Household carbon dioxide emissions from peasants and herdsmen in northwestern arid-alpine regions, China

This study assessed household CO2 emissions (related to the consumption of necessary and luxury goods and services) of peasants and herdsmen households in arid-alpine regions in Gansu, Qinghai and Ningxia provinces, China. We also explored whether agriculture types, family income and family size have played any role in household CO2 emissions. In order to address these issues, we: (i) developed assessment indicators for household emissions; (ii) conducted semi-structured questionnaire household surveys; and (iii) employed input-output analysis (IOA). The results showed that, the average household CO2 emission per capita is 1.43 tons (t) CO2; the proportion of subsistence emissions (related to the consumption of necessary goods and services) accounts for 93.24%, whereas luxury emissions (generated due to consumption of specific goods and services that are consumed only when household income improves) only account for 6.76%t. Moreover, household CO2 emissions increase with family income and family size, but per capita emissions are inversely related to family size. The highest average household emissions were found in the alpine agricultural and pastoral region (6.18 t CO2), followed by the irrigated agricultural region (6.07 t CO2) and the rain-fed agricultural region (5.34 t CO2). In consideration of insignificant amount of household emissions from these poor and vulnerable groups of the society, this study suggests to follow the principle of fairness while making energy conservation, emission reduction and adaptation policies. & 2013 Elsevier Ltd. All rights reserved.


Introduction
Climate change adaptation and greenhouse gas (GHG) mitigation measures are inevitable, but they create additional costs for society. These measures are critical to the human right to survival and development (Ding et al., 2009;Gao, 2006;Ge et al., 2010), especially, for poor and vulnerable groups in society. On the one hand, these groups could not afford climate change adaptation costs due to poor socio-economic conditions (Robert et al., 2006;Stern, 2006). On the other hand, living costs will increase and living standards will decline due to climate change mitigation and adaptation actions (Jacoby and Reiner, 1997). Although newly released data show historic declines in poverty worldwide, there are still 1.29 billion people living below 1.25 USD per day (The World Bank, 2012). Therefore, it is crucial to know GHG emission levels of these poor and vulnerable peoples for climate change negotiations and to make fair and rational emission reduction and energy conservation policies and plans.
Currently, the vulnerability of the poor and sensitive areas to climate change is being reviewed under an international climate policy framework (Niklas et al., 2003;Parry et al., 2009;Weitzman, 2009). However, GHG emission assessment mainly focuses on the analyses of inter-national and intra-national (regional) emissions based on macroeconomic data. Results from these data only reflect inequality of GHG emissions among different nations or regions. For example, per capita emissions are often used as an indicator to compare emissions amongst countries (Heil and Wodon,1997;Qu et al., 2010;Zhang, 2008). However, this indicator, based on macroeconomic data, cannot accurately reflect the emissions of poor and vulnerable groups whose emission levels may be much lower than the national and regional averages .
The family is the basic unit of society and consumer of social programs. However, families have differences in lifestyle, size, income, and consumption of goods and services. Therefore, the study of household level CO 2 emissions is able to better reflect inequalities than per capita emissions. There are some studies conducted along this line, beginning with Vringer and Blok 0 s research on direct and indirect household energy demand (1995). After their study, some scholars have applied inputoutput analysis to measure household emissions (Biesiot and Noorman, 1999;Bin and Dowlatabadi, 2005;Lenzen, 1998;Munksgaard et al., 2000). But these papers are based on macroconsumption data for statistical analysis and have the same shortfalls as described above.
This article examines inequality in per capita CO 2 emissions from direct and indirect energy consumption at the household scale by calculating emissions for the whole family based on data from field surveys conducted in arid-alpine regions in northwest China and communities typical of less-developed groups in western China. Economic inequality within China has been and will be in existence for a long time, which is the main cause of inequality in income, expenditures and CO 2 emissions per capita in China (Golley et al., 2008;Weber and Matthews, 2008). Household inequality between the different developmental regions is obvious (Clarke-Sather et al., 2011). Therefore, it is particularly necessary to carry out an investigation on household level CO 2 emissions to make an appropriate climate policy and to allocate the right to emit carbon to poor and vulnerable people.
In order to address these issues: we (i) developed assessment indicators for household CO 2 emissions; (ii) conducted questionnaire-based household surveys; and (iii) employed the IPCC 0 s reference approach and input-output analysis (IOA). Household CO 2 emissions, which are related to the consumption of necessary and luxury goods and services, is further divided into two categories: (i) subsistence household CO 2 emissions which refer to the necessary emissions from the fundamental consumption of the household, including the CO 2 emissions due to use of coal, LPG, gasoline and diesel oil, and the CO 2 emissions from the production, transportation and service processes of goods and services, such as electricity, food, clothing, medicine and medical care; and (ii) luxury household CO 2 emissions, which are generated due to using specific goods and services that are used when household income improves (to some extent) and families can afford these goods and services, such as CO 2 emissions due to the consumption of education, recreation, transportation and communication services. Although consumption of these services looks basic for many developed communities, the luxury emissions in our research are defined as all emissions excluding basic emissions for subsistence. This paper proceeds as follows: Section 2 discusses the concept of household carbon dioxide emissions. Sections 3 and 4 address the methodology and the results and discussion, respectively. Finally, Section 5 presents some concluding remarks.

The concept of household carbon dioxide emissions
Household CO 2 emissions (HHEs) are defined as the emissions of individuals or their families in order to meet the demands of their existence and development under certain socio-economic conditions, which includes both direct and indirect emissions. Direct emissions come from energy use including fossil fuels and for this research electric power. If CO 2 emissions occur before or after a product or service is in use, they are called indirect CO 2 emissions, in this research indirect emissions mainly come from household expenditures on products and services. Indirect emissions are related to the energy consumption and associated emissions embodied in consumer goods and services such as food, clothing, medicine, education, transport and communication. From an economic perspective, as noted above, the HHEs are divided into subsistence household CO 2 emissions (SEs) and luxury household CO 2 emissions (LEs) (Wang et al., 2010). In this study, we conducted intensive field surveys and assessed HHEs, SEs, and LEs for 2008-2009. The HHEs are used to reflect shared emission levels for the limited emission space and the fairness of allocating levels of interpersonal carbon emissions for individuals in different regions. Since the family and its members are the basic units of human society, HHEs are regarded as the assessment units of individual emissions. This paper adds to current emissions studies by identifying the amounts of CO 2 emissions that are absolutely necessary for poor and vulnerable groups in society.

A brief snapshot of the study area and data collection
This paper purposively selected rural households in Gansu, Qinghai and Ningxia Provinces, located in northwest China, as a study area. This study area is characterized by rich landforms (the transition zone of the Mongolian Plateau, Loess Plateau and Tibetan Plateau), fragile ecosystems, less developed economies and high vulnerability to the impacts of climate change (Fig. 1).
The annual average minimum temperature is generally lower than 10 1C. The altitudes of the household samples are from 1300 to 3400 m. So they usually choose coal for heating. The per capita consumption expenditures of rural household in this area are about 3000 RMB per year, and most of consumption expenditures are for food, education and medicine. In order to recruit households to participate in the survey we initially gave numbers to each household in the study area. Then we used simple random sampling techniques to select the samples. We selected 125 rural households in 49 villages of 32 townships distributed in 22 counties. Among them, even with repeated attempts we could not find people in two selected households. Therefore, our sample size is limited to 123, with a sampling ratio of 1/9425. This sample size looks small but given the homogeneity of the lifestyle and living standards in the study area, these response rates are enough for a benchmark study.
We have collected four types of data for this research: (i) household expenditure data (including consumption of food, clothing, medical care, education, entertainment, transportation and communication); (ii) energy consumption data (amount of electricity, coal, LPG, gasoline and diesel oil used); (iii) income, family size and education level data; (iv) geographical location data (altitude and coordinates). All of these four types of datasets were collected from field surveys in the summers of 2008 and 2009. The fifth type of data ''CO 2 emission factors for different goods and services'' was taken from China 0 s input-output table and the China Energy Statistical Yearbook published by the National Bureau of Statistics of China (2008). On the basis of these data, we estimated the HHEs for the surveyed households and also explored whether altitude, family income and family size have influenced CO 2 emissions levels from these households.
Here a pertinent question may arise whether the collected date are of good quality. In the study area, people generally: buy coals every year; pay electricity bills every month; do transport occasionally; and receive Medicare service when badly needed. They kept most of the receipts. If the data were unexpected, we tried our best to verify them from receipts and or asking some indirect and explanatory questions. For example, how many members of the family were sick, where you got treatment, how many times you traveled and where etc. Therefore, the quality of data is reasonably good.

Household CO 2 emissions
Calculating total HHEs was conducted in two steps. First, direct CO 2 emissions from household energy use were analyzed using the IPCC 0 s reference approach. Second, indirect CO 2 emissions were estimated using the input-output method. Thus, the total HHEs were estimated using Eq. (1): where E i (tons (t) CO 2 ) is the total HHEs from the ith household, E fi (t CO 2 ) is the direct CO 2 emissions from the combustion of fossil fuels in the ith household, E ei (t CO 2 ) is the CO 2 emissions from the electricity consumption in the ith household, and E gi (t CO 2 ) is the indirect CO 2 emissions from the goods/services consumption in the ith household.

Direct CO 2 emissions
Direct household CO 2 emissions are related to the combustion of fossil fuels including coal, gasoline, diesel oil, LPG, and the consumption of electricity in the household (Wang et al., 2010;Liu et al., 2011;Munksgaard et al., 2000). Though biomass is also used as a fuel source in the study area, it has not been considered in this study. It is difficult to compare with other research that does not analyze biomass carbon emissions, since biomass is part of a natural cycle that will not release additional emissions. Direct CO 2 emission estimates for each household were obtained using the IPCC 0 s reference approach as in Eq. (2) (IPCC, 2006).
where n is the number of fuel types, F fi (t) is the fuel consumption of the fth fuel in the ith household, e f (TJ/Gg) is the Net Calorific Value (NCV) of the fth fuel, p f (Kg C/GJ) is the Carbon Emission Factor (CEF) of the fth fuel, O f is the fraction of carbon oxidized (COF) for the fth fuel, and 44/12 is the ratio of molecular weight of CO 2 C. The coefficients of each fuel type are presented in Table 1. Eq.
(3) showed the estimate approach for the emissions(t CO 2 ) from the electricity consumption: where P ei (kWh) is the electricity consumption in the ith household, C e (t CO 2 /MWh) is the CO 2 emission factor of the electricity sector in the study area. In this paper, we use 1.0246 t CO 2 /MWh as the emission factor which is calculated from the weighted average of the emission factors in 2005-2007 in Northwestern grid of China 1 .

Indirect CO 2 emission
Usually, indirect CO 2 emissions from household consumption are measured by input-output analysis (Wang et al., 2010;Liu et al., 2011;Vringer and Blok, 1995) and the consumer lifestyle approach (Bin and Dowlatabadi, 2005;Wei et al., 2007). Due to the lack of availability of certain goods 0 life cycle data in China, we applied the basic input-output method to calculate indirect household CO 2 emissions as suggested in China 0 s input-output table and the China Energy Statistical Yearbook published by the National Bureau of Statistics of China (2008). Then the household CO 2 emissions factors were multiplied by household consumption to evaluate the indirect CO 2 emissions (Eq. (4)): where n is the number of household consumption sectors, S gi (RMB) is the gth household goods/services consumption in the ith household, C g (kg CO 2 /RMB) is the CO 2 emissions from the consumption of goods/service of gth household. The CO 2 emissions factors for various goods and services are given in Table 2.

Total household CO 2 emissions
In this study, all the household emissions related to the use of: electricity; coal; gasoline and diesel oil; LPG; food; clothing; medicine and medical care services; education, cultural and recreation services; and transportation and communication services were estimated. Emissions from these uses were considered to estimate the total HHEs. Table 3 shows the average HHEs in the whole study area and in three geographical regions. The inequalities in CO 2 emissions at household levels between the regions are noticeable. The highest HHE (16.95 t CO 2 ) is more than 68 times higher than the lowest value (0.25 t CO 2 ). However, 85.37% of the HHEs concentrated within the range of 2.489.94 t CO 2 , and average HHEs in the study area were 5.99 t CO 2 , of which the direct emissions are 4.95 t CO 2 and indirect emissions are 1.04 t CO 2 , respectively. Overall, the direct and indirect emissions account for 82.64% and 17.36% of total emissions, respectively.
The total average emissions decrease from the highest values in the alpine agricultural and pastoral region (6.18 t CO 2 ) to irrigated agricultural region (6.07 t CO 2 ) and then to rain-fed agricultural region (5.34 t CO 2 ). The contribution of emissions from coal consumption is the highest for each region, accounting for 60.68%, 62.16%, 64.27% of the total emissions in the alpine agricultural and pastoral region, irrigated agricultural region and rain-fed agricultural region, respectively, followed by electricity emissions (9.74%, 8.15% and 11.05%). In addition, emissions from three consumption categories in the alpine agricultural and pastoral region, including electricity, clothing and transportation and communication services, are higher than that in the other two regions (Fig. 2). In the alpine agricultural and pastoral region, coal consumption contributes more to the total emissions due to low temperatures in winter and at night in the study area.

Per capita CO 2 emissions
Based on the total CO 2 emissions from household consumption and family size 2 , and assuming equity in household consumption levels for every family member in a household, we calculated per capita CO 2 emissions. In the study area it is about 1.43 t CO 2 , far lower than the macro data-based per capita emissions for China (5.24 t CO 2 ) (Boden et al., 2011). These values, for the study area and national level emissions, are far lower than the per capita emissions of America (17.97 t CO 2 ) and Australia (18.96 t CO 2 ) (Boden et al., 2011). In the study area, coal consumption accounts for the highest part of per capita emissions (0.90 t CO 2 or 62.71%), followed by gasoline and diesel oil (10.23%), electricity (9.67%), food consumption (5.29%), education, cultural and recreation services (4.91%), clothing (3.14%), medicine and medical care services (2.00%), transportation and communication services (1.91%) and LPG (0.13%) (Fig. 3).  Note: HHE referrers to household emissions. There is a discrepancy between the total household CO 2 emissions and the sums of direct CO 2 emissions and indirect CO 2 emissions due to rounding error.
(footnote continued) org/wordpress/wp-content/uploads/2011/02/rule_consistency_of_grid_emission_ factors_published_by_CDM_host_country_authorities_14_Feb_2011-.pdf According to a study conducted by Ironmonger et al. (1995), economies of scale in energy use and household expenditures exist among adult-only households. This means, as the number of adults in an adult-only household increases, the per capita energy use and thus CO 2 emissions decrease. In this study, we found that the larger the family size, the lower the per capita CO 2 emissions. Our result is in line with the conclusions from other studies such as Weber and Matthews (2008). Therefore, it is not necessary to have adult-only households to gain economies of scale. Even a mixed-age family could achieve economies of scale if they lived together. This finding fits well with the traditional Chinese culture approach to extended family in which four generations live together (see Fig. 4).
As noted, there is a reasonable homogeneity of the lifestyle and living standards of the people in the study site. Therefore, the range of per capita emissions for each household sizes are not that large. However, in case of family size two (2 people per family) the range of per capita emissions seems to be very high (Fig. 4). There are two reasons: (1) the highest emitting household of this category has used a lot of coals than the other households; and (2) in this particular year, he used a lot of medicines and Medicare services, as both of them were sick.
A plot showing how household income per capita and family size are related is given in Fig. 5. There is a negative relationship between household income per capita and family size (r ¼ À0.315). This means when family size increases, household income per capita decreases and so do the emissions per capita. This is because income per capita and emissions per capita are positively correlated.

Subsistence and luxury CO 2 emissions
As noted above, household CO 2 emissions (HHEs) include subsistence household CO 2 emissions (SEs) and luxury household CO 2 emissions (LEs). The proportion of SEs (which is impossible to avoid for subsistence) reflects the quality of life for household residents. Table 4 shows the SEs and LEs in the study area. The average value of SEs in the whole study area is 5.58 t CO 2 per household, which is about 93.24% of the total average HHE. This reflects that the expenditure and emissions from households in less developed regions in China are mainly occupied by the consumption of basic goods and services, and that their living standards are low as compared to other regions in the world. Some similar research has found that US households emit the highest amount of CO 2 with 48 t CO 2 per year (Jones and Kammen, 2011), followed by UK households with 33.2 t CO 2 per year (Gough et al., 2011). Although the CO 2 emissions per household are lower in the Netherlands, Norway and Sweden with 19, 13.6 and 12.2 t CO 2 per year respectively (Kerkhof et al., 2009), they are much higher than the poor and vulnerable groups studied.

Household CO 2 emissions per capita at different income levels
There is a direct relationship between income level and household CO 2 emissions. We divided households into ten different income levels, and then calculated per capita CO 2 emissions from households for different income levels ( Table 5).
The per capita CO 2 emissions and the emissions from coal and food consumption increase with the rise of per capita household income. The per capita emissions (1.85 t CO 2 /person) for the highest-income households (44000 RMB), is 2.18 times higher than those at the income level range of %1000 RMB, and 1.70 times higher than those at the income level range of 1000-1500 RMB. The higher CO 2 emissions of households at higher income levels mean that they use more goods and services to meet their increasing demands. Emissions per capita from households at the income level range of 2500-3000 RMB were a little higher possibly caused by the pursuit of improving their diet. Our findings are consistent with the findings of Weber and Matthews (2008), Golley et al. (2008) and Liu et al. (2011).  By investigating the economic variables of LEs per capita, we concluded that the LEs per capita were strongly correlated with household income per capita, with an R 2 value of 0.767 (Fig. 6). The households with the highest income level have LEs more than four times that of households within the lowest income level. Fig. 7 shows that the percentage of LEs increases with the household income on a per capita basis.

Conclusions
This paper makes a first step toward understanding the interhousehold CO 2 emissions in northwest China and in particular in arid-alpine regions based on data derived from field surveys. This paper presents annual HHEs, including direct and indirect emissions, and disaggregates the emissions and their characteristics by geographical locations, income level and household size.
First, based on the findings presented in this paper, we conclude that there is inequality in inter-household CO 2 emissions per capita with the average value of 1.43 t CO 2 . The majority of the HHEs per capita (0.90 t CO 2 ) come from the coal consumption. This is mainly because of high altitudes and the need for heating during the chilling winter. If we deduct the portion of emissions resulting from climate related factors, energy use for   heating which is unavoidable, their household level emissions are negligible. Second, the SEs (related to the consumption of necessary goods and services) accounts for 93.24% of the total emissions, whereas LEs (related to consumption of luxury goods and services) accounts for 6.76%. This means, most of the households in the study area are consuming only those goods and services that are absolutely necessary for their subsistence and thus the reduction of emissions below these levels is almost impossible. Therefore, the low-income family will face more difficulties in the international/national environment of carbon emission reductions resulting in higher energy prices. These poor and vulnerable groups should be given special consideration when a new climate change mitigation policy is designed.
Third, the HHEs increase with household income and family size. However, per capita emissions decrease as the household size increases, indicating that households where large families and specifically extended family living together present a promising way for saving energy and reducing CO 2 emissions. Moreover, when family size increases household income per capita decreases and so do the emissions per capita.
Forth, per capita incomes of the people in the study area are rising. Therefore, their demands for the means of livelihood are expected to grow rapidly. We believe that the electricity and LPG will gradually replace the coal, and the petrol consumption for private travel will grow rapidly. Moreover, more money will be paid for cultural, and recreational activities, and educational and communication services. Therefore, the household CO 2 emission of the people in the study area will keep increasing. The SEs and LEs will both increase, and generally, the LEs will increase more rapidly.