(b). Village and household survey
The study focuses on three key livelihoods determinants: road access, forest proximity and caste. We used a stratified random sample frame to get a representative sampling of villages in districts with significant forest cover. We purposively selected the three main forested districts of the state (West Singhbhum, Ranchi, and Palamu districts), of a total of 24 districts. Using a forest cover map, road map and village map, we used GIS to classify all villages into one of three zones, according to their proximity to forest (high forest defined as 35% forest cover in the vicinity of the village) and road access (low access defined as greater than 10 km to an all-weather road). This process generated a list of all villages in the three districts, classified into: high forest + low access (HFLA); high forest + high access (HFHA), and; low forest + high access (LFHA). The fourth possible zone, low forest + low access, did not exist. We then randomly selected three villages from each zone in each district for a total sample of 27 (3 × 3 × 3) villages. As such, the overall data set is representative of the range of village situations within the more forested districts of Jharkhand, and is expected to have many similarities to forested districts in neighboring states.
Field work was conducted between April and October 2006. Survey tools included village-level focus group discussions using participatory rural appraisal techniques such as participatory mapping, cropping calendar discussions, and a checklist of key issues for discussion (Cornwall and Pratt, 2011 and Lynam et al., 2007). These were designed to learn about village context and conditions: socio-economic conditions, income and employment patterns, resource availability and use trends, recent developments, and forest management and use. The participatory mapping yielded a list and location of all households in the village. Households were numbered and randomly selected for the household survey. Forty-five household heads were interviewed in each village. If a selected household was unavailable or unwilling to be interviewed, a replacement household was selected randomly to meet the minimum target of 45 households per village.
The household survey collected data on household demographic characteristics, caste, assets, and detailed information about income by source; e.g., quantities produced/collected, quantities consumed or sold, and prices. The interviewers referred to the cropping calendar and used a checklist of key activities and products generated by the village survey to aid recall of income by source over the 12-month reference period. Income categories included: agriculture; livestock; forest products (products harvested from state forest land); agroforest products (products harvested from private land); forest business; trade, and salary (wage, salary, trade, and business other than forest-based); labor; and remittances.
The data were checked for completeness, uniformity of units and scale, and internal consistency by the survey team and again by the analysis team. Sixteen cases were excluded due to missing or inconsistent data, leaving a total sample of 1203 households.
Asset and income data were converted to rupee values using prices at the village level for products that are traded for cash. Even products that are primarily used as subsistence products are occasionally traded and respondents were able to provide price estimates. Fuelwood is an important subsistence product; it is mainly collected and used within the household, with little local cash trade. Respondents were asked during group discussions to estimate the value of fuelwood. An average of price of Rs. 2/kg, or US$0.046/kg (US$1 = Rs. 43.3) was used for the whole sample. While this price may overestimate the true market value, it is the best estimate available.
Households were subdivided into four income classes using multiples of the Indian rural poverty line (Rs. 3312/capita/year, or US$76.50): Q1 < Rs. 1656; 1656 < Q2 < 3312; 3312 < Q3 < 6624; Q4 > 6624. Note that Rs. 6624 is still less than 34% of the international poverty line of $1.25/capita/day.
(c). Analysis
Descriptive statistics were used to explore relationships between income, assets, source of income, caste, and location. We compared overall mean subsistence, cash and total income by zone, and then used a more detailed analysis by income category to address questions about different livelihood strategies relating to remoteness and forest proximity. We further disaggregated the data by income class to compare income sources in both absolute and relative terms. We then compared cash and subsistence income by income category by caste and by zone for evidence of caste-specific livelihood strategies. We used Anova Mean test and Kruskal–Wallis test to analyze the differences in mean or median across groups or classes. The Kruskal–Wallis test was used to accommodate the non-parametric distribution of households of different castes in the sample.
Regression analysis was done using both Ordinary Least Squares (OLS) and multi-level (household and village level) analyses to assess the relationships between relevant income categories (1. total income per capita; 2. cash income per capita. 3. forest income per capita) as dependent variables. Household-level explanatory variables were socio-economic characteristics (age, caste) and assets (agricultural land area, number of trees owned), values of total household assets and overall income portfolio. Village level variables were forest access class. Multilevel regression was applied using MLWin v. 2.02. The variance partition coefficient (VPC) indicates the percentage variance explained by adding the second (village) level in the Multilevel Regression Model.
4. Results and discussion
(a). Livelihood status and strategies by zone
The survey results show high levels of poverty in the study area, with 48% of households living below the Indian rural poverty line of Rs. 276 per person per month (India Watch, 2007), and all households below the commonly used international poverty line of US$1.25 per person per day; even the wealthiest households cannot be considered rich in absolute terms.
Income is distributed highly unequally in most villages. The village income Gini coefficients range from 0.28 to 0.74, with a mean of 0.50. Villages with the highest mean incomes tend to have the greatest inequality, as a few relatively wealthier households both raise the mean income and income inequality.
Surprisingly, there are no significant differences between zones in terms of cash and total income. Proportion of subsistence relative to total income is significantly lower in LFHA than in HFHA households (Figure 2).