Domestic use of dirty energy and its effects on human health: empirical evidence from Bhutan

ABSTRACT

Use of dirty fuels such as fuelwood, charcoal, cow dung and kerosene is common in developing countries, which adversely affects the health of people living in the dwellings, especially children and women. Using the data from a comprehensive and nationally representative Bhutan Living Standard Survey 2012, the present study examines the effects of dirty fuels on human health and household health expenditure. The result from propensity score-matching approach indicate that households using dirty fuels have a higher incidence of respiratory disease by 2.5–3% compared to households using cleaner fuels. The chances of household contracting tuberculosis are higher for households using dirty fuel in the range of 5–6%. It is also observed that the incidence of eye diseases and health expenditures among households using dirty fuels is higher. Hence the policy should focus on providing access to clean sources of energy to wider population.

KEYWORDS:

• Bhutan
• propensity score matching
• dirty fuel
• health

JEL CLASSIFICATION:

• C21
• O13
• O53
• Q42

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Matching is unambiguously preferred to standard regression methods for two reasons. First, matching estimators highlight the problem of common support, since treatment effects can only be estimated within the common support. Where there is poor overlap in support between the treated and non-treated this raises questions about the robustness of traditional methods relying on functional forms to extrapolate outside the common support. Second, matching does not require functional form assumptions for the outcome equation (that is, it is non-parametric). Regression methods impose a form on relationships (usually linear) which may or may not be accurate and which PSM avoids: this is valuable since these functional form restrictions are usually justified neither by economic theory nor the data used (Dehejia and Wahba 1998; Smith and Todd 2005). If there is no common support for a substantial proportion of participants, the treatment effect is not being estimated for those individuals.

2. The most common matching method proposed in literature are nearest neighbor matching, kernel-based matching, radius matching, spline matching and mahalanobis metric matching.