Study area: The study area included three villages: Ran Bijoypur, Sundarghona, and Bagmara covering two unions, Shatgumbuj and Bagmara of Bagerhat Sadar Upazila (subdistrict), located between 22°35′ and 22°50′ N and between 89°38′ and 89°53′ E (Bangladesh Bureau of Statistics 1992). Agroecologically, the area is part of the Low Ganges River Floodplain. The soil types are silt loams and silty clay loams (Bangladesh Bureau of Statistics 2000). The study area has a hot summer and a mild winter. The mean temperatures in the hottest and coolest months are 35.6°C and 14.6°C, respectively. The rainy season commences in the middle of June and continues up to the end of September. The level of humidity rises above 95% in the monsoon season and the average annual rainfall is 1312mm (Bangladesh Bureau of Statistics 2003a). Agriculture is the main economic activity followed by business and nonagricultural labor. The main agricultural crops include local and high-yield variety (HYV) paddy, wheat, pulses, mustard, potato, vegetables, spices, sugarcane, and other minor crops. A wide variety of trees, shrubs, and clumps of bamboos form the groves that surround the village homesteads (Bangladesh Bureau of Statistics 2003a). The total number of households and population in the study villages are 1756 and 8992, respectively. The literacy rate is 49.9%. The unemployment rate is 25.0% and 36.1% of working people are engaged in household work. Nonfarming activities include manufacturing, wholesale and retail trade, restaurants and hotels, and other community and social services. Sawmilling, small-scale woodworking, basket making, coconut oil, and betel nut processing are the main enterprises that use homestead-grown raw materials (Bangladesh Bureau of Statistics 1992).
Sampling method and data collection Bagerhat Sadar Upazila was purposefully selected to represent the Low Ganges River Floodplain in the southwest region of rural Bangladesh. Selection of the Upazila was based on typical rural location, accessibility of the study area for data collection, and temporal and financial constraints. Based on data published in the Bangladesh Bureau of Statistics (2003a), households were classified into four distinct groups according to landholding size: marginal (<0.4ha), small (0.41–1.01ha), medium (1.02–3.03ha), and large (>3.03ha). A two-stage sampling approach was adopted, in which three villages were randomly selected in the first stage. In the second stage, 30 households were randomly selected from each village given proportionate weight to landholding size classes. Thus, in total 42 households from the marginal landholding class, 27 households from the small landholding class, 15 from the medium landholding class, and 6 from the large landholding class were selected, providing a sample size of 90 households. Data collection was carried out in two stages. Because little is known about homestead forestry of the study area, exploratory information was collected regarding various aspects of homestead forestry through group discussion with farmers and interviewing key informants. In this exploratory stage, three group discussions, one in each village, and nine informal interviews with key informants were made using a semistructured questionnaire. In the survey stage, collection of detailed household information and measurement of tree parameters for woody plants with girth at breast height (gbh) of 10cm or greater were made in the sample households. For socioeconomic information, interviews were made with the head of the household (irrespective of sex) using a structured questionnaire developed with feedback from the first-stage discussions. It was assumed that in rural Bangladesh, the head of the household usually makes the economic decisions; however, during interviews respondents were encouraged to consult family members if necessary. To collect the data, a four-member group was formed with the help of local forest department officials. Local unemployed graduates, three men and one woman, who were familiar with the villages, were recruited to form the group. A field training session was organized for the group to explain and demonstrate the various aspects of data collection work. The first, fourth, and fifth authors participated and supervised the data collection work in the field. Also, local forest department staff provided substantial help during data collection. Data were collected from February through to May 2004. One-way analysis of variance (ANOVA) was used to compare the means of various variables across the landholding size classes, after transforming the data by natural logarithm. Stepwise regression analysis and Mann-Whitney U-Test were employed to examine the statistical significance of relationships between different variables.