کشش تقاضا برای سوخت انرژی و سیاست های زیست محیطی: شواهد بررسی نمونه هند

India has been running large-scale interventions in the energy sector over the last decades. Still, there is a dearth of reliable and readily available price and income elasticities of demand to base these on, especially for domestic use of traditional fuels. This study uses the linear approximate Almost Ideal Demand System (LA-AIDS) using micro data of more than 100,000 households sampled across India. The LA-AIDS model is expanded by specifying the intercept as a linear function of household characteristics. Marshallian and Hicksian price and expenditure elasticities of demand for four main fuels are estimated for both urban and rural areas by different income groups. These can be used to evaluate recent and current energy policies. The results can also be used for energy projections and carbon dioxide simulations given different growth rates for different segments of the Indian population.

The main objective of this paper is to estimate the income and price elasticities of household demand for different kinds of fuels in India. There are a number of motivations for this. Energy is an important necessity for any household. In India the households need to choose not only how much but also which fuel to use. These decisions can have important consequences for the household budget, time allocation and health. They can also lead to negative environmental externalities at local, regional or global level. Price and income elasticities of demand are important for the choice of domestic energy policies. They are also useful in the context of energy policies for greenhouse gas abatement. Given the policy importance of these elasticities, it is striking that there is such a dearth of reliable and readily available estimates.2 Of course, several studies have examined the elasticities of commercial fuels like electricity and LPG in India. However, though some studies have examined the elasticities of fuelwood and also the substitution between fuelwood and commercial fuels, surprisingly few studies have done rigorous analysis. Earlier studies that attempted to analyse fuel demand in India ranged from large-scale macro planning exercises to local household case studies. There was interest in estimates of elasticities for different kinds of fuels as part of macro planning exercise, such as the Energy Survey of India Committee (1965), The Working Group on Energy Policy (1979), The Advisory Board on Energy (1985), The Energy Demand Screening Group (1986), The Rajadhyaksha committee of power sector planning (Gadgil et al., 1989) and the Planning Commission (1998). However, the main limitation of all the studies at macro level was that the projections that were made only took into account the aggregates such as population growth rate, increase in GDP, urbanisation and technological advancements. The fundamental problem with these studies is that although macro factors can influence energy consumption patterns indirectly, the actual determinants of household energy consumption are found at the household level. Aggregate fuel demand is made up by the day-to-day decisions at the household level. These decisions are affected by budget and time constraints of the household, their opportunity costs of time, the relative accessibility of fuels (relative prices) as well as social and cultural factors. Given such a perspective, it is obvious that it is e.g. not only GDP growth that matters but also its distribution. A second group of studies estimated the consumption of biofuels mostly for rural regions (e.g. Joshi et al., 1992). Although surveys of fuelwood consumption at the regional level are an improvement over macro level studies, as the fuel consumption mix is different for different agro-climatic zones, the estimates give only consumption per capita for rural areas. Some studies addressed the urban energy patterns and only some of these studies analysed the determinants of urban energy demand (Ray, 1980, Alam, 1985, Macauley, 1989, Dunkerley et al., 1990, ESMAP/UNDP, 1992 and ESMAP (Energy Sector Management Assistance Programme), 2001). Other studies have looked into various other aspects of urban fuel usage. Reddy and Reddy (1983) made a case study of fuelwood use in Bangalore, India. The studies by Dunkerley et al. (1990) and Bowonder and Unni (1988) did not estimate the demand for fuelwood or other fuels but looked at consumption and prices of fuelwood for Indian cities in the aggregate. Soussan et al. (1990) analysed in a comprehensive study the fuelwood combustion practices in an urban context. Turare (1998) used secondary data to analyse the criteria behind choice of domestic fuel. Alam et al. (1998) too is an investigation into the efficiency aspects of urban domestic fuel choices. Barnes et al. (2005) looked at aggregated energy demand in 46 cities in 13 different countries and is the most comprehensive study of urban fuel in the developing country context to date. A more recent study by Gupta and Köhlin (2006) analysed the preferences for domestic fuel for the Indian city of Kolkata. A third group of studies examined the consumption of fuelwood in different areas by controlling for income, size of households, landholdings, type of profession, agro-climatic zones, season, accessibility of forests etc. While some studies concentrated on the variation in consumption of fuelwood with different income and landholdings, others studied the consumption in different seasons. The studies are scattered across the country and it is very difficult to make meaningful projections for policy analysis. Some studies are based on more formal household models that have the potential to give the elasticities of interest for policy (see for instance, Amacher et al., 1993, Amacher et al., 1996, Pitt, 1985, Bluffstone, 1995, Köhlin and Amacher, 2006 and Heltberg et al., 2000). However, as such studies are very few in number, and only the latter two use data from India, extrapolations cannot be made for the entire country in order to make meaningful policy analyses. To get reasonably accurate fuel elasticities for a country as big and diverse as India, a lot of time and money need to be spent in obtaining information on fuel use and household characteristics, which is a colossal task. In India, the National Sample Survey Organisation (NSSO) collects information on quantity and expenditure on various commodities for a representative sample of the country. Expenditure and quantity of various fuels are among these commodities. One of the advantages of NSSO data is that even fuelwood collected for free is accounted for by imputing some value on it. In countries like India where majority of the rural people collect fuelwood for free, ignoring these values can result in biased assessments. This paper makes use of such a data set in order to estimate the price and income elasticities of fuelwood. Using a sample that truly reflects the whole population of India has made it possible to overcome some of the weaknesses of the previous approaches. The large sample does not only make it more representative, but it also facilitates disaggregation of the analysis to relevant sub-samples such as different income groups and for urban and rural areas separately. This gives us the opportunity to investigate energy transition in general, and the energy ladder hypothesis in particular, for the country with the highest domestic consumption of bio-energy in the world by estimation of expenditure elasticities of demand. We also analyse the own-price elasticities of different income groups and address the scope for energy substitution by estimation of cross-price elasticities of demand for various fuels. The estimations are made using the linear approximate Almost Ideal Demand System (LA-AIDS), proposed by Deaton and Muellbauer (1980), on household data for the year 1999. Instead of income we consider the total household expenditure as a proxy. The advantage of the LA-AIDS model is that the demand system is linear in the structural parameters. The LA-AIDS model has been widely used for analysing demand for various commodities in India as well as in other countries. In this study we use a two-stage budgeting process to obtain the elasticities of different categories of fuels. In the first stage it is assumed that the household decides how much to spend on fuel and non-fuel commodities and in the second stage they allocate expenditure to different categories of fuel. Such two-stage budgeting has been used earlier to analyse demand for meat (Ealas and Unnevehr, 1988 and Gao et al., 1996), fish (Cheng and Capps, 1993 and Dey, 2000), demand for nondurable commodities (Carpentier and Guyomard, 2001) etc. However no study has used such an approach to estimate fuel elasticities. This study is thus an empirical contribution to the domestic energy literature. The plan of this paper is as follows: Section 2 presents the two-stage budgeting model. Section 2 lays out the empirical specification. Section 3 discusses the issues in estimation of the model including the methodology to account for zero expenditure. Section 4 describes the data used in the study. Section 5 presents the empirical results and Section 6 concludes with the policy implications.

Energy is a necessity. Given its importance for household welfare, public investments and environmental considerations it is surprising that not more formal analyses have been carried out for developing countries to analyse income, own-price and cross-price elasticities of demand for a full set of domestic energy sources, including fuelwood. The proliferation of national sample surveys might help to address this gap. In this paper we have shown that the Indian National Sample Survey Organization data, that include quantities and values of self-collected fuelwood, can be used for this purpose. We have also shown that further insights can be sought by disaggregating the data into relevant sub-samples, in this case in urban and rural samples, which were sub-divided into expenditure classes. The results from this exercise can be used in a number of ways, depending on the policy objective in mind. For a country like India, with a tradition of implicit and explicit government interventions that affect the prices of domestic fuels, the impact of such interventions on demand can be analysed based on own- and cross-price elasticities of demand. Similarly, if the desired policy objective is transition towards clean fuels (like LPG and electricity) due to the health impacts or local and global pollution, then these elasticities can prove useful in identifying the most cost-efficient policy. Another area of application is simulation for energy planning. As was indicated in the introduction, there has been a number of energy planning exercises in India over the last decades. Such exercises could be made more realistic and accurate if they were based on the kind of analysis provided in this paper. Domestic energy demand in India is not only of domestic interest today, given the population of India, this has become a global concern. For example, global greenhouse gas emission models would benefit from such estimates. The same holds for simulations of GHG emissions given implementation of a global carbon tax. The analysis has already given rise to a wealth of empirical information. The probit analysis highlighted the rural–urban differences in adoption of modern fuels and indicated that scheduled caste settlements in rural areas might be deprived of electricity with potentially alarming health implications as a result. We also identified significant regional differences in adoption. The estimation of the full LA-AIDS model gave further evidence of these differences and reminded us also of the significance that resources such as forests have in shaping domestic energy demand. Finally, the expenditure elasticities informed us that dependence on fuelwood will continue for a long time and that when we simulate future demand we will need to be careful in considering not only population and income growth, but also the distribution of this growth. There are still a number of improvements that could be made to this approach. Cross-sectional sample surveys are notoriously difficult to use in order to estimate price elasticities, primarily due to the lack of variation in price and the potential confounding with quality effects (Deaton, 1990). For future fuel related estimations, panel data analysis is probably worth attempting combined with more flexible functional forms. Sample surveys combined with collection of fuel prices would of course be ideal. The estimation would probably also be greatly improved if it were possible to combine the household expenditure data with exogenous information regarding accessibility of different sources of fuels. A geographical information system would be useful to structure the physical information for such analysis. Such rich data sets could also be used for much more disaggregated analysis than what has been made here.