Impact of feed shortages on livestock and crop production in Ethiopia: implications for rural poverty reduction

Abstract

Livestock productivity is low in Africa, primarily due to feed shortages. We quantify the impacts of the feed shortage experience on livestock and crop production, and its implications on poverty using an instrumental variable approach. Results revealed that the feed shortage experience increased the value of livestock deaths (14 per cent), increased production expenses (77 per cent) and reduced the value of crop production (4 per cent) by reducing investment in modern inputs and increasing the likelihood of livestock being affected by diseases. The income lost due to feed shortages would have lifted 2.57 per cent of affected households out of poverty. Increasing feed availability could, therefore, be pro-poor.

1. Introduction

Livestock provide income, food, employment, energy, draught power, social value and manure for rural livelihoods in sub-Saharan Africa (Acosta, Nicolli and Karfakis, 2021; Alary, Corniaux and Gautier, 2011; Herrero et al., 2013; Nilsson et al., 2019). These animals also serve as a buffer against economic shocks (Behnke and Wolford, 2010; Hänke and Barkmann, 2017; Herrero et al., 2013; Kato et al., 2011; Smith et al., 2013). However, the productivity of the sector remains low (Mayberry et al., 2017, 2018). Feed shortage is arguably one of the major constraints that slow livestock productivity (Duncan et al., 2023; Mayberry et al., 2017, 2018; Paul et al., 2021). However, the impact of feed shortage has not been quantified rigorously.

This paper estimated the economic impacts of the feed shortage experience on livestock and crop production among smallholder farmers in Ethiopia. The study first investigated the household-level impacts of experiencing feed shortage on the value of livestock deaths, livestock production expenses, and the value of crop production. Household-level economic losses and the poverty implications of experiencing feed shortage were then estimated. The quantification of these various impacts can provide insights for cost–benefit analyses and related policy interventions.

Feed shortages impact farmers’ welfare by reducing livestock and crop production in several ways (Hadush, 2019, 2020; Liversage and Rota, 2020). For example, such shortages may lead to conflicts among households and communities over feed resources (Banjade and Paudel, 2008; Liversage and Rota, 2020; Schilling, Opiyo and Scheffran, 2012). The feed shortage experience also pushes farmers and their livestock to risky areas in search of feed, exposing them to diseases, predators and accidents, among other things. Moreover, feed shortage experience may lead to poor feeding, inducing farmers to buy feed and medicine to treat poorly fed weak or sick livestock, diverting resources away from education or other productive investments, which may have lasting consequences. Furthermore, feed shortages may impact crop production by reducing the availability of draught power, manure supplies and crop residues, while decreases in the latter two affect soil fertility (Jaleta, Kassie and Shiferaw, 2013). Time taken searching for feed could compete with time needed for crop production, off-farm activities, household care and leisure (Hadush, 2019). By reducing the income that farmers generate from livestock production, feed shortages can simultaneously reduce investment in modern inputs such as improved seeds, pesticides and inorganic fertilisers. Finally, feed shortages could lead to stiff competition between livestock and crop production as they may induce farmers to repurpose land allocated to crop production for fodder production.

Despite its significance, the impact of feed shortages on livestock and crop production is often overlooked. An exception is Hadush (2019, 2020), who estimated the economic loss of time spent in search of feed and water in Tigray, Ethiopia. He found that reducing the time spent on collecting feed and water increased food production. Other studies also suggested that the shortage of feed was a major challenge that smallholder farmers faced (Amsalu and Addisu, 2014; Hassanuur, Netsanet and Merga, 2020; Tsegaye, Tolera and Berg, 2008; Zewdie and Yoseph, 2014). However, to the best of this study’s authors’ knowledge, no research currently exists that directly quantifies the impacts of feed shortages on livestock and crop production or their economic and poverty implications. This study, therefore, presents empirical evidence on the scale of economic losses due to feed shortages. A unique and large sample size was used, namely a balanced panel of 5,725 smallholder farmers surveyed over three rounds in four major regions of Ethiopia.¹ The resultant panel data enabled controlling for unobserved time-invariant household-level heterogeneity.

Given the growing interest in addressing feed shortages, the findings of this study provide valuable insights about investing in livestock feed (Herrero et al., 2010; Nabarro and Wannous, 2014; Ravichandran et al., 2020). In a sub-Saharan African context, where per-capita land ownership is declining significantly, addressing feed shortages among smallholders is vital not only for meeting the growing demand for meat in the region, but also for reducing the welfare impact of feed shortages on smallholder farmers (Bachewe and Minten, 2023; Jayne, Mather and Mghenyi, 2010; Latino, Pica-Ciamarra and Wisser, 2020; Mensah et al., 2021; Steinfeld et al., 2006). The findings of this study could also contribute to policy design and programme development in respect of livestock feed in the region by highlighting the economic losses of feed shortages. Furthermore, the results may provide insight into the costs and benefits of interventions that support fodder production (Bediye, Nemi and Makkara, 2018).

Empirically, this study relates directly to those studies by Hadush (2019, 2020). However, there are differences between this study and its predecessors. To estimate the impacts of feed shortages, Hadush (2019, 2020) used the time that farmers took to look for feed and water as a proxy for feed shortages. However, some farmers might succeed in meeting their feed demand in this way, but others might not. Hence, searching for feed (and water) does not directly measure the impacts of feed shortages. The current study therefore used farmers’ self-reported experience of feed shortages, and an objective indicator of such shortages defined herein as a feed gap to estimate their impact. The feed gap is the difference between feed demand and feed produced by the household. By directly addressing what the various impacts that experiencing a shortage of feed have on livestock and crop production, this paper contributes to the scant literature about livestock feed challenges. Another valuable contribution of the study is its use of panel data to address time-invariant heterogeneities. A third addition to the literature was the use of an instrumental variable (IV) approach to control for the potential endogeneity problem arising from reverse causality between feed shortage experience and the outcome variables of interest.

The rest of this paper is structured as follows: Section 2 presents the study context and conceptual framework, while Section 3 describes the data. The estimation methods used to quantify the economic effects of the feed shortage experience are discussed in Section 4. The results are presented in Section 5, and Section 6 offers some concluding remarks.

2. Context and conceptual framework

2.1. Livestock production in Ethiopia

Ethiopia possesses abundant livestock resources, with about 70 million head of cattle, 57 million domesticated birds, 53 million goats, 43 million sheep, 11 million donkeys, 8 million camels, 2 million horses and 0.38 million mules in 2020 (CSA, 2021). This places the country among the top 10 holders of livestock globally (FAO, 2019). The livestock sector contributes significantly to Ethiopia’s economy, accounting for around 45 per cent of the agricultural gross domestic product and 20 per cent of exports, while providing a means of livelihood for millions of people (Behnke, Wolford and Metaferia, 2011). However, the sector faces various challenges, including feed shortages and high morbidity and mortality rates (CSA, 2021).

The shortage of feed remains a critical challenge in the country (Balehegn et al., 2020; Bediye, Nemi and Makkara, 2018; Management Entity, 2021; Yibeltal, 2023). The causes of such shortages are many and varied; they include bad weather and other shocks such as death or the serious illness of a family member as well as social and political conflicts (Cheng, McCarl and Fei, 2022; Hannah, 2022; Yibeltal, 2023), communal and private grazing land shortages (Haile, 2016), farmers’ preference to utilise crop residues for feed or other purposes (Jaleta, Kassie and Shiferaw, 2013) and farmers’ preferences in allocating time to look for feed and for doing other things.

Commercial livestock feeds are expensive for many smallholder farmers (Business Info Ethiopia, 2022; Gebremedhin, Hirpa and Berhe, 2009; Lawrence et al., 2008; Tegegn, 2021). Consequently, despite the over 240 million head of livestock in the country, farmers still employ traditional feeding practices but enjoy only limited livestock extension services (MOANR, 2017). Typical traditional feeds are green fodder (grazing) and crop residues (Figure 1). Less than 1 per cent of farmers use improved feeds such as desmodium, alfalfa and brachiaria, which could partly explain why the productivity of the sector is low and why millions of livestock units die every year.

2.2. Conceptual framework

The dominant agricultural practice in Ethiopia is mixed farming by smallholders who rear livestock and produce crops. These farmers’ production is affected by feed shortages through various channels (Figure 2). For example, feed shortages may lead to livestock being poorly fed, their productivity being reduced or the increased likelihood of them dying from organ failure or low immunity to diseases. Another channel is that feed shortages may induce farmers to consider risky areas for grazing, such as those exposed to diseases or physical hazards. Moreover, feed shortage experience may lead to poor feeding, inducing farmers to buy feed and medicine to treat poorly fed weak or sick livestock. There are also other channels through which feed shortages affect livestock production as shown in the figure.

Similarly, feed shortages affect crop production through various channels, since livestock play a key role in crop production in Ethiopia as a source of draught power, manure and revenue to purchase inputs (Behnke, Wolford and Metaferia, 2011). Feed shortages may induce farmers to reallocate land used for crop production to grow feed instead. Another possible channel is that farmers may overutilise crop residue to use as feed (Jaleta, Kassie and Shiferaw, 2013), which could reduce soil fertility. Moreover, farmers who experience feed shortages may have to allot more labour to look for feed (Hadush, 2019) and more money to purchase feed. This reduces both labour and money that could be dedicated to crop production.

3. Data and descriptive statistics

3.1. Data sources

This study utilised panel data collected by Ethiopia’s Agricultural Growth Program (AGP-I).² The associated surveys were conducted between 2011 and 2017 by a consortium of partners in Ethiopia, namely the Central Statistical Agency (CSA), the Ministry of Agriculture, the International Food Policy Research Institute, and the Ethiopian Development Research Institute (currently known as the Policy Studies Institute). The data were collected from 93 districts (woredas) from four regions: Amhara, Oromia and Tigray, as well as the Southern Nations, Nationalities and Peoples’ Region (see the sample distribution on Fig. 3). The data represented 15 million smallholder farmers in the country.

The sample size included 7,929 randomly selected households in 2011, 7,503 in 2013 and 7,117 in 2017. The selected households were smallholder farmers who practiced mixed farming. After accounting for attrition and households who did not own livestock, a balanced panel with a sample size of 5,725 households was used for the current study. Poultry production was excluded from all analyses as it is less likely to be directly affected by feed shortages. The potential biases associated with the dropped households are addressed in Section 4.

Farmers were asked whether or not they had experienced a shortage of feed in the 12 months preceding each respective survey. They were also asked to assess the severity of feed shortages according to three categories, i.e. the household was (i) not affected, (ii) moderately affected or (iii) severely affected by feed shortages in the 12 months preceding the survey concerned. This self-reported feed shortage experience was used as a key independent variable of interest for the analysis. To check the robustness of the self-reported feed shortage, the quantity of feed gap, which is an objective indicator of the feed shortage experience, was computed. The feed gap was calculated using methods proposed in the existing literature (Assefa, Nurfeta and Banerjee, 2013; Ayele et al., 2021; Debela, Animut and Eshetu, 2017; Demeke, Mekuriaw and Asmare, 2017).

The Feed gap variable was calculated as the difference between feed demand and feed supply. Feed supply for a household was computed from crop residues, fallow land, private grazing land, and land covered by permanent trees and privately owned by the household. Feed supply from crop residues was obtained by first dividing grain production by the weight assigned for each crop type and then subtracting the grain production using the harvest index method proposed by Lal (2005).³ The average weight assigned was 0.35 for cereals, 0.36 for pulses and 0.61 for tubers (Lal, 2005). Other conversion factors used include 3 t/ha for private grazing land, 1.8 t/ha for fallowed private land and 1.2 t/ha for private land covered by permanent crops (Ayele et al., 2021). Also, in accordance with Ayele et al. (2021), the annual feed demand by farmers was computed as a product of three parameters: 6.25 kg of dry-matter feed per day per tropical livestock unit (TLU), total number of livestock owned by the household (in TLU) and the number of days feed was needed for the animals in the prior 12 months (365 days) concerned.

The data set contains detailed information on livestock and crop production. Households reported livestock deaths, livestock production expenses and crop production during the 12 months preceding each survey year, as well as the number of livestock they had at the end of each survey year in question. Furthermore, the data set contains information on each household’s socio-economic and socio-demographic characteristics, which allows for controlling associated heterogeneities. All the monetary values in this paper were computed using the 2011 price to control for differences due to price changes over time.

3.2. Descriptive statistics

Overall, 39, 33 and 36 per cent of the households experienced a feed shortage in 2011, 2013 and 2017, respectively (Figure 4). Conditional on reporting experiencing a feed shortage, farmers were asked about the severity of the feed shortage experience. The percentage of farmers who experienced a severe feed shortage slightly decreased across the survey years, declining from 17 per cent in 2011 to 12 per cent in 2013 and to 10 per cent in 2017, whereas the percentage of households who experienced moderate feed shortage increased in 2017.

Table 1 reveals the persistence of experiencing a feed shortage across the relevant survey years. About 71 per cent of the farmers had experienced a feed shortage at least once in the three years in question. This result could indicate that feed shortage is a major livestock production constraint for the studied households. Moreover, about 7 per cent of the households had experienced a feed shortage in all three survey years. In two of the survey rounds, about 25 per cent of the households said they had experienced such a shortage. However, some 40 per cent of households reported in one specific survey round that they had experienced a shortage. Furthermore, about 33 per cent of the farmers reported having experienced a severe feed shortage in the 12 months preceding at least one of the survey years, while 53 per cent of the households had experienced a moderately severe shortage in at least one of the survey years.

Figure 5 presents sample farmers’ feed demand, own feed production and the feed gap. The feed gap (demand minus production) shows that farmers were unable to meet feed requirements for their livestock from their own production. This unmet demand could result in low livestock productivity and high mortality unless it is covered by purchases (a rare practice in Ethiopia: see Figure 1) or via open grazing land. However, the latter may not fill all the unmet demand for feed because land is already scarce due to high population growth.

The study also found a remarkable consistency between the self-reported feed shortage and the feed gap estimate (Figure 6). Throughout the two distributions, the feed gap was higher for farmers who had experienced a feed shortage than for farmers who had not. When this difference was subjected to the first-order stochastic dominance test (Kaplan, 2019), the result was statistically significant (p-value = 0.0001), indicating the consistency between the self-reported feed shortage and the objectively measured feed gap.

Table 2 presents the descriptive statistics of the outcome variables. The results show that there is a statistically significant correlation between a feed shortage experience and the key outcome variables. For instance, farmers who experienced feed shortages were more likely to lose livestock due to death. Moreover, the value of livestock deaths is higher for farmers who reported having had a feed shortage experience than for those who had not. The value of livestock deaths and livestock production expenses also increased with the severity of the feed shortage experience, whereas the value of crop production decreased with such severity. Overall, the results show that experiencing a feed shortage is negatively correlated with livestock and crop production outcomes.

4. Empirical estimation strategy

4.1. Specification

By exploiting the panel nature of the data, the study estimated the effects of experiencing a feed shortage on the outcome variables as follows:

where |${Y_{it}}$| denotes the outcome variables, i.e. value of livestock deaths (defined as the total heads of livestock that died in the 12 months prior to the survey date multiplied by the price of each livestock type), out-of-pocket livestock production expenses in the 12 months prior to the survey date and value of crop production (defined as the quantity of each crop produced in the 12 months prior to the survey date multiplied by the corresponding price of each crop) of household |$i$| at time |$t$|⁠. The value of livestock deaths and production expenses were subjected to the inverse hyperbolic sine (IHS) transformation to account for zero values (Bellemare and Wichman, 2020). The IHS transformation was required because about 77 per cent of the households in the sample reported no livestock deaths, while 49 per cent of the households did not incur expenses in livestock production. As a robustness check, when the Tobit model was considered without transforming the dependent variables, consistent results were found (see Table A1 in supplementary data at ERAE online). |${G_{it}}$| is a dummy variable that equals one if household |$i$| experienced a feed shortage at time |$t$|⁠, and zero otherwise. In a separate estimate, |${G_{it}}$| represents a vector denoting the severity of the feed shortage experience, namely whether the household was not affected (comparison group), was moderately affected, or was severely affected by a feed shortage. |${X_{it}}$| is a vector of explanatory variables, selected separately for each outcome variable, in accordance with the literature and considering the context of the study (Abro et al., 2023; Do, Nguyen and Grote, 2017; Huttner et al., 2001). |${S_{it}}$| denotes a dummy variable that equals one if household |$i$| experienced shocks such as flood, drought, crop production being affected by shocks and livestock being affected by diseases at time t; the dummy variable equals zero otherwise. |${T_i}$| is a vector of time fixed effects. |${\psi _i}$| denotes household-level time-invariant unobserved heterogeneities. The parameters |$\alpha $|⁠, |$\beta $|⁠, |$\delta $|⁠, |$\,\gamma $| (a vector) and |$\zeta $| are parameters to be estimated, while |${\varepsilon _{it}}$| represents the error term, clustered at district (woreda) level. The summary of the variables used in the estimation are reported in Table 3.

The primary interest in Equation (1) is |$\beta $|⁠. If the experience of a feed shortage has an impact on values of livestock deaths and livestock production expenses, |$\beta $| will be positive and statistically significant. If the experience of a feed shortage has a negative impact on crop production, |$\beta $| will be negative and statistically significant.

Moreover, by exploiting the presence of detailed information in the data on input used for crop production, the study also estimated the channels through which experiencing a feed shortage affected livestock and crop production. The following model was estimated to understand the impact that experiencing a feed shortage had on these channels:

In Equation (2), |${Z_{it}}$| denotes channels through which a feed shortage experience affects livestock and crop production. These channels include the probability of oxen deaths (oxen are key draught animals in Ethiopia), the probability of equine deaths (horses, mules and donkeys, for example, are a key means of transport and draught power in rural Ethiopia), the probability that livestock are affected by diseases, the amount of inorganic fertiliser used in crop production, the amount of money spent on purchasing improved seeds, the number of labour days used for crop production and the probability of using pesticides in crop production. All the other variables are as defined in Equation (1).

4.2. Identification strategy

Equations (1) and (2) assume that the experience of a feed shortage is exogenously determined. However, it could be endogenously determined for various reasons. Firstly, the higher the crop production, the higher crop residue production—the second most important feed source in Ethiopia (Figure 1)—and, hence, the less likely that farmers would experience a feed shortage. On the other hand, a feed shortage experience affects crop production through different channels, as discussed in Section 2.2 (Figure 2). Therefore, there could be a reverse causality between a feed shortage experience and the outcome variables.

Secondly, unobserved heterogeneities could also cause problems in respect of identifying the impact of a feed shortage experience on the outcome variables of interest. For instance, a farmer may simultaneously decide on the share of resources including crop residues, land and labour to be allocated between livestock and crop production (Jaleta, Kassie and Erenstein, 2015). Such decisions could affect both the likelihood of a feed shortage experience and the outcome variables of interest simultaneously.

Thirdly, an error could arise when farmers measure and report their feed shortage experience as moderate or severe.⁴ Even though there was consistency between self-reported feed shortage experiences and the objectively measured feed gap (see Figure 6, but also Figure A1 in supplementary data at ERAE online), farmers may intentionally overstate the severity of that experience to get any support on livestock feed even though household-level government (or other stakeholders) supports for livestock feed have been rare in the Ethiopia context, except community-level supports when there is severe drought or flood. These and similar unobserved heterogeneities affect both the likelihood of a reported feed shortage experience and the outcome variables. These decisions may vary across the survey years, meaning that the fixed effects model may not capture such unobserved heterogeneities. This argument is corroborated by the evidence in the data, namely that the percentage of farmers who allocated land for grazing increased over time from 6 per cent in 2011 to 22 per cent in 2013 and to 26 per cent in 2017. To address these potential endogeneity problems, the study used an IV.

The share of households who experienced a feed shortage in an enumeration area (EA) was used as an IV for a household's feed shortage experience.⁵ Within an EA, the share of the feed shortage experience was calculated by excluding a given household’s own experience of such a shortage. This is because the households in an EA are more likely to experience a similar feed shortage problem since they share similar agroecological conditions and often share similar communal grazing land. Moreover, following the theory of group behaviour, households who belong to the same EA tend to make similar farming decisions (Angrist, 2014; Manski, 1993). Thus, it was hypothesized that the share of farmers who experienced a feed shortage in an EA would be strongly correlated with a given individual farmer’s reported experience of such a shortage in that EA. Following the literature (Di Falco, Veronesi and Yesuf, 2011), the validity of the IV was tested by including it in both the first and second stages of the regressions. The findings revealed that the share of farmers who had experienced a feed shortage in the EA could be considered as a valid IV: the first-stage estimate showed a statistically significant coefficient of the IV on the likelihood of a feed shortage experience, but it was not a statistically significant driver of the outcome variables (see Table A2 in supplementary data at ERAE online).

A falsification test was also conducted, namely by investigating the impacts of a feed shortage experience on the value of poultry deaths.⁶ As stated earlier (Section 3.1), poultry were excluded from the analysis because the experience of a feed shortage was asked in respect of non-poultry livestock in the current study data. To empirically validate this argument, the effect of experiencing a feed shortage was estimated with regard to the value of poultry deaths. As expected, the results showed that the variable experience of a feed shortage was not statistically significant (see Table A3 in supplementary data at ERAE online).

Moreover, the study not only controlled for land size, various socio-economic and sociodemographic characteristics of the sample households as well as livestock and crop production shocks, but it also controlled for heterogeneities that could correlate with both the feed shortage experience and the error terms in Equation (1). For instance, better-educated farmers might manage their livestock and crop production better, which includes managing a feed shortage. Failure to control for these heterogeneities among farmers may still bias the estimates. To mitigate these heterogeneities, the study controlled for several variables, including households’ socio-economic and socio-demographic characteristics and their farming practices (Table 3). Furthermore, Equations (1) and (2) were estimated using the fixed effects regression to control for time-invariant unobserved heterogeneities. In addition, as a robustness check, Equation (1) was re-estimated using the feed gap instead of farmers’ self-reported experience of a feed shortage. As later sections will show, the estimated coefficients of the self-reported experience of feed shortage and feed gap were consistent with each other.

Finally, as mentioned in Section 3, the study findings could be biased because of the excluded households due to attrition and not having livestock at the data collection time. Thus, the potential bias associated with the excluded households due to attrition and non-participation in livestock and crop production was also examined. For this, a three-step approach was used (Chamberlin and Ricker‐Gilbert, 2016). The first entailed running a probit model to estimate the effects of households’ socio-economic and sociodemographic characteristics on the likelihood of being re-interviewed in the second and third rounds of the surveys. The second step involved obtaining the predicted probabilities (P_i) and computing the inverse probability weighting (IPW) as (1/P_i). In the third and final step, the models in Equations (1) and (2) were re-estimated using the IPW as weights. The findings revealed that the estimated coefficients using the IPW as weights were similar to those without the weights. This indicates that the excluded observations did not bias the results. Hence, the results obtained without the weights are reported here, in the main text, while the results with the IPW are reported in Table A4 in supplementary data at ERAE online.

5. Econometric results

This section first presents what impact the experience of a feed shortage had on the value of livestock deaths and livestock production expenses (Subsection 5.1), and the value of crop production (Subsection 5.2). Thereafter, the impact channels through which a feed shortage experience affected these outcome variables are presented (Subsection 5.3). Finally, the estimated economic losses due to a feed shortage experience are discussed, along with what implications such economic losses hold for poverty (Subsection 5.4). In all specifications, the Hausman test comparing fixed and random effects models favoured the former; hence, the results are from a two-way fixed effects model.

5.1. Impact of a feed shortage experience on the value of livestock deaths and production expenses

Table 4 presents the impacts of a feed shortage experience on the value of livestock deaths and on livestock production expenses. The results were estimated both with and without the IV. Columns (1), (2), (4) and (5) report the estimated results without the IV, while columns (3) and (6) present the IV results. To save space, the table only offers the results of variables of interest; the full regression results are available in Table A5 in supplementary data at ERAE online.

Although we cannot definitively identify causal effects, we find strong and consistent associations between a feed shortage experience and the value of livestock deaths and livestock production expenses. The results reveal that experiencing a feed shortage had broadly consistent, substantial and statistically significant impacts on the value of livestock deaths and livestock production expenses. Furthermore, the results (column (1)) show that households who had experienced a feed shortage lost 19 per cent more value in terms of livestock deaths than households who had not experienced such a shortage. When we disaggregate the experience of a feed shortage by its severity, the results further confirm that experiencing a severe feed shortage tended to increase the value of livestock deaths by 48 per cent (column (2)). The coefficient for a moderate feed shortage was positive, but not statistically significant. However, the result should be taken cautiously since the magnitude of the feed shortage experience coefficient drops to 14.0 per cent (column (3)) when the IV is used.

The results also reveal that experiencing a feed shortage tends to increase livestock production expenses. The results from both the non-IV (columns (4) and (5)) approach and the IV (column (6)) approach consistently illustrate that a feed shortage experience substantially increased livestock production expenses. Households who experienced a shortage of feed incurred livestock production expenses that were 74.8 per cent higher than those of households who did not report such shortages. The magnitude of the coefficient is even higher in the IV approach (76.5 per cent), and for households who experience a severe feed shortage (111 per cent).

5.2. Impact of a feed shortage experience on the value of crop production

This subsection presents the impacts a feed shortage experience has on the value of crop production. Column (1) of Table 5 presents these impacts on the value of crop production without controlling for the use of modern agricultural inputs such as inorganic fertiliser, improved seeds and pesticides, while column (2) reflects the results after controlling for those inputs. Columns (3) and (4), respectively, present the severity-disaggregated impacts of a feed shortage experience without controlling for such inputs, and after controlling for them. The last two columns contain the regression results from an IV estimation without controlling for those inputs (column (5)) and after controlling for them (column (6)). Controlling for the use of modern agricultural inputs may reduce the statistical significance and magnitude of the impact of a feed shortage experience on the value of crop production since they are one of the channels through which feed shortages affect crop production. Controlling for inputs made the coefficients of experiencing a feed shortage statistically insignificant (column (2)) and reduces the magnitude of the severe feed shortage coefficient from 7 (column (3)) to 5.7 per cent (column (4)). Furthermore, the magnitude of a feed shortage experience coefficient declined once inputs are controlled, though marginally, i.e. from 4.3 to 4.2 per cent. The full regression results are available in Table A6 in supplementary data at ERAE online.

The results consistently show negative correlation between a feed shortage experience and the value of crop production. In most specifications, the coefficients of a feed shortage experience are statistically significant. The results from the non-IV estimation show that, compared with farmers who had not experienced a feed shortage, farmers who had done so obtained 4.1 per cent less values of crops produced (column (1)). The results from the IV estimation are similar, in that households who experienced a feed shortage produce 4.2–4.3 per cent less value of crop production. As expected, experiencing a severe feed shortage led to 5.7–7.0 per cent less value in respect of crop production in comparison with not experiencing such a shortage. The results are consistent with other findings, which highlighted how time allotted to searching for feed impacted crop production in Tigray, Ethiopia (Hadush, 2017, 2018). Specifically, the author found that reducing the time that the farmers spend in search of grazing land and straw by 1 per cent increases food production by 0.279 and 0.328 per cent, respectively, in Tigray (Hadush, 2018). Similarly, he found that a 1 per cent increase in minutes travelled to search for feed source and straw collection reduce per capita food consumption by 0.102 and 0.092 per cent, respectively (Hadush, 2017).

In addition, this study examined the robustness of the self-reported feed shortage experience indicators by way of an alternative variable—Feed gap (Section 3). Table 6 presents the impacts of this variable on the key outcome variables (see full regression results in Table A7 in supplementary data at ERAE online). All the coefficients of the Feed gap variable are statistically significant. Moreover, the results in respect of the self-reported feed shortage experience were consistent: the higher the feed gap, the higher are the value of livestock deaths and livestock production expenses.⁷

5.3. Impact channels

Table 7 presents the channels through which a feed shortage experience potentially affects the outcome variables of interest (see full regression results in Tables A8 and A9 in supplementary data at ERAE online). A key channel through which the experience of a feed shortage could affect livestock deaths is by increasing the likelihood of catching diseases. The results provide evidence that households who had experienced a feed shortage were 33.7 per cent more likely to report their livestock having been affected by diseases (column 2). Also, farmers who had experienced a feed shortage were 1.1 per cent more likely to have lost equine animals than farmers who had not had that experience. Moreover, farmers who had experienced a feed shortage used less inorganic fertiliser (4.4 kg) and allotted less money to purchasing improved seeds (10.5 ETB) than farmers who had not reported a feed shortage. The impact of experiencing a feed shortage was statistically insignificant in respect of oxen deaths, pesticide use or the quantity of labour days used for crop production. However, Hadush (2017) found that that a 1 per cent increase in searching times of grazing and collecting straw reduces the time spent on crop production by 0.0929 and 0.0992 per cent, respectively, in Tigray.

5.4. Economic loss and poverty impact of a feed shortage experience

The results presented in Subsections 5.1–5.3 consistently show that experiencing a feed shortage had negative and statistically significant impacts on livestock and crop production. To put the negative impacts of a feed shortage experience in perspective, the study calculated the economic losses due to livestock deaths (Equation (3)), increased livestock production expenses (Equation (4)) and a reduced value of crop production (Equation (5)). These calculations were followed by computing the number and percentage of households who would have been above the national absolute poverty line had they earned the lost income due to a feed shortage experience free of cost. The methodology employed here was borrowed from Abro et al. (2023). The sample household-level economic losses due to a feed shortage experience were computed as follows.

The economic loss from livestock deaths due to a feed shortage experience (LL) was calculated as follows:

where |${\beta _l}$| denotes the effect of a feed shortage experience on the value of livestock deaths, as obtained from Equation (1); FS is a dummy variable taking the value of one for households who experienced a feed shortage, and zero otherwise; and LD is the value of livestock deaths due to all causes in the last 12 months preceding each of the surveys.

Economic loss from increased livestock production expenses due to a feed shortage experience (⁠|$LE$|⁠) is given by the following equation:

where |${\beta _e}$| denotes the impact of a feed shortage experience on the value of livestock production expenses, as obtained from Equation (1); |$FS$| is as defined before; and C is the mean value of livestock production expenses that a household incurred annually.

Economic loss from a reduced value of crop production due to a feed shortage experience (⁠|$LC$|⁠) is given by the following equation:

where |${\beta _c}$| denotes the impact of a feed shortage experience on the value of crop production, as obtained from Equation (1); |$FS$| is as defined before; and VC is the value of the crop produced by a household in the last 12 months preceding each of the surveys. In Equations (3)–(5), the coefficients used are those estimated using the IV approach.

Hence, the total economic loss (TL) from livestock deaths, livestock production expenses and the reduced value of crop production due to a feed shortage experience can be expressed as follows:

The percentage of households who would have earned above the national poverty line, had they earned the lost income presented in Equation (6), was then computed. Using the growth elasticity of poverty approach (Abro et al., 2020, 2023; Alene et al., 2009; Fan et al., 2005), the study extrapolated the equivalent amount of poverty reduction had the households mitigated the feed shortage as follows:

where |$n$| is the number of people who could potentially have been lifted above the poverty line, and TL is the total economic loss associated with a feed shortage experience.|$\,AGDP$| stands for the agricultural gross domestic product in Ethiopia in 2016 (ETB 292 billion, roughly estimated using CSA (2017a, 2017b) livestock and crop production surveys and the 2016 price of agricultural products), while e is the elasticity of poverty for AGDP, which equals −1.66 (the only estimate found in the literature) (Diao, Hazell and Thurlow, 2010). |$N$| denotes the estimated 21.90 million people who lived below the poverty line in 2016 (Federal Democratic Republic of Ethiopia National Planning Commission, 2017).⁸

The results are presented in Table 8. On average, the sample households each lost ETB 258 a year due to a feed shortage experience. The lost income is 2.3 and 6.6 per cent of the total annual income of the sample and the affected (by feed shortages) households, respectively. Had the households earned this income, 0.94 per cent of the poor households in the sample and 2.57 per cent of the affected households would have been above the poverty line had they accessed sufficient feed.⁹ The mean annual income loss for households who experienced feed shortages (i.e. 36 per cent of the sample) was ETB 711 per household per year. The estimates using Equation (7) show that the lost income would have lifted 628,952 people above the poverty line at the national level.

This result indicates that feed shortages could be an important determinant of the poverty status of the rural population. Perhaps due to its huge multiplier effects, the potential poverty reduction effects of investing in animal feed are much larger than addressing specific diseases such as trypanosomiasis. Controlling the latter can reduce poverty by only around 200,000 persons per year in Ethiopia (Abro et al., 2023). However, the potential economic losses and their associated poverty impacts could be underestimated because the study estimates do not include potential income losses on milk production, fertility or meat production. Moreover, feed shortages may even get worse because of unabated population growth and climate change (Gebrechorkos, Hülsmann and Bernhofer, 2020; Zeleke et al., 2017).

The study findings clearly demonstrate the need for sustainable interventions to address feed shortages. So far, farmers have usually used open grazing land and crop residues as their main sources of livestock feed (Figure 1). However, the results here show that such sources are insufficient, resulting in loss of livestock, increased livestock production expenses and reduced crop production. For years, farmers have clearly been attempting to complement this insufficiency by reallocating land from growing crops to grazing. Farmers who made such reallocations increased by 20 per cent between 2011 and 2017. A similar trend has been observed in the literature, in that Worku et al. (2021) found that cultivation land, wetlands, natural forests and unproductive bare land have been replaced for other purposes such as grazing lands in northern Ethiopia since the last two decades.

However, smallholder farmers in Ethiopia already only have very small plots of land (less than a quarter of the land available per capita), and even that has been declining substantially over time (Bachewe and Minten, 2023; Jayne, Mather and Mghenyi, 2010). Hence, leaving the livestock feed problem for the smallholders alone without major intervention from the government and other stakeholders may not be a sustainable solution. Other studies show that policy interventions that improve livestock feed by reducing feed trade barriers and by providing subsidies for forage production increase household income (Komarek, Waldron and Brown, 2012). The types of subsidies may include integration of fodder production and feed system in the agricultural development extension system, importing and scaling of improved fodder types. In addition, interventions such as investments in rural market infrastructure increase competitiveness and reduce rural poverty (Bahta and Malope, 2014). Introducing modern livestock production technologies and other innovations to smallholder livestock producers can also improve feeding practices (Ravichandran et al., 2020).

Thus, addressing the feed shortage problem requires short-term, medium-term and long-term solutions that may include exploring alternative feed sources (Quintero-Herrera et al., 2023; Sánchez-muros, Barroso and Manzano-agugliaro, 2014; Wang et al., 2024) and structural transformation. The short- and long-term solutions could involve integrating fodder production more effectively into the existing agricultural development extension programmes, for example. However, while stakeholders have tried expanding the adoption of improved crop seeds and other technologies (Porteous, 2020; Ruzzante, Labarta and Bilton, 2021), the expansion of improved and commercial fodder production has been limited in Ethiopia (Dejene et al., 2014). Another long-term solution may be to include mechanized farming so that farmers do not need livestock for crop production. Also, low livestock productivity could be attributable to farmers rearing livestock as draught animals for crop production because they focus on livestock types suited to farming instead of looking at their productivity and profitability.

6. Conclusion

Livestock remains the key source of food and nutrition, fuel, organic fertiliser, draught power and means of transport in many countries in sub-Saharan Africa. However, the region could not fully exploit the livestock sector’s production potential. Arguably, and as this study has shown, one of the most important reasons for the low production and productivity of livestock in the region is the shortage of feed. Indeed, many studies have reported that farmers face a severe feed shortage problem in the region. However, the economic cost of feed shortages has rarely been systematically quantified.

The purpose of this study, therefore, was to estimate the impacts of a feed shortage experience on livestock deaths, livestock production expenses and the value of crop production. Two-way fixed-effect and IV approaches were employed on a balanced panel data comprising three rounds of surveys conducted among 5,725 smallholder farmers from rural Ethiopia. After calculating the total economic cost of a feed shortage experience, the potential poverty reduction effects of addressing the feed shortage problem were also estimated. Although we cannot definitely identify causal effects, we find a consistent and strong association between a feed shortage experience and the outcome variables of interest. The findings reveal that feed shortages increased the value of livestock deaths by about 14 per cent and livestock production expenses by about 77 per cent. Moreover, such shortages reduce the value of crop production by about 4.3 per cent. The latter outcome is not surprising, since around 80 per cent of the farmers depend on livestock to plough their land and thresh and transport their products (Behnke, Wolford and Metaferia, 2011). The current study’s results highlight that feed shortages have a substantial impact on livestock and crop production that expenses 6.6 per cent of the affected households’ annual income.

Notably, the estimates presented in this paper may underestimate the total impacts of a feed shortage experience as other potential losses—such as decreased animal fertility, reduced livestock products and lower prices for livestock because of their thinness or weakness—were not accounted for, nor were the impacts of feed shortages considered along the value chain. However, even despite this potential underestimation, the lost income could have had a significant impact on reducing poverty among smallholders. Had the income not been lost in this way, it would have lifted 0.94 per cent of the sample households and 2.57 per cent of affected households above the poverty line annually.

This paper provides direct empirical evidence by estimating the household-level economic cost of feed shortages in the context of sub-Saharan Africa. Given the scarcity of information for feed-related economic analysis in the region, the unique data set used assisted in documenting the direct and indirect effects of feed shortages. The study findings can provide policy insights to the government, the private sector and other stakeholders. Given the projected increase in population and erratic rainfall in the region, the feed shortage problem will probably escalate. The recurrent drought and overall climate change are also affecting livestock production.

The study data also show that smallholders have been attempting to curtail the feed shortage problem by reallocating land from crop production to grazing. However, this is less likely to be a sustainable solution not only since the per capita land in rural areas has declined substantially over time due to population growth, but also because smallholders only have about 1 ha of land per household and about 0.25 ha of land per capita. Hence, the problem needs major interventions from governments and other stakeholders. The results of this study demonstrate that developing policies and strategies to mitigate the impact of a feed shortage is likely to have a positive effect on poverty reduction for countries like Ethiopia, which lose significant numbers of livestock during drought periods.

Finally, this study is not without limitations. Even though a battery of observable and time-invariant unobserved characteristics was controlled for and an IV approach was employed, the findings should be interpreted cautiously: there may be unknown time-variant factors for which the study did not control. Of particular importance in this regard are time-variant environmental factors and human-driven actions, which may affect both the outcome variables and the likelihood of a feed shortage experience, biasing the estimated coefficients.

Acknowledgements

The publication was produced in collaboration with the International Food and Policy Research Institute and World Vision under the Strengthen PSNP Institutions and Resilience II Program, Cooperative Agreement Number 720BHA21CA00036-IFPR, funded by the U.S. Agency for International Development. Z.A. and M.K. gratefully acknowledge the financial support for this research by the Swedish International Development Cooperation Agency, the Swiss Agency for Development and Cooperation, the Australian Centre for International Agricultural Research, the Federal Democratic Republic of Ethiopia, and the Government of the Republic of Kenya. The views expressed herein do not necessarily reflect the official opinion of the donors. The authors are so grateful for insightful, detailed and constructive comments received from three anonymous reviewers and Dr Salvatore Di Falco, one of the editors of the journal, which improved the quality of the paper substantially. Moreover, constructive comments were also gratefully received from Professor Stefan Dercon, Professor of Economic Policy at the Blavatnik School of Government and the Director of the Centre for the Study of African Economies at the Department of Economics, University of Oxford. We also greatly appreciate the comments we received from the ICAE 2024 conference participants who attended the session.

Supplementary data

Supplementary data are available at ERAE online.

Footnotes

References

Author notes