Printer Friendly

An economic approach to household collection of gum arabic from the wild/Une approche economique a la cueillette sauvage de la gomme arabique/Analisis economico de la recoleccion familiar de goma arabica en estado silvestre.

INTRODUCTION

Gum arabic is a natural exudate of shrubs and trees of several species of the legume genus Acacia (Nishinari and Doi 1993), which naturally grows throughout the arid and semi-arid lands (ASALs) of sub-Saharan Africa. The gum-producing species grow most densely in the so-called gum arabic belt, spanning an area several hundreds of kilometres wide to the south of the Sahara, stretching from Senegal in the West to Somalia in the East (White 1983, FAO 1985, ICRAF 1992, Islam et al. 1997). In Kenya gum is collected from both Acacia seyal and Acacia senegal. The latter is more common and found as a leguminous tree, deciduous shrub or shrub tree species. It belongs to the subgenus Aculeiferum (Arce and Banks 2001). Four varieties within the species produce gum arabic: var. senegal, kerensis, rostrata and leiorhachis (Brenan 1983, Fagg and Allison 2004). Acacia senegal var. senegal is most valued for quality (Lelon et al. 2010) and var. kerensis forms the basis of the international export market. A. senegal var. kerensis is also the most commonly found variety in Kenya (Booth and Wickens 1988, Chretin et al. 2008).

Gum producing regions are typically sparsely populated and inhabited by poor households. Although gum originates in poor dry land areas of the developing world, most of its uses are industrial. Gum arabic is used in the food industry as a stabilizer, in soft drinks, and in candy; in the pharmaceutical industry; and in industries such as printing, ceramics and textile (Anderson and Weiping 1992, ICRAF 1992, Cunningham et al. 2008). The chemical structure of the gum is so complex that no perfect synthetic substitute has yet been discovered. In fact, it is such a crucial input to industry that the Gum Arabic Company, the largest exporter in the world, had an exemption from the trade embargo of the United States against Sudan even though it was known that 70% of the shares of this company belonged to Osama Bin Laden (Van Dalen 2006).

This industrial importance contrasts sharply with realities faced by gum-collecting households. In Kenya and elsewhere, gum arabic is traditionally gathered in the wild by poor nomadic and semi-nomadic pastoralists (Chikamai and Odera 2002, Chretin et al. 2008, Gachathi and Eriksen 2011) and primarily collected for subsequent sale in the market (Wekesa et al. 2010) although it is also valued for its medicinal properties and used to relieve backache and painful joints (Chikamai and Odera 2002, Obua et al. 2006, Watson and Van Binsbergen 2008). Other Non-Timber Forest Products (NTFPs) are also used by poorer households for subsistence or cash income (Neumann and Hirsch 2000, Shackleton and Shackleton 2004) or as a safety net (Belcher and Schreckenberg 2007). Sunderlin et al. (2005) explain the dependence of poor households on NTFPs by referring to their open-access nature and the low skill and capital required for their collection. This situation certainly applies to gum arabic collection in Kenya, where lands are communally owned and specific tools or skills are not required and hardly used (Wekesa et al. 2010).

The literature on commercialization of NTFPs focuses on direct costs and benefits of collection and marketing, and does not explicitly take into account the interaction between income sources. Especially in a context where returns to income-earning activities are subject to substantial uncertainty and variability, and switching costs between activities are low, such a restricted approach offers an incomplete explanation for observed household behaviour. For this paper we studied gum arabic collection in northern Kenya, where cattle herding provides an alternative source of cash income. Mainly poor households are involved in gum collection (Chretin et al. 2008, Gachathi and Eriksen 2011) and it is available in the dry season when other sources of income fall short (Oni and Gbadamosi 1998). And yet, it is commonly understood that the resource is underutilised (Chikamai and Odera 2002, Chretin et al. 2008, Gachathi and Eriksen 2011). In this paper an economic model is proposed that takes into account multiple income sources and shows the apparent trade-off that households make under different climatic and security pressures.

In the next section the research area and data collection are described. The employed model and how it was tested is described next, followed by a description of how the data were used. The used dataset is special in its focus on the socio-economics of gum collection and marketing and therefore discussed in some detail in the results section, before showing the results of the tests of the theoretical model.

MATERIALS AND METHODS

Research area and data collection

The study was conducted in Turkana, Marsabit, Samburu, Isiolo and Mandera Counties (1) in northern Kenya. These regions have a high natural population of Acacia senegal and Acacia seyal from which local communities actively collect gum arabic (Beentje 1994, Chikamai and Gachathi 1994, Chikamai and Odera 2002).

These five counties are among the poorest in Kenya, with the share of people living below the national poverty line ranging from 71.6% in Samburu to a staggering 94.3% in Turkana (GOK 2007). They are also sparsely populated, with population densities ranging from 4.1 in Marsabit to 39.4 inhabitants per square kilometre in Mandera (GOK 2010). The area is inhabited by several tribes, the largest of which are the Turkana, Samburu, Rendille, and Somali, traditionally nomadic pastoralists.

No accurate map containing locations of all villages in the area was available, which is partly explained by the regular moving of entire villages. Therefore, communities were selected by driving through areas where gum was being collected and asking along the road whether particular branches were leading to settlements. Care was taken to select villages at different distances from main roads, some at over an hour and a half driving away. When a village was encountered, but residents indicated nobody was involved in arabic gum collection, it was excluded from the survey. A total of 20 villages were located this way and included in the study (Figure 1). Names of villages are provided in the appendix (Table 5).

Household-specific data were collected in August-September 2009 as part of the EU funded ACACIAGUM project, using a pre-tested semi-structured questionnaire. Collected data cover the long dry season of 2009 and marketed quantities and received prices during the preceding short dry season, which were obtained by recall. Several collectors were interviewed per village, giving a total of 201 respondents. These respondents were asked for information on their own collecting practices, as well as characteristics of the household to which they belonged. All interviews were conducted by the second author with assistance from the Kenya Forestry Research Institute (KEFRI).

Household-specific data were complemented by secondary GIS data on vegetation cover and rainfall. Vegetation cover data was obtained from a study (2) by the German development organisation (GTZ) (3), which was conducted as an extension of the UNESCO-funded Integrated Project for Arid Lands (IPAL). Although these data are quite old, this does not affect its usefulness for our purposes, as vegetation types are determined by landscape and climate characteristics, factors which do not change within a time span of decades (White 1983). Rainfall data was obtained from the USAID-funded Famine Early Warning Systems Network (FEWS NET), which combines rain gauge data from local weather stations, satellite precipitation estimates, and cloud-top temperature precipitation estimates (4).

Theoretical model

An economic model was used to analyse households' labour allocation to income-generating activities with the goal of maximising household income, given traditional roles. To simplify the analysis, we assumed households could allocate dry-season labour to only two activities: livestock herding and gum arabic collection. Livestock herding was the primary source of income of households, and hence the effect of livestock keeping on gum arabic collection was expected to be larger than the effect of other income sources.

The production possibility frontier (PPF) shows potential output combinations that households can choose (Figure 2). Total household output is higher when a mixture of outputs is produced. Income from livestock herding is shown on the horizontal axis and income from gum arabic collection on the vertical axis. In a normal year, the PPF is represented by the solid line and the return to labour in livestock herding is higher than in gum collection. The level of output the household chooses depends on relative prices of livestock and gum, represented by the slope of the line PP. Given these prices, households' income maximising output in a year of normal rainfall is at point A.

In a drought year, the PPF shifts inwards to the position of the dotted line: the return to household resources falls in both livestock herding and gum arabic collection. The fall in productivity is expected to be more extreme for livestock: not only is there less forage and water available, but also larger distances need to be covered to reach them (Schwartz et al. 1991, Coppock 1994, Schwartz 2005, Sadler et al. 2010). Cattle are especially sensitive to rainfall, and produce significantly less milk in periods of little rainfall, sometimes not producing any milk at all (Schwartz and Schwartz 1985, Lesorogol 2008, Sadler et al. 2010). When droughts are severe enough, even camels may stop producing milk (Field 1979, Sperling 1987) (5). Rainfall shortages also reduce gum-productivity of trees (Dione 1989, Ballal et al. 2005), thereby lowering return to labour in gum arabic collection. At constant prices, production should move to point B, with a sharp drop in livestock income and a slight increase in gum income.

However, during droughts livestock prices fall as a consequence of oversupply caused by distress sales (Fafchamps and Gavian 1997, Barrett et al. 2003). Falling livestock prices cause the relative price line to rotate to PP'. Now, optimal production is at point C, where more gum is collected than in a year with normal rainfall, even though the absolute return to gum collection is lower.

The model was tested using cross-sectional data. Because of the cross-sectional nature of the data, controlling for household-specific factors was not only necessary to allow unbiased estimation of the effects of shifts of the PPF, but also provided an insight into the relative importance of factors hypothesised to influence its shape.

Data were analysed in Stata 10 using linear regression analysis. Ordinary least squares (OLS) estimation was used to obtain parameter estimates (Wooldridge 2006). The dependent variable was the natural logarithm of marketed quantity in kilograms per household, which more closely approximated the normal distribution than the unmodified variable.

y = [alpha] + [[beta].sub.1][X.sub.1] + [[beta].sub.1][Z.sub.1] + [[beta].sub.2][X.sub.2] + [[beta].sub.3][X.sub.3] + [[beta].sub.4][X.sub.4] + [[gamma].sub.2][Z.sub.2] + [epsilon]

Where y is the natural logarithm of marketed quantity of gum arabic from Acacia senegal and A. seyal in kilograms per year, [alpha] is a constant, x indicates an individual variable, and z a vector of variables. Betas ([beta]) and gammas ([gamma]) indicate coefficients and vectors of coefficients respectively. Finally, epsilon (s) is an error term. [x.sub.1] was price; [z.sub.1] was a vector of household-specific variables hypothesised to influence the shape of the production possibility frontier of gum collection and livestock keeping. It included household size, wealth, the number of female and male collectors, distance to preferred collection plot, and two binary variables equal to one when households used tapping or migrated with livestock. [x.sub.2] was the area containing Acacia trees; [x.sub.3] was rainfall shortage; [x.sub.4] was insecurity; and finally, [z.sub.2] was a vector containing tribe dummies. These tribe dummies were included to control for regional or cultural differences that might influence the shape of the PPF. The next section gives a detailed description of these variables.

A major concern regarding reliability of estimated coefficients was omitted-variable bias. No data were available on the number of collectors in each region. Exclusion of this variable induces omitted-variable bias if competition for the resource is important, i.e. when the availability constraint would be binding. (6) Such omitted-variable bias cannot be directly tested for, but is one of the main causes of heteroskedasticity. All regressions were therefore tested for heteroskedasticity using the Breusch-Pagan test (Breusch and Pagan 1979).

Variables used in the regression model

The marketed quantity of gum arabic depends on its relative price, and the position and shape of the PPF. Its relative price increases when market prices of gum increase, or when livestock prices fall. Because no accurate livestock prices were available, only actual selling prices of gum were included in the model (7). Ceteris paribus, higher gum prices should induce higher marketed quantities.

The position of the PPF is household-specific and depends on household size. In order to compare households of different composition, an adult equivalent (AE) scale was constructed, which scales the consumption level of household members of different ages to the equivalent consumption level of an adult. Following Deaton and Zaidi (1999), AE = (A + [alpha]K)[theta], where A refers to adults, K to children, and [alpha] and [theta] are weights. For low-income households, which spend most income on food items, Deaton and Muelbauer (1986) and Deaton (1997) recommended setting [alpha] = 0.3 and [theta] = 0.9. The resulting variable is a measure of household size. Larger households require more food, and are expected to market larger quantities of gum.

Household wealth is a good indicator of the degree to which households rely on NTFPs (Timko et al. 2010). Because wealth is held in a variety of assets, an asset index was constructed to facilitate inter-household comparability. The value of the index for household i is the weighted sum of the different assets owned by the household: [Index.sub.j] = [m.summation over (j=1)][w.sub.j][a.sub.ij], where [w.sub.j] are weights and [a.sub.ij] are assets j = 1, ... ,m. Sahn and Stifel (2000, 2003) suggested factor analysis to determine weights (8), which is particularly useful in the absence of accurate price information. In the study area, the majority of households were pastoralists, and livestock their most important and valuable asset (McPeak 2004, Lybbert and McPeak 2011). Livestock holdings were aggregated into Tropical Livestock Units (TLU) (FAO 1972) (9) and added as a single asset. Other assets included were radios, transport means, telephones, and number of houses (10). Furthermore, years of education were included as a measure of human capital. A priori, the effect of wealth on marketed quantity is ambiguous. Some authors find a negative relationship (Timko et al. 2010), whereas other authors find a positive correlation (Ambrose-Oji 2003, Gauli and Hauser 2011). This ambiguity might indicate a non-linear relationship, and hence a squared term was included.

The number of female and male collectors, distance from preferred plot in minutes, and whether or not a household used tapping or migrated with livestock were directly measured in the survey. Not every household member is involved in gum collection, and traditionally there are stark differences in tasks based on gender, although in general households with more collectors should market larger quantities. Distance increases the opportunity cost of collection, which should have a negative effect on marketed quantities. Tapping is a technique where part of the bark of the tree is removed to stimulate gum exudation. It has been reported that tapping improves gum yields of A. senegal (Wekesa et al. 2009) and A. seyal (Fadl 2011). Therefore, households that use tapping are expected to produce and hence market more gum than those that rely on natural exudation. Migration is a labour-intensive process and in many cases involves a move of the entire village. The type of migration differs between tribes and villages, but is expected to have an overall negative effect on the time allocated to gum collection and consequently marketed quantities.

To complement information from the household level survey, gum availability and rainfall variables were constructed using secondary GIS data. Gum availability was not observed directly, but constructed from data on vegetation cover. These data distinguished 77 distinct vegetation classes in northern Kenya, twelve of which contained Acacia senegal (Appendix, Table 6). For each vegetation class, the ratio of the area it covered to the total area within a fifteen-kilometre radius (11) around each surveyed community was constructed using ArcMap 10. To establish the gum availability in these areas, these twelve ratios plus two dummies on tapping and insecurity were regressed on the quantity individuals in the community expected to collect per day (Appendix, Table 7). Vegetation classes 8.3, 14.8, 16.5, 21.1, 22.13 and 22.17 had a significant effect on the quantity an individual expected to collect daily. The sign on all of these significant vegetation variables was positive, indicating that presence of any of these vegetation types increased the daily quantity an individual expected to collect. Next, a variable called Acacia area was constructed by taking the un-weighted sum of the significant vegetation variables, in effect assuming the density of gum-producing trees to be the same in each of these vegetation classes.

Rainfall shortage was measured as the difference between actual rainfall during the rainy seasons preceding the short and long dry seasons of 2009 and long run average rainfall in these periods. Actual rainfall was subtracted from average rainfall for an area of ten kilometres around each village using ArcMap 10, such that a positive value of the variable indicated rainfall shortage in millimetres.

Insecurity and tribe dummies were obtained directly from the survey. The insecurity situation prevailing in parts of the study area was expected to lower the return to both gum collection and livestock keeping, causing the PPF to shift inwards, and hence lower the quantity of gum collected for marketing. Tribe dummies equalled one when a respondent belonged to a particular tribe. There are no expectations regarding the effect of belonging to the Turkana, Rendille, Samburu, or Somali tribe on marketed quantities; the variables were merely included to control for regional or cultural differences.

RESULTS

Household and regional data description

Descriptive statistics on household characteristics related to gum arabic collection are presented in Table 1. On average, sample households made 2,082 Kenyan Shilling (KSH) from marketing gum in 2009. Based on the 2006 rural food poverty level of 988 KSH per person per month (GOK 2007) and the average adult equivalent family size in the sample (3.3), gum arabic revenues alone were insufficient to raise household income above even the rural food poverty level. Sampled households marketed an average of only 44.3 kilograms of gum arabic in 2009, which would have taken an individual collector around 14 days to collect. In most households, several people were involved in gum collection, with an almost equal share of female and male collectors. Average distance to preferred natural stands of A. senegal and A. seyal was two and a half hours. These distances are long given the extreme climate, with temperatures reaching 40[degrees]C in the dry season. Because most collectors preferred not to spend the night in the field, these long travel distances severely restricted the amount of time available for collection. Tapping was not common. In fact, only 14% of households tapped the gum producing trees, and out of this 14%, only few used specialised tapping tools. Several households indicated damaging the tree by hitting it with a stone while others used their knife. The low occurrence of tapping might be explained by the unrestricted access to trees. All interviewed households indicated that the plots where they collected gum were communally owned. Because it takes several days for gum to exudate and dry after having made the cut, the risk of other collectors taking the gum is substantial, especially due to the presence of migrating herders.

The aridity of the region forces herders to almost continuously move around with their livestock. Of the interviewed households, 56% practised some form of migration, and one household even migrated year-round. Most livestock migration takes place in the dry season to find available pasture and water, thus directly competing for labour with gum arabic collection. Traditionally, men and boys are responsible for cattle and camels, while women and girls take care of smallstock--sheep and goats. Migration therefore reduces both male and female household labour available for gum arabic collection. On average, 21% of households indicated insecurity was their primary concern when collecting gum. In Turkana, where cattle raiding is part of the culture (Pike 2004) and in Mandera, a region bordered by southern Somalia and Ethiopia, where Al-Shabaab--a Jihadist organization with international ambitions--has staged killings and kidnappings since the 1990s (Marchal 2009), households more frequently mentioned security concerns.

A summary of household asset ownership is given in Table 2. Average asset holdings were small. The only asset every household was found to own was a house. As polygamy is practiced by all tribes except the Rendille, and it is customary to provide every wife with her own house, some households owned multiple houses. Livestock ownership, measured in TLU, was low at 4.73 per household and 1.43 per adult household member. Pratt and Gwynne (1977) estimated that 4.5 TLU per adult household member could provide a household with just sufficient food for survival. Based on that measure, average livestock holdings in the sample were far below subsistence level. Given average household size, the majority of households would need more than three times as much livestock to reach this subsistence level. The constructed wealth index was normalized to zero and ranged from -0.536 to 3.269. Wealth was unequally distributed. Almost 80% of households had an index score below zero.

Data were also collected on the main sources of household income (Table 3), although time and resources did not allow for quantification. In the study region most households depend on pastoralism, which over 75% of respondents named as their primary source of income. Arabic gum and wood products had the second-largest contribution to income. This importance of gum arabic appears to suffer from 'social desirability bias' (Grimm 2010), as the average quantity of gum arabic marketed by households reporting it to be their primary source of income was 47.5 kilograms, insufficient for subsistence and not significantly different from the average of households not reporting it as their primary source of income. Similarly, obtaining an income from wood products such as charcoal is likely considered socially undesirable, which would result in underreporting.

The predominance of pastoralism as a source of income is confirmed by other authors on the region (Broch-Due 1999, Lewis 1999, Fratkin 2001, Sadler et al. 2010). This importance of livestock keeping fits well with our model. Relief, mostly in the form of food aid but also through other means such as cash for work schemes, is common in northern Kenya and has taken structural forms. In 1999, half the people in Turkana County are estimated to have received food aid, a situation which is expected to continue (Lind 2005). Given this institutionalisation of relief, it is surprising that few interviewed households indicate a substantial dependence on food aid. Another surprising finding is that none of the households mentioned agriculture as an important source of income, especially given the context of on-going sedentarisation in the region (Little et al. 2001). This finding might be explained by the fact that home consumption was not specifically accounted for, thereby excluding home gardens from being classified as agriculture. Moreover, commercial agriculture may simply not be feasible in areas where arabic gumyielding tree varieties are found, which typically have poor soils with low moisture content and are rocky in nature (Chiveu et al. 2008).

Regression results

Results of regressions on marketed quantity are shown in Table 4. In the first regression only explanatory variables from survey data were included, as these were expected to suffer less from measurement error, and thereby provide a means of comparison for results of subsequent regressions. In the second regression tribe dummies were included to check for the existence of regional and cultural differences on gum arabic collection. In the third regression variables constructed from GIS data, Acacia area and rainfall shortage, were included. All regressions were estimated using OLS and checked for heteroskedasticity using Breusch-Pagan/CookWeisberg tests. Based on these test none of the regressions appeared to suffer from heteroskedasticity problems and hence these statistics are not included in tables nor further elaborated upon in the text.

The effects of price, household-specific characteristics, and insecurity on marketed quantities are shown in the first regression. The coefficient on price is not significant, which is commonly the case in cross-section estimates and might also be related to the non-inclusion of livestock prices. Household size in adult equivalent units is a scaling variable. Larger households have higher consumption and more labour available, and therefore their PPF should be higher for both gum collection and livestock keeping. Because marketed quantities were measured at the household level, the coefficient on household size was expected to be positive. This hypothesised relationship was not significant. Wealth had a significant positive effect on the marketed quantity of gum. The strength of this relationship declined for increasing levels of wealth, which is evident from the negative sign on the squared term. At values of the wealth index above one, marketed quantities started to fall with increasing levels of wealth. Only around 10 per cent of sample households fell into this category. Households on either side of this wealth threshold were compared using independent sample t-tests. The two groups differed significantly only in their sources of income. Households with a wealth index score above one were more likely to derive their income from commerce or wage labour. Wealth index and its square were jointly significant in all regressions (p = 0.0049, 0.0170, and 0.0299). This relationship was not affected by the inclusion of additional explanatory variables (regression 2 and 3) and the size of the coefficients was comparable. The size of the coefficients cannot be interpreted directly, because of the way the wealth index was constructed. The number of female collectors had a positive effect. One additional female collector increased the marketed quantity collected by around 16%. However, this relationship was not very robust and disappeared when the regional dummies were included. The number of male collectors, the distance to the plot where gum was most frequently collected, and whether tapping was used did not have significant effects on marketed quantities either. Migration with livestock had no significant effect in the first regression. Insecurity was highly significant in the first regression and had a large negative effect on marketed quantities. The size of the coefficient should be interpreted with some caution. It appears insecurity caused a fall in marketed quantity by over 60%. However, insecurity was a very regional problem, primarily affecting Turkana and the border region with Somalia, causing the variable to pick up some regional variation. If these regions differed in more aspects than just the level of security, the coefficient might have been biased in the first regression.

In the second regression differences between regions were explicitly taken into account by including the tribe dummies Turkana, Samburu and Somali, with Rendille as the base group. These dummies measure both cultural as well as other regional differences, as the tribes lived in clearly defined geographical regions (Figure 1). The dummies were jointly significant (p = 0.0032), justifying their inclusion. By their inclusion, migration became significant, and insecurity insignificant. The latter effect was most likely due to the regional nature of insecurity, which was only mentioned as a concern by Turkana and Somali collectors. The significance of migration once tribe dummies were included can be explained by the stark difference in migration patterns within tribes, resulting in different labour requirements. Migration patterns depend not only on cultural differences but also on vicinity of dry season pastures and water points. Migration reduced marketed quantities, as migrations are labour intensive, leaving less time to collect gum. Although some gum is collected during migrations, most of this is consumed on the spot by collectors.

The third regression included the GIS variables Acacia area and rainfall shortage. The overall measure for the explanatory power of the regressions improved strongly. The coefficient on Acacia area was positive and highly significant. In areas where Acacia senegal is abundant, households market more gum. When larger areas containing gum-yielding trees are found around the village, more gum can be collected per unit of time, increasing the return to labour in gum collection. Rainfall shortage had a positive and significant coefficient, indicating that below average rainfall increases marketed quantities--despite the lower availability--as predicted by the theoretical model. If rainfall shortage reduced the return on livestock keeping more than the return on gum collection, the attractiveness of the latter increases. The importance of this effect was large: in regions where rainfall fell short by the sample average of 54 mm, marketed quantities were 75% higher. This finding corroborates the hypothesis that gum collection is especially important when other sources of income fall short. Sign, size and significance of the household-specific variables were comparable to earlier regressions. Insecurity was significant in the third regression, albeit only at 10%. This variable, because of its localised nature, appears to have picked up a lot of spatial variation. Finally, the tribe dummies are jointly highly significant, indicating substantial cultural or regional variation.

DISCUSSION

Earlier studies on NTFPs found them to contribute substantially to income (Neumann and Hirsch 2000, Shackleton and Shackleton 2004). This study was the first to quantify actual income generated by gum arabic in Kenya, and our results indicate that its contribution to household income is low. In Ethiopia, gums and resins, including frankincense, myrrh, and haggar, were found to contribute around 30% to household income (Woldeamanuel et al. 2011). Although these other resins were also marketed in our study region, this was on a much smaller scale (Chikamai and Odera 2002). Low observed marketed quantities could imply only few days were spent collecting. An alternative explanation for the small quantities of gum is offered by Chretin et al. (2008), who find that gum collection is usually combined with other household activities, such as collecting firewood, herding livestock or looking for water, which would result in actual daily quantities collected to be much lower than reported expected quantities. On average, households had almost the same number of male and female collectors, a finding which contrasts with earlier studies, which found women and girls to be the main collectors (Wekesa et al. 2010, Gachathi and Eriksen 2011). Although gum productivity of trees can be improved by tapping, this technique was hardly employed. When households did tap, they often used stones or their own knife rather than specialised tools, a finding in line with results from Wekesa et al. (2010). Low observed quantities are likely caused by a combination of low expected returns, given current market prices, combined with substantial effort required for collection, due to high temperatures, insecurity and long travel distances.

The most important source of cash income of interviewed households was pastoralism, followed by gum arabic and wood products. Traditionally, pastoralism was the primary source of income in the region as was the case in our sample, even in the context of on-going sedentarisation (Heald 1999, Fratkin 2008, Little et al. 2008). Although interviewed gum collectors indicated to depend primarily on pastoralism, observed livestock holdings were generally far below subsistence levels, explaining the use of multiple income sources. Income was complemented by forest products, both wood products and gum arabic. The importance of wood products for income generation is cause for concern, as it is an indicator of deforestation, which is especially a threat in fragile environments with population pressure and without property rights (Geist and Lambin 2002, Sunderlin et al. 2005). Respondents indicated all plots where gum arabic was collected had unrestricted access. Although free access is important for non-timber forest products to provide a green social insurance (Cunningham et al. 2008), it also encourages overuse (Hardin 1968). Deforestation threatens the productivity of livestock, as trees are the most important source of fodder during the dry season (Barrow 1990). Moreover, decreasing tree cover reduces the availability of gum arabic, which has the potential to contribute substantially and sustainably to household income (Gachathi and Eriksen 2011).

Rainfall shortage was found to have a significant and large effect on marketed quantities of gum arabic. The effect of rainfall shortage on marketed quantities was positive--where rainfall was below its long-run average, households marketed larger quantities of gum. The model suggested in this study offers an explanation for this observation: marketed quantities depend on the return in gum collection relative to the return in other income generating activities. Although rainfall shortages lower gum productivity of trees (Dione 1989, Ballal et al. 2005), if rainfall shortages cause a sharper drop in the return to alternative activities such as livestock keeping than in the return to gum collection, they increase the relative attractiveness of gum collection, and hence observed marketed quantities. The price of gum arabic in monetary terms did not have any discernible effect on marketed quantities. Although the theoretical model was in terms of relative prices, the monetary price would have an effect in itself if the sample contained a large share of opportunists, which are increasing in importance in the region (Gachathi and Eriksen 2011). Our data do not support the importance of this group since most gum arabic appeared to be collected by households in need. The almost universal fall in productivity of livelihood activities during rainfall shortages lowers total household income, which is evident from the upsurge in food aid dispensed to the region in such periods (Lind 2005). The concurrent increase in marketed quantities of gum arabic found in this study support earlier work on the important safety net function of NTFPs (Byron and Arnold 1999, McSweeney 2004).

Wealth and whether or not a household migrated with livestock were the most important household-specific determinants of marketed quantity. Even though no monetary values for wealth were available, low observed asset ownership shows most interviewed households were poor, a finding in line with earlier work by Timko et al. (2010). Marketed quantities varied with wealth: they increased with wealth until a threshold, after which they decreased. The increase is in line with findings in a recent study on non-timber forest products in Nepal (Gauli and Hauser 2011) and earlier work in Cameroon (Ambrose-Oji 2003). This result might be explained by the thin market for gum arabic in Kenya, where trader visits are infrequent and uncertain (Chretin et al. 2008). Poor households have less capacity to cope with such market uncertainty than wealthier households, limiting the extent of their market participation, indicating increased market access would enhance poverty alleviation capacity (Belcher and Schreckenberg 2007). The decrease in marketed quantities for household above a certain wealth threshold can be explained by low returns to gum collection, which makes participation only relevant for households in need. This is supported by our finding that households above the wealth threshold were more likely to be involved in wage labour or commerce as these activities are less susceptible to the negative effect of droughts. Migration was also found to have a significant effect on marketed quantities. Migrating households marketed less gum. Because most migration takes place in the dry season (Schwartz et al. 1991) it competes directly with gum collection for labour, reducing marketed quantities.

Insecurity had a significant and large negative effect on marketed quantities. It was highly localised and mainly encountered in Turkana and along the Somali border. In Turkana, Pike (2004) found cattle raiding to have a strong influence on herding strategies, limiting access to pastures, especially in the dry season, thereby increasing vulnerability to drought. Whereas traditionally these raids were small and aimed at stealing cattle, in recent decades their size and impact on the human population have increased dramatically. Attempts at privatising land tenure and offering large tracts of land to private investors including mining companies further decreased resource availability and increased conflict (Mkutu 2007). Along the Somali border, insecurity is caused by Al-Shabaab staged killings and kidnappings (Marchal 2009). By limiting pasture availability, insecurity limits the ability of local pastoralists to cope with weather risk and result in livestock losses. Losses of livestock due to raids or droughts were shown to increase dependence on gum arabic (Woldeamanuel et al. 2011). This study shows that gum arabic marketing is also threatened by insecurity, further reducing the food security of households in the region. In fact, findings probably underestimate the true impact of insecurity on marketed quantities, because the most insecure regions were avoided during data collection.

CONCLUSION

The economic model proposed in this study established a relationship between the return to gum arabic collection and the return to alternative sources of income. The main predictions of the model are supported by our data. Rainfall shortages decrease the return to gum arabic collection, but have an even stronger negative effect on the return to alternative allocations of labour such as livestock rearing, causing an increase in the relative return to gum collection and consequently, marketed quantities. Increasing relative returns are caused by increases in relative productivity and relative price. Although the overall effect was correctly predicted, in this study no distinction could be made between productivity and price effects, and hence their relative importance could not be established.

The increase in marketed quantities of gum arabic when other sources of income fall short, demonstrate its safety net function, implying low expected returns to collection. Expected returns are lowered by the high uncertainty involved in gum marketing. Although gum collection does not require any upfront capital investments, the uncertainty about the time of sale caused by market thinness requires an upfront investment of labour, a risk which is more easily borne by wealthier households. Another factor currently limiting marketed quantities of gum is the insecurity situation prevailing in Turkana and along the Somali border during the study period, which was found to have a direct and large negative effect.

The established importance of gum arabic to households in northern Kenya, especially as a safety net, justifies policy interventions aimed at increasing its benefits to collectors. Policies to promote increasing the return to collection and improving the current insecurity situation might be a good starting point. However, further research in these areas is needed to more accurately define the costs and benefits of specific policy interventions.

Little is known about the socio-economics of gum arabic collection, giving tremendous scope for further research. First, most research on gum arabic collection, including this study, focused on current rather than potential collectors. Increased comprehension of the determinants of the participation in gum arabic collection would provide a more complete understanding of the impact of policy measures. Second, the model proposed in this study was tested using cross-sectional data. Using time-series data would allow controlling for household and village-specific characteristics which are hard or costly to measure directly, such as skill of household members or productivity of local pastures, allowing for more accurate testing. The model proposed in this study could serve as starting point to guide such future endeavours.

APPENDIX

TABLE 5 Names of villages where the survey was conducted

Number    Village name

1         Kakilai
2         Kariabur
3         Loritit
4         Namoroputh
5         Lorengippi
6         Lokiriama
7         Kasuroi
8         Nakukulas
9         Kakongu
10        Kurkum
11        Kargi
12        Ngurunit
13        Laisamis
14        Sereolipi
15        WestGate
16        Ngarendare
17        ElDanaba
18        Gither
19        Takaba
20        ShimbirFatuma

Note: Village cursive correspond to numbers in Figure 1.

TABLE 6 Vegetation types containing Acacia senegal

Vegetation                                                Physical
Class        Most commonly encountered species            Class

8.3          Lintonia--A. Senegal                        Grassland
14.15        A. Senegal--Commiphora--Boswellia           Bush land
14.2         A. Senegal--A. Mellifera--Commiphora        Bush land
14.8         Commiphora--A. Senegal--Grewia              Bush land
16.5         Tetrapogon--Aristida--A.                    Grassland
               Tortilis--A. Senegal
16.6         Aristida--Tetrapogon--A.                    Grassland
               Tortilis--A. Senegal--A. Reficiencs
20.17        A. Reficiens--A. Senegal                    Shrub land
20.8         A. Senegal                                  Shrub land
21.1         Sporobolus--misc. Acacia species            Grassland
22.13        Aristida--Indigofera--A. Senegal            Grassland
22.15        Aristida Indigofera--A.                     Grassland
               Tortilis--A. Senegal--A. Reficiens
22.17        Aristida--Indigofera--A. Tortilis--A.       Grassland
               Senegal--A. Reficiens

Data source: GTZ

TABLE 7 Expected daily quantity of gum Arabic

Variable             Coeff.

Vegetation 8.3       107.953 **
                     (2.55)
Vegetation 14.15     -59.779
                     (-0.82)
Vegetation 14.2      -10.200
                     (-1.24)
Vegetation 14.8      38.626 ***
                     (2.73)
Vegetation 16.5      67.422 **
                     (2.27)
Vegetation 16.6      -8.554
                     (-1.41)
Vegetation 20.17     -16.026
                     (-1.57)
Vegetation 20.8      24.004
                     (0.72)
Vegetation 21.1      13.505 *
                     (1.77)
Vegetation 22.13     57.756 **
                     (2.11)
Vegetation 22.15     10.860
                     (1.15)
Vegetation 22.17     12.587 *
                     (1.96)
Tapping              0.618
                     (1.15)
Insecurity           -0.836
                     (-1.57)
Constant             2.800 ***
Observations         (6.59)
                     201
Adjusted R-squared   0.103

Note: Dependent variable is expected quantity per day in kilograms.

T-statistics in parentheses: * Significant at 10%;
** significant at 5%; *** significant at 1%


ACKNOWLEDGEMENTS

We are grateful to the INCO-DEV ACACIAGUM Project (032233) and the European Union, through which this research was funded. Special thanks go to KEFRI for their invaluable support with facilitating the fieldwork. The valuable suggestions made by anonymous referees is gratefully acknowledged.

REFERENCES

AMBROSE-OJI, B. 2003. The Contribution of Ntfps to the Livelihoods of the 'Forest Poor': Evidence from the Tropical Forest Zonde of South-West Cameroon. International Forestry Review 5(2): 106-117.

ANDERSON, D.M.W. and WEIPING, P.W. 1992. Gum Arabic (Acacia Senegal) from Uganda: Characteristics N.M.R. Spectra, Amino Compostions and Gum/Soil Cationic Relationships. International Tree Crop Journal 7(3): 167-179.

ARCE, L.R. and BANKS, H. 2001. A Preliminary Survey of Pollen and Other Morphological Characters in Neotropical Acacia Subgenus Aculeiferum (Leguminosae: Mimosoideae). Botanical Journal of the Linnean Society 135(3): 263-270.

BALLAL, M.E., EL SIDDIG, E.A., ELFADL, M.A. and LUUKKANEN, O. 2005. Relationship between Environmental Factors, Tapping Dates, Tapping Intensity and Gum Arabic Yield of an Acacia Senegal Plantation in Western Sudan. Journal of Arid Environments 63(2): 379-389.

BARRETT, C.B., CHABARI, F., BAILEY, D., LITTLE, P.D. and COPPOCK, D.L. 2003. Livestock Pricing in the Northern Kenyan Rangelands. Journal of African Economies 12(2): 127-155.

BARROW, E. 1990. Usufruct Rights to Trees: The Role in Dryland Central Turkana, Kenya. Human Ecology 18(2): 163-176.

BEENTJE, H.J. 1994. Kenya Trees, Shrubs and Lianas. Nairobi, National Museums of Kenya.

BELCHER, B. and SCHRECKENBERG, K. 2007. Commercialisation of Non-Timber Forest Products: A Reality Check. Development Policy Review 25(3): 355-377.

BOOTH, F.E.M. and WICKENS, G.E. 1988. Non-Timber Use of Selected and Zone Trees and Shrubs in Africa. Rome, Italy, Food and Agriculture Organisation.

BRENAN, J.P.M. 1983. Manual on Taxonomy of Acacia Species. Present Taxonomy of Four Species of Acacia (A. Albida, A. Senegal, A. Nilotica, A. Tortilis). Rome, Food and Agriculture Organisation of the United Nations.

BREUSCH, T.S. and PAGAN, A.R. 1979. A Simple Test for Heteroscedasticity and Random Coefficient Variation. Econometrica 47(5): 1287-1294.

BROCH-DUE, V., Ed. 1999. Remembered Cattle, Forgotten People: The Morality of Exchange and the Exclusion of the Turkana Poor. The Poor Are Not Us: Poverty and Pastoralism in Eastern Africa. Oxford, James Currey.

BYRON, N. and ARNOLD, M. 1999. What Futures for the People of the Tropical Forests? World Development 27(5): 789-805.

CHIKAMAI, B. and GACHATHI, F.N. 1994. Gum and Resin Resources in Isiolo District, Kenya: Ethnobotanical and Reconnaissance Survey. East African Agricultural and Forestry Journal 59(4): 345-352.

CHIKAMAI, B. and ODERA, J.A., Eds. 2002. Commercial Plant Gums and Gum Resins in Kenya: Sources of Alternative Livelihood and Economic Development in the Drylands. Nairobi, Executive Printers.

CHIVEU, C.J., DANGASUK, O.G., OMUNYIN, M.E. and WACHIRA, F.N. 2008. Genetic Diversity in Kenyan Populations of Acacia Senegal (L.) Willd. Reveauled by Combined Rapd and Issr Markers. Arfrican Journal of Biotechnoloby 7(14): 2333-2340.

CHRETIN, M., CHIKAMAI, B., LOKTARI, P.E., NGICHILI, J., LOUPA, N., ODEE, D. and LESUEUR, D. 2008. The Current Situation and Prospects for Gum Arabic in Kenya: A Promising Sector for Pastoralists Living in Arid Lands. International Forestry Review 10(1): 14-22.

CUNNINGHAM, A.B., GERMAN, L., PAUMGARTEN, F., CHIKAKULA, M., BARR, C., OBIDZINSKI, K., VAN NOORDWIJK, M., DE KONING, R., PURNOMO, H., YATICH, T., SVENSSON, L., GAAFAR, A.M. and PUN TODEWO, A. 2008. Sustainable Trade and Management of Forest Products and Services in the Comesa Region: An Issue Paper. Bogor, Indonesia, Centre for International Forestry Research.

DEATON, A.S. and MUELLBAUER, J. 1986. On Measuring Child Costs: With Applications to Poor Countries. Journal of Political Economy 94(4): 720-744.

DEATON, A.S. and ZAIDI, S. 1999. Guidelines for Constructing Consumption Aggregates for Welfare Analysis. Working Papers 217, Princeton University, Woodrow Wilson School of Public and International Affairs, Research Program in Development Studies.

DIONE, M. 1989. Periode De Saignee Et Potentialites En Gomme Arabique De Quelques Localites De La Zone Gommiere Du Senegal. 3eme Symposium sur le Gommier et la Gomme Arabique (SYGGA III). St. Louis, Senegal: 117-126.

FADL, K.E.M. 2011. Effect of Methods of Tapping on Gum Yield of Acacia Seyal Var. Seyal in South Kordofan State, Sudan. Gum Arabic. J.F. Kennedy, G.O. Philips and P.A. Williams, Royal Society of Chemistry: 372.

FAFCHAMPS, M. and GAVIAN, S. 1997. The Determinants of Livestock Prices in Niger. Journal of African Economies 6(2): 255-295.

FAGG, C.W. and ALLISON, G.E. 2004. Acacia Senegal and Gum Arabic Trade. Oxford, Oxford Forestry Institute.

FAO. 1972. Production Yearbook 1971. Rome, Italy, Food and Agriculture Organisation of the United Nations.

--. 1985. Sand Dune Stabilisation, Shelterbelts and Afforestation in Dry Zones. Rome, Italy, Food and Agriculture Organisation.

FRATKIN, E. 2001. East African Pastoralism in Transition: Maasai, Boran, and Rendille Cases. African Studies Review 44(3): 1-25.

--. 2008. Pastures Lost: The Decline of Mobile Pastoralism among Maasai and Rendille in Kenya, East Africa. Economies and the Transformation of Landscape. L. Cliggett and C.A. Pool, Rowman Altamira: 327.

GACHATHI, F.N. and ERIKSEN, S. 2011. Gums and Resins: The Potential for Supporting Sustainable Adaptation in Kenya's Drylands. Climate and Development 3(1): 59-70.

GAULI, K. and HAUSER, M. 2011. Commercial Management of Non-Timber Forest Products in Nepal's Community Forest Users Groups: Who Benefits? International Forestry Review 13(1): 35-45.

GEIST, H.J. and LAMBIN, E.F. 2002. Proximate Causes and Underlying Driving Forces of Tropical Deforestation. Bioscience 52(2): 143-150.

GOK. 2007. Kenya Integrated Household Budget Survey 2005/06. Nairobi, Government of Kenya.

--. 2010. 2009 Population & Housing Census Results. Nairobi, Ministry of State for Planning, National Development and Vision 2030.

GRIMM, P. 2010. Social Desirability Bias. Wiley International Encyclopedia of Marketing, John Wiley & Sons, Ltd.

HARDIN, G. 1968. The Tragedy of the Commons. Science 162: 1243-1248.

HEALD, S. 1999. Agricultural Intensification and the Decline of Pastoralism: A Case Study from Kenya. Africa: Journal of the International African Institute 69(2): 213-237.

ICRAF. 1992. A Selection of Useful Trees and Shrubs for Kenya: Notes on Their Identification, Propagation and Management for Use by Farming and Pastoral Communities. Nairobi, International Centre for Research in Agroforestry.

ISLAM, A.M., PHILLIPS, G.O., SLJIVO, A., SNOWDEN, M.J. and WILLIAMS, P.A. 1997. A Review of Recent Developments on the Regulatory, Structural and Functional Aspects of Gum Arabic. Food Hydrocolloids 11(4): 493-505.

LAWLEY, D. and MAXWELL, A. 1971. Factor Analysis as a Statistical Method. London, Butterworth & Co.

LELON, J.K., JUMBA, I.O., KETER, J.K., WEKESA, C. and ODUOR, F.D.O. 2010. Assessment of Physical Properties of Gum Arabic from Acacia Senegal Varieties in Baringo District, Kenya. African Journal of Plant Science 4(4): 95-98.

LEWIS, I.M. 1999. A Pastoral Democracy: A Study of Pastoralism & Politics among the Northern Somali of the Horn of Africa, James Currey Publishers.

LIND, J. 2005. Relief Assistance at the Margins: Meanings and Perceptions of Dependency in Northern Kenya.

LITTLE, P.D., MCPEAK, J., BARRETT, C.B. and KRIST JANSON, P. 2008. Challenging Orthodoxies: Understanding Poverty in Pastoral Areas of East Africa. Development and Change 39(4): 587-611.

LITTLE, P.D., SMITH, K., CELLARIUS, B.A., COPPOCK, D.L. and BARRETT, C. 2001. Avoiding Disaster: Diversification and Risk Management among East African Herders. Development and Change 32(3): 401-433.

LYBBERT, T.J. and MCPEAK, J. 2011. Risk and Intertemporal Substitution: Livestock Portfolios and Off-Take among Kenyan Pastoralists. Journal of Development Economics 97(2): 415-426.

MARCHAL, R. 2009. A Tentative Assessment of the Somali Harakat Al-Shabaab. Journal of Eastern African Studies 3(3): 381-404.

MCPEAK, J. 2004. Contrasting Income Shocks with Asset Shocks: Livestock Sales in Northern Kenya. Oxford Economic Papers 56(2): 263-284.

MCSWEENEY, K. 2004. Forest Product Sale as Natural Insurance: The Effects of Household Characteristics and the Nature of Shock in Eastern Honduras. Society & Natural Resources 17(1): 39-56.

MKUTU, K.A. 2007. Small Arms and Light Weapons among Pastoral Groups in the Kenya-Uganda Border Area. African Affairs 106(422): 47-70.

NEUMANN, R.P. and HIRSCH, E. 2000. Commercialisation of Non-Timber Forest Products: Review and Analysis of Research. Bogor, Indonesia, CIFOR.

NISHINARI, K. and DOI, E., Eds. 1993. Food Hydrocolloids: Structures, Properties and Functions, Springer.

OBA, G. 1992. Traditional Grazing Systems of the Rendille and Ariaal Pastoralists: Changing Strategies and Options in the Southern District of Marsabit, Northern Kenya. Marsabit, Marsabit Development Program, GTZ.

OBUA, J., AGEA, J.G., NAMIREMBE, S. and EGADU, S.P. 2006. The Potential of Acacia Senegal for Dryland AgroForestry and Gum Arabic Production in Uganda. Journal of the Drylands 1(2): 186-193.

ONI, O. and GBADAMOSI, A.F. 1998. Progeny Variation in Seedlings of Dacryodes Edulis G. Don. Journal of Tropical Forest Resources 14(1): 38-47.

PIKE, I. 2004. The Biosocial Consequences of Life on the Run: A Case Study from Turkana District, Kenya. Human Organization.

PRATT, D.J. and GWYNNE, M.D. 1977. Rangeland Management and Ecology in East Africa /Edited by D. J. Pratt and M. D. Gwynne ; from Contributions by J. R. Blackie ... [Et Al.]. London: Hodder and Stoughton.

SADLER, K., KERVEN, C., CALO, M., MANSKE, M. and CATLEY, A. 2010. The Fat and the Lean: Review of Production and Use of Milk by Pastoralists. Pastoralism 1(2): 291-324.

SAHN, D.E. and STIFEL, D. 2000. Poverty Comparisons over Time and across Countries in Africa. World Development 28(1): 2123-2155.

SAHN, D.E. and STIFEL, D. 2003. Exploring Alternative Measures of Welfare in the Absence of Expenditure Data. Review of Income and Wealth 49(4): 463-489.

SCHWARTZ, H.J., SHAABANI, S.B. and WALTHER, D. 1991. Range Management Handbook of Kenya: Marsabit District. Nairobi, Republic of Kenya, Ministry of Livestock Development.

SHACKLETON, C.M. and SHACKLETON, S.E. 2004. The Importance of Non-Timber Forest Products in Rural Livelihood Security and as Safety Nets: A Review of Evidence from South-Africa. South African Journal of Science 100: 658-664.

SMITH, K. 1998. Sedentarization and Market Integration: New Opportunities for Rendille and Ariaal Women of Northern Kenya. Human Organization 57(4): 459-468.

SPENCER, P. 1973. Nomads in Alliance: Symbiosis and Growth among the Rendille and Samburu of Kenya. London, Oxford University Press.

SUNDERLIN, W.D., ANGELSEN, A., BELCHER, B., BURGERS, P., NASI, R., SANTOSO, L. and WUNDER, S. 2005. Livelihoods, Forests, and Conservation in Developing Countries: An Overview. World Development 33(9): 1383-1402.

TIMKO, J.A., WAEBER, P.O. and KOZAK, R.A. 2010. The Socio-Economic Contribution of Non-Timber Forest Products to Rural Livelihoods in Sub-Saharan Africa: Knowledge Gaps and New Directions. International Forestry Review 12(3): 284-294.

VAN DALEN, D. 2006. Arabische Gom. Amsterdam, Bert Bakker.

WATSON, D.J. and VAN BINSBERGEN, J. 2008. Livelihood Diversification Opportunities for Pastoralists in Turkana, Kenya. Ilri Research Report. Nairobi, International Livestock Research Institute. 5: 43.

WEKESA, C., MAKENZI, P.M., CHIKAMAI, B., LELON, J.K., LUVANDA, A.M. and MUGA, M.O. 2009. Gum Arabic Yield in Different Varieties of Acacia Senegal (L.) Willd. In Kenya. African Journal of Plant Science 3(11): 263-276.

WEKESA, C., MAKENZI, P.M., CHIKAMAI, B.N., LUVANDA, A.M. and MUGA, M.O. 2010. Traditional Ecological Knowledge Associated with Acacia Senegal (Gum Arabic Tree) Management and Gum Arabic Production in Northern Kenya. International Forestry Review 12(3): 240-246.

WHITE, F. 1983. The Vegetation of Africa. Paris, UNESCO.

WOLDEAMANUEL, T., ARTS, B., LEMENIH, M. and BONGERS, F. 2011. Dryland Resources, Livelihoods and Institutions: Diversity and Dynamics in Use and Management of Gum and Resin Trees in Ethiopia. Forest Ecology and Forest Management Group. Wageningen, Wageningen University. PhD.

WOOLDRIDGE, J.M. 2006. Introductory Econometrics: A Modern Approach. Mason, OH, Thomson/South-Western.

(1) These counties were called districts before an administrational change in 2009. Currently, each county consists out superfluous of several districts.

(2) See Oba (1992) for a complete description of the study.

(3) Deutsche Gesellschaft fur Technische Zusammenarbeit.

(4) A more detailed description and data are available online at http://earlywarning.usgs.gov/fews/

(5) In addition, rainfall shortage causes a drop in productivity of the rain-fed home gardens, which are becoming more common in the region due to increased sedentarisation (Smith 1998, Little et al. 2001).

(6) In fact, several authors found that in Kenya only a small part of available gum arabic is collected (Chikamai and Odera 2002, Chretin et al. 2008), ameliorating such concerns.

(7) Omitting livestock prices from the model could also cause omitted variable bias, which is (indirectly) tested for using Breusch-Pagan tests.

(8) See Lawley and Maxwell (1971) for an explanation of the technique.

(9) TLUs are constructed based on the metabolic weights of animals. It was found to be a good base of comparison for animals of different species, whether the variable of interest was feed intake, manure produced, or product produced. The following conversion factors were considered appropriate for the Sahel region: 1 TLU = 1 cattle = 1.25 camels = 10 smallstock (Dahl and Hort 1976). For donkeys, a factor of 1 TLU = 2 donkeys was used (Houerou and Hoste 1977).

(10) Polygamy is common, and because each wife traditionally has her own house, several households had multiple houses in their homestead or 'Manyatta' (Spencer 1973).

(11) In the sample, 73% of households stayed within three hours walking distance from the village centre.

W. VELLEMA [a,b], G. MUJAWAMARIYA [b], M. D'HAESE [a] and K. BURGER [b]

[a] Department of Agricultural Economics, Ghent University, Coupure Links 653, 9000 Gent, Belgium

[b] Department of Development Economics, Wageningen University, Hollandseweg 1, 6706 KN Wageningen, The Netherlands

Email corresponding author: w.vellema@ugent.be

Other emails: rgaudiose@gmail.com, marijke.dhaese@ugent.be, kees.burger@wur.nl

TABLE 1 Household-specific descriptive statistics

Variable                     Mean      SD

Income                      2082.0   3558.4
Price (KSH)                  38.8     7.6
Yearly quantity (kg)         44.3     19.8
Exp. daily quantity (kg)     3.2      2.4
Adult equivalent HH size     3.3      1.2
Women collecting             1.2      1.1
Men collecting               1.1      1.1
Distance in minutes         151.0    112.5
% tap                        14%       -
% migrating                  56%       -
Duration of migration *      4.3      2.2
% mentioning insecurity      21%       -

Note: n = 201, * of migrating households. Data source: Acaciagum
Project

TABLE 2 Household asset ownership

Variable              Mean    SD     Min    Max

Houses                1.08   0.36     0      4
Radio                 0.15   0.37     0      2
Telephone             0.09    -       0      1
Transport means       0.07    -       0      1
Years of education    0.59   2.02     0      12
TLU                   4.73   7.65     0      62
TLU per adult (AE)    1.43   2.39     0      17
Wealth index          0.00   0.80   -0.54   3.27

Note: n = 201. Data source: Acaciagum project

TABLE 3 Reported sources of income

Income source         Primary   Secondary

Commerce                 5          6
Agriculture              0          0
Pastoralism             154        32
Wage                     1          4
Wood products            2         74
Arabic gum              33         68
Remittances              0          0
Relief                   9         19
Non-wood products        0          1

Note: n = 201. Data source: Acaciagum Project

TABLE 4 Determinants of marketed quantities

                                          Regressions

Variables      Unit             (1)           (2)          (3)

Price          KSH per kg    -0.007       -0.011        -0.011
                             (-0.78)      (-1.21)       (-1.31)

Household      Adult         0.006        0.007         -0.002
size           equivalents   (0.08)       (0.11)        (-0.03)

Wealth         Index         0.816 ***    0.694 ***     0.555 **
                             (3.15)       (2.68)        (2.23)

Wealth         Index         -0.404 ***   -0.351 ***    -0.306 ***
squared                      (-3.29)      (-2.89)       (-2.64)

Female         Number        0.162 *      0.085         0.049
collectors                   (1.96)       (1.01)        (0.61)

Male           Number        -0.109       -0.059        -0.092
collectors                   (-1.33)      (-0.72)       (-1.17)

Distance       Minutes       0.000        0.000         0.001
                             (0.56)       (0.58)        (1.12)

Tapping        Dummy         -0.322       -0.238        -0.102
                             (-1.31)      (-0.99)       (-0.43)

Migrating      Dummy         -0.202       -0.391 **     -0.441 **
                             (-1.12)      (-2.12)       (-2.50)

Acacia         Ratio                                    2.305 ***
area                                                    (2.71)

Rainfall       Millimetres                              0.014 ***
shortage                                                (4.36)

Insecurity     Dummy         -0.673 ***   -0.298        -0.398 *
                             (-3.06)      (-1.26)       (-1.72)

Turkana        Dummy                      -0.912 ***    -0.820 ***
                                          (-3.68)       (-3.04)

Samburu        Dummy                      -0.229        -0.580 ***
                                          (-1.05)       (-2.61)

Somali         Dummy                      -0.408        -0.524
                                          (-0.95)       (-1.29)

Constant                     3.914 ***    4.525 ***     3.748 ***
                             (7.43)       (8.39)        (6.94)

Observations                 201          201           201

Adjusted                     0.082        0.134         0.218
R-squared

Note: Dependent variable is the natural logarithm of marketed
quantity per household.

T-statistics in parentheses: * Significant at 10%;
** significant at 5%; *** significant at 1%
COPYRIGHT 2013 Commonwealth Forestry Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2013 Gale, Cengage Learning. All rights reserved.

 
Article Details
Printer friendly Cite/link Email Feedback
Author:Vellema, W.; Mujawamariya, G.; D'Haese, M.; Burger, K.
Publication:International Forestry Review
Article Type:Report
Geographic Code:6KENY
Date:Jun 1, 2013
Words:9526
Previous Article:A choice experiment approach for assessing preferences to forest law configuration and compliance: the case of NTFP traders in Cameroon/Choix d'une...
Next Article:Restoring Tropical Forests: A practical guide.
Topics:

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters