It is worth noting that the dose-response function estimated by Pope III et al. is non-linear and it has support in exposures of from 0.18 to 0.90 mg/day and then above 18 mg/day so we need to rely on the linear interpolation of the values up to 18mg/day. This caveat does not seem a strong weakness since the linearization is highly accurate in this neighborhood, with R-squared values of 0.99 , 0.96 , 0.86 , and 0.80 . The changes in exposure are associated with a decrease in relative risk of lung cancer from 4.0 to 3.1 for the female head, nearly a 25% reduction. The relative risk for the male head falls by 33%, while the reduction for female child is 25% and for the male child is almost 30%. The estimated reductions in the relative risk of ischemic heart disease, cardiovascular disease, and cardiopulmonary disease are between 3 to 4%. Consistent with the results in lung cancer, these changes are higher for adult males, but the differences at these levels of exposure are relatively small. As we have argued earlier, access to electricity should unleash a series of changes. In fact, some recent evidence suggests that electricity may increase female labor supply or improvements in educational outcomes, consumption, and income . However, not all studies find impacts beyond lighting . Besides these mixed results, there is almost no evidence on the mechanisms that drive the changes pointed above. Some changes, like improvements in indoor air quality, are expected to be present in most settings, but in most cases impact will depend heavily on household and context characteristics. This is because before electricity can impact income or expenditure, households need resources to invest in new tools and complementary inputs, knowledge on how to operate these technologies, and demand for the goods and services produced with this new method. Constraints in access to credit or inputs, insufficient demand, or lack of know-how can prevent these changes from being reflected in economic outcomes like labor supply,cannabis drying rack income or expenditure. In addition, the effects on human capital could be negative if electrification increases child labor.

A major challenge in the literature of the effects of electrification is the identification of causal effects because electrification potentially unleashes a number of changes through a complex chain of causality. The identification of causal effects is further complicated because the changes interact with each other, sometimes increasing the exacerbating the effects and others attenuating them. Since the electric grid cannot be expanded randomly, recent studies use time variation and instrumental variables , mainly geographic variables, to deal with the endogeneity of connection. Studies have used land gradient , distance to hydroelectric dams , distance to the electricity line , and distance to power generating plants and baseline electrification rate in the locality . The first-stage relationship in this type of studies is usually clear: since land gradient affects the cost of grid expansion, it is correlated with the probability of grid connection. The exclusion restriction is usually more difficult to justify. Land gradient, for instance, plausibly affects the cost of building and maintaining other types of infrastructure, like roads, schools or hospitals, thus potentially affecting transportation costs and access to markets as well as education and health outcomes. Land gradient also may affect the crop varieties that can be grown in a region , thus influencing directly economic activity and income flows.2 Thus, the exclusion restriction requires the observed variation in land gradient to be in a range that does not affect other types of infrastructure, crops, or other economic activities; or, alternatively, it requires the variation in said variables generated by variation in land gradient to have no effect on the outcomes of interest. This may perhaps be not too far from reality in some settings, but it is ultimately an assumption that cannot be directly confronted with the data. Randomized Encouragement Designs offer an appealing alternative. This approach, originated by Imbens and Angrist , consists in randomly allocating incentives to connect to the grid, and using them as instruments in an IV estimation. It has been used extensively in other contexts , but Bernard and Torero is the first study to implement a RED approach to study electrification in developing countries. We implemented an approach similar to theirs. In our study setting, households were required to pay a $100 fee for a security inspection in order to get an electric connection.

We randomly allocated discount vouchers for 20% and 50% off the inspection fee, thus generating exogenousvariation in the connection cost. The problem with the RED approach is empirical: the complier sub-population is small , which is associated with low predictive power in the first stage, which in turn results in noisy IV estimates. To avoid relying on noisy estimates, we report are the “reduced form” effect of voucher allocation on the outcome of interest, also known as intent to treat estimates. ITT estimates are of inherent interest, since they correspond to what would be observed if our experiment were to be replicated. We rely on the “first stage” relationship between voucher allocation and grid connection to argue that vouchers influenced outcomes through their effect on electrification . Having investigated in chapter 2 the effect of electrification on indoor air quality and the implied health effects, we turn to investigate other mechanisms through which electrification can unleash the changes: time use, and electronic appliance ownership, and we show that through these channels, electricity can unleash effects on human capital and income. Next, we study the changes in time use by household members, uncovering increases on investment in education among children and increased non-farm work among adults. First, school-age children from voucher recipient households increase time studying at home, both in the intensive and extensive margin. Since this time is being spent in an environment with lower pollutant concentration, in a smokeless, better illuminated room, each minute of studying is more productive as well. Accordingly, we find evidence of improvement in a math skills index among voucher recipients. As a placebo test we show that there are no effects on years of schooling. We do not find systematic changes in time allocation among adult females, but adult males adjust their work activities, reducing time in independent farm work and increasing time in other work. Similarly, the probability of having engaged in non-agricultural independent work in the four weeks leading to the survey increased by 13 percentage points among voucher recipients. This change seems to be concentrated among 30-40 years old, although the standard errors are too wide to arrive to conclusions.

These change towards independent non-farm labor activities is arguably responsible for income gains: annual per capita income increased by $186 among voucher recipients . These income changes had some distributional effects, with voucher recipients being 10 percentage points less likely to have income below the median, but not more likely to reach the highest income quartile. Finally, we examine changes in appliance ownership. We find increases in ownership of television sets, stereos, refrigerators and blenders. Access to refrigeration haspotentially important effects on food storage, food safety, and nutrition. If materialized, this would reinforce the health gains from reduced indoor air pollution. The time savings generated by blenders and other kitchen appliances could have further welfare effects. Television and stereos primarily improve leisure, but television has also been shown to increase access to information, political participation,grow trays 4×4 and even contraceptive awareness. The remainder of this document is organized as follows. Section 2 describes the conceptual framework that guides our study. Section 3 presents the study setting and discusses the data. Section 4 presents the econometric approach. We discuss the main results in section 5, and section 6 concludes. Down the chain of causality, electricity may impact income. However, the channels through which this would happen are unclear. It may be that electrification opens the door to new, more profitable activities, but it is also possible that agents work longer hours in the same activities to afford new appliances. Access to electricity affects the marginal value of labor and leisure among adults. As we argued in the previous section, electricity facilitates access to television, increasing the marginal value of leisure. Banerjee and Duflo point out the critical importance of leisure time among in poor households’ utility, and the role that television plays in it. Television has other benefits, since it also increases awareness about news, politics, birth control, aspirations, etc. An increase in the marginal value of leisure should pull down labor supply. On the other hand, these new appliances need to be paid for. TV sets, refrigerators, and other appliances are not free to buy or to operate. Banerjee and Duflo argue that poor households leave profitable opportunities untouched because the extra income would not make a salient impact in their lives. In the face of electrification, household members may decide to work more in order to afford new appliances. However, electricity can also increase the marginal productivity of labor. Independent workers may now use power tools, farmers may use water pumps, shopkeepers may offer refrigerated goods. But all these changes require access to the tools, pumps, and refrigerators, which require the use of savings, access to credit or other sources of income. At baseline 17% of households had access to credit , but 70% reportedly believed to have access. By round 4, 87% believed to have access to credit, but only 20% had take up some.

This increase is far from sufficient to explain the increase in asset ownership we find below. If they can’t be paid for with credit, households would need transfers from family or to work more to increase income. The literature suggests that, after lighting, the first appliance bought by households are television sets . As discussed above, television increases the marginal value of leisure but also access to information. Refrigerators save time by reducing the frequency of trips to the market, but they also increase the variety of foods that can be consumed by the household and provide better food storage. Thus, access to refrigeration is another channel through which electrification can impact health. Appliances like blenders and microwaves probably have little impact on the final product, but can save time in household chores. For a given set of household chores, this is equivalent to relaxing the time constraint, which is unequivocally welfare increasing. The additional time may be allocated to leisure or work, for instance sewing, or cooking for sale. It may also replace the male in some farm tasks, while the male works in other things. Additional work, at least in principle, should lead to higher income which should lead to higher welfare. Electricity also facilitates access to cellphones, since they can be now charged at home, more cheaply. Cellphones have been shown to increase access to information and generate income gains . Other devices, like DVD players or stereos are primarily to increase the marginal utility of leisure, which can also enhance welfare.The improvements in indoor air pollution studied in chapter 1, for instance, are almost immediate, given that PM2.5 concentration will decrease a few minutes after putting out a kerosene lamp. In addition, the effect should be permanent. As long as other PM2.5-generating activities are not altered by electrification and households do not revert to traditional lighting, overnight PM2.5 concentration will remain at the new low level. Both conditions seem to hold in this case. In chapter 1 we saw that cooking practices – the main source of PM2.5 – are unaffected by electrification. Reversion to traditional lighting seems highly unlikely, since at least in the first four years of our study we don’t see households dropping their electric connections. Other effects take time, for a variety of reasons. Before an individual starts new income generating activities some conditions must be met. First, they need to realize there is a new opportunity, which is not always obvious. They also need to overcome any potential liquidity or credit constraint in order to gain access to human and physical capital, secure access to a steady stream of inputs, and find a demand that is not only steady, but also large enough to make the new enterprise profitable, net of all costs and risks.