reforms in 2011.

The next step involves a multivariate statistical analysis to uncoverthe changes in the factors explaining childcare attendance and hoursof attendance. The selection of the explanatory factors has been guidedby the goals of the reforms, as were highlighted in the previous sectionof the paper. Thus, wealth of household is measured by a continuouswealth index supplied in the data set. Here, we see that the higher thevalue of the index, the wealthier the household. The mother havingattended higher education is a dummy variable which has a value of 1if themother has higher education and 0 if mother does not have higher

16 N. Habibov / Children and Youth Services Review 40 (2014) 13–19

education. Russian-speaking household is a binomial variablewhich hasa value of 1 if themain language used in the household is Russian, and 0if otherwise. We use an array of dummies for the socio-economicregions of Kazakhstan. The omitted variable is for the Astana region(which includes the capital of Astana), and for the Almaty region(which includes the former capital and the largest city in the country).Additionally, rural is a dummy variable which has a value of 1 if thehousehold resides in a rural area and 0 if it resides in an urban area.

In addition, we also control for the age and gender of child, as wellas mother's age since previous studies have found significant age andgender bias in the neighbouring transitional countries of Central Asia(UNESCO, 2006, 2008; World Bank, 2000). Thus, male child is adummy variable which has a value of 1 if the child is male, and 0 iffemale. The child's age and mother's age are continuous variableswhich represent the age of the child and mother in years.

As shown above, we have two outcome variables of interest. Thefirst outcome variable is childcare attendance, which is measured as abinomial variable—attended versus not attended. Hence, to assess theeffects of the explanatory characteristics on childcare attendance, weestimate binomial logistic regression. The second outcome variable ishours of attendance which is measured as a continuous variable—number of childcare hours attended last week. One can estimate OLSregression inasmuch as the outcome is continuous. However, we canreasonably expect large numbers of children for whom the number ofchildcare hours has not been observable since they do not attendchildcare. If the number of children who do not attend childcare isconsiderable, and we do not explicitly take this situation into account,our estimations of OLS would yield inconsistent results. To adjust forthe considerable number of children with zero hours of attendance,we estimate the tobit regression model (Cameron & Trivedi, 2005;Wooldridge, 2002). This model explicitly takes into account the consid-erable number of children forwhomnohours of attendance in childcarehave been observed.

To uncover the changes in factors explaining childcare attendanceand hours of attendance from 2005 to 2011, we estimate binomiallogit and tobit regression models for each year separately. Followingthis, the test of equality of coefficients is used to assess whether coeffi-cients are statistically significant between the years. A significant χ2 ofthe test indicates that coefficients are significantly different betweenyears from a strictly statistical point of view. To facilitate interpretation,regression coefficients are shown in the form ofmarginal effects (Baum,2006). Similarly, to ensure a direct comparison of the effects of theexplanatory variables within the model, fully standardized regressioncoefficients are shown in italics (Long & Freese, 2006). All estimationsare adjusted for the clustering of observations.

4. Findings

We commence with descriptive and bivariate results reported inTable 1. Panel A of the table shows that out of 1549 children surveyedin 2005, only 242 children (approximately 15%) attended childcare.In contrast, out of 1906 children surveyed in 2011, 759 children

Table 1Descriptive analysis.

Panel A: Attendance

Years Attended Not attended

Observations Observ. % Observ. %

2005 1594 242 15.18 1352 84.822011 1906 759 39.82 1147 60.18Pearson χ2(1) = 258.0p-value = 0.000

Note: Data are rounded.

(approximately 40%) attended childcare. Pearson χ2, reported at thebottom of the panel, suggests that the across-time difference in atten-dance is statistically significant.

Panel B demonstrates that from the 242 children who attendedchildcare in 2005 these children attended childcare for an average of17.80 h per week. In comparison, from the 759 who attended childcarein 2011, these children attended an average 10.46 h per week. Thet-test of means reported at the bottom of the panel indicates that theacross-time difference in hours of attendance is statistically significant.

Panel B also confirms our expectations that hours of childcareattendance were not observable for a considerable number of children.Specifically, number of hours is not observable for 85% of children in2005 and for 60% in 2011. Under these circumstances, an estimationof OLS will provide inconsistent results. Consequently, tobit regression,which explicitly takes into account the considerable number of childrenfor whom no hours of attendance were observed, should be estimatedinstead.

The results of binomial logitmodels on the determinants of childcareattendance in 2005 and 2011 are reported in Table 2. The effect of sev-eral important explanatory variables dropped in 2011, as comparedwith 2005. Thus, having a mother with higher education is associatedwith an 85 percentage point increase in the probability of childcare at-tendance in 2005. In comparison, the effect of an educated motherwas reduced to 0.16 percentage points in 2011. However, a test of theequality of coefficients between years fails to reject the hypothesisthat these coefficients are different from a strictly statistical point ofview (p-value= 0.41). Similarly, being from a Russian-speaking house-hold is associated with a 64 percentage point increase in probability ofchildcare attendance in 2005. In contrast, the effect of being from aRussian-speaking household became non-significant in 2011. The testof the equality of coefficients suggests that these coefficients differ in astrict statistical sense (p-value = 0.04). Likewise, being from a wealth-ier household is associated with an 89 percentage point increase in thelikelihood of childcare attendance in 2005. In comparison, this effectplummeted to 10 percentage points in 2011. These significant resultsof the equality of coefficients test indicates that these coefficients differin a strict statistical sense (p-value = 0.00). Also, children living in theWestern Kazakhstan region had a lower probability of enrolment inchildcare in 2011 than in 2005, with this difference being statisticallysignificant. The same effect can be observed in theMangystau region, al-though the test of equality of coefficients fails to reject the hypothesisthat these coefficients are different across years from a strictly statisticalpoint of view.

The effect of several important explanatory variables increased in2011 as compared with 2005. Thus, the negative effect of women'sage on the likelihood of their child's attendance in childcare slightly in-creased in 2011 as compared with 2005. This increased is statisticallysignificant (p-value 0.037). However, in fairness, we should point outthat the magnitude of the effect is rather negligible. The increase inprobability of childcare enrolment in Atyrau, Kostanai, Pavlodar, andEastern Kazakhstan regions became significant in 2011, while it was in-significant in 2005. Again, the test of equality of coefficients suggests

Panel B: Hours of attendance

95% Confidenceintervals

Observ. Mean Std. Dev. Std. Err. Min Max

242 17.80 17.21 1.10 15.62 19.99759 10.46 12.69 0.46 9.56 11.37t-test of means = 7.1458p-value = 0.000

Table 2Determinants of childcare attendance.

Variables 2005 2011 Test of equality

χ2 p-value

Male child 0.226 −0.029 3.41 0.064(0.158) (0.024)

0.052 −0.029Child's age 0.137 0.007 0.48 0.487

(0.156) (0.003)0.032 0.045

Woman's age −0.001 −0.004* 4.33 0.037(0.001) (0.002)

−0.027 −0.054Higher education of mother 0.859*** 0.164*** 0.68 0.411

(0.170) (0.028)0.166 0.154

Russian-speaking household 0.644*** 0.050 3.32 0.048(0.187) (0.034)

0.119 0.039Wealth of household 0.891*** 0.101*** 11.47 0.000

(0.120) (0.012)0.397 0.291

Rural 0.076 0.059 0.40 0.528(0.235) (0.033)

0.017 0.060Aktobe region −0.082 0.085 0.71 0.398

(0.406) (0.060)−0.009 0.038

Atyrau region −0.333 0.212*** 7.85 0.000(0.349) (0.054)

−0.041 0.101Western Kazakhstan region 0.949** 0.392*** 0.00 0.048

(0.360) (0.049)0.094 0.173

Zhambyl region 0.518 0.140* 0.02 0.883(0.323) (0.056)

0.065 0.070Karagandy region 0.374 0.235*** 2.30 0.129

(0.319) (0.053)0.035 0.112

Kostanai region −0.0213 0.415*** 15.32 0.000(0.376) (0.049)

−0.002 0.182Kyzylordao region −0.035 0.141** 2.01 0.156

(0.368) (0.051)−0.004 0.079

Mangystau region −0.318 −0.190*** 2.15 .0142(0.343) (0.041)

−0.038 −0.123Pavlodar region 0.219 0.260*** 3.27 0.042

(0.378) (0.057)0.019 0.111

Southern Kazakhstan region −0.018 −0.079 0.74 0.390(0.328) (0.048)

−0.003 −0.053Northern Kazakhstan region 0.371 0.314*** 4.90 0.432

(0.382) (0.057)0.034 0.134

Eastern Kazakhstan region 0.139 0.282*** 5.01 0.025(0.407) (0.057)

0.013 0.122Number of observations 1,582 1,897Log likelihood −544.9 −1085.0Likelihood ratio test χ2 260.27 380.10Probability Nχ2 0.000 0.000

Note: Data are rounded.Outcome variable is childcare attendance. Binomial logit regression is estimated.Coefficients are shown asmarginal effects. Standard errors are adjusted for clustering andshown in parenthesis. Fully standardized regression coefficients are in italics.Significance: * p b 0.05, ** p b 0.01, *** p b 0.001.

17N. Habibov / Children and Youth Services Review 40 (2014) 13–19

that these differences are statistically significant. The same effect can beobserved in Zhambyl, Karagandy, Kyzylordao and Northern Kazakhstanregions, although the test of equality of coefficients fails to reject thehypothesis that these coefficients are different across years in a strictlystatistical sense.

It is also instructive to compare the changes in the magnitude of theeffects of the explanatory variables using fully standardized coefficients.In 2005, the highest magnitude of the effect on the probability ofchildcare attendance among all explanatory variables could be observedfor household wealth and higher education of mother with standard-ized coefficients of 0.39 and 0.16. The same coefficients had the highestmagnitudes among all explanatory variables in 2011, but their valueswere reduced to 0.29 and 0.15, respectively. The third highest magni-tude in 2005 is observed for Russian-speaking households—0.18.However, this variable ceased to be a significant predictor in 2011.

The results of the tobit model on the determinants of childcare at-tendance hours in 2005 and2011 are reported in Table 3. The significantpositive effect of having a mother with a higher education and beingfrom awealthier household dropped from 2005 to 2011. The significantresults of the equality of coefficients test indicate that these coefficientsdo differ across time in a strict statistical sense. The positive effect ofbeing from a Russian-speaking household became non-significant in2011—thus we see an over-time change which is confirmed by signifi-cant results of the test for equality of coefficients.

The effect of living in the Karagandy region was non-significant in2005, but became positive and significant in 2011 as confirmed bythe significant results of the test for equality of coefficients. In con-trast, the effect of living in Aktobe, Western Kazakhstan, Kostanaiand Northern Kazakhstan regions became positive and significantin 2011, as compared with 2005, but the test of equality of coeffi-cients fails, from a strictly statistical point of view, to reject thehypothesis that these coefficients are different across the years.

To directly compare the changes in magnitude of the effects of theexplanatory variables on childcare attendance hours, we use fully stan-dardized coefficients. As we can see, the highest magnitudes among allthe explanatory variables in 2005 are observed for household wealthand higher education of the mother, with standardized coefficients of0.44 and 0.16. The same coefficients also have the highest magnitudesamong all explanatory variables in 2011, but their values have droppedto 0.28 and 0.12, respectively. The third highest magnitude in 2005is observed for Russian-speaking households—0.08. However, thisvariable is no longer a significant predictor in 2011.

5. Discussion

The objective of this paper is to assess the results of Kazakhstan'sreforms, which aimed at improving access to childcare. Specifically,we attempt to evaluate how childcare attendance has changed, as wellas the ways in which the determinants of childcare attendance havechanged in the period following the reforms?

To answer the first research question of this study, we found that, ingeneral, the reformswere successful in increasing childcare attendance,andmost specifically, childcare attendancemore than doubled since theinception of the reforms. Indeed, it increased from 15% in 2005 to 39% in2011. Such an increase is an undeniable accomplishment of the reformgiven that the increase was achieved over a rather short period of just5 years. However, the results of the reforms fell short of the targetswhich had been set up. At the beginning of the reforms, the governmentplanned to increase childcare attendance to at least 75% by 2010. Thefindings of this paper demonstrate that this target was not achieved.The real attendance in 2011 was much lower than target indicatorset up by the government. Therefore, full success of the reform cannotbe claimed.

Another interesting finding of this paper is that the mean hoursof childcare attendance dropped significantly following the reforms.In 2005, the mean rested at about 18%, while by 2011, it had gonedown to approximately 10%. This finding suggests that the increase inoverall childcare attendance was achieved through a significant in-crease in the share of children only attending part-time. On the onehand, even part-time attendance is better when compared with noattendance at all. As well, an increase in part-time time attendance is a

Table 3Determinants of hours of attendance.

Variables 2005 2011 Test of equality

χ2 p-value

Male child 0.709 −0.276 3.25 0.000(0.459) (0.269)

0.055 −0.025Child's age 0.536 0.001 1.39 0.000

(0.454) (0.041)0.042 0.001

Woman's age −0.001 −0.033 2.09 0.000(0.003) (0.022)

−0.031 −0.037Higher education of mother 2.644*** 1.469*** 9.41 0.002

(0.587) (0.317)0.160 0.123

Russian-speaking household 1.425* 0.683 2.50 0.013(0.620) (0.381)

0.085 0.048Wealth of household 2.964*** 1.068*** 36.10 0.000

(0.356) (0.141)0.444 0.281

Rural 0.927 0.343 0.98 0.322(0.673) (0.368)

0.072 0.031Aktobe region 1.638 5.013*** 0.40 0.527

(1.227) (0.948)0.056 0.168

Atyrau region 0.358 0.902 0.02 0.882(0.986) (0.642)

0.014 0.039Western Kazakhstan region 2.087 4.060*** 0.01 0.921

(1.354) (0.931)0.063 0.133

Zhambyl region 1.177 0.204 1.20 0.273(1.057) (0.621)

0.047 0.009Karagandy region −0.0617 1.722* 1.46 0.026

(0.984) (0.680)−0.002 0.071

Kostanai region 0.492 4.491*** 1.98 0.159(1.151) (0.971)

0.016 0.141Kyzylordao region 0.325 0.302 0.02 0.897

(1.072) (0.575)0.014 0.015

Mangystau region 0.523 −2.387*** 3.59 0.048(0.998) (0.481)

0.020 −0.140Pavlodar region 0.981 1.114 0.07 0.794

(1.211) (0.704)0.027 0.042

Southern Kazakhstan region 0.171 −0.911 0.63 0.428(0.937) (0.526)

0.009 −0.056Northern Kazakhstan region −0.178 1.929* 1.24 0.264

(1.236) (0.791)−0.005 0.069

Eastern Kazakhstan region 1.186 1.310 0.11 0.736(1.309) (0.734)

0.036 0.050Number of observations 1582 1897Log likelihood −1392.0 −3646.5Likelihood ratio test χ2 243.24 322.73Probability N χ2 0.000 0.000

Note: Data are rounded.Outcome variable is hours of childcare attendance. Tobit regression is estimated.Coefficients are shown asmarginal effects. Standard errors are adjusted for clustering andshown in parenthesis. Fully standardized regression coefficients are in italics.Significance: * p b 0.05, ** p b 0.01, *** p b 0.001.

18 N. Habibov / Children and Youth Services Review 40 (2014) 13–19

sign of the greater flexibility of childcare providers who madeaccommodations for variation in demand for childcare. However,the observed increase in part-time attendance means that for asignificant share of children, daily and weekly participation in earlychildhood development education starts too late and finishes too

early during a week. Reduction in hours of attendance renders thepsychosocial and cognitive development of children through earlychildhood education somewhat abrupt and limited. As a result, theprogression of children to their early school education years maybe seriously jeopardized (UNESCO, 2005).

In response to the second research question of this study, wefound that the reforms significantly reduced the role of householdwealth as a barrier to both childcare attendance and hours of atten-dance. Nevertheless, the wealth of the household remained an im-portant determinant of childcare attendance both before and afterthe reforms. In fact, the wealth disparity was, and continues to bethe most significant determinant explaining variation in childcareattendance. Our findings suggest that even after reform, childrenfrom poorer households have a significantly lower likelihood of attend-ing childcare when compared to children from wealthier households.Even in the cases that they are able to attend childcare, children frompoorer households also log a significantly lower number of attendancehours than do children from wealthier households.

Our findings suggest that the existing childcare system in Kazakhstanis disproportionately skewed towards benefitting children fromwealthier households, most of whom may also have a head start athome. In comparison, the current system discriminates against childrenfrom poorer households whose families may not offer an environmentwhich encourages learning and development at home. Childcare pro-vides psychosocial and cognitive interaction, learning, and life experi-ences as unique inputs into the early development of the childrenfrom poorer households. Research has repeatedly demonstrated thatoverall, children from low-income households benefit more from thechildcare than do children from better-off households (Danzinger &Waldfogel, 2000; Young & Richardson, 2004).

Several supply and demand-driven strategies could be used to miti-gate the negative effects of wealth with respect to childcare attendance.The government needs to implement a strong policy in terms of the dis-tribution of childcare facilities. Poorer areas must be explicitly targetedfor building new or restoring older facilities. The geographic targetingof childcare towards the poorest communities could be an effectivemechanism through which to increase childcare attendance (Ruel, dela Briere, Hallman, Quisumbing, & Coj, 2002). At the same time, such asupply-driven strategy could prove to be insufficient in creating signifi-cant change if both official and unofficial fees are likely to be the mostserious barriers to childcare attendance for the poorer-households.

Hence, a supply-support strategy should be supplemented by ademand-driven strategy. One such strategy could involvemaking accessto childcare dependent on an income-test (UNESCO, 2005). As such,children from poorer households may be offered subsidized access tochildcare, while children from the poorest householdsmay be providedwith free childcare. Another strategy could be to link childcare atten-dance, including hours of attendance, to the take-up of targeted socialassistance benefits (Habibov, 2012b). Under this scenario, there wouldbe no need to create a new special mechanism of mean assessmentwhich would have to be administered by childcare facilities. Rather,the existing targeted social assistance which is presently administeredby the Ministry of Labour and Social Protection could be used instead.The take-up of social assistance benefits could be linked to a numberof positive behavioural outputs such as childcare attendance, regularhealth check-ups for children etc. A combination of the above-discussedapproaches would also be possible.

Another important factor explaining variation in attendanceinvolves regional disparities. The disparities in both attendance andhours of attendance remain strong despite the reforms. A numberof factors have contributed to this high variation in attendance. Onesuch factor is that regions have a wide range of financial resources forchildcare, and as such, some regions are better-off financially than areothers and so can provide more resources for childcare. To complicatethe situation, some regions experience more demand for childcarethan do others. For instance, the Astana region, which includes the

19N. Habibov / Children and Youth Services Review 40 (2014) 13–19

capital Astana, and the Almaty region,which includes the former capitaland the largest city in the country have more resources for childcare.Nevertheless, the demand for childcare is also much higher in themore industrialized areas of the Astana and Almaty regions. Conse-quently, the shortage of space in childcare facilities in these regions isconsiderable and the waiting lists are long (UNESCO, 2004). To add tothis already high variation is the subjective factor that during thereforms, the national government laid the responsibility for childcareon the shoulders of the heads of regional administrations, who thenbecame personally accountable for childcare development. However,it has been noted that there has beenwide variation among the regionaladministrative heads with respect to their support for childcareinasmuch as someof themhave beenmore supportive than have others.

To mitigate regional disparities in attendance, Kazakhstan shoulddevelop a new budgetary system (Habibov, 2010a, 2010b). This systemought to include a careful needs assessment which would reveal theactual needs of the regions and identify the existing gaps. This needsassessment would then serve as a baseline. Following this, limitedchildcare funds would be distributed according to the assessed needs.Thus, the new system would allow for the effective matching ofpriorities with available resources. As has been suggested by ourfindings, without such a new system, reform will fail to reduceregional disparities.

Turning to the effects of language, we can observe some dramaticchanges which took place after the reforms. Being from a Russian-speaking household was a significant predictor of childcare attendanceand hours of attendance in 2005. However, this characteristic ceased tobe a significant predictor in 2011. This finding indicates that the reformscreated a more equal environment for childcare attendance for allchildren, regardless of language spoken.

Similarly, we can observe drastic changes in the effect of themother's education. The positive effect of mothers' education withrespect to childcare attendance and hours of attendance droppedafter the reforms were put in place. It appears that mothers withhigher educational levels may have higher confidence, and thatthey are better able to nurture children in their homes than theywould be nurtured in childcare. It is also possible that motherswith higher education have more flexible hours in the labourmarket,and hence may have more time to nurture their children at home.

Finally, contrary to previous studies (UNESCO, 2006, 2008; WorldBank, 2000) we did not find age and gender of child to be significantpredictors of either childcare attendance or hours of attendance. Thisfinding suggests the absence of an age and gender bias in Kazakhstan.

6. Conclusion

In 2005, Kazakhstan implemented dramatic reforms in order toincrease childcare attendance and to make it more equitable. In termsof an assessment of the reforms, we found that although attendancesignificantly increased, it failed to achieve numerical targets. Thereforms reduced the effects of household wealth, language spoken inhousehold, and education of mother, although wealth and educationof mothers are still significant predictors of attendance. The reformsalso failed to reduce geographical disparities in attendance.

This study provides an initial assessment of the reforms. Futurestudies should assess the influences of programmatic factors suchas quality of childcare programs and distance to childcare facilities.

Future studies should also focus on the perspectives of administrators,teachers, and parents with respect to the reforms. Finally, the continuedmonitoring of Kazakhstani reforms is required since the reforms havecontinued after 2011, and their outcome is of great interest of policy-makers and social administrators in transitional countries.

References

Anderson, K., & Heyneman, S. (2005). Education and social policy in Central Asia: Thenext stage of the transition. Social Policy and Administration, 39(4), 361–380.

Anderson, K., Pomfret, R., & Usseinova, N. (2004). Education in Central Asia during thetransition to a market economy. In S. Heyneman, & A. DeYoung (Eds.), The challengesof education in Central Asia (pp. 131–152). Greenwich, CT: Information Age Publishers.

Baum, C. (2006). An introduction to modern econometrics using STATA. College Station, TX:Stata Press.

Cameron, A., & Trivedi, P. (2005). Microeconometrics: methods and applications. NY:Cambridge University Press, New York.

Danzinger, S., & Waldfogel, J. (2000). Securing the future: Investing in children from birth tocollege. New York: Russell Sage Foundation.

Giddings, L., Meurs, M., & Temesgen, T. (2007). Changing preschool enrolments in post-socialist Central Asia: Causes and implications. Comparative Economic Studies, 49(1),81–100.

Government of Kyrgyzstan (2007). Poverty reduction strategy paper—Country DevelopmentStrategy (2007–2010). Bishkek: The Government of Kyrgyz Republic.

Government of Tajikistan (2010). Poverty reduction strategy paper. Dushanbe:The Government of the Republic of Tajikistan.

Government of Uzbekistan (2008). Poverty Reduction Strategy Paper. Welfare improvementstrategy of Uzbekistan. Tashkent: The Government of the Republic of Uzbekistan.

Habibov, N. (2010). An intertemporal evolution of inequality in Azerbaijan: 1995–2002.Problems of Economic Transition, 52(9), 51–77.

Habibov, N. (2010). Understanding the over-time evolution of living standard determi-nants in transitional countries: Evidence from Azerbaijan. Journal of ComparativeSocial Welfare, 26(01), 43–63.

Habibov, N. (2012). Early childhood care and education attendance in Central Asia.Children and Youth Services Review, 34(4), 798–806.

Habibov, N. (2012). Does childcare have an impact on the quality of parent–childinteraction? Evidence from post-Soviet Kyrgyzstan, Tajikistan, and Uzbekistan.Children and Youth Services Review, 34(12), 2367–2373.

Lokshin, M. (2004). Household childcare choices and women's work behavior in Russia.Journal of Human Resources, 39(4), 1094–1115.

Long, S., & Freese, J. (2006). Regression models for categorical dependent variables usingSTATA. College Station, Texas: StataCorp LP.

Mertaugh, M. (2004). Education in Central Asia, with particular reference to the KyrgyzRepublic. In S. Heyneman, & A. DeYoung (Eds.), The challenges of education in CentralAsia (pp. 153–181). Greenwich, CT: Information Age Publishers.

Penn, H. (2004). Childcare and Early Childhood Development Programmes and Policies:Their relationship to eradicating child poverty. CHIP Report No. 8. London: ChildhoodPoverty Research and Policy Centre.

Ruel, M. T., de la Briere, B., Hallman, K., Quisumbing, A., & Coj, N. (2002). Does subsidizedchildcare help poorworking women in urban areas? FCND briefs 131.Washington, D.C.:International Food Policy Research Institute (IFPRI).

Stewart, K., & Huerta, M. (2009). A share of new growth for children? Policies for the veryyoung in non-EU Europe and the CIS. Journal of European Social Policy, 19, 160–173.

UNESCO (2004). The status of preschool education in the Republic of Kazakhstan. TheBackground Report of Kazakhstan. UNESCO/OECD Early Childhood Policy ReviewProject. Paris: UNESCO.

UNESCO, & UNESCO (2005). Policy review report: Early childhood care and education inKazakhstanEarly Childhood and Family Policy Series N° 12 – 2005. Paris: UNESCO.

UNESCO (2006). Regional overview: Central and Eastern Europe and Central Asia. Paris:UNESCO.

UNESCO (2008). Education for all by 2015. Will we make it? Education for all globalmonitoring report 2007. Paris: UNESCO.

UNICEF (2004). UNICEF’s Executive Directive (Ex.Dir.CF/EXD/2009–002). Retrieved onNovember 30, 2013 at. http://www.childinfo.org/mics4.html

Wooldridge, J. (2002). Econometric analysis of cross-section and panel data. Cambridge,MA: MIT.

World Bank (2000). Republic of Tajikistan poverty assessment. Washington, D.C.: WorldBank.

Young, M., & Richardson, L. (2004). Early child development: From measurement to action.Washington, D.C.: The World Bank.