INTRODUCTION
The world trade in non-timber forest products (NTFPs) (for example forest fruits, medicines, aromatics and resins) is worth billions of dollars (Iqbal Reference Iqbal1993). Many have advocated that NTFPs and other forms of forest products be promoted to provide increased income opportunities for forest dwellers and users (Counsell & Rice Reference Counsell and Rice1992; Wollenberg & Ingles Reference Wollenberg and Ingles1998). However, the initial optimism that the twin goals of conservation and economic development could be linked through NTFP extraction seems to have diminished (Ruiz-Pérez & Arnold Reference Ruiz-Pérez and Arnold1996; Arnold & Ruiz-Pérez Reference Arnold and Ruiz-Pérez2001). There have been naive assumptions behind many marketing plans, and historical trends in NTFP use indicate that negative outcomes are common (Gray Reference Gray1990, Dove Reference Dove1993).
However, many millions of households continue to harvest forest products to enhance their livelihoods (Byron & Arnold Reference Byron and Arnold1999). Better understanding of why some households harvest forest goods while others do not may help explain some of the problems encountered in NTFP promotion, such as whether the poor or rich are more likely to benefit from commercialization schemes (Neumann & Hirsch Reference Neumann and Hirsch2000; Marshall et al. Reference Marshall, Newton and Schreckenberg2003; Belcher et al. Reference Belcher, Ruiz-Pérez and Achidiawan2005). Recent research has highlighted key socioeconomic characteristics of forest-dependent households that can play roles in explaining forest use. For example, in a study in the Philippines, elderly people were more likely to collect forest goods because of their more extensive knowledge of forest plants and wildlife (Lacuna-Richman Reference Lacuna-Richman2002). Elsewhere, younger households are more dependent on wild-collected products, as they set out to start families and have lower agricultural assets than older better-established households (Coomes et al. Reference Coomes, Barham and Takasaki2004; McSweeney Reference McSweeney2004).
Another key variable of interest is the relationship between income and forest use. Siebert and Belsky (Reference Siebert and Belsky1985) found the households with the lowest level of rice self-sufficiency relied most on rattan harvesting for income in the Philippines. Gunatilake et al. (Reference Gunatilake, Senaratne and Abeygunawardena1993) found that contributions of NTFPs to incomes in Sri Lanka declined as incomes rose. Similar arguments have been made elsewhere that the poor are more dependent on forest goods than better-off households (Cavendish Reference Cavendish2000; Hegde & Enters Reference Hegde and Enters2000; Mahapatra & Tewari Reference Mahapatra and Tewari2005), and the poor particularly rely on forest income in times of particular need (McSweeney Reference McSweeney2002, Reference McSweeney2004).
Other studies indicate that medium-income or richer households are, in some situations, more likely to have forest income than the poor, owing to high labour requirements or elite capture of valuable resources (Godoy et al. Reference Godoy, Brokaw and Wilkie1995; Wickramasinghe et al. Reference Wickramasinghe, Perez and Blockhus1996; Ambrose-Oji Reference Ambrose-Oji2003; de Merode et al. Reference de Merode, Homewood and Cowlishaw2004). Often the role of income class depends on what variable is measured. In a meta-analysis of 54 different studies, absolute forest income increased as total household income increased and ‘was thus important not only for poor communities’, but forest income as a share of total income decreased, indicating the poor were more dependent on forest income (Vedeld et al. Reference Vedeld, Angelsen, Sjaastad and Berg2004, p. xiv). This finding was echoed in Ethiopia, where wealthier households received more absolute cash income from forest produce, while poorer households were dependent on forests for a larger percentage of their income (Mamo et al. Reference Mamo, Sjaastad and Vedeld2007). When comparing forest use in South Africa, poor, average and rich households did not differ in terms of the number of NTFPs used or the proportion of households using them (Shackleton & Shackleton Reference Shackleton and Shackleton2006). However, the poorer households did use more NTFPs per person in terms of volume when both income and subsistence purposes were considered.
Other studies have noted the importance of land tenure. The landless and land-poor are often more dependent on forest product collection than the land-rich (Lacuna-Richman Reference Lacuna-Richman2002; Pandit & Thapa Reference Pandit and Thapa2003). For those who have no access to land for agriculture, NTFPs can provide a much needed source of support, especially when they are collected from common or open lands. In Orissa (India) dependence on forest income was strongly correlated with size of land holdings, with the landless being most dependent (Fernandes & Menon (Reference Fernandes and Menon1987).
Other social variables may also influence forest use. In one study, NTFP exploitation was positively correlated with household debt, labour availability and male to female ratios and negatively correlated with income, education, distance to forest, involvement in non-agricultural activities and incorporation into the market (Gunatilake Reference Gunatilake1998). Factors such as the size and labour capacity of households (Mamo et al. Reference Mamo, Sjaastad and Vedeld2007), migration status (Lacuna-Richman Reference Lacuna-Richman2006), opportunity costs of collection and the substitutions of forest products by market purchased goods (Senaratne et al. Reference Senaratne, Abeygunawardena and Jayatilake2003), and the strength of markets for forest produce (Ruiz-Pérez et al. Reference Ruiz-Pérez, Belcher, Achidiawan, Alexiades, Aubertin, Cabellero, Campbell, Clement, Cunningham, Fantini, de Forestra, Gracia Fernandez, Gautam, Martinez, de Jong, Kusters, Kutty, Lopez, Fu, Alfaro, Nair, Ndoye, Ocampo, Rai, Ricker, Schreckenberg, Shackleton, Shanley, Sunderland and Youn2004; Bista & Webb Reference Bista and Webb2006) may also be important. Previous studies having highlighted heterogeneity even within smaller forest-extracting communities (Coomes et al. Reference Coomes, Barham and Takasaki2004; Vedeld et al. Reference Vedeld, Angelsen, Sjaastad and Berg2004), more studies are needed to comprehensively account for use of forest products across a range of ecological locations and social situations.
In Vietnam, millions of rural people live in close proximity to forests, yet there has been little published on NTFP use, the research there predominantly focusing on ethnic minorities, who comprise around 13% of the national population and live in mountainous areas with higher rates of forest coverage (Wetterwald et al. Reference Wetterwald, Zingerli and Sorg2004; Dang Viet Quang & Tran Nam Anh Reference Quang and Anh2006; Hilfiker et al. Reference Hilfiker, Zingerli, Sorg and Lothi2006). There has been much less attention to forest use among ethnic Vietnamese and those in lowland areas. The present study attempts to remedy this through a case study of forest extraction by Vietnamese households living in lowland and midland areas of north central Vietnam.
The study aimed to build on experiences garnered from previous studies on NTFP use and environmental income, and attempted to follow the ‘best practices’ of Vedeld et al. (Reference Vedeld, Angelsen, Sjaastad and Berg2004). First, I examined all plant and animals extracted from forests, both NTFPs and wood products, to establish their relative importance so that different sources of forest income could be clearly compared. Second, I collected information on all other types of household income, both subsistence and in cash, in the study area, so that forest environmental income could be put in the context of overall household livelihoods. Third, I worked with a number of households with diverse socioeconomic backgrounds, including from all income classes and with households from both the land poor and land rich, thus accounting for factors often ignored in other studies which frequently focus only on poor or landless households (Bista & Webb Reference Bista and Webb2006). Fourth, I worked with migrants and local-born populations to see what effect the household's background and history played in their forest use decisions. The main objective of the project was to determine which of a number of socioeconomic factors had the strongest relationship with the use of forest produce and forest environmental income dependency.
METHODS
Research setting
The study was conducted in rural areas of Ha Tinh province, approximately 300 km south of the national capital Hanoi (Fig. 1). Ha Tinh had an estimated population of 1.286 million in 1999 and is among the poorest areas of Vietnam (Department of Planning and Investment Ha Tinh 2003). The province has an area of 6055 km2 divided into 11 districts and a further 259 communes, the lowest level of state administration in Vietnam. Below the commune most households group into villages, although these are not officially recognized as an administrative unit.
Map of the study area.
Ha Tinh is characterized by low coastal plains bordering the South China Sea, rising to high mountains in the Annamite chain separating Vietnam from Laos. Two major nature reserves, the Vu Quang Nature Reserve and the Ke Go Nature Reserve (KGNR), have been demarcated in the past 15 years to protect what are seen as high levels of biodiversity, particularly of mammals and birds (Eames Reference Eames1996; Le Trong Trai et al. Reference Le Trong Trai, Dung, Cu, Cham, Eames and Chicoine1999). The KGNR was established in 1996 with > 35 000 ha, primarily to protect populations of two endemic and endangered pheasant species. The Reserve was described at the time of founding as having one of the ‘largest remaining blocks of broadleaf evergreen forest in the level lowlands of central Vietnam’ (Le Trong Trai et al. Reference Le Trong Trai, Dung, Cu, Cham, Eames and Chicoine1999, p. vii). However, more than 75% of the forest has been classified as heavily disturbed due to past logging by state-owned timber companies, and only at higher elevations are areas of lightly disturbed forest found. The topography of the Reserve is mostly low, gently sloping hills, with altitudes of 50–500 m and it supports at least 46 mammal, 270 bird and 562 plant species (Le Trong Trai et al. Reference Le Trong Trai, Dung, Cu, Cham, Eames and Chicoine1999).
Officially, most protected areas in Vietnam consist of a strictly protected inner core in which almost all anthropogenic activities are banned (ICEM [International Centre for Environmental Management] 2003). Within national parks and nature reserves in particular, it is prohibited to ‘log, exploit (excluding activities related to forest cleaning and rehabilitation), hunt animals, collect specimens under any means and forms. . .Strict protection areas within national parks and nature preservation areas should be protected strictly. Every activity that causes negative impacts to forest is not allowed’ (MARD [Ministry of Agriculture and Rural Development] 1997). Any commune sharing a border with a protected area is considered to be a ‘buffer zone’ in state law, although in reality this creates no significant restrictions on land use (Gilmour & San Reference Gilmour and San1999). Many buffer zones, however, have been able to attract projects, such as integrated conservation and development projects (ICDP), in order to reduce dependence of residents on protected area resources; one such project was set up in Ke Go, funded through the IUCN (World Conservation Union), to encourage the planting of domesticated NTFPs like rattan and medicinal plants in home gardens.
When the KGNR was demarcated the boundaries were deliberately drawn to exclude human settlements, which fell into a buffer zone of 22 000 ha. While no households were living inside the actual boundaries of the Reserve at the time of the present study, approximately 40 000 people lived in the buffer zone, spread over eight communes in the district of Cam Xuyen, one of the poorer districts. While not all residents in these communes were involved in forest extraction activities, those in areas closer to the Reserve were often actively engaged. The most accessible areas within the KGNR used by nearby villages were dominated by secondary forest and scrub growth where timber, fuelwood and a variety of NTFPs could be harvested, and buffalo and cattle were occasionally grazed as well. The borders of Ke Go were only sporadically patrolled by small numbers of rangers; while in the law any exploitative extraction of goods from a nature reserve was illegal, rangers primarily focused interdiction efforts on timber, charcoal extraction, and hunting, while ignoring infractions of other NTFPs or fuelwood harvesting for the most part. The typical punishment for illegal logging was confiscation of timber and any equipment used to cut it, and a monetary fine up to 500 000 VND (c. US$ 34), while charcoal makers were subject to lower fines. In addition to these restrictions on extraction, national law also stated that no land tenure certificates would be granted to people farming inside the official boundaries of any protected area, and any encroachment of agriculture onto the Reserve would be stopped through fines and resettlement.
Study sample
Five villages in the KGNR buffer zone were chosen based on stratified random sampling to include those with good access and those with poorer access to the KGNR forest. The approximate distance households in each village had to walk to arrive at the natural forests that make up the KGNR were used to classify villages as either close (<3 km away) or far (>3 km away). Once households reached the KGNR border, it was usually an additional two or more kilometres walk into Reserve areas with sufficient forest cover to collect the most common products. The general topography and ecology was approximately the same for all villages; altitudes were <500 m and all villages were located on the east or north-east side of the KGNR. I visited the villages regularly from November 2000 until October 2001.
Data collection
I randomly selected one-fifth of the households on each village's annual census roll and interviewed them in Vietnamese with a standardized survey over the course of several hours up to several days. Households were defined by the official Vietnamese government classification of family members living and eating together (known as a ho in Vietnamese). In most cases, both husband (traditionally the household head) and wife were interviewed together to provide the most comprehensive recall on answers. Household level information collected included household size, income, migration status and history, educational levels and position in the village, along with major household assets, such as motorcycles, water pumps and tractors. Respondents were also asked about land tenure holdings and access, and local knowledge of forest species. They were asked to estimate both the quantity and cash income raised from all forest product extraction activities in the previous 12 months, along with questions about the seasons and labour needed. All answers were based on informant recall. Checklists of the main categories of forest goods and species used were compiled from group meetings and used as prompts to aid in memory recall (for example respondents were specifically asked ‘Did you collect any of species X in the past 12 months?’ for all known products, as well as being allowed to add any additional products collected).
A market survey was conducted at the main commune and district markets to assess prices of goods throughout the year and check against prices reported by respondents. Qualitative interviews were also held with key informants, and for each major type of forest product collected from the KGNR (palm leaves, medicinal plants, wildlife, timber, rattan and aromatics), focus groups were conducted with self-identified collectors. More than 300 plant voucher specimens were taken and deposited at the Institute of Biogeography's herbarium in Hanoi, and identified by a botanist specializing in central Vietnam.
Data analysis
A number of variables were used in the analysis. Income equivalents were derived for all products that were sold, through use of average market prices and by informant recall. Some products, however, were primarily used for subsistence purposes (such as fuelwood, collected by 76% of households but sold by only 35%). It is difficult to attach values to subsistence forest goods (Godoy & Lubowski Reference Godoy and Lubowski1992; Wollenberg Reference Wollenberg2000; Gram Reference Gram2001), but in this study, most products had a local market (only few edible plants and fruits did not), aiding construction of income equivalents for subsistence goods. However, subsistence equivalent values should be taken as approximations only, as recall on consumption was more difficult for informants than for cash incomes raised.
All forest products collected by households in the past 12 months were listed and identified by species, through voucher specimens for plants and by consultation of illustrated guidebooks for animals. The survey distinguished between forest goods collected in natural forests and income raised from plantation forests (such as pine resin tapping) or planted gardens (where some fuelwood was harvested). Domesticated or planted trees were included as agricultural or plantation income, and only income derived from natural forests was counted as ‘forest income’. Although the collection of many forest goods was technically illegal, households appeared quite open in discussing patterns of use, primarily because enforcement of the KGNR border was fairly low and there was no threat of punishment by speaking with the researcher. Furthermore, because this survey was combined with ethnographic presence in the villages over the course of several months, households were more open about forest use activities. Multiple visits to households and multiple visits to forest harvest sites with collectors were used to verify the data obtained in the survey alone.
Following Vedeld et al. (Reference Vedeld, Angelsen, Sjaastad and Berg2004), Sjaastad et al. (Reference Sjaastad, Angelsen, Vedeld and Bojo2005) and Vedeld et al. (Reference Vedeld, Angelsen, Bojo, Sjaastad and Berg2007), environmental income was defined as ‘value added’, in other words ‘gross benefit net the costs of capital consumption and intermediate inputs’ (Vedeld et al. Reference Vedeld, Angelsen, Sjaastad and Berg2004, p. 6). This definition does not require the subtraction of labour costs, as a definition of rent might. The use of value added income models is appropriate for this site in Vietnam in which labour markets were very underdeveloped; a day collecting forest goods was not a day taken away from other wage activities, but rather was an activity undertaken when there was a lack of other work opportunities, such as during the agricultural slack season. Furthermore, there was little to no capital outlay needed for any forest environmental activities in Ke Go; for example, logging was carried out with hand saws that the household often possessed anyway, not with capital intensive chainsaws as is common in other areas of Asia. Further, there was little to no processing of forest products beyond drying leaves or burning charcoal. Thus no processing costs for goods needed to be included in income figures, due to the low levels of value-adding to harvested products.
All other sources of household income were also calculated from informant recall of amounts of goods produced and income earned for a number of sectors. In agriculture, the main crop planted was rice, along with a number of non-irrigated crops such as corn, potatoes, sesame and beans. The total amount of rice produced in the previous year was assessed, which was multiplied by the average market price per kg of rice to reach a value of ‘subsistence rice production’. However, because only a minority of households (39%) actually sold any rice, the ‘cash rice income’ was also calculated. Cash income raised from the sale of all non-rice crops was also assessed; the majority of households raised beans, corn and sesame for sale, not subsistence, while cassava and potatoes were primarily grown for livestock. Livestock income was reported for cash earnings from the sales of meat, milk, eggs or young animals; however, subsistence income was difficult to calculate and therefore was underestimated for livestock. Any garden products sold such as vegetables, fruits and herbs were assessed as garden income, but as much garden produce was consumed and not marketed, and as subsistence figures were very difficult for households to recall exactly, garden income was likely also underreported here.
All other cash income earned by household members in various activities was enumerated, including unskilled wage labour employment (primarily in construction, road building and brick factories), as well as skilled labour, including providing services (such as renting ploughs, fixing motorbikes, etc), salaried employment, and government retirement or pensions. Government subsidy payments, business income and migrant remittances were all added to cash income to form the category ‘cash wages’.
Once total annual income was calculated for each household, income terciles were created, using both measures of only cash income and cash plus subsistence income. For comparisons of cash income, ‘low income households’ (n = 34) were those with total household income below 2.8 million VND (US$ 193) (US$ 1 = 14 500 VND in 2001), ‘average income households’ had incomes of 2.8–5.5 million VND (US$ 193–379) and ‘high income households’ were those with incomes > 5.6 million VND (>US$ 386). For total cash and subsistence income, ‘low income households’ were those with incomes <4.7 million VND (<US$ 324), ‘average income households’ had incomes of 4.8–7.61 million VND (US$ 324–525) and ‘high income households’ had incomes > 7.62 million VND (>US$ 525).
Landholdings were assessed using local land categories. Agricultural lands included rice fields, both irrigated and non-irrigated, and agricultural plots for field crops such as cassava and corn. Around the household compound, home gardens were usually found, and some households also had hill gardens located further away. Some households' land holdings included government-allocated forest land, which were plots of land with planted trees (mainly acacia, pine and eucalypts) or which were slated for reforestation, and which had been given to households for long-term protection rights. Landholdings were measured in sao, the local land measurement unit (1 sao = 500 m2 or 20 sao = 1 ha). Most households were clear on the number of sao they had land rights to as such information was listed on all household land tenure certificates, which had been recently issued.
Household expenses were calculated from informant recall, a standard tool in most household living standards surveys done in Vietnam and elsewhere (World Bank 2001). While this is an imperfect method, households often did keep receipts for many purchases, which could be checked and verified, and when exact amounts of money spent could not be verified, approximations based on total number of items purchased and current market prices were used. However, because of the difficulty of recall, these numbers should be taken as relative assessments that provide approximate, not exact, amounts of expenditures. Households were asked to estimate their expenses for the previous 12 months in a number of categories, including agriculture (buying seedlings, pesticides, fertilizers, irrigation, harvesting and transport), forestry, house maintenance (electricity, repair and upgrade, household goods), food costs, schooling costs, health costs, agricultural and land taxes and miscellaneous travel, ceremonial and other expenses. These expenses were compared with income figures to provide general estimates of household welfare.
Non-parametric t-tests (Mann-Whitney U tests) and non-parametric one-way analysis of variance (ANOVA) (Kruskal-Wallis tests) were used to interrogate the variations between households in their forest use and income to determine characteristics of forest-dependent and non-forest-dependent households. The non-parametric Spearman rank correlation was used to study associations among variables, and ordinary least squares (OLS) multiple regressions were used to build models of the household characteristics associated with forest income earnings, both in terms of absolute forest income and the relative share of overall household income.
RESULTS
Household characteristics
Household size was relatively even across the five study villages, with a mean size of 4.8 members (SD 1.5). Household heads on average had completed 6.6 years of schooling, and were 45 years old. Sixty-three per cent of the households identified themselves as ‘migrant households’, meaning the household head had been born other than at the current place of residence, with most having moved 200 km or less. The mean annual income in cash for households was 4 710 031 VND (US$ 325) while the absolute income including subsistence activities was 6 408 938 VND (US$ 442). Average landholding was 15.4 sao (0.77 ha), of which 62% was used for agriculture and the rest for residences and forestry.
Forest products collected
Eighty-eight per cent of households had harvested some sort of forest product from around their villages in the past year. Households on average collected 5.5 different wild species (SD 7.0). Of 10 major forest product categories, fodder was the most difficult to quantify, as several households indicated they let their animals graze in the Reserve for several months per year, but it was not possible to quantify this in terms comparable with other plant extraction activities. Thus, total income figures are underestimations, as they do not include fodder that was not cut and brought to the home (which could be measured and recalled).
In most cases, forest products were collected for both subsistence and commercial use, with the exception of charcoal, which was produced solely for commercial sale (Table 1). The most commonly collected forest product for subsistence was fuelwood, which 76% of households obtained from natural forests, other households sourcing it from private gardens or purchase. The most common commercial products were fuelwood and leaves, both sold by 35% of households. The most lucrative income-generating product collected from forests was charcoal, raising on average over 1 million VND yr−1 (US$ 69) and accounting for 25% of the income of charcoal-making households (18% of the sample). Overall, 57% of households surveyed raised at least some cash income from collecting forest products. On average, all non-wood products contributed 190 952 VND (US$ 13) in cash value per year to all surveyed households, while wood products (fuelwood, timber and charcoal) contributed 469 173 VND (US$ 32) in cash. If we exclude those households who reported no cash forest income at all, the mean annual non-wood forest product income was 336 593 VND (US$ 23) and the mean wood income was 827 016 VND (US$ 57). Adding subsistence goods, the values are even greater: the mean forest environmental income was 1 137 649 VND (US$ 78) from all sources averaged across the sample, and was US$ 90 for only those households reporting some forest use.
Average cash income generation from forest-based sources.
While absolute forest income may seem small, as a percentage of total income it was a significant contributor to households in Ke Go, who reported very low total incomes overall (Table 2). Forest environmental income was the second-highest in absolute terms among all household cash income sources, and was third among all income sources in number of people benefiting. On average, households received 20% of their total overall incomes from natural forest exploitation, and 18% of their cash incomes.
Sources of household cash income in Ke Go.
Factors influencing forest environmental income
Households that derived some cash income benefit from forests (hereafter ‘forest cash income (FCI) households’) differed from ‘non-forest cash income (non-FCI) households’ (Table 3). Non-FCI households had significantly more income from wage labour and livestock than FCI households, while the latter were younger and more likely to live nearer the forest. Agricultural production did not differ between the two groups, although landholdings of the FCI households were larger than non-FCI households (p = 0.042). Migration status did not differ between FCI and non-FCI households, nor did overall levels of income of the household.
Comparison of forest cash income (FCI) households with non-forest cash income (non-FCI) households. * = significant (p < 0.05), ** = highly significant (p < 0.01).
Household age
Young households (household head ≤ 30 years), middle-aged households (31–55 years) and households over 56 years old (commensurate with Vietnam's mandatory retirement age of 55) differed in the amount of rice produced (Kruskal-Wallis p = 0.001), the cash obtained from selling wood products (p =0.012), the income obtained from employment (p = 0.029) and in landholdings (p = 0.004). Young households depended more on forest extraction, while older households were more likely to have income from wages, salaries or pensions, and less income from forests, as well as smaller landholdings.
Migration status
Migrants had higher incomes from the sale of wood products (Mann-Whitney p = 0.049), but not from other forest goods. Migrants' overall percentage of cash income derived from forests was significantly higher than non-migrants, indicating they were more dependent on forest products (p = 0.023). Migrants also had much lower incomes from employment than did non-migrants (p = 0.029). Overall, however, migration status had less effect than other socioeconomic factors on forest environmental income.
Relationship of overall household wealth to forest environmental income
Absolute forest environmental income
The poorest households were not those with the highest absolute (AFI) or relative forest income (RFI) (Table 4). When households were ranked by their overall cash income, the cash-poorest household tercile had much lower levels of cash AFI (378 265 VND) than did middle income households (1 060 200 VND) and rich households (533 857 VND) (χ2 = 11.05; p = 0.004) (Fig. 2). Poor households also differed from average and rich households in other ways; in general, poor households produced lower amounts of cash income from agricultural goods and had fewer livestock, lower incomes from cash wages and smaller overall landholdings.
ANOVA of socioeconomic variables, comparing households ranked by cash income sources. * = significant (p < 0.05), ** = highly significant (p < 0.01).
Box plot of total cash income from forest-based sources by income classes. AFI = absolute forest income, HH = household. Thick black line = median value, grey box = 25th–75th percentile of values, whisker line = largest value within 1.5 box lengths.
When households were ranked by a combination of both cash and subsistence income, the differences in AFI became less significant: the poorest households received 884 510 VND in all forms of forest environmental income, while the middle households received 1 321 511 VND and rich households received 1 199 694 VND (χ2 = 3.92; p = 0.141). However, it should be remembered that fodder is undervalued in these figures, for reasons noted earlier, and if it were included, it would probably increase the subsistence incomes of the average and rich households more than the poor, as the poor were less likely to own buffalo or cattle (χ2 = 5.41; p = 0.067)
However, because subsistence forest income did not differ significantly between income classes (poor households 428 099 VND, medium households 461 582 VND and rich households 541 476 VND), there was a fairly standard ‘baseline’ of forest subsistence consumption (primarily fuelwood use) across all classes. When households exceeded this baseline and sold additional products for cash, it was middle income, not the poor households, who most benefited.
Relative forest income and dependency
Absolute income is not always the best measure to explain the importance of forests to households. Rather, it is often the percentage of household income from forests that reflects how dependent households are on natural resources. There was a wider range of variation in RFI among economic classes than there was in AFI (Fig. 3). Both the poor and average income terciles had several households with high RFI, including some households who obtained 100% of their families' annual cash income from forests. Yet, overall RFI corresponded with AFI; poor households were less dependent on forest income than average households. On average, the poor raised 21% of their cash from forests, middle income households raised 26% and rich households raised 7% (χ2 = 11.95; p = 0.003).
Percentage of household cash income contributed by forest-based sources by income classes. RFI = relative forest income, HH = household. Thick black line = median value, grey box = 25th–75th percentile of values, whisker line = largest value within 1.5 box lengths.
Further ANOVA analysis showed that high dependency households (those obtaining 50% or more of their cash income from forests, n = 16) earned less from wage labour (p = 0.035) and livestock (p = 0.007) than low-dependency (n = 42) and non-dependent (n = 46) households. This suggests that it is insufficient to analyse whether poorer or richer households are dependent on forest income, without additionally focusing on the specific income streams that correlate with lower household dependency. In the Ke Go case, it appeared that households may use forest environmental income to make up for lower income from other sources like wages or livestock.
Factors affecting income from forests
Combining all the previously considered independent continuous variables (age, income sources and landholdings), and adding in binary dummy variables such as migration status, distance to the forest and income status (poor and non-poor), allows analysis of the relative importance of these variables in regression models (Table 5 and 6). OLS regression to test the effects of the various independent variables on the dependent variable of AFI (Table 5) indicated there was a clear relationship between forest income and social variables such as the age of the household head, and whether it was a poor household; the model indicated younger and medium-rich households had higher AFI. AFI was also related to specific income streams: those who derived a higher percentage of their cash income from wages and livestock had negative relationships with AFI, as did those households which produced more rice. A household's migration status, its distance to the forest and size were not significant in the AFI regression model.
OLS regression with absolute forest income (AFI) as the dependent variable (n = 104, R = 0.713, adjusted R2 = 0.461, SE = 750374.889, F =10.799, p = 0.000). * = significant (p < 0.05), ** = highly significant (p < 0.01).
OLS regression with RFI as the dependent variable (Table 6) indicated variables such as income from wages and livestock and rice productivity and landholdings were negatively associated with RFI. However, unlike AFI, age was not important in predicting RFI while distance was: households closer to the forest had a higher RFI. Migration status and household size did not affect RFI.
OLS regression with relative forest income (RFI) as the dependent variable (n = 104, R = 0.698, adjusted R2 = 0.438, SE = 0.19268, F = 9.920, p = 0.000). * = significant (p < 0.05), ** = highly significant (p < 0.01).
DISCUSSION
The study supported the initial supposition that socioeconomic variables can potentially be used to identify and even predict likely forest–income generating and forest-dependent households. While subsistence use of forests was fairly evenly spread across all households, FCI households were generally closer to the forest, younger households and had fewer wage labour opportunities and low income from livestock. When all variables were combined in regression models however, R2 values were modest, indicating the difficulties of developing a robust model of forest dependency or forest income, given the heterogeneity in the sample and number of variables involved. Nonetheless, age, distance, income class and sources of income were all significantly linked to the household AFI and RFI, and these variables clearly warrant further research (Vedeld et al. Reference Vedeld, Angelsen, Sjaastad and Berg2004).
This study indicates household age should be included in forest income analysis, however only 14 out of 54 cases in a recent meta-analysis of forest environmental income included household age, and statistics did not reveal variation in age to be a significant factor in explaining forest incomes (Vedeld et al. Reference Vedeld, Angelsen, Sjaastad and Berg2004). Other work on poverty in Asia has posited that land-based income, such as that from forests, is increasingly less important for young households who have more off-farm opportunities (Rigg Reference Rigg2006), however this study failed to confirm this. Some young people had migrated out of the province and sent remittances to their families (included in the household income totals, and averaging about 1.9 million VND [US$ 131] per migrant). However, the majority of young people in Ke Go did not migrate and had no alternative to continued on-farm work, including forest exploitation. Younger households reported much lower access to prime government jobs and local wage labour than middle-aged and older households, and concomitantly higher reliance on forest environmental income as a result. Age was particularly important in explaining levels of AFI, with young households registering high AFI, likely due to the lack of other alternative income streams and smaller landholdings, and their ability to undertake physically demanding forest work.
The findings of the study on the links between forest environmental income and household welfare in Ke Go raise intriguing questions, as they appear to contradict much of the literature on poverty and forest use (Angelsen & Wunder Reference Angelsen and Wunder2003; Belcher Reference Belcher2005). For example, in Zimbabwe, Cavendish (Reference Cavendish2000) found poor households derived more than 40% of their income from forests, a higher percentage than the rich, while here the poor had significantly lower RFI levels than average income households, although all household classes used approximately the same amounts of subsistence forest products. Why did the poor not collect more products in the KGNR to sell? This would seem to be a logical welfare strategy, given that cash-poor households also produced less rice and sold less rice. They also derived significantly less of their income from wages and livestock. The poor also had less land, especially the most productive irrigated rice lands. Given these disadvantages, why did poor households not raise more FCI?
There are several possible explanations. First, poorer households had fewer overall landholdings, and may have been more likely to expend limited household labour on inputs into the small agricultural landholdings they did have, rather than collecting forest products. It is well known that wet rice production can be raised substantially, even in the absence of cash inputs like pesticides and fertilizer, by increasing inputs of labour, a process first termed ‘agricultural involution’ (Geertz Reference Geertz1963). This contrasts with other cropping systems, in which labour inputs peak at some point and produce diminishing returns per increase in labour expended. Because increasing labour inputs for wet rice farming may often be the only choice available for capital-deficient households, this can result in labour shortages that may be especially acute for the cash-poor. Any additional labour spent on rice probably reduced labour available for forest collection activities for poor households; qualitative interviews with the poor confirmed that this was a concern for many.
Second, most FCI in Ke Go was raised from the protected forests of the KGNR. However, because extraction there was technically illegal, under state law rangers had the right to confiscate all produce leaving the Reserve. In practice, rangers rarely imposed fines or meted punishments for NTFPs, but timber and charcoal were increasingly being targeted for interdiction. Households reported that an ‘informal payment’ could often be offered to rangers to allow them to continue to take produce out of the KGNR; the average payment needed to remove timber was c. one-third of the timber's retail value. Poorer households would likely have less available cash for these bribes, and thus may have been less able to afford the cost of access to the KGNR. Either poorer households entered the forest less frequently due to concerns bribes could not be paid, or the poor took less produce from the KGNR in the hope of avoiding larger fines. Future studies on the impact of forest law enforcement among different income classes are required to resolve this question.
CONCLUSIONS
Poor households in Ke Go show lower levels of reliance on forest produce in both absolute and relative terms than wealthier households. Development interventions around protected areas often assume that poorer households will be more dependent on forests, and thus these households are often targeted for development assistance in integrated conservation and development projects (ICDPs) (Salafsky & Wollenberg Reference Salafsky and Wollenberg2000; Hughes & Flintan Reference Hughes and Flintan2001). This has certainly been in case in Vietnam, where poor households living around protected forests have been prime objects of ICDP interventions, such as providing alternative income sources like garden plants or livestock (Sage & Nguyen Cu Reference Sage and Cu2001; ICEM 2003). Yet this study indicates that if conservation projects wish to target high forest-using households to induce them to abandon forest exploitation, it is middle income households that need to be targeted. Poor households may also be targeted to reduce poverty, but a concomitant conservation outcome would not be as clear, since the poor use less forest produce.
While specific household targeting may appear infeasible or unwieldy, pro-poor targeting is well-established in many projects in Vietnam, both in conservation and in other sectors. Villages keep extensive lists of which households meet national poverty criteria, and these households are often eligible for ID cards that enable them to receive reduced prices on some subsidized goods, free health care, free schooling and other benefits (AusAid 2002). Recent studies on forests and poverty in Vietnam have advocated better links between the two sectors and the extension of pro-poor policies and poverty targeting to distribution of land and forest resources (Muller et al. Reference Muller, Epprecht and Sunderlin2006). Although these pro-poor targeting efforts are expanding rapidly, they may not necessarily be applicable to conservation work, which may need to focus on middle-income households. Simply identifying where poor communes and districts intersect with high forest cover (Muller et al. Reference Muller, Epprecht and Sunderlin2006) may not be as useful as identifying household level factors that influence forest use. Even in areas of general poverty like Ke Go there were large variations in the relative levels of poverty, and these appear to strongly influence forest use patterns among households.
Forestry could contribute to poverty alleviation in Vietnam (Sunderlin & Huynh Thu Ba Reference Sunderlin and Ba2005; Sunderlin Reference Sunderlin2006). For example, there may be poverty outcomes that could be addressed by providing poor households with greater access to goods now harvested illegally from places like the KGNR, along with better access to markets for harvested goods and value-added processing to raise prices for goods sold. However, in terms of conservation outcomes, projects to reduce the impact of households on protected areas should pay close attention to households who use protected forests the most, and these may not be the poorest households.
Further, ICDPs and other conservation interventions often rely on community participation to help determine who benefits from projects. Older households may be more likely to be invited to meetings than younger ones, and older community leaders (such as village headmen or shamans) may be recruited to lend their local respect and prestige to ICDP activities. Yet this study suggests that older households are least likely to exploit the forest, thus meetings and projects would do well to also explicitly target younger households, as they are more likely to be forest-users.
Forest dependency seems to be lowest in households with alternative wage labour opportunities, while relationships between agricultural production, for example in gardens, and forest use were less clear. KNGR households with steady cash wages were the least likely to have forest income, while households with greater access to gardens and alternative crops did not differ in lower forest use. This suggests that instead of improving agriculture, through garden development or alternative crops in rural areas near protected reserves (the most common approach in Vietnam and many other places; see examples in SNV [Netherlands Development Organization] 2000), conservation interventions should focus on increasing wage labour opportunities, such as providing job training for unskilled workers or creating waged positions in forest protection, for example rangers. If funded, such alternatives are likely to be feasible given that a number of households surveyed were already seeking such opportunities through migrant remittances. If more local waged positions were available, it is likely that more households would be able to reduce their forest use, given access to appropriate income substitutions.
ACKNOWLEDGEMENTS
Research funding was generously provided by the Wenner Gren Foundation for Anthropological Research and a National Science Foundation Dissertation Improvement Grant. I would like to thank my research sponsor in Vietnam, the Center for Natural Resources and Environmental Studies of Vietnam National University and, in particular, Dr Vo Quy, Dr Truong Quang Hoc and Vo Thanh Giang, for supporting this research. I am grateful to Dr Le Tran Chan for assistance in identifying specimens. I would also like to thank Dr. J.E.M. (Mike) Arnold for inspiring my interest in NTFP research many years ago.
