Linking Soil Suitability and Crop Productivity in Africa: Evidence from Maize Farmers in Uganda. Sydney Gourlay, Talip Kilic 1

Transcription

1 Linking Soil Suitability and Crop Productivity in Africa: Evidence from Maize Farmers in Uganda Sydney Gourlay, Talip Kilic 1 ABSTRACT Despite the ubiquitous impacts of agricultural productivity on various indicators of wellbeing, realized productivity is generally well below its potential. While productivity is hindered by a multitude of factors, this paper focuses on the suitability of a given crop, maize, for a given agricultural plot. Farmers cultivating maize on land that is not agronomically suitable for maize will face limited yield potential. This paper, therefore, sets out to determine the magnitude of forgone production due to cultivation on less than suitable land and to identify which groups of farmers are bearing the burden of this constrained productivity, ultimately allowing for greater targeting of agriculture-based poverty reduction policies. We estimate a multi-dimensional measure of soil suitability, based on existing standards, for maize across a sample of approximately 900 households/plots, spanning 4 districts in Eastern Uganda; the leading maizeproducing region in the country. This is made possible by collecting and lab-testing plot-level soil samples based on international best practices in the context of a methodological household survey experiment. This contribution is key to cultivating a more nuanced understanding of micro-level variations in specific soil constraints, hence the required interventions for achieving higher yields, that are otherwise likely to be missed in coarser, geospatial analyses. The sampled plots are classified into three suitability classes, namely highly-suitable, moderatelysuitable, and marginally-suitable based on existing standards. Leveraging plot-level crop cutting-based maize yield measures, the distribution of observed maize yields is compared across suitability class. We extend this analysis by estimating stochastic frontier models of maize yield separately for each suitability class to understand differences in (1) returns to factors of production, (2) technical efficiency, and (3) potential yield measures. Preliminary results suggest that 12 percent of farmers in the sample are cultivating maize on land that is marginally suitable for maize, while only 13 percent are cultivating maize on land that is highly suitable for maize, and the bulk of the sample is somewhere in between highly and marginally suitable, classified as moderately suitable. The observed mean maize yield for farmers in this marginally suitable category is 51 percent of the mean yield realized by farmers with highly suitable soil, while the productivity potential of plots with marginally suitable soil is only 40 percent of the potential yield on highly suitable land. On average, farmers cultivating highly suitable land can potentially increase their already superior yields by as much as 106%, while farmers cultivating maize on marginally suitable land can only increase productivity up to 61%. Thus, preliminary findings suggest that farmers cultivating maize on land that is less than highly suitable for maize production are operating significantly closer to their production frontier. KEYWORDS: Agricultural Productivity; Maize; Soil Fertility; Africa; Smallholder Farming Draft prepared for submission to the 2018 Annual Conference of the Centre for the Study of African Economies (CSAE). 1 Ph.D. Candidate, Department of Economics, The American University, Washington, DC, and Survey Specialist, Living Standards Measurement Study (LSMS), Survey Unit, Development Data Group, World Bank. sgourlay@worldbank.org Senior Economist, Living Standards Measurement Study (LSMS), Survey Unit, Development Data Group, World Bank, Rome, Italy. tkilic@worldbank.org.

2 I. Introduction Agricultural production is at the root of rural livelihoods in developing economies. Yet, despite the ubiquitous impacts of agricultural productivity on various indicators of wellbeing, realized productivity is generally well below its potential. Increasing agricultural productivity is paramount to feeding the growing population, providing sufficient nutrient intake for rural populations, and alleviating rural poverty. Productivity is hindered by a multitude of factors including lack of knowledge and extension services, poor soil quality, market failures (labor, credit, and food), inadequate use of, or suboptimal distribution of, improved inputs. This paper focuses on the suitability of a given crop, maize, for a given agricultural plot. Farmers cultivating maize on land that is not agronomically suitable for maize will face limited yield potential. This paper, therefore, sets out to determine the magnitude of forgone production due to cultivation on less than suitable land and to identify which groups of farmers are bearing the burden of this constrained productivity, ultimately allowing for greater targeting of agriculture-based poverty reduction policies. The importance of agricultural productivity on economic outcomes and wellbeing is undeniable, particularly in the developing economy context. The effects of the structure and efficiency of agricultural systems are pervasive throughout multiple sectors, including food security, health, economic growth, income inequality, poverty and vulnerability, and gender equality, among others. With the majority of hungry poor residing in rural areas where agriculture is the primary livelihood, the linkages between agricultural productivity and food security require little explanation (IFAD, 2016). Higher agricultural productivity has been linked to poverty reduction and improved nutritional outcomes (De Janvry & Sadoulet (2002) and Kennedy & Bouis (1993), for example). Increases in the quantity of production lead directly to increased consumption, with the potential for an increase in cash income from sales of surplus harvest. Increases in cash income can, in turn, lead to increased savings and reductions in vulnerability. Increases in food security and income can also encourage crop diversification, including a shift to cash crops (Fafchamps, 1992). Both theory and evidence point to a strong, positive relationship between agricultural productivity and economic outcomes, yet, large productivity gaps exist. Sizable gaps exist between agronomically feasible yields and those realized by smallholder farmers, even when climate conditions are considered. The Global Yield Gap Atlas, for example, suggests that the potential yield for rainfed maize in Uganda is roughly 6.9 tons per hectare, yet they estimate realized yields at 1.6 tons per hectare ( Productivity gaps are more nuanced than simple differences from potential, however. Larson, Savastano, Murray, & Palacios-López (2015) summarize that there are also gender differentials in productivity, ranging from 4 to 40 percent in sub-saharan Africa. The complexity of agricultural productivity, and the varying levels of factors that influence it, can limit the effectiveness of independent policy actions if the context is not carefully considered. Seemingly straightforward policy actions like the promotion of improved agricultural inputs or practices, which have been proven effective in Asia s Green Revolution, have experienced limited success in sub-saharan Africa, for example. According to the FAO, the use of sustainable soil management techniques could boost production by as much as 58 percent (FAO, 2015). The use of improved crop varieties and chemical input use has been shown to improve productivity and/or resilience exponentially (Cassman, 1999; Duflo, Kremer, & Robinson, 2008). Limited uptake of improved inputs, such as inorganic fertilizers, and sustainable soil management practices, such as zero tillage, contribute to the productivity gaps observed in the region but do not explain it in full. Labor market failures, lack of rural infrastructure and connectivity, lack of human capital, inadequate inputs, and gender norms can all contribute to 2

3 the observed productivity gaps. Finally, what will be the main focus of this paper is the cultivation of crops that are ill suited for the land and climate on which they are farmed. As will be illustrated below, the production potential is capped by the suitability of soil for a given crop, in this study, maize. There is a multiplicity of potential barriers to optimal crop selection, in terms of matching agronomic suitability, ranging from market failure (in labor, input, food, credit, and/or crop sales markets) to infrastructure to human capital. Infrastructure can present the most visible barrier, with lack of rural road networks preventing access to markets or lack of storage facilities resulting in high post-harvest loss, thereby dissuading investment in increasing productivity or cultivating crops with short shelf lives. Omamo (1998) succinctly illustrates that access to markets, and related transaction costs, has direct effects on crop selection, crop diversification, and cultivation patterns, with those further from markets making agricultural choices that result in lower aggregate returns. Incomplete labor markets are shown by Fafchamps (1993) and others to negatively affect productivity. Similarly, failures of the food market encourage staple food production, rather than cash crop production, constraining gains from agriculture (for example, de Janvry, Fafchamps, & Sadoulet, 1991). Education and farming knowledge, often acquired through extension services, may be insufficient to inform specific crop needs. Reliance on subsistence farming for consumption can directly influence the crop that is cultivated, with risk aversion factoring in the more vulnerable a household. Fafchamps (1992) explains this phenomenon with the belief that food security is best assured by food self-sufficiency. Cultural preferences may also come into play if a population is accustomed to consuming a particular staple food. Intertwined with the decision between cash and staple crop cultivation are gender norms. The existence of female and male farming systems, in which women are largely responsible for food crop production and men cash crops, has been noted as early as the 1960s by Boserup (2007), More recent research by Doss (2002) suggests that, in the Ghanaian context, there are few crops that are strictly farmed by males (and none strictly by females), but that there are gender asymmetries in the probability of farming many crops. If the land is more suitable for a given a cash crop, for example, female productivity will be constrained by the sub-optimal crop selection. Even within cultivation of a given crop, gender differentials persist. Doss and Morris (2000) suggest that within maize farmers in Ghana, significant differences exist in the use of improved varieties and other non-labor inputs utilized by males and females. Identification of which barriers are most binding in a particular context can aid in policy formation targeting agricultural productivity, food security, and poverty. To date, the literature on crop-land suitability is concentrated on large scale analysis utilizing geospatial data (for example, Ahamed, Rao, & Murthy, 2000; Hall, Wang, & others, 1992; Reshmidevi, Eldho, & Jana, 2009; Wang, 1994). While this research is useful for identification of suitable crops on a large scale, it fails to consider the relationships between household characteristics, gender relations, local market functionality, crop prices, and intra-household consumption and dietary diversity needs. Additionally, these studies fail to quantity the forgone production that results from farming crops on land that is less than suitable agronomically, and for which types of farmers these losses are most pronounced. This paper sets out to quantify the cost, in terms of productivity, of cultivating maize on land that is less than optimal for maize production, and analyze what characteristics are associated with farming maize on land that does not possess the desired agronomic properties and how this affects the production potential for these farmers. Unique data from Uganda, which includes plot-specific soil analysis coupled with objectively measured maize production, a questionnaire on agricultural practices, plot manager and household characteristics, and geo- 3

4 referenced plot locations allows for this original contribution. Results suggest that despite the fact that the Eastern Region is Uganda s leading maize producing region, only 13 percent of farmers are cultivating soil that is highly suitable for maize production, while the vast majority are cultivating only moderately suitable lands. The key binding constraints differ across suitability classes, suggesting that soil-based interventions need to be carefully considered for the specific context in which they take place. The relationship between soil suitability, observed yields, and yield potential, is not insignificant. Findings suggest that farmers cultivating highly suitable land have the potential to increase their observed yields by as much as 106 percent, up to 3,327 kg/ha, while those at the opposite tail of the suitability distribution, those with marginally suitable land, operate closer to the efficient frontier and thus can only seek to increase observed yields by up to 61 percent, up to 1,332 kg/ha. The paper is organized as follows: Section II discusses the theoretical framework surrounding agricultural productivity, Section III describes the data, Section IV presents the methodology, Section V discusses the key results, and Section VI concludes. II. Theoretical Framework The theoretical framework is founded on the assumption that households are net revenue maximizing, and therefore utilize inputs efficiently. In the case of Uganda, and much of sub-saharan Africa in general, smallholder agricultural households are not only producers, but rather they are often primarily consumers of their production with few (or no) sales. Despite the fact that agricultural production is often the primary means of food consumption, culture in sub-saharan Africa often dictates that men and women cultivate different crops and/or with different levels of inputs. That is, rather than allocate agricultural inputs at Pareto-efficient levels, inputs are often distributed at sub-optimal levels depending on the individual who manages the plot. Udry (1996) elegantly illustrates this finding with data from Burkina Faso. This information is useful in conveying the context, but is only tangential to the analysis presented here as the focus is on understanding who is cultivating maize on illsuited land and the negative effects of this suboptimal combination on production potential, with analysis conducted at the plot level. The inputs to agricultural production are many. Physical inputs, including organic and inorganic fertilizers, land area, land quality, and seed are critical for successful harvest. Labor inputs come in the form of household labor, hired labor, or exchange labor, with the data often particularly noisy. Intangible inputs, such as farm management practices and human capital of the plot manager, are also essential and result in varying levels of technical efficiency. As illustrated by Udry (1996), the gender of the plot manager can play a role in the intensity with which the land is farmed, and, therefore, can arguably be considered an input into production. Rainfall and other climate-related variables are inputs into the agricultural production function, especially when the incidence of irrigation is minimal as in the Ugandan context. Access to public infrastructure is also believed to affect agricultural production (Ekbom and Sterner 2008). Kelly et al. (1995) clearly illustrate the alternative approaches to agricultural productivity analysis. Average measures of productivity, including partial and total factor analysis, can be used to create a single statistic but the methods require high quality price data that are often hard to come by in rural agricultural contexts with thin markets. Alternatively, marginal productivity analysis can be conducted with more direct policy-related takeaways. Cobb Douglas functions, and variations of Cobb Douglas, are commonly used (Deininger et al., 2007; Sherlund, Barrett, and Adesina, 2002; Odhiambo and Nyangito, 2003). 4

5 The limitation of a simple linear production function in this context is that it assumes all farmers to be performing at optimal levels, without explaining the deviations between the observed and attainable (predicted) output levels. Therefore, stochastic frontier analysis will be utilized in an attempt to analyze the technical efficiency of each farmer, thereby better explaining the aforementioned deviations. The empirical methodology used to model the theory above will be examined in detail following a discussion of the data. III. Data The data utilized come from the Methodological Experiment on Measuring Maize Productivity, Soil Fertility, and Variety (MAPS). MAPS is a two-round household panel survey aimed at testing alternative methods for measuring key agricultural inputs, including soil fertility, maize variety, and plot area, as well as maize production. The MAPS project was implemented through collaboration between the World Bank s Living Standards Measurement Study (LSMS), the Uganda Bureau of Statistics, the World Agroforestry Centre, the FAO Standing Panel on Impact Assessment (SPIA), and Stanford University. MAPS Round I was fielded in 2015, while Round II was fielded in The second round of the study did not include soil analysis, and therefore, this paper will utilize only MAPS Round I data. Sampling for MAPS Round I was completed in a multi-stage process. First, three strata were identified in the primary maize-growing regions of Eastern Uganda; Serere district, Sironko district, and a 400km 2 area spanning Iganga and Mayuge districts. From each stratum, enumeration areas were randomly selected with probability proportional to size (15 from Serere and Sironko each, and 45 from the Iganga/Mayuge stratum). In each selected enumeration area, a full household listing was conducted, identifying households who cultivated at least one maize plot and in what cultivation pattern (intercropped versus pure stand). Finally, 12 households were selected from each enumeration area, with an effort to stratify based on cultivation pattern. The sampling strategy aimed to have a sample of 50% intercropped plots and 50% pure stand plots because intercropping confounds the already complex measurement of agricultural productivity, both in terms of measuring yields and through biological interactions that could help or hinder production. 2 Due to the low incidence of pure stand maize plots, and cases in which plots identified as pure stand in the household listing phase were intercropped at the time of the first interview, the final sample was made of up 385 pure stand maize plots and 515 intercropped maize plots (43 percent and 57 percent, respectively). Therefore, the sample is made up of 900 maize plots, from 900 households, each from a different household. Fieldwork was implemented over three household visits, namely post-planting, crop-cutting, and post-harvest. The majority of the effort took place during the first visit, the post-planting visit, as it involved the administration of a questionnaire to the household as well as the objective measurements of plot area, the demarcation of cropcutting subplots, and the collection of leaf and soil samples (discussed below) on the randomly selected maize plot. The post-planting questionnaire included a standard individual-level module on household composition and basic characteristics (age, gender, education, etc.), a durable assets module, a farming assets module, questions on 2 Reporting of yields on intercropped plots is complicated by the fact that the denominator, planted area, could arguably be less than the full plot area or equal to the full plot area. For example, if maize and beans are planted in rows, one might argue (and a farmer might report) that only 50% of the plot area should be used in the denominator. Alternatively, one could argue that you could not have planted more maize on the plot even if beans were absent (assuming the maize plants are optimally spaced) and therefore the full plot area should be used. 5

6 the use and availability of agricultural extension services, and finally parcel and plot-level details. 3 The plot-level modules made up the bulk of the post-planting questionnaire, with questions on tenure status, cultivation status, which household members manage the plot, what farm implements were used, what farm management practices were employed (for example, tillage, crop rotation, etc.), post-planting labor inputs, and most importantly, farmer assessment of plot area, soil quality, and seed usage. It is critical to note that farmer assessment was made prior to any objective measurement so as to not influence the farmer response. In the second visit, the crop-cutting visit, enumerators harvested the demarcated subplots which were set during the post-planting visit in order to obtain objectively measured production quantities for the crop-cutting subplots, which are subsequently extrapolated to the full plot area. The final household visit took place following completion of all maize harvests. At this time, farmers were administered an additional questionnaire, which asked for the estimated total maize production per plot as well as fertilizer inputs and harvest labor inputs. The MAPS project focused on methodological validation of measurement of (i) quantity of maize production, (ii) agricultural plot area, (iii) soil fertility, and (iv) seed variety. While the primary objective of the MAPS project was a validation of subjective estimates of these indicators against objective measures, this paper will take advantage of the highly unique combination of objective measures only. The objective of this paper is not to compare methods, but rather, use the best method available. Descriptions of the methods used for maize production, plot area, and soil fertility follow: Maize production: A 4x4 meter subplot (divided into four 2x2 meter quadrants) and a separate 2x2 meter subplot were laid on the randomly selected maize plot during the post-planting visit following a strict protocol to ensure the location of the subplots was random. The subplots were roped off until harvest, when the enumerators were alerted and completed the harvest with the assistance of the farmer and a local assistant. The shelled maize from each 2x2 meter subplot was weighed and barcoded separately. The maize was then dried at a central, monitored location in Kampala until moisture content was in the range of 12 to 14 percent. Once desired dryness was met, the maize was re-weighed and the dry weight and final moisture content recorded. For analysis, all maize weights were normalized to 12 percent moisture content. Plot area: After walking the perimeter of the plot with the farmer to identify the proper boundary, enumerators re-paced the perimeter and measured the area with a Garmin etrex 30 handheld GPS device. The area was recorded on the questionnaire in square meters and the raw GPS track outline was stored. 3 Smallholder agricultural questionnaires in Uganda are structured such that there is a parcel of land, and within that parcel there may be multiple plots. The level of interest in this paper is the plot level. In the study, a parcel was defined as a contiguous piece of land with identical (uniform) tenure and physical characteristics. It is entirely surrounded by land with other tenure and/or physical characteristics or infrastructure e.g. water, a road, forest, etc. A plot was defined as a contiguous piece of land within a parcel on which a specific crop or a crop mixture is grown. A parcel may be made up of one or more plots. 6

7 Soil fertility: Analysis of soil fertility was done in partnership with the World Agroforestry Center (ICRAF). Plot level soil samples were collected from each of the selected fields following a protocol carefully designed to maximize the representativeness of the samples while maintaining feasibility of implementation. From each plot, four samples were collected from the top-soil (0-20cm depth) and combined to create one composite top-soil sample. Additionally, a single subsoil sample (20-50cm depth) was collected from the center of the plot. After being processed locally, the samples were shipped to ICRAF Nairobi, where all samples were subject to spectral soil analysis and approximately 10 percent were subject to conventional wet chemistry testing. A portion of this 10 percent sample was used to calibrate the prediction models, while the remainder was used to verify the predictions made onto the spectral data. For details, see (Shepherd & Walsh, 2002). The final results from the soil analysis include key indicators of soil fertility such as ph, texture analysis (percent sand, percent clay), cation exchange capacity, and the concentration of multiple elements and micronutrients, such as carbon, nitrogen, and potassium. Because MAPS was a small-scale methodological validation study, great care was take to ensure no missingness in these key variables, therefore, there are no concerns of missing data. There were, however, circumstances that required the sample to be restricted to 841 from 900. Plots which did not have any soil fertility measurement (due to mismatching of soil sample labels) or no crop-cutting (due to non-compliance of households) are excluded. 4 Table 1 presents descriptive statistics of key variables. Socio-economic indicators, plot manager characteristics, and agricultural variables are included, as all will be relevant for the analysis that follows. <<<< Table 1 about here >>>> IV. Methodology The following three-step empirical approach is employed: (1) estimate crop suitability measures at the plot-level; (2) analyze correlates of cultivating maize on less than suitable land; and, (3) stochastic frontier analysis to estimate production frontiers for each class of maize suitability. The contribution of this paper comes from the ability to execute each of these steps on the same sample, and from being able to do so with objectively-measured soil properties and crop production. i. Estimating maize suitability measures Estimating aggregate crop suitability measures requires comparing a vector of optimal soil properties against the levels of said properties on each plot. Crop suitability cannot be reduced to a single soil property, as several 4 The missingness of soil measurement is likely independent of production on the plot as the missingness stems from errors by the enumerator or laboratory. It could be argued, however, that non-compliance by the household (in which they harvest the crop-cutting subplot before the enumerator s arrival) could be a systematic problem in which households with fewer resources cannot afford to forgo the maize on the crop-cutting subplot. 7

8 Preliminary Draft Not For Citation properties affect plant growth simultaneously, and soil property requirements vary by crop. The crop suitability framework set forth by FAO (1976), and illustrated in Figure 1, will be used to gauge the level of suitability at the crop-soil property level. The maize suitability analysis completed here includes ph, cation exchange capacity (CEC), organic carbon, salinity (soil electrical conductivity), and plot slope (percent).5 <<<< Figure 1 about here >>>> After identifying the suitability class of each soil property individually, based on the critical values borrowed from Naidu (2006) and further reviewed and modified with input from the World Agroforestry Centre, a fuzzy membership method is used to construct a membership grade for each suitability class, allowing for identification of the suitability class that best approximates the soil sample overall. The fuzzy membership method is commonly employed in land suitability analysis with GIS data (Ahamed et al., 2000; Ceballos-Silva & López-Blanco, 2003; Hall et al., 1992; Kalogirou, 2002). The method is applicable to the plot-level MAPS data, however, as the data includes precise measures of soil parameters that are often extrapolated from lower resolution geospatial data. In this study, the unit of analysis is the plot rather than the pixel as in geospatial analysis. The fuzzy membership method, drawn heavily from (Ahamed et al., 2000; Hall et al., 1992), begins with an identification of the similarity, or Euclidean distance, between the vector of soil properties on each plot, x, and the representative vector for a given suitability class. The distance measure is constructed as follows:!!!!,!! =!!!!"! (1)!!! where:! = (!!,!!,,!! ) is the vector of soil parameters on a given plot; and!! = (!!!,!!!,,!!" ) is the representative vector of soil properties that corresponds to suitability class, c. Equation (1) results in a distance measure for each suitability class, where a higher score reflects greater divergence (less similarity) between the properties on a given plot and the respective suitability class. Subsequently, a membership grade is computed to allow for comparison of the degree to which the properties on a given plot belong to each suitability class.!!"#$! = 1!!!,!! 1!!!!!!!,!! (2) 5 Multiple variations of the soil suitability framework were created, each containing a different combination of key soil properties. Ultimately, the framework that was chosen was based on its superior predictive power in bivariate regression on yields. 8

9 Equation (2) results in a plot-level membership grade for each suitability class based on a given crop s representative vectors for each class. A plot will then be assigned the overall suitability class of that with the highest membership score. It is important to note that the method above assumes equal weights for each of the soil properties, which may be a strong assumption considering agronomic needs. Use of weights will be left for future analysis. To summarize, the distance measure is an absolute measure of the difference between the soil properties on a given plot and a specific suitability class, while the membership grade is a relative score, ranging from zero to one, indicating the relative fit of a plot into each suitability class. The membership grades for S1, S2, S3, and N, therefore, sum to one for each plot. ii. Correlates of cultivating sub-optimal land for maize Following the construction of plot-level maize suitability measures, analysis can shift to identifying the correlates of cultivating sub-optimal maize land. The following OLS model will be estimated:!!! =! +!!! +!!! +! (3) where!!! is the distance from the representative vector of properties for high maize suitability, S1. As the intention is to identify which types of farm households (or plot managers) are cultivating maize on sub-optimal land, a matrix of household characteristics, H, is included. These characteristics include age of the household head, highest education attained within the household, household size, female headship, the use of agricultural extension services, an index of agricultural implements, and an index of asset wealth. Property rights are also controlled for H. Districts effects are controlled for in matrix D. Parameters to be estimated include the constant,!, as well as vectors!! and!!.! is an error term, assumed to be normally distributed with mean zero. iii. Stochastic frontier analysis Aigner, Lovell, and Schmidt (1977) lay out the potential problems in minimizing the sum of squares of a simple production function, such as Cobb Douglas, in estimating the maximum output for a given level of inputs. The authors argue that this method of estimation inadequately explains observed deviations from the maximum output for given levels of inputs. In their proposed stochastic frontier model, they attempt to explain the variation in deviations from the modeled maximum output, or the production frontier, and predict an observation-level measure of technical inefficiency. Much of the literature on stochastic frontier models assumes a translog production function, in which inputs into the production function are also interacted (see Greene (2008), Sherlund et al. (2002), and Ekbom and Sterner (2008)). This can, however, result in an explosion of parameters to be estimated in the case of many inputs, such as in this agricultural model. Rather than the translog function, this paper assumes a log-linear Cobb Douglas model, following the seminal work of Aigner, Lovell, and Schmidt (1977) and the agricultural examples set forth by Deininger et al. (2007), Kilic et al. (2009), and others. The estimated stochastic frontier model is as follows: 9

10 Preliminary Draft Not For Citation! ln!! =! +!! ln!!" +!! (4)!!!!! =!! +!! (5) where!! is total maize grain output (in kilograms) on plot i, and! and!! parameters to be estimated. X is a vector of traditional economic inputs, including land area, household and hired labor inputs, and inorganic fertilizer usage. The distance measure (from highly suitable soil) is included in X for analysis conducted on the full sample. Because both pure stand and intercropped plots are included in the sample, a dummy for the cropping pattern and a continuous variable for the seeding rate are included in the X vector.6 Indicator variables for the administrative district are included in an attempt to control for large-scale differences in rainfall and duration of the agricultural season. The error term,!!, is disaggregated into a symmetric disturbance term,!!, and a non-positive disturbance,!!. The symmetric disturbance is assumed to be independently and identically distributed with! 0,!!!. It is assumed to be independent of!! and results from measurement error, climate-related shocks that affect production, and other exogenous shocks. The non-positive term,!!, represents the technical inefficiency of the household cultivating the plot, or the distance from the potential production frontier. It is assumed to be from a! 0,!!! truncated at zero (Aigner, Lovell, and Schmidt, 1977). Furthermore,!! is modeled as a linear function of variables that are believed to explain a household s technical efficiency or ability (Deininger et al., 2007; Kilic et al., 2009):!!! =! +!!!!" +!! (6)!!! Z is a vector of covariates used to explain technical efficiency, which includes age and gender of the head of household, the highest level of education achieved by household members, household size and dependency ratio, an asset-based wealth index and and agricultural implements index, the distance of the plot from the household, and a dummy variable indicating whether or not the household received agricultural extension services. Indicators of access to credit and public infrastructure could theoretically be included in Z, although actual indicators are more difficult to come by. The error term,!!, is assumed to be normally distributed with mean zero and truncated at (! +!!!!!!!!" ), such that!! remains non-positive. Technical efficiency and the parameters from equation (4) are estimated jointly using maximum log likelihood. The model, which substitutes equations (5) and (6) into equation (4), is estimated four times: (i) including all plots; (ii) including only plots classified as highly suitable (S1); (iii) including only plots classified as moderately suitable (S2); and, (iv) including only plots classified as marginally suitable (S3)7. Technical efficiency scores are 6 Pure stand plots are those on which only maize is grown. Intercropped plots are plots on which maize and at least one other crop is grown. The seeding rate is a ratio of the quantity of maize seed used on the plot to the quantity of seed that would have been used had the farmer planted only maize. The seeding rate is, therefore, bounded to (0,1] and equals 1 for all pure stand plots. The seeding rate is included in addition to the dummy variable for cultivation pattern because it is believed that some combinations of crops could improve potential maize yields. 7 No plots were classified as non-suitable (N). 10

11 computed following Battese and Coelli (1988). The technical efficiency scores can then be used to compute potential production and productivity for the given level of inputs. 8 V. Results i. Maize suitability Table 2 summarizes the maize suitability classifications as measured using the fuzzy set method described above. Classifying each plot based on the highest membership grade results in only 12.5 percent of plots being highly suitable, while the majority (75.5 percent) are moderately suitable. Twelve percent of plots are only marginally suitable. None were classified as not at all suitable. Note that classification into a specific group does not suggest that the plot-level soil properties fit that category in full. Rather, they are most closely aligned with that class relative to the other classes. A unique capability of this using this method with this data is the ability to identify specific constraints to suitability for each class. Table 3 identifies the limiting factors for each suitability class separately, enabling an assessment of what interventions would be most effective in increasing the suitability of plots from one level to the next. While highly suitable plots are most commonly constrained by ph, moderately suitable plots are constrained primarily by cation exchange capacity, and marginally suitable plots by cation exchange capacity and plot slope. The suitability classification is consistent with expectations with respect to agricultural productivity. Figure 2 illustrates the distribution of maize yields (kg/ha) by suitability class. Highly suitable plots realized an average of 1,614 kg/ha, while moderately and marginally suitable plots realized an average of 1,014 and 828 kg/ha, respectively. <<<< Table 2 about here >>>> <<<< Table 3 about here >>>> <<<< Figure 2 about here >>>> ii. Correlates of cultivating sub-optimal land for maize Results from OLS regression analysis on the correlates of maize suitability are reported in Table 4. The dependent variable, distance measure from S1, decreases with increased maize suitability. Therefore, a negative and significant coefficient suggests that the respective covariate is associated with improved maize suitability, while a positive and significant coefficient suggests the opposite. Controlling for district effects, few household characteristics exhibit significant explanatory power. The age of the household head is positively correlated with maize suitability, which may be the result of greater farming experience and ability to match crops and soils. Education of the household head is negatively correlated with maize suitability, however. Parcels that are leased 8 Potential output is computed as: observed output/ technical efficiency score. This follows Namonje (2015). 11

12 in are significantly more suitable for maize than parcels acquired by other means, a finding which may reflect positively on the ability of land markets to match soil suitability with farmer needs. Notably, the index of assetbased household wealth is not significantly correlated with the suitability of soils. The specified model fails to capture the correlates of maize suitability to any useful degree. The fit of the model, as measure by the R 2, is extremely low (R 2 = 0.067). As the aim of this model was to examine the socioeconomic, demographic, and geographic determinants of farming maize on land that is less than suitable, agricultural inputs were not controlled for. Doing so may improve the predictive power of the model. <<<< Table 4 about here >>>> iii. Stochastic frontier analysis a. Output elasticities The stochastic frontier analysis, results of which are presented in Table 5, immediately reveals differences in the output elasticities on plots with highly suitable (S1), moderately suitable (S2), and marginally suitable (S3) land. Most notably, the output elasticities with respect to inorganic fertilizer application and labor vary with suitability class, with highly suitable plots benefiting more greatly from inorganic fertilizer application and labor days than moderately or marginally suitable plots. Additionally, the output elasticity with respect to plot area is increasing with decreasing suitability. The coefficients suggest that returns to land area are greater on plots that are less than highly suitable. b. Technical efficiency <<<< Table 5 about here >>>> Factors associated with technical efficiency are also presented in Table 5. The coefficients model technical inefficiency, therefore a positive coefficient implies greater inefficiency and a negative coefficient greater efficiency. For plots categorized as highly suitable, only the agricultural implements index and household size exhibit significant coefficients. For these S1 plots, a greater agricultural implements index is associated with greater technical efficiency, as expected, while a larger household size decreases technical efficiency (an admittedly puzzling result). For S2 plots, where the majority of the sample is classified, no observable factors that are controlled for significantly correlate with technical efficiency. Finally, the technical efficiency of S3 plots is affected significantly by the agricultural implements index, in the same direction as the S1 plots, and the dependency ratio. The coefficient on dependency ratio for S3 plots suggests that as the dependency ratio of a household increases, so too does the technical efficiency, likely through labor channels. Technical efficiency scores, which represent the distance from the distance from the potential production frontier, are computed following Battese and Coelli (1988). Figure 3 presents the distribution of technical efficiency scores under each suitability class subgroup while Table 6 summarizes the scores, the potential production (in kilograms), and the potential yields (kilograms/hectare). Farmers cultivating S3 plots exhibit the highest technical efficiency scores, indicating they are operating most closely to the production frontier given their soil suitability. Farmers cultivating S1 and S2 plots, have higher realize yields and lower technical efficiency scores, suggesting 12

13 they have the potential to more greatly increase their maize production. The gap between mean realized maize yield and the potential yield (as estimated here) is 1,713 kg, or 106 percent of the realized mean yield, on S1 plots. On S3 plots, the difference is only 504 kg, or 61 percent of the realized mean yield. Figure 4 illustrates the efficient frontier with respect to the distance measure from S1, high suitability. It is clear that the production potential is constrained by soil suitability, and that the untapped potential is greatest on the most suitable plots. <<<< Figure 3 about here >>>> <<<< Table 6 about here >>>> <<<< Figure 4 about here >>>> VI. Conclusions We estimate a multi-dimensional measure of soil suitability, based on existing standards, for maize across a sample of approximately 900 households, spanning 4 districts, namely Serere, Sironko, Iganga and Mayuge, in Eastern Uganda; the leading maize-producing region in the country. This is made possible by collecting and labtesting plot-level soil samples based on international best practices in the context of a methodological household survey experiment. This contribution is key to cultivating a more nuanced understanding of micro-level variations in specific soil constraints, and hence the required interventions to achieve higher yields, which are otherwise likely to be missed in coarser, geospatial analyses. Classifying the sampled plots into three suitability classes, namely highly-suitable, moderately-suitable, and marginally-suitable, based on existing standards, and leveraging plot-level crop cutting-based maize yield measures allows for comparison of the distributions of observed maize yields by suitability class. We extend this analysis by estimating stochastic frontier models of maize yield separately for each suitability class to understand differences in (1) returns to factors of production, (2) technical efficiency, and (3) potential yield measures. Compared to the observed yields, the potential yield estimation provides a unique overview of maximum yield gains that can be achieved in each suitability class by increasing the efficiency with which the current set of inputs into agricultural production are utilized. Results clearly illustrate the production penalties for cultivating maize on land that is not highly suitable for maize production. Farmers cultivating only marginally suitable land are operating with higher technical efficiency and, thus, have less room for improvement than farmers cultivating more agronomically suitable land, given the condition of their soil. This result has implications for agriculture-based poverty reduction and food security policies. Effectively, by cultivating maize on land that is only marginally suitable rather than highly suitable, farmers limit their production potential by as much as 1,995 kg/ha, or 150 percent. The majority of the sample, 76 percent, was classified as cultivating moderately suitable land, for which the difference between mean observed yield and potential yield is 685 kg/ha, indicating a potential gain of 68 percent over current observed yields. Focusing policy, then, on either more appropriate matching of crop selection to soil suitability and/or addressing specific soil deficiencies that render the land unsuitable for a given crop can result in great gains in agricultural productivity. 13

14 The analysis above, however, fails to adequately identify what types farmers systematically suffer from this soil suitability penalty. Further analysis is necessary to determine who is being constrained by growing maize on less than highly suitable land. This, as well as additional analysis focused on simulating soil-specific agricultural interventions and extending the frontier analysis to take advantage of the panel nature of the data, is intended to be completed prior to the CSAE conference in March Also left for future research, which is aimed to be completed prior to the conference, is a replication of the above analysis relying strictly on geospatial data for soil properties rather than plot-level soil testing results. The intention of such a replication would be to examine the feasibility of utilizing geospatial data to conduct crop-specific soil suitability analyses, in a context in which plotlevel results are available to serve as a benchmark. 14

15 VII. References Ahamed, T. N., Rao, K. G., & Murthy, J. S. R. (2000). GIS-based fuzzy membership model for crop-land suitability analysis. Agricultural Systems, 63(2), Aigner, D.J.; Lovell, C.A.K.; Schmidt, P. (1977) Formulation and estimation of stochastic frontier production functions. Journal of Econometrics, 6: Battese, George, E., and Timothy Coelli. (1988). Prediction of firm-level technical efficiencies with a generalized frontier production function and panel data. Journal of Econometrics, vol. 38 (3), pages Boserup, E. (2007). Woman s Role in Economic Development. Earthscan. Cassman, K. G. (1999). Ecological intensification of cereal production systems: yield potential, soil quality, and precision agriculture. Proceedings of the National Academy of Sciences, 96(11), Ceballos-Silva, A., & López-Blanco, J. (2003). Evaluating biophysical variables to identify suitable areas for oat in Central Mexico: a multi-criteria and GIS approach. Agriculture, Ecosystems & Environment, 95(1), de Janvry, A., Fafchamps, M., & Sadoulet, E. (1991). Peasant Household Behaviour with Missing Markets: Some Paradoxes Explained. Economic Journal, 101(409), De Janvry, A., & Sadoulet, E. (2002). World poverty and the role of agricultural technology: direct and indirect effects. Journal of Development Studies, 38(4), Deininger, Klaus, W., Calogero Carletto, and Sara Savastano. (2007). "Land Market Development and Agricultural Production Efficiency in Albania." European Association of Agricultural Economists 104 th Seminar. Doss, C. R. (2002). Men s crops? Women s crops? The gender patterns of cropping in Ghana. World Development, 30(11), Doss, C. R., & Morris, M. L. (2000). How does gender affect the adoption of agricultural innovations? Agricultural Economics, 25(1), Duflo, E., Kremer, M., & Robinson, J. (2008). How high are rates of return to fertilizer? Evidence from field experiments in Kenya. The American Economic Review, 98(2), Ekbom, A., & Sterner, T. (2008). Production Function Analysis of Soil Properties and Soil Conservation Investments in Tropical Agriculture. Environment for Development Discussion Paper Series. Retrieved from Fafchamps, M. (1992). Cash Crop Production, Food Price Volatility, and Rural Market Integration in the Third World. American Journal of Agricultural Economics, 74(1), Fafchamps, M. (1993). Sequential Labor Decisions under Uncertainty: An Estimable Household Model of West- African Farmers. Econometrica, 61(5), FAO. (1976). A Framework For Land Evaluation. In Soils Bulletin (p. 72). Rome: FAO. FAO. (2015). Healthy soils are the basis for healthy food production, I4405E/1/ Available from: (Accessed 07/21/2015). Greene, W. H. (2008). The Econometric Approach to Efficiency Analysis. In H. O. Fried, C. A. K. Lovell, & S. S. Schmidt (Eds.), The Measurement of Productive Efficiency and Productivity Change (pp ). Oxford University Press. Retrieved from chapter-2 15