Changing preschool enrolments in post-socialist Central Asia: causes and implications.
Giddings, Lisa ; Meurs, Mieke ; Temesgen, Tilahun 等
INTRODUCTION
Preschool is often analysed in its childcare, or custodial,
function, as facilitating women's labour force participation
(Blumberg, 1981; O'Connor, 1988; Craig, 1981). But for
preschool-aged children, preschool can also contribute importantly to
human capital development, especially among poor children (Danzinger and
Waldfogel, 2000).
By international standards, preschool enrolments were high under
socialism, although rates in Central Asia lagged those in other areas
(Riazantsev et al., 1992, p. 27). Since the collapse of the Soviet
Union, preschool enrolment rates have plummeted in Central Asia (UNICEF,
2003, p. 81). This may leave large numbers of children without
preparation necessary to succeed in school and to be included in the
region's post-socialist development.
To better understand the factors underlying declining preschool
enrolments in Central Asia, we examine preschool enrolments in
Kyrgyzstan over the period 1993-1998. Kyrgyzstan is one of the poorest
countries in Central Asia, and has seen the preschool enrolments fall to
the lowest level among Central Asian countries, with the exception of
Tajikistan (UNICEF Transmonee, 2005). We evaluate demand- and
supply-side explanations for low enrolments, and use data from the
Ryrgyzstan Poverty Monitoring Survey to develop a logistic model of
preschool enrolment. On the demand side, we examine the impact of
households' need for custodial care, household ability to pay for
care, and cultural/social factors (mother's education and ethnic
background). This last variable builds on a previous finding in Bulgaria
of significant differences in preschool enrolment dynamics across ethnic
groups. On the supply side, we consider the impact of preschool
availability and quality. Through this model, we hope to broaden
understanding of household decisions to invest in preschool, and to
assist policy makers in improving preschool attendance as a form of
school preparation and early human capital accumulation, especially
among poorer households.
Current policy recommendations often target access as a key to
raising preschool enrolments (eg Henscher and Passingham, 1996). While
national-level data currently show excess supply of preschool places, we
find that access to a preschool is the most important determinant of
enrolment. However, we also find that demand-side factors including the
availability of non-working adults to provide home care, household cash
income per capita, and social background have a significant impact on
enrolment.
THE ROLE OF PRESCHOOL
Preschool services can be oriented towards achieving different
goals, and we begin by distinguishing between custodial and human
capital-enhancing childcare (Connelly et al., 1996, p. 9). Preschool is
probably most often understood in its custodial form, as a service that
facilitates women's labour force participation. Childcare advocates
highlight both the distributional and efficiency implications of
providing such a service. By expanding women's work options,
preschool 'accompanies and facilitates women's
citizenship' and is thus a prerequisite for equality between the
sexes (Blumberg, 1981; O'Connor, 1988; Craig, 1981). By enabling
the expansion of women's labour force participation, custodial
childcare may also enhance economic efficiency by allowing women to move
into higher productivity occupations (O'Connor, 1988, p. 24).
Preschools provide more than custodial care, however. Early
childhood education can improve cognitive development and other outcomes
(Danzinger and Waldfogel, 2000), thus increasing human capital
formation, labour productivity and growth. Studies of the impact of
state-subsidised preschool education in France, for example, reveal a
strong correlation between preschool attendance and performance in the
first grade, especially among children from poorer backgrounds
(Bergmann, 1996, p. 33). Since children have been found much more likely
to drop out if they receive poor grades or repeat a grade in early
years, early childhood education may be an effective means of improving
overall human capital acquisition (Currie and Thomas, 1995, p. 359-360).
In the US, the Head Start preschool programme has been used to improve
learning and social skills, as well as the health status of poor
children. A recent review that focused only on well-structured studies
of publicly funded US preschool programmes found large, positive, and
significant long-term effects on schooling and earnings, and negative
effects on problems such as crime and participation in various welfare
programmes, including Temporary Aid to Needy Families (TANF) (Garces,
Thomas and Cuttle, 2000).
A recent, similarly well-structured study in Germany found positive
impacts of early childhood education on schooling outcomes of the
seventh graders, but only among families of immigrant (disadvantaged)
children (Spiess et al., 2003, p. 17). Although studies of preschool
impact in developing countries are scarce, a recent study of a
government preschool programme in Bolivia, based on a large data set and
careful matching techniques, found strong, positive impacts of preschool
on tests of health, cognitive and psychosocial development among
children who attended at least 7 months (Behrman et al., 2004). As many
of the studies suggest that positive impacts of preschool are
concentrated among children from more disadvantaged backgrounds,
expanded preschool enrolments may enhance equity as well as efficiency
and growth.
If preschool is valued mainly for custodial purposes, then the
recent radical declines in enrolments across the former socialist
countries may not be the cause for concern. Where unemployment is rising
and wages falling, there may be less of an efficiency argument for broad
use of custodial care. Reduced availability of custodial care may have
implications for gender equity, but with many unemployed males, even
these implications are not certain.
A more troubling interpretation of fails in preschool enrolments
during transition results if we consider the implications for human
capital formation. Declining shares of preschool-aged children in
childcare may result in lower attendance and success in primary school
and beyond, and represent declining investments early human capital,
which will have dire consequences for future labour productivity.
Indeed, the UNDP reports that the number of children dropping out of
school in Central Asia is increasing, and that this problem is
underestimated in official government statistics (2001b). Similarly, a
study of Kazakhstan reports a negative impact of declining kindergarten enrolments on primary schooling, through the inclusion of more poorly
prepared pupils in the crowded classrooms of poorly funded primary
schools (Henscher and Passingham, 1996). Reduced access to childcare may
therefore contribute to further increases in inequality, in a region
where inequality is already rising rapidly.
PRESCHOOL UNDER SOCIALISM
Former socialist economies were characterised by high rates of
labour force participation for both men and women. Female labour force
participation rates ranged from about 65 per cent in Poland to 90 per
cent in Bulgaria in the mid-1980s (UNICEF, 1999, p. 24; European
Commission, 1995). In Kyrgyzstart, about 85 per cent of working age men
and 75 per cent of working age women were employed in 1989 (UNICEF,
1999).
The high labour force participation rates resulted in fairly
widespread need for childcare. Most socialist countries had a two-tier
system of state-run childcare. This included a system of nurseries for
children from birth to 2 years of age, and a system of preschools, for
children aged 3-6 years. These facilities were organised through the
Ministries of Health (nurseries) or Education (preschools) or by state
enterprises. While relatively few children from 0 to 2 years attended
nurseries, preschool enrolments were strong across many of the socialist
economies. In 1988, in the European Republics of Russia, Byelorussia and
Moldova, enrolment rates among preschool-aged children were around 70
per cent (Cornia, 1995, p. 69). Preschool enrolments in Central Asia
lagged significantly behind those in other Soviet Republics. Notable
progress was made in many areas in the decades between 1970 and 1990,
with enrolments in Kazakhstan rising from 30 per cent in each age group
in 1970 to 53 per cent in 1988. In Kyrgyzstan, enrolments rose from 18
per cent to 30 per cent over this period (Riazantsev et al., 1992, p.
27). In 1989, about 31 per cent of preschool-aged children attended
preschool in Kyrgyzstan (UNICEF, 1999).
In addition to the impact of the above-mentioned high labour force
participation rates, high enrolments can be attributed to highly
subsidised fees in the centres, and reasonably good quality of care.
Although the socialist states used the preschools to instil the
characteristics of obedience and ideological adherence, they were also
used to assure early childhood education and school preparation
(Henscher and Passingham, 1996), as well as public health. Jean Ipsa, in
her book Child Care in Russia in Transition (1994), documents the
provision of a safe, stable, and loving environment, in which teachers
actively promoted the development of gross and fine motor skills,
cooperative behaviours, ecological awareness, and basic math and reading
skills. In addition, all students attending the preschools were
regularly examined by a nurse and received a regular schedule of
vaccinations. The relatively strong health and education indicators seen
today in former socialist countries can be attributed, in part, to the
preschool policies of the era (UNICEF, 1999; Anderson et al., 2004).
The relatively lower enrolments in Central Asia may be attributed
to a number of factors. They were clearly related to the lower supply of
childcare facilities. The 1980s were characterised by excess demand for
places, with preschools enrolling 140 students for each 100 places
(compared to 107 children per 100 places in the USSR as a whole)
(Cornia, 1995, p. 36), and still long waiting lists persisted (Klugman
et al., 1997). The resultant crowding may also have reduced quality. In
Kyrgyzstan, 74 per cent of preschools were run by state enterprises in
1991, and these were characterised by higher quality of care as compared
to those run by municipalities (Klugman et al., 1997, p. 189). Finally,
the continued dispersion of population in rural areas may have
contributed to the reduction in access to available centres (65 per cent
of Kyrgyz residents continue to live in rural areas) (UNDP, 2001a, p.
26).
RECENT DEVELOPMENTS IN KYRGYZSTAN
Former socialist countries have seen a drastic decline in
enrolments of preschool-aged children in childcare since 1991, and
Central Asian countries are no exception. In Kyrgyzstan, net enrolment
rates for children 3-6 years fell from 31 per cent in 1989 to 8.7 per
cent in 1998, while enrolment rates fell from 53 per cent to 12.4 per
cent in Kazakhstan, and from 36.8 per cent to 16.1 per cent in
Uzbekistan. In Tajikistan, where enrolments started at a lower level (16
per cent), they fell less dramatically, but to the lowest level in
Central Asia, 6 per cent (Transmonee Database, 2005).
Below, we analyse factors underlying preschool enrolment in
Kyrgyzstan. To examine household choice to send children to preschool,
we draw on recent literature on household choice of the form of care for
young children (Meurs and Giddings, 2006; Anderson et al., 2004; Klugman
et al., 1997; Fong and Lokshin, 2000). We develop a simple model that
includes both supply- and demand-side factors. The two supply-side
factors considered are availability and quality of preschools. Even if
the supply of preschools is adequate, however, households may choose not
to enrol children for a variety of reasons. We consider three factors
underlying household demand for preschool. Households may value
preschool only for its custodial function. If there are adults available
in the household to provide care (grandparents or working age adults not
employed outside the home), then households may choose to keep children
at home. Alternatively, parents may view preschool as superior to home
care as an investment in their children's futures (Connelly et al.,
1996; Fong and Lokshin, 2000). In this case, ability to pay fees and
transport costs, rather than the availability of substitute care, may
underlie decisions about preschool attendance. Thirdly, cultural factors
may influence the households' perspectives on the value of
preschool.
Access to preschools has declined over the post-socialist period.
Centres previously run by state enterprises, which had been
characterised by higher quality care than that found in municipal-run
centres, closed as enterprises were privatised or liquidated. The number
of municipal centres also declined, as financing responsibilities were
transferred to municipalities, but these have become a relatively more
important source of care, increasing from 26 per cent to 55 per cent of
all centres in Kyrgyzstan by 1994 (Klugman et al., 1997, p. 189).
Overall, the number of preschools fell dramatically--from 1,696 in 1990
to about 468 in 1998 (NSC, 1995, p. 93; Ministry of Education, 2001, p.
2) (Table 1).
Still, the decline in available centres does not appear to explain
the overall decline in enrolments. While the number of preschools fell
to 28 per cent of the 1990 level by 1998, enrolments fell more rapidly
to 22 per cent of previous levels (Ministry of Education, 2001, p. 2).
The number of enrolled children per available 100 spaces fell from 140
in 1990 to 87 in 1993 to 75 in 1995, before rising again to 86 in 1998
(Table 1). By 1993, Kyrgyzstan experienced an overall excess capacity in
preschool, and the number of excess spaces increased into the 1990s
(Klugman et al., 1997, p. 190; NSC, 1995, p. 93). (1) Of course, excess
capacity in some areas co-exists with lack of access in others. In a
model of kindergarten availability in Kyrgyzstan, Anderson et al. (2004,
p. 141) found that in 1993 and 1997 kindergartens were significantly
less prevalent in poorer regions, while more likely to be found in areas
with better education, more heavily Slavic populations, with more
developed manufacturing bases.
As in other post-socialist cases (Meurs and Giddings, 2006; Klugman
et al., 1997), lower preschool enrolments may be an outcome of lower
demand. There have been dramatic, post-socialist declines in employment
and wages. Kyrgyzstan, an early reformer, experienced a rapid collapse
of output from 1991 to 1995. This was followed by rapid growth
thereafter, but by 1998 GDP had achieved only 60 per cent of 1989
levels. With declining production, average real monthly wages fell to
49.6 per cent of 1990 levels in 1993, before falling further and then
rising back to 49.1 per cent in 1997 (about $40 at 1993 exchange rates)
(Transmonee Database, 2005; Pomfret, 2003, p. 454). For many households,
a large (and increasing) share of their income comes from subsistence production and is not, consequently, available for cash expenditures
such as childcare. Additionally, in 1995 many firms also paid wages with
significant delays (Anderson and Pomfret, 2000, p. 507), leaving
households with less available for such expenditures.
At the same time, earnings inequality increased significantly in
Kyrgyzstan, with the Gini coefficient rising from 0.26 in 1989 to 0.43
in 1998 (UNICEF, 2003, p. 93), leaving those at the bottom of the
distribution with severely reduced earnings. Female labour force
participation remained high and the gender wage gap narrowed, but men
continued to earn more than women as wages fell. There were also
significant ethnic wage gaps, with Russians earning more than ethnic
Kyrgyz. Uzbeks earned less than Kyrgyz in 1993, but by 1997, they earned
significantly more. Those living in the urban north of the country
earned more over the period than those in other regions (Anderson and
Pomfret, 2000, p. 512).
At the same time that earnings declined, the fees for use of
preschools (and related transport) increased and the collection of fees
has become better enforced. Some households may simply be unable to
afford preschool fees and other related expenses, including transport,
even if they see the preschools as providing an important boost to
children's human capital accumulation. The UNDP reports that the
number of children dropping out of school is related to declines in
family income and increases in the costs of education and related
expenses (transport) (UNDP, 2001b, p. 13, 77).
Declines in employment have lagged behind the declines in wages.
The number of officially unemployed persons jumped radically from only a
couple of thousands of workers (0.2 per cent of the labour force) in
1993 to 3.1 per cent in 1998 (NSC, 2002, p. 58; UNICEF, 2003, p. 92).
Clearly, effective levels of unemployment are higher than suggested by
these modest official figures, with workers clinging to
'employed' status in order to retain other firm-related
benefits (including access to preschool). At the same time,
'employed' workers have faced large arrears in wage payment
and limited demands on their time.
More women were recorded as unemployed in this period (women were
58 per cent of the unemployed) and their share of the paid labour force
fell somewhat, from 51 per cent to 46 per cent from 1993 to 1996 (UNDP,
2001a, p. 61). Some researchers report women taking on more of the
income-generating responsibilities as male employment and earnings fall
(Narayan and Petesch, 2002, p. 292). As a result, women may be shifting
to a greater extent into the informal economy. Overall, the apparently
modest declines in the need for custodial care seem unlikely to provide
a full explanation for the radical drop in preschool enrolments.
However, in households where parents see no human capital advantages to
preschool over home care, unemployed adults available for childcare may
result in a corresponding decline in enrolment in preschool.
There is also evidence that the quality of preschools has declined
over the period, and this may affect parental choice. Decentralisation of government has shifted some of the financial responsibility for
education onto local governments, which have not received an equivalent
transfer of financial resources (UNDP, 2001a, p. 40). With few resources
available for maintenance, supplies, or even meals and snacks (Klugman
et al., 1997), parents must provide most materials themselves (Henscher
and Passingham, 1996). Under these conditions, centres may be an
unappealing option both in terms of custodial care and school
preparation. Quality of transportation may also affect attendance, as
many routes have been discontinued (Narayan and Petesch, 2002).
Finally, households may be expressing preferences for home care,
which were repressed during the socialist period. There might be a
cultural preference for home care in general, or there may be concerns
among minority ethnic groups regarding increased use of the Kyrgyz
language, and cultural teachings, in schools. Studies of other countries
have found that parental education has a significant impact on
households' valuation of preschool, presumably by affecting the
preference of parents for schooling (Strauss and Thomas, 1995). More
educated parents are more likely to invest in preschool. In the section
below, we examine the impact of these economic and cultural factors on
household decisions to send children to preschool.
The declines in preschooling in post-socialism have been the
subject of some previous analysis. However, none of these analyses have
considered the relative role of supply, custodial needs, ability to
invest, and cultural factors as we do here. Klugman et al. (1997) and
Fong and Lokshin (2000) examine the changing preschool enrolments
somewhat indirectly, through their impact on women's labour force
participation. Klugman et al. (1997) find evidence that custodial needs
and cultural factors (the presence of non-employed adults in the
household and mother's education) influence attendance in
Kazakhstan. Fong and Lokshin find that ability to pay affects enrolment
in Romania (2000). Similarly, modelling enrolment outcomes in Bulgaria,
Meurs and Giddings (2006) find that ability to pay, need for custodial
care, and cultural (ethnic) factors all influence enrolment. Anderson et
al. (2004) find that both access and cultural factors (mother's
education) are significant.
EXPLAINING ENROLMENTS IN KYRGYZSTAN
The data
In the 1980s, the World Bank designed a survey to measure the
living standards of the population in developing countries. This survey
has become the Living Standards Measurement Survey (LSMS) and has been
conducted in more than 40 developing countries. Five rounds of such
surveys, the Kyrgystan Poverty Monitoring Survey (KPMS), have been
conducted in the Kyrgyz Republic on behalf of the National Statistical
Committee of the Kyrgyz Republic, with technical assistance from
Research Triangle Institute.
The KPMS is designed to be a nationally representative survey
capable of measuring the standard of living in the Kyrgyz Republic. The
KPMS surveys were conducted in fall 1993, and then annually from fall
1996. We use the 1993 and 1998 (the first and the latest that is
publicly available) surveys here to examine post-socialist preschool
enrolment dynamics. As seen in Table 1, the greatest drop in preschool
enrolments occurred in the period 1989-1993, the period during which
Kyrgyzstan, an early reformer, experienced radical changes in
production, income, and social service provision (Pomfret, 2003; UNDP,
2001b). Since survey data are available only in the beginning in 1993,
we cannot examine causes of this change in enrolments. Instead, we
examine the decision to enrol children (or not) during the period
1993-1998, attempting to isolate the factors related to that decision.
A stratified sample of 2,000 households was randomly selected for
the 1993 KPMS. The 1998 KMPS had a sample size of 2,979 households. Both
surveys contain information on the composition of the household,
economic activities, health, education, migration, and labour. The
implementation of the household questionnaires was accompanied by the
implementation community questionnaires (Population Point Surveys),
which collected information about local infrastructure, services, and
other amenities. These can be matched to the household data, to examine
the impact of community-level variables on household outcomes.
These surveys provide the only nationally representative,
probabilistic sampling of households for Kyrgyzstan. It is difficult to
say, without other survey data for comparison, anything definitive about
the data quality. Angus Deaton (1997, pp. 35-40)argues that feedback
from users of LSMS (of which the KPMS is an example) in general has been
positive, and the Kyrgyz data is typically in line with that collected
by other methods by the National Statistical Office.
In order to examine preschool attendance in the Kyrgyz Republic, we
selected a sub-sample of children between the ages of 3 and 6 years
(inclusive). The 1993 sample consisted of 963 children. However,
complete data for the regressions was available for only 674 children.
The 1998 sample consisted of a sample of 1,601 children in this age
range, of whom 1,476 had complete data for inclusion in the regressions.
Preschool attendance
In 1993, the majority of surveyed children 3-6 years of age were
cared for only by household members in the 7 days prior to when the
questionnaire was administered. Only 43 children between the ages of 3
and 6 years (six per cent of the sample) attended kindergarten, nursery,
extended school day group, or the like. The situation remained
relatively unchanged in 1998, when 88 children (six per cent of the
sample) were reported to attend a kindergarten or nursery. These numbers
conform approximately to those provided by the National Statistical
Committee for 1998, but are significantly below officially reported
numbers (13 per cent) for 1993 (NSC, 1998). In 1998, all children
attended nearly full time, with daily hours in care ranging from 7 to
12.
There are three types of institution that a child can attend:
state-owned, enterprise-owned and privately owned centres. In 1993, over
70 per cent of children attended kindergartens or nurseries that
belonged to the state, while most others attended those belonging to a
ministry or enterprise, and one individual attended a privately owned
childcare institution. By the 1998 survey, we saw a slight increase in
the role of state or public institutions over enterprise-run
institutions, but use of private kindergartens had not increased.
Underlying factors
Building on the model outlined above, we begin by examining the
factors underlying enrolment choice separately. We then develop a
logistic model of preschool attendance to examine the relative impact of
these factors.
Even if households recognise the importance of preschool as an
investment in early human capital, economic hardship may prevent
enrolment. The Kyrgyz survey data suggest no simple relationship between
a household's economic status and attendance, however. In 1993,
households whose preschool age children attended preschool had a mean
per capita monthly household income approximately equal to that of
households whose children do not attend (47.9 soms per month, compared
to 49.8 (2)) (about $8 at 1993 exchange rates). Considering only cash
income, which could be used to pay for things like preschool, households
of children not attending had significantly lower incomes, however (10
soms compared to 26 soms) (Table 2). (3) Income may not have been
particularly important to schooling decisions in 1993, because preschool
was not costly for those households that used it. The majority of
households with children attending in 7 days prior to the survey claimed
that they had not paid for the service. This supports a 1995 finding by
ABD Associates of widespread non-payment for childcare services (Klugman
et al., 1997, p. 186).
By 1998, however, households with children attending preschool
reported having significantly higher average per capita incomes than
those whose children were not attending (613 soms, approximately $29 at
1998 exchange rates, compared to 461 soms). This result does not seem to
be the result of the endogeneity of mothers' incomes. Excluding
mothers' incomes, households not enrolling children still had lower
per capita incomes--439 soms compared to 527 soms for households with
children enrolled in preschool. As before, the gap in cash income
exceeds the gap in overall income and differs significantly between
groups (Table 2). In 1998, no households claimed not to pay for the
service. Clearly, by 1998 the cost of preschool had become an issue for
some households: when asked why a child who previously attended
preschool had ceased to attend, 33 per cent of such cases reported that
it was too costly. Over this period, preschools appear to change from a
subsidised service to a costly service afforded only by better-off
households.
With rising official unemployment between 1993 and 1998, we see a
declining apparent need for custodial care. In 1993, there was an
average of 1.6 working adults in households of children 3-6 years, while
the number of non-working adults averaged 1.1. In 1998, households had
an average of 1.7 working members, and 1.9 non-working adult members.
However, the availability of this labour varied widely. In 1998, 22 per
cent of households had no non-working adult present, while another 26
per cent had only one non-working adult. In both years, households with
children not attending preschool reported approximately twice as many
non-working adults (Table 2).
Cultural factors may also drive decisions about preschool. There is
little evidence of cultural preferences for educating one sex rather
than the other in Kyrgyzstan. In both years, girls were slightly more
likely to attend kindergarten than boys (Table 3) (6 per cent of girls
attend, compared to 8 per cent of boys in 1993, and 5 per cent of girls
and 6 per cent of boys attended in 1998) (Table 3).
We do see significant differences in preschool attendance by
ethnicity, however. Kyrgyzstan's ethnically mixed population is
about 60 per cent Kyrgyz, 16 per cent Russian, and 14 per cent Uzbek
(UNDP, 2001a, p. 60). The rest of the population was divided among a
large number of other ethnicities. Russian and other children of Slavic
nationalities (Ukrainian, Belorussian) were most likely to attend
preschool (22 per cent of Slavic 3-6 year olds attended in 1993, 24 per
cent in 1998), while Kyrgyz children had a lower likelihood of attending
(4 per cent in 1993, 5 per cent in 1998). Over the period, Uzbeks appear
to have radically altered their use of kindergartens, becoming more like
Kyrgyz in their behaviour (about 7 per cent attended in 1993 and 1996,
but by 1998, only 2 per cent of Uzbek children were reported to attend)
(Table 3).
The ethnic differences in attendance may be related to differences
in incomes between the groups. Per capita income of Slavic households is
significantly higher than that for households from other ethnic groups.
Ethnic differences in attendance might also be related to ethnic
differences in preferences regarding childcare, or to other regional
differences like preschool accessibility, since ethnic groups tend to be
concentrated in particular regions of the country (with Russians more
concentrated in urban areas with a higher concentration of preschools).
Considering the impact of parental education on preschool
attendance, we used dummy variables for primary school or less, beyond
primary to complete secondary school, and beyond secondary school. (4)
These measures capture significant differences in the population, with a
little over 50 per cent of the mothers reporting at least primary but no
more than secondary school in both years, and approximately another 40
per cent reporting more than secondary in both years. In both years,
children of mothers with tertiary education were much more likely to
attend preschool than those of mothers with primary or secondary
education only (Table 2). In 1998, the year for which years of schooling
are reported, mothers of enrolled children had significantly more years
of schooling (12.3 years, on average) than those of unenrolled children
(10.8 years).
Finally, access and quality vary significantly. We matched
population point data to household data in order to examine the impact
of local availability of preschools on enrolment decisions. This is not
a perfect measure. Households living near the border of the population
point may have access to a preschool in a neighbouring population point,
even if their own population point does not have one. The measure
provides the best available indication of local availability, however.
We see an apparent decline in access to preschools over this
period. From 1993 to 1998, the share of population points reporting
state-run preschools fell from 80 per cent to 53 per cent, while the
share reporting the presence of a private facility fell slightly from 3
per cent to 2 per cent. If we look at the number of survey households
affected by these differences, we find that an even larger share of
surveyed households lived in population points without access. In 1993,
66 per cent of households lived in areas served by preschools (state or
private). In 1998, only 45 per cent of households lived in a served
area. One per cent reported having a privately run preschool in their
population point. When asked why a child had stopped attending
preschool, in 1998 33 per cent of such households responded that the
kindergarten had closed.
Quality of preschool is of course very difficult to measure, and
parents were not asked to specifically evaluate quality of preschools.
In the 1998 survey, however, households were asked to evaluate whether
schools (not only preschools) in the area had adequate buildings,
blackboards, books, and other materials. Using these measures of
adequacy of six different school conditions, we created a proxy for
school quality in 1998, an index varying from 0 to 6. In 1998,
households with children attending preschool gave a significantly higher
quality ranking than did households without children attending preschool
(Table 2).
To examine the relative impact of these factors, we develop a
logistic model of whether or not a particular child is attending
preschool at the time of the survey. (5) In this model, we are
attempting to predict the probability of preschool enrolment. To do
this, we examine the impact of the household's (lack of) need for
custodial care (number of non-working adults in the household, Adult Not
Working), ability to pay for preschool (cash income per capita, PerCap
HH Income, in hog values), (6) and whether there is preschool available
in the population point (School in PSU). To capture the impact of
cultural factors, we also consider the impact of the sex (male = 1) and
ethnicity of the child. We use three dominant ethnic groups (Kyrgyz,
Slavic, which includes both Russian and other Slavic nationalities, and
Uzbek), with Kyrgyz as the excluded category. (7) We also examine the
impact of mother's education on attendance (Mother's
Education). (8) Because mother's education was coded differently in
the 1993 and 1998 surveys, we use dummy variables for primary education
or less, and more than secondary education, as compared to the reference
category of some or complete secondary education. Finally, we control
for age. (9) Means and standard deviations of the variables are
presented in Table 2.
We first estimate the model separately for 2 years, then pool the
data and estimate the model for the period 1993-1998. For the pooled
data, we include a year dummy (1993 = 1). We expect the ability to pay,
centre availability and quality, and mother's education to be
positively related to attendance, while the availability of non-working
adults in the household is expected to reduce the likelihood of
attendance.
The results of this model are as seen below in Table 4. We report
marginal effects (Delta P), which reflect the marginal effects of a
one-unit change in the dependent variable on the outcome, at the mean.
Overall, we see a set of very consistent relationships over this period,
although the model performs slightly better in 1998 than 1993. Pseudo [R.sup.2]'s varied from 0.22 in 1995 to 0.29 in 1998. These values
are relatively strong results for this type of cross-sectional data and
indicate that the model does well at explaining enrolment.
In all three logit regressions, the availability of a preschool had
the largest significant impact on attendance, although the magnitude of
the impact declines slightly from 1995 to 1998. Having a preschool in
the Population Point led to almost a 6 percentage point increase in the
likelihood of attendance in 1993 and 5 per cent in 1998. Ethnicity also
has a significant impact on attendance, even controlling for (regionally
differentiated) access to centres and household income. Children in
Slavic households were significantly more likely (5 percentage points)
to be attending preschool than were Kyrgyz children, whereas Uzbek
children were statistically indistinguishable from Kyrgyz.
Both household cash income per capita and the number of non-working
adults available for childcare also had a significant impact. The impact
of cash income was very small, and surprisingly consistent between the 2
years, despite the apparent change in the enforcement of preschool fees.
A one-unit change in log income increased the likelihood of enrolment by
only 0.3 percentage points. The impact of available non-working adults
was also small, although much larger than the impact of an increase in
income, and also remained fairly stable over the period. One additional
adult available for childcare reduced the likelihood of attendance by
about 1.4 percentage points in 1993 and 1 percentage point in 1998.
Finally, having a mother with post-secondary education had a significant
positive impact on a child's attendance. The impact of
mother's education was large, slightly larger than the negative
impact of non-working adults and, like other effects, remained fairly
stable over the post-socialist period. Having a mother with tertiary, as
opposed to secondary, education increased the likelihood of attending
preschool by almost 2 percentage points in both years.
In 1998, when the proxy for preschool quality is available, this
variable does not have a significant impact on attendance. In the pooled
regression, the year dummy (1993 = 1) is negative and significant. As
the economy recovered after 1993, the likelihood of attending preschool
increased.
Finally, the control variables of age and sex did not have a
significant impact on preschool attendance. Parents are not more likely
to enrol children in preschool as the age of school entry approaches,
and they are not significantly more likely to send boys than girls to
preschool.
These results support findings of other work on preschool
attendance in post-socialist cases. Like Anderson et al.'s (2004)
analysis of preschool attendance in Kyrgyzstan, we find that access to
preschools and mothers' education are important factors in
supporting attendance, while household choices about preschooling do not
differ significantly by the sex (or age) of the child. Like Meurs and
Giddings (2006), in their analysis of preschool attendance in Bulgaria,
we find that both households' ability to pay and the availability
of non-employed adults in the household also affect preschool
attendance, as does ethnicity.
Conclusions
Childcare enrolment rates have fallen significantly in Central Asia
since the early 1990s. They seem to have stopped falling by the
mid-1990s and even recovered very slightly, but the very low current
rates may be a cause for concern on both equity and efficiency grounds.
Since investment in preschool can create significant positive
externalities, and economic inequality can create significant negative
externalities, there may be a strong rationale for government
intervention, even in a time of extremely tight budgets (Aslaksen et
al., 2000, p. 98). In this paper, we have attempted to contribute to an
understanding of the factors underlying current low enrolments. This may
help target resources and improve the likelihood of effective government
intervention.
Aggregate national level data show that enrolments have fallen
faster than available preschool places, suggesting that declines in
enrolment are a demand-side, and not a supply-side, phenomenon. This
outcome has been noted by researchers in other post-socialist cases as
well (Meurs and Giddings, 2006; Klugman et al., 1997). Our household
survey data show dramatic declines in households' access to
preschool, however. In our logit regressions, access to a preschool is
consistently the most important significant factor in explaining
preschool attendance. This suggests that, despite what is seen in
aggregate data, supply-side factors play an important role in low
enrolments. Preschool supply is lumpy. A school cannot be established to
serve the demand of only a few households, and many households may go
unserved in areas where demand does not meet a critical level, even as
surplus places exist in other areas.
On the demand side, (low) need for custodial care plays a
significant role in household choice. Many households have adult labour
available for childcare. Ethnic differences are also significant. But
these do not appear driven by fears that minority group children will be
poorly served by government schools (as seen in the Bulgarian case, for
example). In fact, ethnic Kyrgyz and Uzbek children are significantly
less likely than Slavic children to attend Kyrgyz government-run
preschools. Krygyz and Uzbek children are more likely than Slavic
children to be living in less-developed, rural areas, but the ethnicity
effect persists when controlling for centre availability and income.
Further investigation into Kyrgyz and Uzbek families' use and
non-use of preschools will be needed if there is to be effective policy
to boost attendance. For the moment, specific policies to encourage the
enrolment of girls do not appear necessary.
Household income is also a significant constraint on enrolment, but
after controlling for other factors the impact of income is small,
despite increases in the enforcement of fee payment. There may therefore
not be strong grounds for targeted childcare subsidies. Rather, direct
support for establishing and maintaining centres in poor areas and
further investigations into population preferences regarding preschool
seem like the most effective manner of addressing current low
enrolments.
The question of what drives household decisions about preschool
enrolment is of particular importance because the effects of poor school
preparation will be felt in the labour force for years to come. As more
children from disadvantaged backgrounds enter school poorly prepared and
perform poorly in their first years of primary school, drop-out rates
are likely to be exacerbated, undermining Central Asia's high
literacy rate and contributing to an overall downward trend in
development level.
Acknowledgements
We wish to thank Sibel Selcuk and Zamira Satarkulova for research
assistance, and participants in our 2004 EEA conference panel and an
anonymous reviewer for helpful comments on the paper.
REFERENCES
Anderson, K and Pomfret, R. 2000: Living standards during
transition to a market economy; The Kyrgyz Republic in 1993 and 1996.
Journal of Comparative Economics 28(3): 502-524.
Anderson, K, Pomfret, R and Usseinova, N. 2004: Education in
Central Asia during the transition to a market economy. In: Heyneman, S
and DeYoung, A. (eds). The Challenges of Education in Central Asia.
Information Age Publishers: Greenwich, CT. pp. 131-152.
Aslaksen, I, Koren, C and Stokstad, M. 2000: The effect of
childcare subsidies: A critique of the Rosen Model. Feminist Economics 6(1): 95-103.
Behrman, J, Chen, Y and Todd, P. 2004: Evaluating preschool
programs when length of exposure to the program varies: A nonparametric
approach. Review of Economics and Statistics 86(1): 108-132.
Bergmann, B. 1996: Saving our Children from Poverty. Russell Sage Foundation: New York.
Blumberg, RL. 1981: Rural women in development. In: Black, N and
Baker Cottrell, A. (eds). Women and World Change. Sage Publications:
Beverly Hills, CA. pp. 32-56.
Connelly, R, DeGraff, D and Levinson, D. 1996: Caring for Preschool
Children in Urban Brazil Labor 10(3): 583-608.
Cornia, GA. 1995: Ugly facts and fancy theories: Children and youth
during transition Innocenti Occasional Papers, EPS 47. UNICEF: Florence,
Italy.
Craig, JE. 1981: The expansion of education. Review of Research in
Education 9: 151-210.
Currie, J and Thomas, D. 1995: Does head start make a difference?
American Economic Review 85 (3): 341-364.
Danzinger, S and Waldfogel, J. 2000: Investing in our children:
What do we know?. In: Danziuger, 8 and Waldfogel, J. (eds). Securing the
Future: Investing in Children from Birth to College. Russell Sage
Foundation: New York. pp. 1-18.
Deaton, A. 1997: The Analysis of Household Surveys: A Microeconomic Approach to Development Policy. Johns Hopkins University Press:
Baltimore, MD.
Deti v Stranakh. 2001: CIS Statistical Committee, Moscow. Cited
from EastWest Database (online.eastview.com/index.jsp_blank).
European Commission. 1995: Employment Observatory: Central and
Eastern Europe 6.
Fong, M and Lokshin, M. 2000: Child care and women's labor
force participation in Romania. World Bank Working paper no. 2400. World
Bank: Washington.
Garces, E, Thomas, D and Currie, J. 2000: Longer-term effects of
head start. American Economic Review 92(4): 999-1112.
Grootaert, C and Braithwaite, J. 1998: Poverty correlates and
indicator-based targeting in Eastern Europe and the Former Soviet Union.
World Bank Policy Working paper no. 1942. World Bank: Washington.
Henscher, M and Passingham, S. 1996: The impact of economic
transition on kindergartens in Kazakhstan: Problems and policy
questions. Compare 26(3): 305-313.
Ipsa, J. 1994: Child Care in Russia in Transition. Bergin and
Garvey: Westport, CT.
Klugman, J, Marnie, S, Micklewright, J and O'Keefe, P. 1997:
The impact of kindergarten divestiture on household welfare in Central
Asia. In: Falkingham, J, Klugman, J, Marnie, S and Micklewright, J.
(eds). Household Welfare in Central Asia. MacMillan Press: New York. pp.
183-200.
Meurs, M and Giddings, L. 2006: Declines in preschool use in
post-socialist societies: The case of Bulgaria. Journal of European
Social Policy 16(2): 155-166.
Ministry of Education and Culture. 2001: The Report on Deuelopment
of Education in the Period 1991-2001. Ministry of Education and Culture:
Bishkek.
Narayan, D and Petesch, P. 2002: Voices of the Poor. From Many
Lands. World Bank: Washington, DC.
NSC (National Statistical Committee). 1995: Statisticheski
Ezhegodnik 1995. National Statistical Committee: Bishkek.
NSC. 1998: Statisticheski Ezhegodnik 1998. National Statistical
Committee: Bishkek.
NSC. 2002: Women and Men in the Kyrgyz Republic. National
Statistical Committee: Bishkek.
O'Connor, S. 1988: Women's labor force participation and
preschool enrollment: A cross-national perspective 1965-80. Sociology of
Education 61: 15-28.
Pomfret, R. 2003: Economic performance in Central Asia since 1991:
Macro and micro evidence. Comparative Economic Studies 45(4): 442-465.
Riazantsev, A, Sipos, S and Labertsky, O. 1992: Child welfare and
the socialist experiment: Social and economic trends in the USSR,
1960-1990. Innocenti Working Paper EPS 24, UNICEF: Florence, Italy.
Spiess, CK, Buchel, F and Wagner, G. 2003: Children's school
placement in Germany: Does kindergarten attendance matter? IZA Working
Paper 722, February 2005 (www.iza.org).
Strauss, J and Thomas, D. 1995: Human resources: Empirical modeling
of household and family decisions. In: Srinivasan, TN and Behrman, JR.
(eds). Handbook of Development Economics. Vol. 3. Elsevier Science
Publishing, North-Holland: Amsterdam, The Netherlands. pp. 1883-2025.
UNDP. 2001a: Democratic Governance: Alternative Approaches to
Kyrgyzstan's Future. UNDP: Bishkek.
UNDP. 2001b: Kyrgyzstan, Common Assessment. UNDP: Bishkek.
UNESCO. 2005: Global Education Digest 2005. UNESCO Institute for
Statistics, Montreal.
UNICEF. 1999: Women in Transition. UNICEF: Florence.
UNICEF. 2003: Social Monitor 2003. UNICEF: Florence.
UNICEF. 2005: Transmonee Database. UNICEF: Florence.
(1) Children per place may not be the best measure of capacity. In
2002-2005, Kyrgyzstan had a relatively high pupil-teacher ratio in
preschool: 21 compared to 12 in Bulgaria and 18 in France (UNESCO,
2005). Using a European measure, no excess capacity would exist.
(2) In fact, households of children attending preschool had
slightly lower incomes, but the difference is not statistically
significant.
(3) These numbers may appear low, but they are very close to those
reported by the UNDP for 1993 (48.3 soms per capita) (2001). Grootaert
and Braithwaite (1998) argue that income may be more systematically
underestimated in household surveys of transition economies than others.
But in 1993, the income data is also distorted by the non-payment of
wages noted above. For the weeks of the survey, households might
therefore report no income at all, but then receive a large sum later,
when back-wages (reduced by inflation) were finally paid. Our numbers
are lower than the expenditure-based figures reported by Anderson and
Pomfret--177 soms per month (2000). But Anderson and Pomfret's
numbers are based on expenditures per adult equivalent in all households
(not just those with pre-school aged children, and thus include
households with older, more experienced workers) and, as Pomfret and
Anderson note, reported expenditures systematically exceed reported
income. We have no reason to believe that any tinder reporting of income
varies in a systematic way across the sample however, so cross-sectional
measures of the impact of income's impact on household decisions
may not be affected by any under-reporting.
(4) Educational coding across the 2 years was not identical, making
it difficult to use the continuous variable years of schooling for 1993.
We also considered the impact of fathers' education, but found it
insignificant.
(5) Anderson et al. (2004) also develop a model of kindergarten
enrolment. They do not consider economic variables such as
households' ability to pay and need for custodial care, however,
but confine their investigation to demographic variables such as
household size, sex and age of the child, ethnicity and parent
education, as well as availability of kindergarten in the locality. We
compare our results with theirs where relevant below.
(6) When the need for custodial care is measured by mother's
employment, income and the need for custodial care are not independent.
However, by using total number of non-working adults as our measure of
the need for custodial care, we separate household income per capita and
the need for custodial care--the use of grandparent labour, for example,
does not imply a reduction in income. We measure cash income here as
regular monetary income: wages, pensions and stipends. Total household
income also includes agricultural income, much of which is imputed from
subsistence production in the survey. Although subsistence production
supports households, it cannot be used to pay for schooling. We have
also estimated the model using a measure of total household per capita
income. The total income measure is not significantly correlated with
enrolment. To take the natural log of the income variable, we
substituted 0.01 for zero in the cash income measure.
(7) Members of other, small ethnic minorities are combined with
Kyrgyz due to coding problems.
(8) Mother's education and per capita cash income are
correlated at a very low level, with a correlation coefficient of
0.09-0.17 in the 2 years.
(9) Although regional differences are significant in Kyrgyzstan, we
did not include a regional dummy here because of the inclusion of other
variables which capture these regional differences, including income,
ethnicity, and access to preschool. A variable indicating number of
siblings was included in previous versions of the equation, but was not
significant.
LISA GIDDINGS (1), MIEKE MEURS (2) & TILAHUN TEMESGEN (3)
(1) Department of Economics, University of Wisconsin-La Crosse,
1725 State Street, La Crosse, WI 54601, USA. 608.785.5297.
http://www.uwlax.edu/faculty/giddings/
(2) American University, 4400 Massachusetts Avenue, NW, Washington,
DC 20016, USA. E-mail: mmeurs@american.edu
(3) The World Bank, 1818 H Street, NW, Washington, DC 20433, USA.
Phone: (202). 473-9181
Table 1: Preschool availability and enrolment in Kyrgyzstan, 1990-1998
1990 1993 1998
Preschools 1,696 998 468
Per cent enrolled 31.3 13.4 8.7
Pupils per place 140 87 86
Sources: NSC, 1995, p. 93; UNICEF, 2003, p. 81; Ministry of Education,
2001, p. 2; Deti v Stranakh, 2001.
Table 2: Descriptive statistics
Total sample Enrolled
Mean S.D. Mean S.D.
1993 n=674 n=43
Sex, M=1 0.51 0.50 0.42 0.50
Age 4.58 1.12 4.7 1.11
PC Cash Income 11.09 25.51 26.29 * 35.74
Adult Not Work 1.06 1.22 0.53 * 0.83
School in area 0.66 0.48 0.98 * 0.13
Mom Primary Ed 0.06 0.24 0 0
Mom Tertiary Ed 0.41 0.49 0.67 0.47
Uzbek 0.14 0.35 0.16 0.37
Slavic 0.09 0.30 0.32 * 0.47
School quality NA NA NA NA
1998 n=1,476 n=1,388
Sex, M=1 0.52 0.50 1.44 0.05
Age 4.49 1.11 4.44 1.04
PC Cash Income 183.43 180.32 384.93 309.77
Adult NW 1.86 1.61 0.83 * 1.02
School in PSU 0.45 0.50 0.90 * 0.30
Mom Primary Ed 0.05 0.21 0.03 0.18
Mom Tertiary Ed 0.42 0.40 0.74 * 0.44
Uzbek 0.07 0.25 0.02 ** 0.15
Slavic 0.06 0.23 0.23 * 0.42
School quality 3.80 1.50 4.10 ** 1.49
Not enrolled
Mean S.D.
1993 n=631
Sex, M=1 0.52 0.50
Age 4.6 1.12
PC Cash Income 10.06 * 24.36
Adult Not Work 1.1 * 1.24
School in area 0.64 * 0.48
Mom Primary Ed 0.07 0.25
Mom Tertiary Ed 0.40 0.49
Uzbek 0.14 0.35
Slavic 0.08 0.27
School quality NA NA
1998 n=88
Sex, M=1 1.49 0.05
Age 4.49 1.11
PC Cash Income 170.66 160.69
Adult NW 1.92 * 1.62
School in PSU 0.42 * 0.49
Mom Primary Ed 0.05 0.21
Mom Tertiary Ed 0.40 * 0.40
Uzbek 0.07 ** 0.25
Slavic 0.05 * 0.21
School quality 3.79 ** 1.49
Source: authors' calculations.
NA = not applicable. * Difference in means significant at P<0.00;
** Difference significant at P<0.10.
Table 3: Per cent enrolled in preschool by group, for groups
represented by dummy variables, Kyrgyzstan
1993 1998
Male 7.6 6.4
Female 5.2 5.5
Slavic 21.5 23.5
Kyrgyz 4.2 5.3
Uzbek 7.3 2.1
Mom primary ed 0.0 4.4
Mom secondary ed 4.0 2.5
Mom tertiary ed 10.4 10.5
Source: authors' calculations.
Table 4: Logistic regression, determinants of child attending
preschool in Kyrgyzstan
1993 1998
n=674 n=1,601
Delta P S.E. Delta P S.E.
School in PSU 0.005 * 0.014 0.051 * 0.389 *
Adult Not Working -0.014 ** 0.006 -0.010 0.123 *
PerCap HH Income 0.012 ** 0.001 0.003 0.098 **
Sex -0.009 0.008 -0.003 0.245
Age 0.001 0.004 -0.002 0.112
Slavic 0.053 *** 0.029 0.029 0.340 **
Uzbek 0.006 0.013 -0.008 0.761
Mom primary ed NA **** 0.005 0.696
Mom tertiary ed 0.82 ** 0.011 0.019 0.288 *
School quality NA -0.001 0.097
Year NA NA
Constant
Pseudo [R.sup.2] 0.22 0.29
Pooled
n=2,150
Delta P S.E.
School in PSU 0.052 * 0.008
Adult Not Working -0.012 * 0.002
PerCap HH Income 0.003 * 0.003
Sex -0.003 0.004
Age -0.002 0.002
Slavic 0.029 * 0.013
Uzbek -0.008 0.007
Mom primary ed 0.005 ** 0.008
Mom tertiary ed 0.019 * 0.006
School quality
Year -0.013 * 0.005
Constant
Pseudo [R.sup.2] 0.27
NA = not applicable.
* Significant at P < 0.01.
** Significant at P < 0.05.
*** significant at P < 0.10.
**** The variable mom primary edu and the cases where
mom primary edu=1 were dropped from the 1993
regression because all mom primary edu=1 cases failed
to attend preschool. In other words, if the mother
had primary education or less, the child had no chance
of attending preschool in 1993. This accounts for
very few cases (41).
Source: authors' calculations.