Few studies have been carried out in the area of smallholder livestock keepers’ breeding choice with largely never discuss anything about these smallholder decisions on choices of the type of cattle to be reared and its implication in poverty reduction, this increases demands for knowledge on the discipline and methodological attributes. However, the available few studies have paid attention to livestock and poverty (Engida et al. 2015; Grace et al. 2016 and Konga, 2014) while others have focused on livestock and food security (IFAD, 2016) while few have discussed breeding preferences (Mutenje et al., 2020 and Martin-Collado, 2018)
However, Mujibi et al. (2019) highlight that nearly 40% of the entire agricultural contribution to GDP in African countries is due to livestock activities which are highly dominated by the smallholder keepers. The contribution is higher in individual countries as it ranges between 30–80%. Additionally, Sub-Saharan African countries alone constitute about 450 million smallholder livestock keepers who are engaged in mixed farming activities; and this group accounts for nearly half of the entire livestock production in the continent although the majority of these livestock keepers have remained to be poor with many more being under extremely poor situations.
Moreover, annual revenues received from the livestock sector in the East African region has reached US dollar 1 billion as receipts from exports from a total of more than 2 billion cattle (Michael et al. 2018). Nonetheless, the general livestock population in the SADC region is projected to be 529,000,000 with 75% of the livestock population being kept under a smallholder traditional farming system (SADC, 2020).
The global share of Tanzania’s cattle population and production stands at 1.40% and 11% respectively (FAO, 2016). The Tanzania Livestock Survey of 2016/17 presents that cattle and goats are the leading animals kept in Tanzania with 28.401 million and 16.67 million respectively while thenumber of sheep and pigs being 5 million and 2 million respectively. At these figures, the recent share of livestock in the GDP has reached 7.4% with the sector being one of the slowest growing annually at the rate of 2.3%. Major constraints associated with the sector includes the types of breeding animals kept by the majority of livestock keepers which has low production and poor diseases resilience that most can not survive during hot and dry seasons as well as high mortality (Michael et al. 2018). Despite its low contribution to the GDP and its low growth rate yet Tanzania livestock’s population is ranked 3rd in Africa after Sudan and Ethiopia (Engida et al. 2015). Furthermore, 99.9% of Tanzania’s livestock are kept by smallholder farmers living the contribution of a large-scale farm very insignificant (Engida et al. 2015).
Table 1
Annual increase in cattle population in Tanzania
Productions zone | 2016/2017 | 2017/2018 | 2018/2019 | 2019/2020 | 2020/21 | 2021/2022 | % Change |
Traditional system |
Central | 13,102,022 | 14,098,320 | 14,632,283 | 15,186,470 | 15,761,646 | 16,358,606 | 20% |
Coastal and Lake | 11,560,207 | 12,301,694 | 12,626,411 | 12,959,700 | 13,301,786 | 3,652,901 | 14% |
Highlands | 3,773,606 | 4,095,903 | 4,288,036 | 4,489,182 | 4,699,763 | 4,920,222 | 26% |
Total | 28,435,835 | 30,495,917 | 31,546,730 | 32,635,351 | 33,763,194 | 34,931,729 | 18% |
Ranching System |
Central | 12,330 | 12,988 | 13,682 | 14,413 | 15,182 | 15,993 | 30% |
Coastal and Lake | 19,297 | 19,525 | 19,755 | 19,988 | 20,224 | 20,463 | 6% |
Highlands | 41,400 | 46,037 | 51,193 | 56,927 | 63,303 | 70,393 | 70% |
Total | 73,027 | 78,550 | 84,630 | 91,328 | 98,709 | 106,848 | 46% |
Cattle in feedlots |
Feedlots | 78,111 | 115,878 | 171,905 | 255,020 | 378,323 | 561,242 | 619% |
Dairy Subsector | 260,293.01 | 315,888.3 | 383,357 | 465,236 | 564,603 | 685,191 | 163% |
Total | 338,404 | 431,765 | 555,261 | 720,255 | 942,924 | 1,246,432 | 268% |
Source: Tanzania Livestock Master Plan (TLMP) 2017/2018 – 2021/2022 |
Nonetheless, due to persistent income poverty among smallholder livestock keepers in Tanzania, only less than 1/3 of all family-owned livestock is vaccinated. Additionally, an average of 60% of all animals is reported to have some type of disease with only 6% of rural livestock holders can hire labour while the rest depends on the family workforce; these facts justify that livestock is highly characterized with poverty in Tanzania (Michael et al. 2018).
To enhance smallholders’ livestock keepers income and food security status number of breeding programs were established in Tanzania (TLMP, 2021). However, most of the smallholder farmers have been found to choose breedings based on the production of the cattle breeds, diseases resilience and environmental adaptability most often on heat and dry season (Kearney and White, 2016).
Studies of non-breeding choices and production show that indigenous or traditional breeds have the lowest milk production (Marshall et al. 2019). However, these traditional breeds have often been crossed to produce breeds with varying (Mujibi et al. 2019). In Tanzania, the main breeds that have the largest share in terms of the number of animals kept and beef are Shorten zebu (80%) and ankole (14%). Generally in Tanzania, the traditional breeds of cattle constitute about 94% of the entire meat produced in a country and commercialized ranches makes only 6% (TLMP, 2021).
Table 2
Traditional and Improved Cattle Productivity in Tanzania
Parameter | Traditional Cattle breed | Improved Cattle breed | Smallholder | Ideal Standard |
Calving rate | 30.00% - 50.00% | 55% - 73% | 40.00% - 50% | 80.00% |
Calving interval (months) | 18.00 – 24.00 | 15.00 – 21.00 | 17.00 – 18.00 | 12.00 |
Age at first calving (months) | 36.00 – 48.00 | 30.00 – 36.00 | 43.00 – 46.00 | 27.00 – 30.00 |
Pre weaning mortality (%) | 25.00% - 40.00% | 4.30% | 5.00% - 6.00% | < 5.00% |
Calf mortality (%) | >25.00% | 3.30% | 5% - 6% | <10.00% |
Adult mortality (%) | 8.00% - 10.00% | 1.30% | <1.00% | <5.00% |
Mature weight (kgs) | 200 – 300 | 250 – 350 | - | 300 – 500 |
Lactation yield | 160.0 – 250.0 | 2800– 3500 | 1500 – 2000 | 2500 – 3500 |
Lactation length | 200.00 | 300.00 | 270.00 – 300.00 | 305.00 |
Source: Tanzania Livestock Master Plan 2017/2018 – 2021/2022 |
Table 2 justifies that enhancing the adaptation of new breeds among smallholders livestock keepers in Tanzania will help in increasing productivity and income among smallholder livestock keepers and hence reducing poverty levels among these groups. Therefore, this paper helps to examine determinants of smallholders livestock keepers choice (decision) on the breeding type of cattle and its implication in the poverty reduction strategies in Tanzania by considering that each breed has unique characteristics in terms of production which are highly associated with the household income and wealth.
Theoretical foundation
The discrete choice model has been used in this paper as an application to the random utility theory in modelling livestock keepers utility-maximizing constraints (Mutenje et al. 2020). The basics for the random utility theory has been adopted from Lancaster Consumer Theory which explains utility that is always derived or obtained from attribute which can never be bought independently. (Lancaster, 1966). Therefore, considering that individuals value differently the quality of a commodity hence it is impossible to determine general utility rather than individual satisfaction feedback. However, these farmers are always rational towards the choice of breeding that gives them the highest utils obtained from utilizing these breeds
To show this, this paper has assumed that smallholder livestock keepers satisfaction depends on choices made from the available set of choices for breeding types of cattle \(\left(j\right)\), which later provide a general utility function that:
$${U}_{ij}={V}_{ij}+{\epsilon }_{ij}, j=\text{1,2},\dots ..j\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots .\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots .1$$
whereas, for a smallholder livestock keeper \(i\), a one will be indifferent at \(U\) based on the cattle breeding choice \(j\). Therefore, this theory divide utility into deterministic part \(\left(V\right)\) and an unobservable part \(\left(\epsilon \right)\) and smallholder livestock keepers (consumers) are regarded as rational as they will always pick the breeds that provide them with the highest utility.
Analytical Model
In order to examine the most likely explanations for the household’s utility of specific types of cattle breeds among different existing breeds, this paper uses the multinomial logit model (Maddala, 1983). The choice of the model has been based on other significant facts like the ability of the MNL to use the cumulative distribution function of the logistic distribution and it has been widely used in similar studies. Consider a regressand variable Y with only two choices (dichotomous) and with its regressor X, and therefore;
$$\pi \left(x\right)= p\left(Y=1\right|X=x)=1-p\left(Y=0|X=x\right)\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots ...2$$
Thus simple model for logistic regression will be given by the equation;
$$Logit\left[\pi \left(x\right)\right]=log\left(\frac{\pi \left(x\right)}{1-\pi \left(x\right)}\right)=\alpha +\beta x \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots 3$$
Whereas the odds will always be given by;
$$Odds=\frac{\pi \left(x\right)}{1-\pi \left(x\right)}\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots .\dots \dots \dots \dots \dots \dots \dots \dots 4$$
Therefore the logarithm of the odds is called logit which is hereby given by
$$Logit\left[\pi \left(x\right)\right]=log\left(\frac{\pi \left(x\right)}{1-\pi \left(x\right)}\right)=\text{log}[\text{exp}(\alpha +\beta x\left)\right]= \alpha +\beta x\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots ..5$$
When there are multiple cattle breeding choices the model can be extended as follows;
Let \(k\) represents the number of predictors of the binary dependent variable \(Y\) that\({ x}_{1}, {x}_{2}, {x}_{3}\dots \dots {x}_{k}\). Hence the model for the log of odds is given by;
$$logit \left[P\left(Y=1\right)\right]=\alpha +{\beta }_{1}{x}_{1}+{\beta }_{2}{x}_{2}+\dots +{\beta }_{k}{x}_{k}\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots .6$$
With the alternatively direct expression being
$$\pi \left(x\right)=\frac{\text{exp}\left(\alpha +{\beta }_{1}{x}_{1}+{\beta }_{2}{x}_{2}+\dots +{\beta }_{k}{x}_{k}\right)}{1+\text{exp}\left(\alpha +{\beta }_{1}{x}_{1}+{\beta }_{2}{x}_{2}+\dots +{\beta }_{k}{x}_{k}\right)}\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots .\dots \dots \dots \dots \dots 7$$
Whereby \({\beta }_{i}\) represents effects explanatory variable \({x}_{i}\) on log-odds that \(Y=1\), while controlling other explanatory variables \({x}_{k}\), therefore \(\text{e}\text{x}\text{p}\left({\beta }_{i}\right)\) become a multiplicative effect on odds of a unit increases on the explanatory variable \({x}_{i}\), when all other variables \({x}_{k}\)are constant.
Therefore, when there are \(n\) observations, \(p\) independent variables, and \(k\) categorical responses in the given function, the ideal behind constructing multinomial logit is by making one of the responses as a base outcome of which all other remaining categories will be constructed relatively to it and all responses are not ordered hence any of them can be a base outcome. To simplify these explanations, consider \({\pi }_{j}\) as a multinomial probability of observations falling into \({j}^{th}\) category with \(p\)explanatory variables,\({x}_{1}, {x}_{2}, {x}_{3}\dots \dots {x}_{p}\)
Therefore, the multiple logistic regression model is given by;
$$log\left[\frac{{\pi }_{j}\left({x}_{i}\right)}{{\pi }_{k}\left({x}_{i}\right)}\right]={\alpha }_{0i}+{\beta }_{1j}{x}_{1i}+{\beta }_{2j}{x}_{2i}+\dots +{\beta }_{pj}{x}_{pi}\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots ..\dots \dots ..8$$
Whereas \(j=1, 2,...,(k-1), i=1, 2, ..., n.\)However, \(\pi \text{'}s\) add to unity therefore the equation is reduced to;
$$log\left({\pi }_{j}\left({x}_{i}\right)\right)=\frac{\text{e}\text{x}\text{p}({\alpha }_{0i}+{\beta }_{1j}{x}_{1i}+{\beta }_{2j}{x}_{2i}+\dots +{\beta }_{pj}{x}_{pi})}{1+\sum _{j=1}^{k-1}\text{e}\text{x}\text{p}({\alpha }_{0i}+{\beta }_{1j}{x}_{1i}+{\beta }_{2j}{x}_{2i}+\dots +{\beta }_{pj}{x}_{pi})}\dots \dots \dots \dots \dots .\dots \dots \dots \dots \dots \dots .9$$
For\(j = 1, 2, \dots , (k-1)\), whereas parameters will be estimated by the use of maximum likelihood.