Regional Disparity and Convergence in Learning amongRural Children: An Analysis of Indian States

December 2023 | Mallika Sinha and Rama Pal

I. Introduction

India has witnessed impressive progress in terms of school enrolments in recent years. However, despite this increasing enrolment, many children are unable to acquire even basic literacy and numeracy skills (Pritchett 2013, Alcott and Rose 2017, Sandefur 2018, Iyer, et. al. 2020). Across countries, literature shows that learning outcomes heavily depend on regional differences in demographic characteristics, school provision, education policies, job opportunities and local infrastructure (Alcott and Rose 2017). This pattern is also observed in the Indian context. As education is on the concurrent list of the Indian constitution, state governments play an important role in the public provision of education. Historic evidence shows that variations in state-specific factors and efforts generate significant differences among educational outcomes across Indian states (Besley, et. al. 2007, Jha, et. al. 2008). Some states exhibit abysmally poor learning outcomes, whereas others have performed remarkably. Given this backdrop, the paper aims to examine the disparities in learning outcomes across Indian states by taking into account some state-specific factors. The ontribution of this study is twofold: (1) we analyse the variations in learning outcomes across Indian states over time and (2) using bivariate and multivariate analysis, we investigate the role of state-specific factors in determining the divergent performance of states in the learning outcomes of children.

Pimary school enrolment in India has been almost universalized (Kingdon 2007). The rise in the net primary enrolment in India has been so impressive over the last few decades that now it exceeds 90 percent in most of the country (Das and Zajonc 2010). However, learning outcomes are dismal (Kingdon 2007, Banerjee, et. al. 2007). Goyal and Pandey (2012) find that in 2007, the ajority of the children in grades four and five could not achieve adequate scores in mathematics, reading comprehension, and word eaning in the Indian states of Uttar Pradesh and Madhya Pradesh. Muralidharan and Zleneik (2013), using a unique panel data set for primary grade children in Andhra Pradesh, find that learning levels in maths and language are not only low but also the learning trajectories are poor over time. This implies that policies effective at expanding school enrolment might not be helpful in raising learning outcomes (Banerjee, et. al. 2007). Also, Muralidharan (2013) points out that one of the reasons that improvement in learning outcomes is not yet at par with improvements in school quality, which is indicated by school inputs, is due to the failure of education policies to prioritize learning outcomes. Education policies and outcomes at the national level mask the differences at the sub-national levels. This paper attempts to study educational policies and factors affecting educational outcomes at the state level.

Literature shows that educational policies and efforts adopted by state governments differ considerably due to social norms, institutions and historical legacies. For instance, Himachal Pradesh witnessed a ‘schooling revolution’ in the early 1970s as an outcome of state action, community participation, public response, parental demand and social equality (Dreze and Sen 2002 and Agarwal, 2014). The resulting changes in the schooling system are responsible for the stellar performance of Himachal Pradesh in learning outcomes (Agarwal 2014). The Education Guarantee Scheme1 (EGS) of Madhya Pradesh and the Shiksha Karmi Project2 of Rajasthan are notable interventions aimed at improving primary education in these states. The states of Gujarat, Karnataka, Tamil Nadu and Kerala reaped the benefits of early recognition and investment in primary education due to the erstwhile policies endorsed by the rulers of Baroda, Mysore, Madras and Travancore regarding the public provision of compulsory primary education
(Mehrotra 2006, Jha, et. al. 2019). At the same time, other states, such as West Bengal and Bihar, lagged behind due to a lack of decentralization, populist policies and ineffective implementation of government policies (Acharya 2002, Mehrotra 2006 and Jha, et. al. 2008).

Differences among education policies manifest themselves in variations in school infrastructure. The school infrastructure has two main components, namely, teachers and physical infrastructure. Tilak (2018) finds that infrastructure and teachers are related to learning outcomes. Chatterjee, et. al. (2018) find large disparities in infrastructure and teacher quality across states. They report better performance of the extreme Northern and Southern states on these parameters as compared to the states in the Northern region and some of the states in the Eastern region. Tilak (2018) notes a good Pupil-Teacher Ratio (PTR) is one of the factors
strongly associated with better school participation rates, the continuation of children in schools and their learning levels. Some studies find that libraries in schools positively impact learning outcomes (Murillo and Roman 2011, Glewwe, et. al. 2011, Chudgar, et. al. 2015). On the contrary, Borkum, et. al. (2012) find libraries in school do not impact the academic achievement of children. However, libraries provide children with a large variety of books that create a fondness for reading and reinforce children’s reading and other learning skills. It also aids children and teachers in supplementing their learning and class teaching activities. (Murillo and Roman 2011). Chakraborty and Jayaraman (2019) find a significant and positive impact of midday meals on learning outcomes in the long run but a negligible impact in the short run in India. Vermeesch and Kremer (2005) and McEwan (2013) do not find the impact of the school feeding program on learning achievements in Kenya and Chile, respectively. Studies show that, along with school infrastructure, parents’ socio-economic factors and educational background affect children’s learning. Children from states with lower female literacy rates are more likely to be first-generation learners than those from states with high female literacy rates. Children who are first-generation learners face immense hardships in accessing school facilities and learning (Govinda and Bandopadhyay 2010). Moreover, they are very less likely to receive the time and attention required for fruitful schooling. Dreze and Sen (2002) enlist some activities parents require, to give time and attention ‘to prepare the child for school in the morning, stimulate his or her interest, help him or her with homework, and establish a rapport with the teachers’. Thus, these children are vulnerable to the risk of silent exclusion from the school system (Govinda and Bandopadhyay 2010).

There exist significant differences in the quality of life by social groups in India. Scheduled Castes (SC) and Scheduled Tribes (ST) are the historically marginalized social groups who persistently suffer from socio-economic oppression (Lastrapes and Rajaram 2016). The exclusion of socially deprived groups in higher levels of learning is reinforced by differences in income, parental education and employment status of families (Deb 2018). According to Desai and Kulkarni (2008), even primary education has the potential to significantly augment the earning prospects for children belonging to the SC and ST communities. This is because primary education gives them the eligibility for lower-level government jobs than depending on scarcely available manual labour jobs in the private sector. Differences of levels of economic development play a critical role in influencing educational outcomes. States with higher levels of economic development are more likely to have a higher capacity to invest in education compared to states with lower levels of economic development. Due to the prevalence of the Zamindari system during the colonial rule in the Eastern states of Bihar, West Bengal and Odisha, their economies suffered significant setbacks, the repercussions of which are felt till today (Jha, et. al. 2019). Jha, et. al. (2019) observe that states with historical underinvestment in social sectors have, with a high share of child population, lower economic capacities and low real per child expenditure on child development and vice versa. To sum up, previous studies point out wide disparities in learning across Indian states. We examine whether such inter-state disparities have changed over
time and whether they persist even today. We analyse the variations in learning outcomes and across Indian states over time. We aim to understand how these disparities may be explained due to differences in school infrastructure and socioeconomic background at the state level. Lastly, we examine the relative importance of the state-level factors in determining the variations in learning outcomes. This paper is organized as follows. Section II illustrates the data and methodology. Section III presents the empirical findings. Finally, Section IV concludes.

II Data and Methodology

Data
We construct a state-level dataset that combines information on child learning, school infrastructure, social and economic factors for Indian states. The sources of data are Annual Status of Education Survey (ASER), District Information for System of Education (DISE), Census, Report on Employment Unemployment Surveys (EUS) and Handbook of Statistics on Indian States. ASER is a household-based survey in rural India that provides information on basic reading and mathematics levels of children aged 5 to 16. It collects
information on child, household and village characteristics. It also collects information on school characteristics by visiting a government school in each sampled village. ASER employs a two-stage sample design. In the first stage, it selects 30 villages from each rural district from the Census directory using Probability Proportion to Size (PPS). In the second stage, it selects 20 households
randomly from each village. ASER gives estimates at the district, state and national levels. While aggregating estimates from district to state and national levels, households must be assigned weights. The weight variable it uses is a household multiplier that
denotes the number of households each sampled household represents in the population. Due to its sampling strategy of PPS in the first stage and Simple Random Sampling in the second stage, all households get an equal chance of being selected at the district level. This means weights assigned to households within a district are the same. Thus, weighted estimates are equivalent to unweighted estimates at the district level. Nonetheless, weights must be included to obtain estimates at the state and national levels as states differ in the number of districts and districts vary by population. (Annual Status of Education Report (Rural 2018). We use ASER household datasets for the years 2009, 2014 and 2018 for the variables learning outcomes and enrolment in government schools. DISE is an administrative dataset which is a census of all recognized schools in India, conducted annually by the Ministry of education. It contains information on school infrastructure, enrolment, teachers, funding and other characteristics of schools. We use data aggregated at the state level from 2008, 2013 and 2016/2017 for the variables PTR, kitchen shed, playground and library.

Outcome Variables
The ASER datasets provide information on learning for rural children in the agegroup 5 to 16 years. The learning levels are reported based on tests of basic literacy and numeracy3. The highest levels of learning tested by ASER align with Std. II on reading (i.e., whether the child is able to read a short story) and Std. IV on mathematics (i.e., whether the child is able to solve a division problem) (ASER, 2018). The same tests are conducted for all children (i.e., age-group 5 to 16). According to NCERT (2017), the expected age to read a short story is 7-8 years and to solve a division problem is 9-10 years. Therefore, we also estimate the learning outcomes for a smaller sample of children i.e., age 10 to 16 as the second specification of the model. This smaller sample consists of older children who are expected to be proficient in the learning tests. To understand the overall learning in the state, we consider the proportion of
children attaining the highest levels of learning considered in the survey. As elaborated in the previous paragraph, these levels indicate foundational reading and mathematical skills that a child is expected to acquire by the end of its primary schooling. We consider the sampling weights, while estimating the proportion of children attaining the highest levels of learning at the state level.

Explanatory Variables
To understand the factors that influence learning outcomes, we consider schoollevel, social and economic factors. School-level factors comprise of enrolment in government schools, PTR, kitchen shed, playground and library. Kitchen shed in school is a proxy variable for midday meals being served in schools. Social factors include female literacy rate, female labour force participation rate, proportion of SC population, proportion of ST population. The Economic factor consists of Net State Domestic Product (NSDP) per capita. Since the data on these state-level background characteristics is not always available for the exact same years as the ASER surveys, we consider the years that are closest to the ASER surveys. At the same time, the difference in the reference years is mostly one year, except for one occasion where it is two years. All the reference years, along with the definitions and sources of variables

II Data and Methodology

Methods
Exploratory Analysis
We aim to explore the correlates of learning outcomes across Indian states graphically. To pursue the graphical analysis, we use simple scatter plots for the initial year, 2009. Then, we analyse the correlation between learning outcomes and some school, social and economic characteristics of Indian states. The school-level characteristics consist of enrolment in government school and PTR. The social characteristics include female literacy rate, female labour force participation rate, proportion of SC population and proportion of ST population. The economic characteristics include NSDP per capita.

Fixed Effects Regression
The data of 2009, 2014 and 20184 is arranged in a panel data structure. We conduct regression analysis to find out how the state-level factors are related to learning outcomes and how important these factors are in predicting children’s performance in learning outcomes. The outcome variables are in terms of proportions. As a result, it is inappropriate to use in the linear regression model, since it may give absurd predictions for extreme values of the regressors (Baum 2008). However, it is possible to use the linear regression analysis if we transform the outcome variables using logit transformation. A logit transformation is the log of the odds ratio (i.e., y/100-y). Using the log-odds ratio as the dependent variable, the fixed effects regression is given as follows:

𝑌it = 𝜃0 + 𝜽𝟏𝑿𝒊𝒕 + 𝛿i + γt + 𝜀it …(1)

Where i are the states varying from 1 to 27, t refers to the years 2009, 2014, 2018. 𝑌 is the logit transformation of learning outcome indicators for state i in year t. We consider Mathematics and Reading indicators for two different age groups, thus giving us a total of four dependent variables. The above model is estimated separately for these variables. 𝑿𝒊𝒕 is the vector of explanatory variables that are given in the previous section. The state and time-fixed effects are indicated by 𝛿 and 𝛾 , respectively.

Absolute and Conditional Convergence
In the last part, we want to understand whether the regional differences among the Indian states decrease over time. For this purpose, we consider the 𝝈 − and 𝜷 − convergence methods. Under the 𝝈- convergence, we estimate the standard deviation for both the learning indicators for each year; 2009, 2014 and 2018. If the variation declines over time, then we say that there is 𝝈- convergence among the states. The 𝜷- convergence method may be used to test both absolute and conditional convergence. This method is based on a regression model for the yearly growth5 in the learning indicators, given as follows:

𝐺𝐸it = 𝛽0 + 𝛽1𝑙𝑛𝐸it– + 𝛿i + γt + 𝜀it …(2)

In equation (2), 𝐺𝐸 , the growth rate in the learning outcome is regressed on 𝐸
the initial level of learning. A negative coefficient indicates that the states with lower initial learning indicators grow at a faster rate. Thus, it implies an absolute convergence. The above model is again estimated using the fixed effects regression.

To test conditional convergence, we include 𝑿𝒊𝒕 𝟏 , the other background characteristics of the states at the initial period. This specification controls for the differences in initial conditions apart from the level of learning and thus, implies conditional convergence.

𝐺𝐸it = 𝛽0 + 𝛽1 𝑙𝑛𝐸𝒊𝒕-𝟏 + 𝜽𝑿𝒊𝒕-𝟏 + 𝛿i + 𝛾t + 𝜀it …(3)

Table 2 describes the mathematical symbols in equations (2) and (3) explaining convergence in learning outcomes.

III Empirical Findings
Literature has reported wide variations in learning across Indian states. We examine whether such variations persist even today. We first analyse variations in the learning outcomes across Indian states for 2009, 2014 and 2018. Next, we aim to understand how these state variations may be explained due to differences in school infrastructure and socio-economic background at the state level. Using simple scatter plots for the initial year, 2009, we analyse the correlations between the learning outcomes and these state-level variables. Lastly, using the panel regression for all three years, we examine the determinants of learning achievement by states.

Regional Variations in Learning Outcomes Over Time

We examine changes in the learning outcomes across states in the past decade. Figure 1 and 2 depict the proportion of children with the highest reading and mathematics levels for 2009, 2014 and 2018. Haryana, Himachal Pradesh, Kerala, Punjab are consistently the best-performing states on learning outcomes in all the years. The main factors reported by studies are state initiatives to improve investment in education and greater opportunities for weaker sections (Sood 2001, Deshpande 2000 and Asadullah and Yalonetzky 2012). Manipur, Mizoram, Sikkim and Uttaranchal also perform well in learning outcomes. The states of Arunachal Pradesh, Assam, Jammu and Kashmir, Jharkhand, Meghalaya, Tripura, Uttar Pradesh have poor learning outcomes. Our findings are in agreement with (Jha and Parvati 2014, Singh and Sarkar 2015, Tilak 2018).

Figure 1: Proportion of Children with Read Level 5 by States (2009, 2014 and 2018)

Figure 2: Proportion of Children with Math Level 5 by States (2009, 2014 and 2018)

Goa and Maharashtra perform remarkably well in reading achievement in all the years. Madhya Pradesh has high reading achievement in 2009 but it deteriorates in subsequent years. These states also exhibit good performance in the Mathematics test with 65.27 per cent, 49.36 per cent and 43.15 per cent of children achieving level 5 in Goa, Madhya Pradesh and Maharashtra in 2009. However, this performance considerably deteriorates over time. This decline in Mathematics achievements is visible in almost all the Indian states. Surprisingly, some of the less-developed states, such as Bihar in mathematics and Odisha in reading, perform well on learning achievement. Despite being a poor state, Odisha has prioritized spending on child development and devised
schemes on children’s pre-education and nutrition, enabling its transition to better status in child development, whereas states such as Gujarat lost the early advantage it had (Jha, et. al. 2019).

Regional Disparity in Learning and State Characteristics

We examine the inter-state variations in learning outcomes by taking into account their background characteristics. These background characteristics comprise of some school-level, social and economic factors of Indian states.

(i) School-Level Factors
Filmer, et. al. (2020) point out that differences in quality of school and school system play a substantial role in students’ learning. Studies show that in many cases, learning differs across government and private schools (Muralidharan and Kremer 2009, Goyal 2009, French and Kingdon 2010, Singhal and Das 2019). Moreover, the quality of education is reflected in other supply-side factors, such as the availability of teachers. Following the literature (Dreze and Sen 2002, Chechi, et. al. 2016) we consider the pupil-teacher ratio as a proxy for the quality of education6.

a) Enrolment in Government Schools
In most of the well-performing states7, such as Himachal Pradesh, Punjab and Madhya Pradesh, a higher percentage of students are enrolled in government schools8 as compared to the other states. For instance, Himachal Pradesh has a high percentage of children (60.19 per cent) enrolled in government schools and a high proportion of children with the given Reading (58.96 per cent) and
Mathematics (54.54 per cent) levels (Figure 3). At the same time, the data does not bring out any definite correlation between
enrolment in government schools and learning outcomes (Figure 3). So, higher percentage of students enrolled in government schools is associated with both better (Himachal Pradesh, Punjab, Madhya Pradesh) and poor (Bihar, West Bengal, Odisha, Assam, Tripura) learning outcomes. These findings support the observations of earlier studies that government schools exhibit substantial
heterogeneity in terms of physical infrastructure and teachers across states (Agarwal 2014). As a result, it is difficult to find any clear relation between government schools and learning.

Figure 3: Learning Outcomes and Enrolment in Government Schools and PTR (Across Indian States9, 2009)

b) Pupil-Teacher Ratio
We find that high PTR in schools is associated with a lower proportion of students achieving basic numeracy and literacy (Figure 3)10. For instance, Jharkhand, Bihar, Uttar Pradesh, West Bengal have high PTR and low learning outcomes. The PTR for West Bengal in 2014 and 2018 is lower compared to 2009. Meghalaya is an outlier with low learning achievement in reading and mathematics in spite of having low PTR. High PTR indicates slow growth of teacher supply resulting in overcrowded classrooms (Jha, et. al. 2008). For instance, the percentage of singleteacher primary schools in 2016 is 45.84 per cent in Arunachal Pradesh, 36.69 per cent in Goa and 28.04 per cent in Jharkhand (DISE, 2016). These states have low learning achievement11.

(ii) Social Factors
Inequalities in educational outcomes persist along lines of gender and caste. (Dreze and Sen 2002, Asadullah and Yalonetzy 2012). Women agency is closely related to female literacy rates and female labour force participation rate (Murthi, Guio and Dreze 1995). Deshpande and Ramachandran (2016) find large caste gaps in learning outcomes between the social groups. We consider female literacy rate, female labour force participation rate, proportion of SC population and proportion of ST population to indicate the social characteristics of the states.

a) Female Literacy Rate
There appears to be a positive correlation between learning outcomes and female literacy rate. For instance, Himachal Pradesh, Kerala, Goa, Maharashtra, Uttarakhand have high learning achievement and high female literacy rates. Jharkhand, Jammu and Kashmir, Uttar Pradesh, Bihar, Rajasthan, Arunachal Pradesh have low learning levels and low female literacy rates. (Figure 4).

Figure 4: Learning Outcomes and Female Literacy Rate and Female Labour Force Participation Rate (Across Indian States, 2009)

b) Female Labour Force Participation Rate
Female labour force participation enhances women’s agency in the households which enables women to focus on and prioritize child care (Sen 1999). Thus, women are able to devote resources and time to their children’s education. Female labour force participation rates are positively related to learning outcomes. Jharkhand, Jammu & Kashmir, Bihar, West Bengal, Odisha, Tripura, Assam, Uttar Pradesh have low learning levels and low female labour force participation rates. Maharashtra, Himachal Pradesh, Chhattisgarh, Mizoram, Andhra Pradesh have high learning levels and low female labour force participation rates (Figure 4).

c) Proportion of SC and ST Population
Figure 5 shows some association between the proportion of disadvantaged population and learning outcomes. For instance, states of Assam, Uttar Pradesh, Jammu and Kashmir, West Bengal, Gujarat have a large proportion of SC population along with low learning outcomes. North-eastern states such as Nagaland, Tripura, Arunachal Pradesh, Meghalaya along with West Bengal have
a high proportion of ST population and low learning outcomes. Whereas Goa, Kerala, and Maharashtra have low proportion of SC and ST population and perform well on the learning outcomes (Figure 5). However, while considering all the states, the findings do not show a strong association between the proportion of SC or ST population and learning.

Figure 5: Learning Outcomes and Proportion of SC and ST Population (Across Indian States, 2009)

One prominent exception to the above-mentioned overall pattern is the state of Himachal Pradesh. Although, Himachal Pradesh has a high proportion of SC and ST population, it has high learning levels. Dreze and Sen (2002) note that caste differences in Himachal Pradesh are less hierarchical and divisive in comparison to other regions. It has relatively egalitarian social norms as well as a strong tradition of cooperative action (Dreze and Sen 2013).

(iii) Economic Factors
We consider NSDP per capita to understand the economic capacity of states.

NSDP Per Capita
States with higher economic capacity are likely to invest more in educational outcomes than states with lower economic capacity. There exists a positive correlation between learning outcomes and NSDP per capita. For instance, Bihar, Uttar Pradesh, Jharkhand, West Bengal, Assam, Tripura, Meghalaya, Jammu and Kashmir have low learning achievement and low NSDP per capita. On the
contrary, Goa, Himachal Pradesh, Uttarakhand, Punjab, Maharashtra, Kerala depict high learning achievement and high NSDP per capita (Figure 6).

Figure 6: Learning Outcomes and NSDP Per Capita (Across Indian States, 2009)

Regression Analysis of Learning Outcomes with State-Level Factors

The simple correlation analysis above shows that school infrastructure and socioeconomic factors indeed have an association with learning outcomes. Now, we want to examine the relative importance of the variables comprising school infrastructure and socio-economic factors in predicting learning outcomes. Using
the data from 2009, 2014 and 2018, we construct a panel dataset to conduct a fixed effects regression analysis to analyse which variables matter more in determining learning outcomes. We regress the learning outcomes of children on school, social
and economic factors. We report two specifications of the model. The first includes all variables covering the entire sample of children in age-group 5 to 16. The second estimates a smaller sample of children in age-group 10 to 16. Table 3 reports the results.

Note: Standard errors are given in parenthesis. Statistical significance. ***p<0.01, **p<0.05, *p<0.1.
Source: Estimation based on three-year longitudinal data for Indian states.

Table 3 shows that out of all the school-level factors considered in the regression, PTR has a statistically significant negative impact on reading achievement of children of age-group 5 to 16. Several studies have pointed out the negative effect of high PTR on learning achievement (Urquiola 2006, Urquiola and Verhoogen 2009, Mulera, et. al. 2017). A Kitchen shed in school has a positive
significant impact on reading achievement of children of age-group 5 to 16 but no impact on mathematics achievement. Our finding is in consonance with Adrogue and Orlicki (2013) that school meal programs led to improvement in language scores and not maths scores in Argentina. Enrolment in government schools, playground and library is insignificant in determining the learning achievement of children.
Female literacy rate has a positive significant impact on the mathematics achievement of children for both the specifications. Female labour force participation rate and proportion of SC population lack significance in predicting child learning. Proportion of ST population is negatively significant in determining reading achievement for children of age-group 5 to 16 and mathematics achievement for both the specifications. Deb (2018) finds that caste status has an adverse impact on education and learning achievement of some states and serves as a source of disadvantage in enrolment, attendance and completion rates of children.
NSDP per capita has a positive significant effect on mathematics achievement for both specifications. For instance, a one percent increase in NSDP per capita increases the odds of better mathematics outcomes by 0.89 percent. Higher NSDP per capita tends to be associated with low poverty levels and thus, reduced risk of drop-outs due to poverty (Tilak 2018).
The negative time trend in mathematics tests for both specifications indicates that learning outcomes in mathematics are declining over time. Shah & Steinberg (2019) point out several reasons for this phenomenon of increasing enrolment and declining learning outcomes after RTE came into effect (i.e., 2009). Firstly, RTE facilitated the influx of first-generation learners into the school system thereby causing negative peer effects in the school. Secondly, lack of exams and automatic promotion of students in primary school has deteriorated the learning levels of children.

Absolute and Conditional Convergence
In order to understand whether regional variations across Indian states decrease over time, we employ the 𝝈 − and 𝜷 − convergence methods. Figure 7 shows the standard deviations of learning indicators i.e., reading and mathematics achievement of Indian states for each year considered in our analysis – 2009, 2014 and 2018. The dispersion in mathematics achievement of states gets reduced
during the examined period, therefore we can say 𝝈 convergence occurs in mathematics achievement among the states. However, we do not find similar signs of 𝝈 convergence in reading achievement of states, particularly for the age-group 10-16 years. This result indicates that the growth rate of proportion of children achieving proficiency in the lagged-behind states for reading achievement is not sufficient to reduce the dispersion over time.

Figure 7: Standard Deviations for Learning Indicators (𝜎-convergence)

To evaluate absolute and conditional convergence, we use the regression model in equations (2) and (3) and estimate using the fixed effects method again. Table 4 provides the absolute and conditional convergence results. Model 1 of Table 4 shows the results for absolute convergence. The negative coefficient of the lagged value of the initial level of reading and mathematics achievement exhibit evidence in favour of absolute convergence in learning indicators among states. States with low initial learning performance are able to catch up with better-off states in the long run. Model 2 of Table 4 shows the results for conditional convergence after controlling for the above-mentioned state characteristics. We find support for the existence of conditional convergence. School-level factors such as enrolment in government schools, library and social factors such as female labour force participation rate, proportion of ST population impact learning indicators of states. Thus, regional variations among states in learning indicators tend to decline over time.

Notes: (1) Standard errors are given in parenthesis. (2) Statistical significance. ***p<0.01, **p<0.05, *p<0.1. (3)

Time fixed effects are included in both the models.
Source: Estimation based on three-year longitudinal data for Indian states.

IV Conclusion
This paper studies the disparities in Indian states in learning achievement and its association with some state-specific factors. In order to uncover the explanation behind regional disparities in learning outcomes, we explore the association of learning with school infrastructure and socioeconomic variables across the Indian states. Among the school-level factors, states with very high PTR have low learning achievement. Among the social characteristics, female literacy rates and female labour force participation are positively associated with learning outcomes. Looking at the economic factors we find that states characterized by high NSDP
per capita have a positive association with learning outcomes.
The regression analysis shows that among the social factors considered, proportion of ST population is negatively significant in determining reading achievement for children of age-group 5 to 16 and mathematics achievement for both the specifications. Female literacy rate has positive significant impact on mathematics achievement. The economic factors such as NSDP per capita have a positive significant effect on mathematics achievement. Among school-level factors, PTR has a negative effect on reading achievement and kitchen shed has a positive impact on reading achievement for children of age-group 5 to 16. Although the disparities exist, the convergence analysis shows some evidence of a reduction in the same. We find σ-convergence in mathematics achievement for Indian states but not in reading achievement. Also, the absolute and conditional βconvergence in the learning indicators among the states holds true. Thus, regional variations among states in learning indicators tend to decline over time.
The regression analysis re-establishes the adverse role of disadvantages faced by the marginalised sections of the population on the basis of social background. Children from socially disadvantaged groups largely attend government schools. Improving the quality of the government schooling system is imperative for improving the learning levels of children equitably. However, most school-level factors have no significant impact on learning achievements across both the specifications. Thus, for all the states taken together, the schooling system does not seem to correct the inequalities in learning based on socio-economic and educational backgrounds. Banerjee and Duflo (2011) rightly argue that since the school curriculum and teaching in developing countries is aligned to cater to the needs of the affluent sections of children rather than the regular ones, school inputs are unable to improve school quality and hence learning. This is also evident from the fact that female literacy and per capita income remain important determinants of the learning outcomes.
A decentralized approach towards education has the potential to yield better educational outcomes. The experience of Madhya Pradesh and Rajasthan highlights the importance of decentralized efforts in education. An in-depth analysis of such success stories may throw light on possible measures that can be taken for improving learning across all Indian states.

Endnotes

1.Education Guarantee Scheme is a community-centred, rights-based initiative to establish primary schooling facility within 90 days within a distance of 1 km owing to demand from at least 25 learners in tribal areas and 40 learners in non-tribal areas (Gopalakrishnan and Sharma 1998).
2.Shiksha Karmi Project aimed at improving the access to basic education by recruiting local youths as teachers in rural remote areas where the primary schools are grappling with teacher absenteeism, dysfunctional schools, community despair (Ramachandran 2001) low enrolment and high dropout rates, especially of girls (Rajagopal 2015). It laid emphasis on women teachers, i.e., Mahila Shiksha Karmis and enabled women empowerment by promoting Mahila Prashikshan Kendras, Mahila Sahyoginis and Women’s groups as well as their representation and active role in the village education committees (Tilak 2018).
3.The learning tests assess learning based on five Reading and five Mathematics levels. The five reading levels are (I) cannot read letters, (II) can read letters but not words, (III) can read words but not paragraph, (IV) can read paragraph but not short story, (V) can read short story. Similarly, the Mathematics levels are also divided into five categories as, (I) cannot recognize numbers 1 to 9, (II) can recognize numbers 1 to 9 but not 10 to 99, (III) can recognize numbers 10 to 99 but cannot solve two-digit subtraction problem with borrowing, (IV) can solve twodigit subtraction problem with borrowing but cannot solve division problem of three-digit number divided by one-digit number, (V) can solve the division problem.
4.We could not take all the years between 2009, 2014 and 2018 and are unable to make it a continuous panel as data on the outcome and all the explanatory variables are not available for all the year in-between. For instance, ASER data is not available for 2015 and 2017. Similarly, data on some other explanatory variables are not available for some other years.
5.As the number of years between the two time-periods, 2009-2014 and 2014-2018 is not the same; while calculating the growth rate of the first time period, we divide by four and for the second, we divide by five.
6.Additional school characteristics are included in the regression analysis as mentioned in the previous section.
7.Goa and Kerala are the exceptions as they exhibit a relatively higher percentage enrolment (56.7 per cent and 42 per cent, respectively) in the private schools (ASER, 2009).
8.Among non-government schools, private schools have the highest enrolment. 18.80 per cent of enrolment was in private schools and only 0.84 per cent enrolment was in remaining nongovernmental schools (ASER, 2009). These non-governmental schools comprise of Madrasa, EGS or ‘other’ non-formal schools in the ASER data.
9.AP=Andhra Pradesh ARP=Arunachal Pradesh AS=Assam BH=Bihar CH=Chhattisgarh GO=Goa GJ=Gujrat HR=Haryana HP=Himachal Pradesh JK=Jammu & Kashmir JH=Jharkhand KT=Karnataka KR=Kerala MP=Madhya Pradesh MH=Maharashtra
MN=Manipur MG=Meghalaya MZ=Mizoram NG=Nagaland OD=Odisha PN=Punjab RJ=Rajasthan SK=Sikkim TN=Tamil Nadu TR=Tripura UP=Uttar Pradesh UK=Uttarakhand WB=West Bengal
10.As a result, to reduce disparity in student learning across socio-economic backgrounds, the Right to Education Act (RTE), 2009 recommends a PTR of 30: 1.
11.Although Goa has done well in mathematics in 2009, its performance has reduced drastically in the subsequent years.

Reference

Acharya, P. (2002), Education: Panchayat and Decentralisation: Myths and Reality, Economic and Political Weekly, 37(8): 788-796.

Agrawal, T. (2014), Educational inequality in rural and urban India, International Journal of Educational Development, 34: 11-19.

Alcott, B. and P. Rose (2017), Learning in India’s Primary Schools: How do Disparities Widen across the Grades?, International Journal of Educational Development, 56: 42-51.

Asadullah, M.N. and G. Yalonetzky (2012), Inequality of Educational Opportunity in India: Changes over Time and across States, World Development, 40(6): 1151-1163.

ASER (2009), Annual Status of Education Report (Rural), Pratham, New Delhi.

———- (2014), Annual Status of Education Report (Rural), Pratham, New Delhi.

———- (2018), Annual Status of Education Report (Rural), Pratham, New Delhi.

Banerjee, A.V. and E. Duflo (2011), Poor Economics: A Radical Rethinking of the Way to Fight Global Poverty, New York, Public Affairs.

Banerjee, A.V., S. Cole, E. Duflo and L. Linden (2007), Remedying Education: Evidence from Two Randomized Experiments in India, The Quarterly Journal of Economics, 122(3): 1235-1264.

Baum, C.F. (2008), Stata Tip 63: Modeling Proportions, The Stata Journal, 8(2): 299-303.

Besley, T., R. Burgess and B. Esteve-Volart (2007), The Policy Origins of Poverty and Growth in India, In Tim Besley, and Louise Cord (Eds.), Delivering on the Promise of Pro-Poor Growth, London, Palgrave Macmillan and the World Bank.

Borkum, E., F. He and L.L. Linden (2012), The Effects of School Libraries on Language Skills: Evidence from a Randomized Controlled Trial in India, National Bureau of Economic Research, Working Paper, w18183.

Census of India (2001), Office of the Registrar General and Census Commissioner, Ministry of Home Affairs, Government of India.

———– (2011), Office of the Registrar General and Census Commissioner, Ministry of Home

Affairs, Government of India.

Chakraborty, T. and R. Jayaraman (2019), School Feeding and Learning Achievement: Evidence from India’s Midday Meal Program, Journal of Development Economics, 139: 249-265.

Chatterjee, I., I. Li and M.C. Robitaille (2018), An Overview of India’s Primary School Education Policies and Outcomes 2005–2011, World Development, 106: 99-110.

Chudgar, A., M. Chandra, R. Iyengar and R. Shanker (2015), School Resources and Student Achievement: Data from Rural India, Prospects, 45(4): 515-531.

Das, J. and T. Zajonc (2010), India Shining and Bharat Drowning: Comparing Two Indian States to the Worldwide Distribution in Mathematics Achievement, Journal of Development Economics, 92(2): 175-187.

Deb, S. (2018), Learning and Education in India: Social Inequality in the States, Social Change, 48(4): 630-633.

Desai, S. and V. Kulkarni (2008), Changing Educational Inequalities in India in the Context of Affirmative Action, Demography, 45(2): 245-270.

Deshpande, A. (2000), Does Caste Still Define Disparity? A Look at Inequality in Kerala, India, American Economic Review, 90(2): 322-325.

Deshpande, A. and R. Ramachandran (2016), The Changing Contours of Intergroup Disparities and the Role of Preferential Policies in a Globalizing World: Evidence from India, Centre for Development Economics, Working Paper, 267.

District Information System for Education (2008), Elementary Education in Rural India: Where do We Stand? Analytical Tables 2008-09, New Delhi, National University of Educational Planning and Administration.

———- (2013), Elementary Education in Rural India: Where do We Stand? Analytical Tables 2013-

14, New Delhi, National University of Educational Planning and Administration.

———- (2016), Elementary Education in Rural India: Where do We Stand? Analytical Tables 2016-

17, New Delhi, National University of Educational Planning and Administration.

District Information System for Education (2017), Elementary Education in Rural India: Where do We Stand? Analytical Tables 2017-18, New Delhi, National University of Educational Planning and Administration.

Dreze, J. and A. Sen (2002), India: Development and Participation, New York, Oxford University Press.

———- (2013), An Uncertain Glory: India and Its Contradictions, Princeton, Princeton University

Press.

Filmer, D., H. Rogers, N. Angrist and S. Sabarwal (2020), Learning-Adjusted Years of Schooling (LAYS): Defining a New Macro Measure of Education, Economics of Education Review, 77: 1-30.

French, R. and G. Kingdon (2010), The Relative Effectiveness of Private and Government Schools in Rural India: Evidence from ASER Data, London: Institute of Education.

Glewwe, P.W., E.A. Hanushek, S.D. Humpage and R. Ravina (2011), School Resources and Educational Outcomes in Developing Countries: A Review of the Literature from 1990 to 2010, National Bureau of Economic Research, Working Paper No. w17554.

Gopalakrishnan, R. and A. Sharma (1998), Education Guarantee Scheme in Madhya Pradesh: Innovative Step to Universalise Education, Economic and Political Weekly, 33(39): 2546-2551.

Government of India (2010), Report on First Annual Employment-Unemployment Survey 2009-10,

Volume I, Chandigarh, Ministry of Labour and Employment, Labour Bureau.

———- (2014), Report on Fourth Annual Employment-Unemployment Survey 2013-14, Volume I,

Chandigarh, Ministry of Labour and Employment, Labour Bureau.

———– (2016), Report on Fifth Annual Employment-Unemployment Survey 2015-16, Volume I,

Chandigarh, Ministry of Labour and Employment, Labour Bureau.

Govinda, R. and M. Bandyopadhyay (2010), Social Exclusion and School Participation in India: Expanding access with Equity, Prospects, 40(3): 337-354.

Goyal, S. (2009), Inside the House of Learning: The Relative Performance of Public and Private Schools in Orissa, Education Economics, 17(3): 315-327.

Goyal, S. and P. Pandey (2012), How Do Government and Private Schools Differ?, Economic and Political Weekly, 47(22): 67-76.

Handbook of Statistics on Indian States (2019), Reserve Bank of India, 2019-20.

Iyer, P., C. Rolleston, P. Rose and T. Woldehanna (2020), A Rising Tide of Access: What Consequences for Equitable Learning in Ethiopia?, Oxford Review of Education, 46(5): 601- 618.

Jha, J. (2003), Primary Schools in West Bengal, Economic and Political Weekly, 38(27): 2839-2840. Jha, J., B.V. Madhusudhan Rao, S. Siddarth, J. Sowmya, P.T. Lekshmi, M.V. Susmitha, K.S. Deepa and S.M. Abraham (2019), Public Expenditure on Children in India: Trends and Patterns,

India, Centre for Budget and Policy Studies and United Nations Children Fund.

Jha, P. and P. Parvati (2014), Assessing Progress on Universal Elementary Education in India: A Note on Some Key Constraints, Economic and Political Weekly, 49(16): 44-51.

Jha, P.K., S. Das, S.S. Mohanty and N.K. Jha (2008), Public Provisioning for Elementary Education in India, India, SAGE.

Kingdon, G.G. (2007), The Progress of School Education in India, Oxford Review of Economic Policy, 23(2): 168-195.

Lastrapes, W.D. and R. Rajaram (2016), Gender, Caste and Poverty in India: Evidence from the National Family Health Survey, Eurasian Economic Review, 6(2): 153-171.

Mehrotra, S. (Ed.) (2006), The Economics of Elementary Education in India: the Challenge of Public Finance, Private Provision and Household Costs, New Delhi, Sage publications.

Mulera, D.M.W., K.K. Ndala and R. Nyirongo (2017), Analysis of Factors Affecting Pupil Performance in Malawi’s Primary Schools based on SACMEQ Survey Results, International Journal of Educational Development, 54: 59-68.

Muralidharan, K. (2013), Priorities for Primary Education Policy in India’s 12th Five-Year Plan,

India Policy Forum, 9(1): 1- 61, New Delhi. National Council of Applied Economic Research.

Muralidharan, K. and M. Kremer (2009), School Choice International, In Paul Peterson and Rajashri Chakrabarti (Eds.), Public and Private Schools in Rural India, Cambridge, MIT Press.

Muralidharan, K. and Y. Zieleniak (2013), Measuring Learning Trajectories in Developing Countries with Longitudinal Data and Item Response Theory, UC San Diego.

Murillo, F.J. and M. Román (2011), School Infrastructure and Resources Do Matter: Analysis of the Incidence of School Resources on the Performance of Latin American Students, School Effectiveness and School Improvement, 22(1): 29-50.

Murthi, M., A.C. Guio and J. Dreze (1995), Mortality, Fertility, and Gender Bias in India: A District- Level Analysis, Population and Development Review, 21(4): 745-782.

NCERT (National Council of Educational Research and Training) (2017), Learning Outcomes at the Elementary Stage, New Delhi, NCERT.

Pritchett, L. (2013), The Rebirth of Education: Schooling ain’t Learning, Washington, Center for Global Development.

Rajagopal, S. (2015), Education and Empowerment in India, In Avinash Kumar Singh (Eds.), Making Policies Work for Education and Social Empowerment: Reflections on the Shiksha Karmi Project in Rajasthan, India, Routledge.

Ramachandran, V. (2001), Community Participation in Primary Education, Economic and Political Weekly, 36(25): 2244-2250.

Sandefur, J. (2018), Internationally Comparable Mathematics Scores For Fourteen African Countries, Economics of Education Review, 62: 267-286.

Sen, A. (1999), Development as Freedom, New York, Alfred A. Knopf, Inc.

Shah, M. and B. Steinberg (2019), The Right to Education Act: Trends in Enrollment, Test Scores, and School Quality, In AEA Papers and Proceedings, 109: 232-38.

Singh, R. and S. Sarkar (2015), Does Teaching Quality Matter? Students Learning Outcome Related to Teaching Quality in Public and Private Primary Schools in India, International Journal of Educational Development, 41: 153-163.

Singhal, K. and U. Das (2019), Revisiting the Role of Private Schooling on Children Learning Outcomes: Evidence from Rural India, South Asia Economic Journal, 20(2): 274-302.

Sood, A. (2001), Himachal Pradesh: Critical Issues in Primary Education, Economic and Political Weekly, 38(25): 2469-2470.

Tilak, J.B. (Ed). (2018), Education and Development in India, Critical Issues in Public Policy and Development, New Delhi, Palgrave Macmillan.

Urquiola, M. (2006), Identifying Class Size Effects In Developing Countries: Evidence From Rural Bolivia, The Review of Economics and Statistics, 88(1): 171-177.

Urquiola, M. and E. Verhoogen (2009), Class-Size Caps, Sorting, And The Regression-Discontinuity Design, American Economic Review, 99(1): 179-215.

Gokhale Institute Of Politics And Economics

Artha Vijnana

Regional Disparity and Convergence in Learning amongRural Children: An Analysis of Indian States

Recommended Articles