Effectiveness of policy regarding power supply to schools through solar energy: Using solar panels to increase enrolment at schools in Pakistan

Background: Climate change has adversarial impact on social and economic development, with negative consequences on living conditions, health outcomes and labour productivity. Although there is ample literature that highlights the use of renewable energy, especially solar energy, in mitigating the effects of climate change, empirical evidence linking benets of renewable energy with educational access and attainment is very limited. Results: By using school level data between 2013-18 from Pakistan, this paper estimates the effectiveness of the policy regarding power supply to schools through solar energy on enrolment. Using a difference-in-difference estimation and controlling for school-specic and infrastructure-specic characteristics, we show that average enrolment per school in Punjab increased by 48.5 students, compared to average enrolment per school elsewhere in Pakistan, after the policy of installing solar panels at schools in Punjab was announced. Furthermore, gender-wise breakup of results suggest that the policy resulted in increased enrolment for boys’ schools, girls’ school and mixed-gender schools. The results are consistent when standard errors are normal, robust, or clustered at the province level. Policy implications: Limitation of the study includes unidentied mechanisms for the policy effect and heterogeneous effects on gender, which should be the focus of future research.

One of the leading causes of climate change is the increased levels of green-house gas emissions resulting in global warming, with the amount of emissions exponentially rising in the recent decades photovoltaic and concentrated solar power plants, the employment effects of solar energy industry in Turkey are determined (Cetin et al., 2011). They conclude that solar energy in Turkey would be the primary source of energy demand and would have big employment effects on the economics. However, the achieved targets can only be achieved with the support of governmental feed-in-tariff policies of solar energy and by increasing research-development funds. Moving on, Wiser et al. (2016) study the environmental and public health bene ts of achieving high penetrations of solar energy in the US. They nd that with feasible cost reductions, solar power can provide major environmental bene ts, along with air quality, public health bene ts and reduced power-sector water withdrawals and consumption.
Moreover, the success of the UK policy to reduce carbon emissions has been partly attributed to successful persuasion of households to become more energy su cient, and to encourage installation of domestic solar systems (Faiers et al., 2006). However, since the nancial, economic and aesthetic characteristics of adopting solar systems are only viewed plausible by a small fraction of the population, emphasis on marketing and development of solar products must be made. Furthermore, by conducting a randomized eld experiment in India to estimate the causal effects of off-grid solar power on electricity access and broader socioeconomic development, Aklin et al. (2017) reveal that daily hours of access to electricity only increased marginally, whereas kerosene expenditure on the black market only decreased slightly. At the same time, no systemic evidence for changes in savings, spending, business creation or time spent working or studying were found. Finally, Aevarsdottir et al. (2017) provide experimental evidence on the impacts of non-grid small scale electri cation by offering randomized subsidies towards a solar lamp with a mobile phone charging point to hosueholds in Tanzania. They nd that lamps, along with positively affecting expenditure on lighting and mobile phone charging, also impacted labour supply, household income and well-being. Additionally, they nd signi cant positive effects on health in the subsample of households that did not previously own a solar lamp.
Using school level data of 21,412 schools across 176 districts of Pakistan from years 2013 till 2018, we examine the effectiveness of the policy regarding power supply to schools through solar panel. By using a difference-in-difference estimation and controlling for school-speci c and infrastructure-speci c characteristics, we estimate whether the policy resulted in differential in enrolment between schools belonging to province of Punjab, where the policy was implemented, and elsewhere in Pakistan, where the policy was not implemented. As a result, our sample size can be divided into 5,444 schools from Punjab being part of the treatment schools, while 15,968 schools from elsewhere in Pakistan are part of the control schools. We provide conclusive evidence that the policy of installing solar panels in schools of Punjab resulted in higher enrolment, compared to enrolment in schools elsewhere in Pakistan. In particular, average enrolment per school in Punjab increased by 48.5 students, compared to average enrolment per school elsewhere in Pakistan, after the implementation of the policy. Gender-wise breakup of results shows that for boys' schools, average enrolment per school in Punjab increased by 24.1 students, compared to average enrolment per school elsewhere in Pakistan, after the implementation of the policy. At the same time, for girls' schools too, average enrolment per school in Punjab increased by 48.5 students, compared to average enrolment per school elsewhere in Pakistan, after the implementation of the policy. Similarly, for mixed-gender schools also, average enrolment per school in Punjab increased by 54.1 students, compared to average enrolment per school elsewhere in Pakistan, after the implementation of the policy. The results are consistent when standard errors are normal, robust, or clustered at the province level. Mekhilef et al. (2011) point out that with the increasing popularity in industrial application, solar thermal is an alternative to generate electricity, process chemicals and even space heating. At the same time, solar electricity can be widely applied in telecommunication, agriculture, textile and water desalination.
They highlight that according to International Energy Agency, by 2050, solar array installations will supply around 45 per cent of energy demand in the world. Mondal et al. (2011) nd the impacts of solar home systems (SHSs) application in Bangladesh and reveal that under the SHSs, women and children bene tted from the improved quality of life for household work and studying in the evening. Moreover, users of SHSs became accustomed to the better quality of light as solar electri cation added to the overall comfort and satisfaction of the households. Jacobson (2007) conclude that while income and work related uses of solar lighting in Kenya are modest, education uses are more signi cant. However, solar electri cation may potentially contribute to sustainable development, but concerns about equity and other social issues need careful attention. In India, Buragohain (2012) examines the impact of installing solar photovoltaic home lighting systems under the 'Remote Village Electri cation Program'. The study reports that the Program led to signi cant improvement in children's education, along with improvement in standard of living. Bene ciaries of the Program spent more time on income generating activities while crime rate also declined due to availability of solar lights in the village.
This paper adds to the debate on the advantages of renewable energy, especially solar power, by exploring a substantial gap in the literature related to the impact of solar energy on educational access and attainment. Literature on the effectiveness of solar energy in lowering barriers to entry to schooling, along with improving supply-side determinants, is very limited and scattered. By using an extensive dataset related to schools, and a uniquely identi ed policy in favour of providing power supply to schools through solar energy, this paper provides quantitative evidence that solar energy can drive enrolment at schools, especially in countries where barriers to entry in education are high. Countries like Pakistan, which are facing severe challenges in achieving Sustainable Development Goal related to universal access to free, quality and equitable schooling, can bene t greatly from technologically-advance alternate methods of power generation to schools, especially in areas with large populations cut-off from the ongrid electri cation. At the moment, due to global warming and extreme weathers, Pakistan is on the trajectory of becoming uninhabitable by the end of the century (Worland, 2017) whereas severe energy crises and electrcitiy shortfall has only hindered social and economic development (Ahmed et al., 2015). These facts point towards immediate inclusion of renewable energy as a viable source of power generation, although there is need to scale-up research in the development of renewable energy in order to better inform energy policies.
The paper is structured as follows: Section II explains the background of the solar power policy implemented in Punjab province of Pakistan. Section III describes the dataset and provides descriptive statistics while Section IV discusses the identi cation strategy applied for our estimations. Results are presented in Section V while Section VI concludes the paper.

Background Of Policy Regarding Power Supply To Schools Through Solar Energy
On his visit to China in early 2014, the Chief Minister of Government of Punjab province, Shahbaz Sharif, met with potential investors belonging to different energy companies, nancial institutions and development agencies, and discussed various proposals for investment in coal, hydel and solar power projects in Punjab (Pakistan Today, 2014). Due to the high energy consumption, the Government of Punjab was exploring various options to increase electricity production to cater to the demand of the province. By late 2014, Government of Punjab had selected 46 sites in the province to set up small-scale solar power projects aimed at reducing power shortages (The Express Tribune, 2014). Power shortages had been linked to having a negative impact on health, agriculture and education sector. A few months later, the Government of Punjab communicated that a strategy was being devised to install solar panels in primary schools to ensure uninterrupted power supply to schools (Geo News, 2014). In early 2015, Shahbaz Sharif o cially announced that 4,000 primary schools in rural Punjab would use solar power under the rst phase of the Off-Grid Solar Solution program (The Express Tribune, 2015). Under the program, immediate installation of solar panels would take place at the chosen schools to ensure continuous provision of electricity. By mid-2015, the Government of Punjab communicated that 10,000 schools were being shifted to solar energy in the province, half of which did not have electricity (Pakistan Today, 2015). With the nancial assistance provided by Asian Development Bank, the project was completed by the end of 2015 (The News, 2016). With the successful completion of the project, the Government of Punjab launched its second program, Khadim-e-Punjab Ujala Program, that would solarize 20,000 schools in the entire province, focusing on remote areas (Associated Press of Pakistan, 2017). The purpose of the project was not only to improve the power supply to schools, but also raise awareness about solar energy as an alternative energy source. With the success of the rst program, along with less than 50 per cent of the population of entire Pakistan connected to the national grid, and Pakistan being heavily dependent on imported electricity, the program attracted support from international nancing agencies such as Asian Development Bank and AFD Bank of France (Climate Action, 2017). In conclusion, the policy of shifting schools on solar energy was formulated, and initiated, in 2015, following which schools were provided electricity from solar energy in academic year 2016 onwards.

Data And Descriptive Statistics
Data used for this study is provided by Annual Status of Education Report (ASER) Pakistan. ASER Pakistan is part of a South-South initiative across India, Kenya, Mali, Mexico, Senegal, Tanzania and Uganda, aimed at addressing quality, accountability and governance in education. ASER Pakistan is the largest citizen led, household based initiative in Pakistan, and is active in all the districts of Pakistan. Each year, 30 villages per district are selected randomly using the Probability Proportional to Size (PPS) technique from the village directory of the 1998 Population Census. From each selected village, one government school is chosen, resulting in an average of 4,500 schools chosen every year across Pakistan. Moreover, each year, 20 villages from the previous year are retained and 10 new villages are added. The data used for this study is from year 2013 till 2018, with the exclusion of 2017 as ASER did not release any statistics related to education in that year.

Enrolment per school in Punjab and elsewhere in Pakistan
We rst provide descriptive analysis of our main outcome variable, enrolment at school, for schools in Punjab, the province where the policy regarding power supply to schools through solar energy was announced, and for schools elsewhere in Pakistan. Our data divides enrolment at school into four different categories i.e. enrolment per school for all schools, enrolment per school for only boys' schools, enrolment per school for only girls' schools and enrolment per school for only mixed schools. Year-wise trend of enrolment at school for all four categories is provided in Table1.  Figure 1.

School-speci c characteristics
Descriptive statistics on school-speci c characteristics are presented in Table 2. School-speci c characteristics include type of school, medium of instruction at school, grade-level of school and total number of teachers at school. Type of school contain information on whether a school is only for boys, or only for girls, or mixed-gender. Similarly, grade-level of school shows whether a school is up till primary-level, or up till secondary-level school, or up till high school, or any other grade-level school.
Moreover, medium of instruction at school demonstrates whether learning at school is in English, or in Urdu, or in regional language Pashto and Sindhi, or in any other regional language. Finally, teachers at school represent total teachers appointed at school. Per cent of schools with other regional language as medium of instruction

Infrastructure-speci c characteristics
Descriptive statistics on infrastructure-speci c characteristics are presented in Table 3. Infrastructurespeci c characteristics are de ned as facilities available at individual schools. These facilities include school having a boundary wall, available drinking water at school, toilet facility at school, electricity availability at school, playground facility at school, available science laboratory at school and availability of books in school library.  Table 3 suggests that the majority of schools selected in the sample have basic facilities such as boundary wall, drinking water, toilets and electricity. Over the years, two-third of the schools in the sample consistently have electricity and toilets, whereas the proportion of schools with drinking water uctuated

Estimation Model
To estimate the effectiveness of the policy regarding power supply to schools through solar energy, we treat the announcement of the policy in 2015 as a stimulus, and apply a difference-in-difference identi cation strategy. This estimation technique relies on the differential changes in enrolment of schools belonging to districts in Punjab, where the policy was implemented, and enrolment of schools belonging to districts from elsewhere in Pakistan, where the policy was not implemented.  The set of control variables X' i,d,t include school-speci c characteristics and infrastructure-speci c characteristics. School-speci c characteristics control for level of teaching at school, type of school, medium of instruction at school, and number of teachers at school. Level of teaching is de ned as an ordinal variable that indicates whether school i belonging to district d in year t is only up till primary-level (equals one), or up till middle-level (equals two), or up till high-school (equals three) or any other classi cation (equals four). Moreover, type of school is de ned as an ordinal variable that indicates whether school i belonging to district d in year t is an only boys school (equals one), or an only girls school (equals two) or a mixed-gender school (equals three). Furthermore, medium of instruction is also de ned as an ordinal variable that indicates whether learning at school i belonging to district d in year t is in English (equals one), or in Urdu (equals two), or in regional language Pashto (equals three) and Sindhi (equals four), or in any other regional language (equals ve). Lastly, number of teachers at school represents the total appointed teachers in school i belonging to district d in year t.
Moving on, infrastructure-speci c characteristics controls for the facilities available at each school i belonging to district d in year t. The controls are binary variables, each variable individually representing whether school i belonging to district d in year t has electricity available, toilet facility, a boundary wall, available drinking water, availability of books in library, playground facility and available science laboratory. Each binary variables takes a value equal to one if the facility is available at school, otherwise the binary takes a value equals to zero.
Finally, our estimation equation also includes district dummies and time dummies as controls for each individual district and year, to account for any variation in enrolment due to district-speci c or yearspeci c characteristics. ε i,d,t is the error term.

Results
We rst provide results for the effectiveness of the policy regarding power supply to schools through solar energy on enrolment of all schools. We examine whether installing solar panels at schools in Punjab resulted in increased enrolment, in comparison to schools elsewhere in Pakistan.  Notes: OLS estimations with enrolment as dependent variable with normal standard errors in Column (1), robust standard errors in Column (2) and clustered standard errors at province level in Column (3).
Province*Year is the independent variable of interest which is de ned as the interaction between province and year. Control variables include level of teaching at school, type of school, medium of instruction at school, total appointed teachers at school and binary variables for whether a school has electricity, toilet, boundary wall, drinking water, books in library, playground and science laboratory. Sample sizes are 13,884 for Columns (1) till (3).
The positive coe cient of Province*Year in Column (1) in Table 4 is signi cant at 1 per cent level, suggesting strong evidence that the policy of installing solar panels at schools in Punjab was highly effective in increasing enrolment at schools. This means, average enrolment per school in Punjab increased by 48.5 students, compared to average enrolment per school elsewhere in Pakistan, after the implementation of the policy. When standard errors are made robust in Column (2) to account for any bias that may be present due to heteroscedasticity, the results are still signi cant at 1 per cent level. Even when standard errors are clustered on the province level in Column (3) to allow for correlation in observations within each province, the results are signi cant at 1 per cent level. This provides conclusive that the policy regarding power supply to schools in Punjab through solar energy was highly effective as it led to an increase in average enrolment per school.
We now provide results for the effectiveness of the policy on enrolment of boys' schools speci cally. Table 5 provides results for the OLS estimations with enrolment as dependent variable and normal standard errors in Column (1), robust standard errors in Column (2) and clustered standard errors at province level in Column (3). Once again, Province*Year is the independent variable of interest, whereas control variables include level of teaching at school, type of school, medium of instruction at school, total appointed teachers at school and binary variables for whether a school has electricity, toilet, boundary wall, drinking water, books in library, playground and science laboratory. Sample sizes are 5,078 schools for Columns (1) till (3).  (1), robust standard errors in Column (2) and clustered standard errors at province level in Column (3).
Province*Year is the independent variable of interest which is de ned as the interaction between province and year. Control variables include level of teaching at school, type of school, medium of instruction at school, total appointed teachers at school and binary variables for whether a school has electricity, toilet, boundary wall, drinking water, books in library, playground and science laboratory. Sample sizes are 5,078 for Columns (1) till (3).
The positive coe cient of Province*Year in Column (1) in Table 5 is signi cant at 5 per cent level, suggesting strong evidence that the policy of installing solar panels at schools in Punjab was highly effective in increasing enrolment at boys' schools. Average enrolment per school for boys' school in Punjab increased by 24.1 students, compared to average enrolment per school for boys' school elsewhere in Pakistan, after the implementation of the policy. When standard errors are made robust in Column (2), and clustered on the province level in Column (3), the results are signi cant at 5 per cent level. This means that policy regarding power supply to schools in Punjab through solar energy was successful in increasing average enrolment preschool for boys' schools.
We now provide results for the effectiveness of the policy on enrolment of girls' schools speci cally. Table 5 provides results for the OLS estimations with enrolment as dependent variable and normal standard errors in Column (1), robust standard errors in Column (2) and clustered standard errors at province level in Column (3). Once again, Province*Year is the independent variable of interest, whereas control variables include level of teaching at school, type of school, medium of instruction at school, total appointed teachers at school and binary variables for whether a school has electricity, toilet, boundary wall, drinking water, books in library, playground and science laboratory. Sample sizes are 5,764 schools for Columns (1) till (3). *** Signi cant at 1% level, ** signi cant at 5% level, * signi cant at 10% level.
The positive coe cient of Province*Year in Column (1) in Table 6 is signi cant at 1 per cent level, suggesting strong evidence that the policy of installing solar panels at schools in Punjab was highly effective in increasing enrolment at girls' schools. Average enrolment per school for girls' school in Punjab increased by 47.9 students, compared to average enrolment per school for girls' school elsewhere in Pakistan, after the implementation of the policy. When standard errors are made robust in Column (2), and clustered on the province level in Column (3), the results are signi cant at 1 per cent level. This means that policy regarding power supply to schools in Punjab through solar energy was also successful in increasing average enrolment preschool for girls' schools.
Finally, we now provide results for the effectiveness of the policy on enrolment of mixed schools speci cally.  Notes: OLS estimations with enrolment as dependent variable with normal standard errors in Column (1), robust standard errors in Column (2) and clustered standard errors at province level in Column (3).
Province*Year is the independent variable of interest which is de ned as the interaction between province and year. Control variables include level of teaching at school, type of school, medium of instruction at school, total appointed teachers at school and binary variables for whether a school has electricity, toilet, boundary wall, drinking water, books in library, playground and science laboratory. Sample sizes are 3,042 for Columns (1) till (3).
The positive coe cient of Province*Year in Column (1) in Table 7 is signi cant at 1 per cent level, suggesting strong evidence that the policy of installing solar panels at schools in Punjab was highly effective in increasing enrolment at mixed schools. Average enrolment per school for mixed school in Punjab increased by 54.1 students, compared to average enrolment per school for mixed school elsewhere in Pakistan, after the implementation of the policy. When standard errors are made robust in Column (2), and clustered on the province level in Column (3), the results are signi cant at 1 per cent level. This means that policy regarding power supply to schools in Punjab through solar energy was also successful in increasing average enrolment preschool for mixed schools.

Conclusion
Climate change is costing the economy billions of dollars each year through business interruption, insured losses, property damage and disruptions in supply chain (Crawford et al., 2020). Going forward, as the impact of climate change worsens, such costs will increase signi cantly. Thus, the use of renewable energy, in particular solar energy, is crucial in mitigating the negative consequences of climate change.
The adversarial effects of climate change on living conditions, health outcomes and labor productivity are evident. So are the bene ts of solar energy in reducing green-house gas emissions, improving air quality, increasing electri cation of remote areas and assisting in rural development. However, there is very limited research on how solar energy can increase access to education, and improve educational outcomes. By reducing barriers to entry, and providing cost effective and environmental friendly supplyside alternates, solar energy could be the technological progress needed in overcoming the challenges in accomplishing the Sustainable Development Goals.
By using school level data between 2013-18 from Pakistan, this paper estimates the effectiveness of the policy regarding power supply to schools through solar energy on enrolment. Using a difference-indifference estimation and controlling for school-speci c and infrastructure-speci c characteristics, we show that average enrolment per school in Punjab increased by 48.5 students, compared to average enrolment per school elsewhere in Pakistan, after the policy of installing solar panels at schools in Punjab was announced. Furthermore, gender-wise breakup of results suggest that the policy resulted in increased enrolment for boys' schools, girls' school and mixed-gender schools. The results are consistent when standard errors are normal, robust, or clustered at the province level.
There are certain limitations to the study. Firstly, while this study controls for a variety of school and infrastructure related control variables, it does not account for household-level or student's individual-level information that may impact the coe cients of our estimations. Secondly, issues related to installation of solar panels and costs, direct and indirect, related to converting schools from conventional energy sources to solar energy are not addressed in this study. Thirdly, the mechanisms through which the policy regarding power supply to schools through solar energy effects enrolment, along with mechanisms responsible for the differential impact of the policy on enrolment of boys and girls are not tested in this study. However, this study paves way for future research to be potentially directed towards addressing the limitations of this study, building upon the results presented.
In summary, this study highlights the importance of policies related to solar energy, which is considered a key energy source of the future. Policymakers must acknowledge that, going forward, reliance on solar energy will be crucial in tackling challenges related to social and economic development. Solar energy is associated with improved standard of living of individuals, in particular, and communities, in general, by providing electricity to regions that are either remote or inaccessible, generating environmental friendly electricity with no carbon emissions, using underutilised land, causing less electricity loss and improving grid security. Even for schools, utilizing solar energy will be critical in lowering barriers to entry and improving educational outcomes. Advocating policies linked to greater dependence on solar energy would be imperative in accomplishing the objectives of the Sustainable Development Goals, which, otherwise, will remain more than just a distant reality.

Declaration
Ethics approval and consent to participate Not applicable

Consent for publication
Not applicable Availability of data and material The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.