## INTRODUCTION

## Source details

**Canonical URL:** [INTRODUCTION](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020095-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020095-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020095-print-pdf.pdf.json)

---

### Context and objectives
- Crop yields are much lower in sub-Saharan Africa, leading to large food imports (UN 2015).
- Enhancing agricultural production and productivity can reduce forex import needs and employ the large number of working-age entrants to the labor force over the next twenty years.
- Low crop productivity is linked to poor soils (CIMMYT, 2015) and limited use of essential inputs: improved seeds, fertilizers, irrigation, pesticides.
- Hypothesis: use of these inputs would boost crop productivity.
- This paper uses household survey microdata from Ethiopia, Rwanda and Uganda to assess the extent to which these inputs improve crop yields and to assess factors mitigating climate-change-related crop damage.

### Data sources and crop focus
- Microdata: Rwanda EICV5 (2016/17); Ethiopia LSMS (2015/16); Uganda national panel (2015/16).
- Samples capture mainly smallholders: median farm sizes:
  - Rwanda: "a quarter of a hectare" (0.25 ha).
  - Uganda: 0.5 hectares.
  - Ethiopia: 1.1 hectares.
- Major crops/legumes analyzed: maize, sorghum, beans, wheat; plus soybeans for Rwanda, teff for Ethiopia, cassava for Uganda.
- These crops correspond to roughly 60 percent of cultivated area in Rwanda and about 86 percent in Ethiopia (2015/16 estimate).

### Aggregate and micro yield comparisons
- Rwanda: yields stable over 2011–17 except a drastic improvement in cassava (better plot-size understanding and control over cassava brown streak disease, FAO 2019). Aggregate and micro (median) yields correspond closely except for beans (micro sample appears more productive).
- Ethiopia: maize, sorghum and wheat yields have improved; wheat yield now close to South Africa. Micro data yields are less than half the aggregate FAO data — possible reasons: smaller farm sizes in micro sample and FAO imputation-based data.
- Uganda: mixed—improvements in maize and pulse yields, deterioration in sorghum; micro and aggregate yields closely correspond except for cassava.
- Relative rankings across crops from micro and aggregate data generally match (except beans in Rwanda; sorghum ranking differs between micro and aggregate in Rwanda).
- Selected aggregate vs Micro yield comparisons (kilograms per hectare — as presented):
  - Rwanda: Maize 1540 (Aggregate) / 1751 (Micro).
  - Ethiopia: Teff 1408 (Aggregate) / 1030 (Micro).
  - Uganda: Cassava 6500 (Aggregate) / 3271 (Micro).

### Determinants of crop yields — methods
- Standard input variables defined as binary dummies: fertilizer, insecticide, improved seeds, irrigation, protection against erosion.
- Controls: region dummies, degree of urbanization, crop dummies, crop failure associated with climate shocks; soil quality, slope and elevation (Ethiopia and Uganda); gender, proprietorial variables (female farmers, land title holders).
- Econometric approaches: OLS and Heckman selection correction to address potential omitted variable bias (farmer quality proxied by literacy/education or by use of agricultural extension).

### Rwanda — principal empirical findings
- Inputs with strong positive and significant effects:
  - Insecticide: raises crop yield by about 47–49 percent (reported coefficients 0.49***, 0.47***, 0.45***, 0.42*** across specifications).
  - Improved seeds: positive and significant (coefficients 0.09***, 0.08***, 0.11***, 0.19***).
  - Irrigation: positive and significant in baseline (0.12**, 0.12**, 0.13*, 0.07).
  - Protection against erosion: significant positive impact at the 95 percent level (0.07** etc.).
- Fertilizer: not significant at standard confidence levels in baseline OLS; becomes significant when sample restricted to those aged less than 66.
- Crop damage: negative impact (approximately -0.12 to -0.13***).
- Location and socioeconomic controls:
  - Urban proximity: strong positive effect on yield (e.g., Urban 0.35***).
  - Female farm owner: negative effect on yields (-0.18***).
  - Land title holder: positive effect (0.15***).
  - Surface area of plot (log): large negative effect on yields (-1.61***).
- Heckman selection test: Chi squared for error term independence insignificant (0.56), suggesting OLS estimates appropriate in main specification.
- R-squared reported around 0.36; number of observations in table variants: 17,332 / 17,332 / 17,332 / 10,062 in one column.

### Ethiopia — principal empirical findings
- Each input variable significant except irrigation (insignificant or negative in some specifications).
  - Fertilizer: 0.29***.
  - Insecticide: 0.36***.
  - Improved seeds: 0.45***.
  - Protection against erosion: 0.11***.
- Soil quality: positive and significant (0.15***).
- Crop damage: large negative impact (-0.36***).
- Elevation effects:
  - Low elevation: negative/insignificant in some specs.
  - High elevation (above 2000 meters break): positive and significant (0.86*** in one specification).
- Surface area of plot (log): negative (-0.16***).
- Female farm owner: positive effect (0.12***).
- Land title holder: positive effect (0.18***).
- Heckman selection test: Chi squared for independence insignificant (0.87), suggesting OLS appropriate given the extension-services proxy.
- R-squared around 0.14; number of observations around 10,925 / 10,925 / 10,525 / 19,041 across columns as presented.

### Uganda — principal empirical findings
- Input variables generally insignificant in Uganda regressions.
- Significant negative impacts:
  - Poor soil quality: -0.17*** to -0.19*** (across specifications).
  - Crop damage related to erosion: negative (e.g., -0.14* to -0.16*).
- Land title holder: positive (0.13**/0.12*/0.13**).
- Surface area of plot (log): large negative effect on yield (-1.5*** to -1.51***).
- Urban: modest positive effect (0.13* to 0.15**).
- Education/literacy:
  - Literacy: positive and significant in some specifications (0.26***).
  - College ed.: negative in one specification (-0.35***), suggesting interpretation complexities.
- Heckman tests: error term independence Chi squared significant (5.77**, 5.73**), indicating selection issues—Heckman model used; input variables remain largely insignificant.
- R-squared around 0.58–0.59; observations ~21,262 / 21,262 / 25,325 / 3,201 in table variants.

### Surface area and yields
- Strong, robust negative relationship between plot yields and surface area of farms across all three countries (scatter analyses shown).
- The negative relationship persists despite potential measurement errors; cleaning methods unlikely to fully eliminate it.
- Surface area coefficients in regressions:
  - Rwanda: Surface area of plot (log) -1.61***.
  - Ethiopia: Surface area of plot (log) -0.16*** (plus surface area of farm holding log -0.07***).
  - Uganda: Surface area of plot (log) -1.5***.

### Use of agricultural inputs and frontier yield estimates
- Usage prevalence (% of households using inputs; Table 6):
  - Fertilizer: Rwanda 42.2; Ethiopia 52.2; Uganda 1.3.
  - Insecticide: Rwanda 27.2; Ethiopia 11.2; Uganda 4.3.
  - Improved seeds: Rwanda 28.5; Ethiopia 5.5; Uganda (not reported numerically in extract).
  - Irrigation: Rwanda 4.1; Ethiopia 2.9; Uganda 1.2.
  - Protection against erosion: Rwanda 73.3; Ethiopia 58.9; Uganda 6.2.
- Fertilizer application in Ethiopia:
  - Wortmann and Sones (2017) recommend up to 80 kilos per hectare for wheat, maize, sorghum, teff; median usage for urea and DAP in Ethiopia is about 90 kilos per hectare.
- Production frontier (“all inputs available” scenario, assumptions):
  - Plot assumed one-third hectare on a 1 hectare farm holding, no crop damage, and farmer-type specifications (Ethiopia: female landowner, literate, college educated; Rwanda: same but male; Uganda: male, high-school educated land title holder).
  - Estimated maize yields per hectare:
    - Ethiopia: 7300 kgs.
    - Rwanda: 7900 kgs.
    - Uganda: 7644 kgs.
  - Estimated wheat yields:
    - Ethiopia: 6700 kgs.
    - Rwanda: 5400 kgs.
  - Conclusion: potential exists to achieve much higher yields assuming input availability.

### Measurement error and welfare implications
- Microdata contain sizeable non-classical measurement errors in both yields and plot surface area (Abay et al., 2019). Biases may cancel when both variables are used in regressions.
- To test information content, regression of household consumption determinants includes average plot yield and farm size.
- Consumption determinants (selected results):
  - Rwanda:
    - Yield coefficient: 0.03*** (Total Cons and Food cons entries: 0.03*** / 0.03*** / 0.06*** / 0.04*** depending on column).
    - Surface area of plot (log): 0.14***.
    - Land title holder: 0.14***.
    - Farm worker: -0.08***.
    - Non-farm worker: 0.15***.
    - Household size: 0.62***.
    - Interpretation: one standard deviation increase in yields raises consumption by 7.5 percent; one standard deviation increase in surface area raises consumption by almost 15 percent.
  - Ethiopia:
    - Yield coefficient implies a one standard deviation increase in yields raises consumption by 3.2 percent; corresponding increase in surface area raises consumption by 14 percent.
- Overall: larger farm holdings provide significantly larger boosts to welfare than higher yields, despite the negative relationship between yields and plot size.

### Climate effects and crop damage
- Crop damage prevalence and causes:
  - About 38–39 percent of farmers affected by crop damage during the agricultural season in Ethiopia and Rwanda.
  - Among those affected, around 70 percent attribute damage to climate change; additional 10–20 percent associate damage with erosion and crop disease.
  - Uganda: figures less dramatic; survey question focused on erosion effects.
- Quantitative impacts of crop damage on yields:
  - Crop damage causes yield reductions between:
    - 13 percent in Rwanda (reported -0.13*** in some specifications).
    - 52 percent in Ethiopia (reported -0.52***).
    - Uganda around 12/13 percent range in summary text.
- Mitigation effects:
  - Protection against erosion reduces the impact of crop damage on yields by 16 percent in Ethiopia (interaction term 0.16** in Table 9).
  - Access to investment finance:
    - Ethiopia: about 27 percent of households have access to loan financing.
    - Rwanda: about 77 percent have access.
    - Access to finance is highly significant in Rwanda and improves yields by almost 25 percent (0.25***), but not significant in Ethiopia in baseline.
    - Interaction of access to finance and crop damage is negative, but the sum of main and interaction terms implies net positive effect for those with access: in Rwanda, families suffering crop damage but with access to finance can mitigate crop damage on yields by 6 percent.
- Input effectiveness under heat/climate stress:
  - Isolating heat exposure shows irrigation importance increases in Rwanda, raising yields impact from 12 percent to 19 percent (irrigation coefficients move from 0.12 to 0.19*** in heat-exposed specs); little change for Ethiopia.

### Conclusions and policy recommendations
- Micro vs aggregate data:
  - Rwanda: aggregate yields at or below micro estimates; Ethiopia: micro yields less than half aggregate FAO data; Uganda: close correspondence except for cassava.
  - Relative yield rankings across crops largely similar across data sources, except beans in Rwanda.
- Effectiveness of inputs:
  - Fertilizer, improved seeds, protection against erosion, and pesticides significantly improve crop yields in the microdata analysis.
  - Irrigation significantly positive in Rwanda; not significant in Ethiopia and Uganda—possibly due to low usage prevalence.
  - With all positive determinants in place, wheat and maize yields could increase up to fourfold in Ethiopia, Rwanda, and Uganda (as suggested by frontier estimates).
- Caveats:
  - Paper does not establish economic viability of scaling up inputs—would require randomized experiments given limited input-cost data in welfare surveys.
- Policy recommendations:
  - Improve seed allocation through an effective seed regulatory framework.
  - Establish a reliable and internationally acceptable seed certification system.
  - Consider ways to boost growth of the domestic seed industry.
  - Promote measures to mitigate climate damage: protection against erosion, access to finance, and switching to crop varieties more resistant to climate-change effects.

*Source: wpiea2020095-print-pdf - INTRODUCTION (household survey analysis for Ethiopia, Rwanda, Uganda) — IMF staff estimates and referenced sources as presented in the unit.*

### INTRODUCTION

### INTRODUCTION

### Context and objectives
- Crop yields are much lower in sub-Saharan Africa, leading to large food imports (UN 2015).
- Enhancing agricultural production and productivity can reduce forex import needs and employ the large number of working-age entrants to the labor force over the next twenty years.
- Low crop productivity is linked to poor soils (CIMMYT, 2015) and limited use of essential inputs: improved seeds, fertilizers, irrigation, pesticides.
- Hypothesis: use of these inputs would boost crop productivity.
- This paper uses household survey microdata from Ethiopia, Rwanda and Uganda to assess the extent to which these inputs improve crop yields and to assess factors mitigating climate-change-related crop damage.

### Data sources and crop focus
- Microdata: Rwanda EICV5 (2016/17); Ethiopia LSMS (2015/16); Uganda national panel (2015/16).
- Samples capture mainly smallholders: median farm sizes:
  - Rwanda: "a quarter of a hectare" (0.25 ha).
  - Uganda: 0.5 hectares.
  - Ethiopia: 1.1 hectares.
- Major crops/legumes analyzed: maize, sorghum, beans, wheat; plus soybeans for Rwanda, teff for Ethiopia, cassava for Uganda.
- These crops correspond to roughly 60 percent of cultivated area in Rwanda and about 86 percent in Ethiopia (2015/16 estimate).

### Aggregate and micro yield comparisons
- Rwanda: yields stable over 2011–17 except a drastic improvement in cassava (better plot-size understanding and control over cassava brown streak disease, FAO 2019). Aggregate and micro (median) yields correspond closely except for beans (micro sample appears more productive).
- Ethiopia: maize, sorghum and wheat yields have improved; wheat yield now close to South Africa. Micro data yields are less than half the aggregate FAO data — possible reasons: smaller farm sizes in micro sample and FAO imputation-based data.
- Uganda: mixed—improvements in maize and pulse yields, deterioration in sorghum; micro and aggregate yields closely correspond except for cassava.
- Relative rankings across crops from micro and aggregate data generally match (except beans in Rwanda; sorghum ranking differs between micro and aggregate in Rwanda).

### Key cross-country data points from tables and figures
- Aggregate vs Micro (selected yield comparisons, kilograms per hectare — as presented):
  - Rwanda: Maize 1540 (Aggregate) / 1751 (Micro) — (table shows various values per crop)
  - Ethiopia: Teff 1408 (Aggregate) / 1030 (Micro)
  - Uganda: Cassava 6500 (Aggregate) / 3271 (Micro)
  - (Full tabulated values and kernel density figures are presented in the source.)

### Determinants of crop yields — methods
- Standard input variables defined as binary dummies: fertilizer, insecticide, improved seeds, irrigation, protection against erosion.
- Controls: region dummies, degree of urbanization, crop dummies, crop failure associated with climate shocks; soil quality, slope and elevation (Ethiopia and Uganda); gender, proprietorial variables (female farmers, land title holders).
- Econometric approaches: OLS and Heckman selection correction to address potential omitted variable bias (farmer quality proxied by literacy/education or by use of agricultural extension).

### Rwanda — principal empirical findings
- Inputs with strong positive and significant effects:
  - Insecticide: raises crop yield by about 47–49 percent (reported coefficients 0.49***, 0.47***, 0.45***, 0.42*** across specifications).
  - Improved seeds: positive and significant (coefficients 0.09***, 0.08***, 0.11***, 0.19***).
  - Irrigation: positive and significant in baseline (0.12**, 0.12**, 0.13*, 0.07).
  - Protection against erosion: significant positive impact at the 95 percent level (0.07** etc.).
- Fertilizer: not significant at standard confidence levels in baseline OLS; becomes significant when sample restricted to those aged less than 66.
- Crop damage: negative impact (approximately -0.12 to -0.13***).
- Location and socioeconomic controls:
  - Urban proximity: strong positive effect on yield (e.g., Urban 0.35***).
  - Female farm owner: negative effect on yields (-0.18***).
  - Land title holder: positive effect (0.15***).
  - Surface area of plot (log): large negative effect on yields (-1.61***).
- Heckman selection test: Chi squared for error term independence insignificant (0.56), suggesting OLS estimates appropriate in main specification.
- R-squared reported around 0.36; number of observations in table variants: 17,332 / 17,332 / 17,332 / 10,062 in one column.

### Ethiopia — principal empirical findings
- Each input variable significant except irrigation (insignificant or negative in some specifications).
  - Fertilizer: 0.29*** (positive and significant).
  - Insecticide: 0.36***.
  - Improved seeds: 0.45***.
  - Protection against erosion: 0.11***.
- Soil quality: positive and significant (0.15***).
- Crop damage: large negative impact (-0.36***).
- Elevation effects:
  - Low elevation: negative/insignificant in some specs.
  - High elevation (above 2000 meters break): positive and significant (0.86*** in one specification).
- Surface area of plot (log): negative (-0.16***).
- Female farm owner: positive effect (0.12***).
- Land title holder: positive effect (0.18***).
- Heckman selection test: Chi squared for independence insignificant (0.87), suggesting OLS appropriate given the extension-services proxy.
- R-squared around 0.14; number of observations around 10,925 / 10,925 / 10,525 / 19,041 across columns as presented.

### Uganda — principal empirical findings
- Input variables generally insignificant in Uganda regressions.
- Significant negative impacts:
  - Poor soil quality: -0.17*** to -0.19*** (across specifications).
  - Crop damage related to erosion: negative (e.g., -0.14* to -0.16*).
- Land title holder: positive (0.13**/0.12*/0.13**).
- Surface area of plot (log): large negative effect on yield (-1.5*** to -1.51***).
- Urban: modest positive effect (0.13* to 0.15**).
- Education/literacy:
  - Literacy: positive and significant in some specifications (0.26***).
  - College ed.: negative in one specification (-0.35***), suggesting interpretation complexities.
- Heckman tests: error term independence Chi squared significant (5.77**, 5.73**), indicating selection issues—Heckman model used; input variables remain largely insignificant.
- R-squared around 0.58–0.59; observations ~21,262 / 21,262 / 25,325 / 3,201 in table variants.

### Surface area and yields
- Strong, robust negative relationship between plot yields and surface area of farms across all three countries (scatter analyses shown).
- The negative relationship persists despite potential measurement errors; cleaning methods unlikely to fully eliminate it.
- Surface area coefficients in regressions:
  - Rwanda: Surface area of plot (log) -1.61***.
  - Ethiopia: Surface area of plot (log) -0.16*** (plus surface area of farm holding log -0.07***).
  - Uganda: Surface area of plot (log) -1.5***.

### Use of agricultural inputs and frontier yield estimates
- Usage prevalence (% of households using inputs; Table 6):
  - Fertilizer: Rwanda 42.2; Ethiopia 52.2; Uganda 1.3 (percent of total household population).
  - Insecticide: Rwanda 27.2; Ethiopia 11.2; Uganda 4.3.
  - Improved seeds: Rwanda 28.5; Ethiopia 5.5; Uganda (not reported numerically in extract).
  - Irrigation: Rwanda 4.1; Ethiopia 2.9; Uganda 1.2.
  - Protection against erosion: Rwanda 73.3; Ethiopia 58.9; Uganda 6.2.
- Fertilizer application in Ethiopia:
  - Wortmann and Sones (2017) recommend up to 80 kilos per hectare for wheat, maize, sorghum, teff; median usage for urea and DAP in Ethiopia is about 90 kilos per hectare (close to recommended).
- Production frontier (“all inputs available” scenario, assumptions):
  - Plot assumed one-third hectare on a 1 hectare farm holding, no crop damage, and farmer-type specifications (Ethiopia: female landowner, literate, college educated; Rwanda: same but male; Uganda: male, high-school educated land title holder).
  - Estimated maize yields per hectare:
    - Ethiopia: 7300 kgs.
    - Rwanda: 7900 kgs.
    - Uganda: 7644 kgs.
  - Estimated wheat yields:
    - Ethiopia: 6700 kgs.
    - Rwanda: 5400 kgs.
  - Conclusion: potential exists to achieve much higher yields assuming input availability.

### Measurement error and welfare implications
- Concern: microdata contain sizeable non-classical measurement errors in both yields and plot surface area (Abay et al., 2019). Biases may cancel when both variables are used in regressions.
- To test information content, regression of household consumption determinants includes average plot yield and farm size.
- Consumption determinants (selected results):
  - Rwanda:
    - Yield coefficient: 0.03*** (Total Cons and Food cons entries: 0.03*** / 0.03*** / 0.06*** / 0.04*** depending on column).
    - Surface area of plot (log): 0.14*** (Total Cons and Food cons variants).
    - Land title holder: 0.14***.
    - Farm worker: -0.08***.
    - Non-farm worker: 0.15***.
    - Household size: 0.62***.
    - Interpretation: one standard deviation increase in yields raises consumption by 7.5 percent; one standard deviation increase in surface area raises consumption by almost 15 percent.
  - Ethiopia:
    - Yield coefficient implies a one standard deviation increase in yields raises consumption by 3.2 percent; corresponding increase in surface area raises consumption by 14 percent.
- Overall: larger farm holdings provide significantly larger boosts to welfare than higher yields, despite the negative relationship between yields and plot size.

### Climate effects and crop damage
- Crop damage prevalence and causes (percent):
  - About 38–39 percent of farmers affected by crop damage during the agricultural season in Ethiopia and Rwanda.
  - Among those affected, around 70 percent attribute damage to climate change; additional 10–20 percent associate damage with erosion and crop disease.
  - Uganda: figures less dramatic; survey question focused on erosion effects.
- Quantitative impacts of crop damage on yields:
  - Crop damage causes yield reductions between:
    - 13 percent in Rwanda (reported -0.13*** in some specifications).
    - 52 percent in Ethiopia (reported -0.52***).
    - Uganda around 12/13 percent range in summary text.
- Mitigation effects:
  - Protection against erosion reduces the impact of crop damage on yields by 16 percent in Ethiopia (interaction term 0.16** in Table 9).
  - Access to investment finance:
    - Ethiopia: about 27 percent of households have access to loan financing.
    - Rwanda: about 77 percent have access.
    - Access to finance is highly significant in Rwanda and improves yields by almost 25 percent (0.25***), but not significant in Ethiopia in baseline.
    - Interaction of access to finance and crop damage is negative, but the sum of main and interaction terms implies net positive effect for those with access: in Rwanda, families suffering crop damage but with access to finance can mitigate crop damage on yields by 6 percent (text statement).
- Input effectiveness under heat/climate stress:
  - Isolating heat exposure shows irrigation importance increases in Rwanda, raising yields impact from 12 percent to 19 percent (irrigation coefficients move from 0.12 to 0.19*** in heat-exposed specs); little change for Ethiopia.

### Conclusions and policy recommendations
- Micro vs aggregate data:
  - Rwanda: aggregate yields at or below micro estimates; Ethiopia: micro yields less than half aggregate FAO data; Uganda: close correspondence except for cassava.
  - Relative yield rankings across crops largely similar across data sources, except beans in Rwanda.
- Effectiveness of inputs:
  - Fertilizer, improved seeds, protection against erosion, and pesticides significantly improve crop yields in the microdata analysis.
  - Irrigation significantly positive in Rwanda; not significant in Ethiopia and Uganda—possibly due to low usage prevalence.
  - With all positive determinants in place, wheat and maize yields could increase up to fourfold in Ethiopia, Rwanda, and Uganda (as suggested by frontier estimates).
- Caveats:
  - Paper does not establish economic viability of scaling up inputs—would require randomized experiments given limited input-cost data in welfare surveys.
- Policy recommendations (as stated in source):
  - Improve seed allocation through an effective seed regulatory framework.
  - Establish a reliable and internationally acceptable seed certification system.
  - Consider ways to boost growth of the domestic seed industry.
  - Promote measures to mitigate climate damage: protection against erosion, access to finance, and switching to crop varieties more resistant to climate-change effects.

*Source: wpiea2020095-print-pdf - INTRODUCTION (household survey analysis for Ethiopia, Rwanda, Uganda) — IMF staff estimates and referenced sources as presented in the unit.*

### REFERENCES

### wpiea2020095-print-pdf - REFERENCES

### Academic articles and working papers
- Adamopoulos, T. and D. Restuccia (2018), “Geography and Agricultural Productivity: Cross-Country Evidence from Micro Plot level data,” working paper
- Alvarez, J. and C. Berg (2019), “Crop selection and international differences in aggregate agricultural productivity, working paper
- Gollin, D., S. Parente, and R. Rogerson (2002), “The role of agriculture in development,” American Economic Review 92(2), 160-164
- Gollin, D., Lagakos, D. and M. Waugh (2014) “Agriculture productivity differences across countries,” American Economic Review Papers and Proceedings 104(5), 165-70
- Wooldridge, J. “On estimating firm-level production functions using proxy variables to control for unobservables,” Economics Letters 2009

### Country- and region-specific studies
- Epule, T. et al., “The determinants of crop yields in Uganda: what is the role of climatic and non-climatic factors?” Agriculture and Food Security, 2018
- Mann M. and J. Warner, “Ethiopian wheat yield and yield gap estimation: a spatially explicit small area integrated data approach,” Field Crops Research, 2016
- Taffesse, A. “Crop production in Ethiopia: regional patterns and trends,” IFPRI working paper 2013
- FAO, Restoring Cassava farming in Rwanda, 2019
- CIMMYT, Poor soils a huge limitation for Africa’s food security,” 2015
- United Nations, The Least Developed Countries Report, New York, 2015

### Fertilizer, inputs, and agricultural practice studies
- Mcarthur, J. and G. Mccord, “Fertilizing growth: agricultural inputs and their effects in economic development, “Journal of Development Economics, February 2017
- Saweda, L. et al. “Maize Farming and Fertilizers: Not a profitable mix in Nigeria,” Agriculture in Africa, World Bank Washington DC
- Wortmann, C. and K. Sones, “Fertilizer Use Optimization in Sub-Saharan Africa,” CABI, Nairobi 2017

*Source: wpiea2020095-print-pdf - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020095-print-pdf.pdf_