JOBS WORKING PAPER Issue No. 73 Does Agricultural Intensification Pay? Ghislain Aihounton Luc Christiaensen DOES AGRICULTURAL INTENSIFICATION PAY? Ghislain Aihounton Luc Christiaensen 2 © 2023 International Bank for Reconstruction and Development / The World Bank. 1818 H Street NW, Washington, DC 20433, USA. Telephone: 202-473-1000; Internet: www.worldbank.org. Some rights reserved This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http://creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: Aihounton, Ghislain, Luc Christiaensen. 2023. “Does Agricultural Intensification Pay?.” World Bank, Washington, DC. License: Creative Commons Attribution CC BY 3.0 IGO. Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. 3 Does Agricultural Intensification Pay? Ghislain Aihounton and Luc Christiaensen1 Abstract Modern inputs and mechanization are promoted across Africa to raise smallholder labor productivity and broker the structural transformation. Yet, adoption has remained low and the implications for returns to labor and labor allocation remain poorly understood. This paper explores the effects of different intensification packages on farm performance, market orientation, and food security using data from lowland rice farmers in Côte d'Ivoire. Employing a multinomial treatment effect model, the findings reveal that intensification increases land and labor productivity, especially when agro-chemicals and mechanized land preparation are combined. Returns to labor double to triple, inducing specialization and greater market orientation as well as greater food security, while productively releasing agricultural labor for other activities. Labor in agriculture becomes more waged. The gender balance remains the same. Child labor input does not decrease. The findings call for greater attention to labor productivity and confirm that agricultural intensification can pay, and enhance rural transformation. Key words: Rural transformation, intensification, farm performance, specialization, food security. JEL classification: N57, O12, O13, Q12, Q15, Q16, Q18 1 Ghislain Aihounton (daihounton@worldbank.org; aihountong@gmail.com) is Evaluation specialist consultant at the Rome Jobs and Labor Mobility Center, Jobs Group, World Bank. Luc Christiaensen (lchristiaensen@worldbank.org) is Head of the Rome Jobs and Labor Mobility Center, Jobs Group, World Bank. Financial support from the World Bank Jobs Multi-Donor Trust Fund is gratefully acknowledged. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. 4 1 Introduction Raising agricultural labor productivity has been widely recognized as an important ingredient to broker the structural transformation and alleviate poverty and food insecurity in developing countries (de Janvry & Sadoulet, 2020; Mellor, 2017). It requires, among others, the increased uptake of modern inputs and technologies such as improved crop varieties, inorganic fertilizer, irrigation and mechanization. They improve the productivity of land and labor and help farmers transition from growing mainly food crops (cereals) for auto-consumption to larger scale farmers growing food (and other) crops primarily for sale (Barrett et al., 2017; Liverpool-Tasie et al., 2017). Yet, in Sub-Saharan Africa (SSA), the adoption of modern agricultural technologies has lagged (Sheahan et al., 2017), the majority of the work continues to be done manually, and a large proportion of the labor force has remained in low-productivity agriculture (Beegle & Christiaensen, 2019). To boost intensification, several studies have investigated the adoption of labor-saving technologies such as farm mechanization2 and agrochemicals3. Recent studies about the performance of these technologies show partly positive, partly negative, and partly insignificant estimates (Magezi et al., 2021; Mano et al., 2020; Paudel et al., 2019). One reason for obtaining mixed results may be that studies have so far mainly focused on evaluating the effects of the different technologies separately, such as the effect of farm mechanisation (Paudel et al., 2019; Mano et al., 2020; Magezi et al., 2021) or yield increasing inputs on their own (Danso-Abbeam & Baiyegunhi, 2018; Sheahan et al., 2017). This ignores potential complementarities and questions the potential of one input/technology to shift productivity and profitability sufficiently on its own. 2 Diao et al., 2014; Ghosh, 2010; Kahan et al., 2017; Keil et al., 2016; Mottaleb et al., 2017; Pingali, 2007; Takeshima et al., 2013. 3 Athukorala et al., 2023; Emmanuel et al., 2016; Liverpool-Tasie et al., 2017; Martey et al., 2019. 5 Bundles of productive technologies have been shown to be needed to improve farm performance (Hörner & Wollni, 2022; Kassie et al., 2015; Oparinde, 2021; Park et al., 2018). Reviewing a vast array of studies on technology adoption and productivity in SSA (including 26 randomized control trials), Suri & Udry (2022, p39) conclude that : 1) no single constraint explains low productivity in African agriculture, 2) packages of interventions may be the most useful approach to foster adoption of new technologies and improve agricultural productivity, but 3) little is known about the effectiveness of multifaceted intervention packages, with their productivity and profitability likely also highly context specific. Much of the evidence and policy dialogue has further focused on land as opposed to labor productivity (Cock et al., 2022). From a structural transformation perspective, labor productivity and the returns to labor are the more relevant metrics. This study helps fill these voids. It examines whether agricultural intensification, involving multiple labor and land saving technologies can improve farm performance (productivity and profitability), increase specialisation and market orientation, and reduce food insecurity of agrarian households, thereby inducing the structural transformation. It further details how farm labor gets reallocated as intensification proceeds, across labor arrangements (self vs wage employed), across age groups and gender (van den Ban, 2011), and across the different production activities (land preparation, sowing, weeding, harvesting). This provides an inside micro picture of the household labor reallocation dynamics at work during the structural transformation which have been described mainly at the macro level so far. Overall, the focus is on the effects of different technology production packages on labor productivity and profitability, not the multiple constraints to adopting these packages, in input, factor and product markets or as related to knowledge or transaction costs. As such, the findings 6 inform about the possibility of profitable intensification, not on the particular interventions or intervention packages needed to broker it. Limited profitability of intensification has been raised regularly as a potential reason for low technology adoption in SSA (Burke et al., 2017; Liverpool- Tasie et al., 2017; Sheahan et al., 2013). Knowing that profitable intensification is possible is a precondition to the design of any intervention package. The empirical application concerns intensification among smallholder rice growers’ households in Côte d’Ivoire. Rice is an important staple crop for the majority of households in SSA, with rapidly expanding demand following income growth and urbanization continuing to exceed more gradually growing domestic supply. As a result, SSA continues to import around 40 percent of its rice consumption. It accounts for about one third of the global rice trade. Rice production in SSA remains largely rainfed, including in lowlands. This also holds in Côte d’Ivoire. Lagging production levels have been explained by delays in sowing time due to lack of mechanization, limited access to modern inputs, poor agronomic practices and lack of working capital among the supply chain actors (Komatsu et al., 2022; Niang et al., 2017; Tanaka et al., 2015, Wopereis, et al., 2013). In particular, baseline data from low-land rice farmers in 60 villages from three different agro- ecological regions in Côte d’Ivoire4 are used to investigate the productivity and profitability effects of four different, currently practiced, production technology packages: (a) a traditional production package (TP) involving manual land preparation, without any chemical inputs, (b) a semi- traditional production package (STP) including manual land preparation and the use of chemical inputs, (c) a semi-modern production package (SMP) involving animal traction for land 4 The data were collected in the context of a World Bank inclusive smallholder rice value chain development pilot (https://www.jobsanddevelopment.org/cote-divoire-country-pilot/). 7 preparation, combined with chemical inputs, and (d) a modern production package (MP) involving the use of a tractor and chemical inputs. Employing a multinomial treatment effect model, the findings reveal that intensification increases land and labor productivity, especially when agro-chemicals and mechanized land preparation are combined. Returns to labor double to triple, inducing specialization and greater market orientation as well as greater food security, while productively releasing agricultural labor for other activities. Labor in agriculture becomes more waged. The gender balance remains the same. Child labor input does not decrease. The paper proceeds as follows. Section 2 lays out the materials and metrics used, i.e., the sampling procedure and the data collected, the production packages applied in the study zone, and the outcome variables studied. Section 3 presents the econometric strategy. Section 4 discusses the findings. Section 5 concludes. 2 Materials and metrics 2.1 Information base and indicators The empirical application uses cross-sectional baseline data from rice growing households collected under a World Bank pilot study on smallholder inclusion in the rice sector in Côte d’Ivoire. Based on the prevalence of smallholder rice production in the regions, the regions’ agro- ecological diversity, and the presence of a medium-sized mill that could develop into an engine for inclusive rice value chain development through rice contract farming, three regions were purposively selected: Poro, Tchologo and Tonkpi. Within a 25 km buffer zone of the rice mill in 8 each region villages were randomly selected among those with lowland for rice cultivation. Proximity to the rice mill was retained as a criterion for the selection of the project pilot villages to ensure sufficient market access, for buying inputs and selling their products. In total, 21, 20, and 19 villages were selected in Poro, Tchologo and Tonkpi, respectively. Households were selected through a stratified random sampling, with each village a stratum. Prior to the survey, a census of all households within each selected village was conducted with the help of the village leaders. Subsequently, a list of rice growing households was compiled, among which only rice growing households with suitable lowland for rice cultivation were retained. From this final list, 15 rice growing households were each time randomly selected for the survey, resulting in a total of 1446 rice growing households interviewed in all three regions. 60 households that did not grow rice during the harvest season 2019 were removed. The final sample consists of 1386 rice growing households. The survey collected information on farmers’ socio-economic and demographic characteristics (e.g., age, sex, contact with extension agents, group membership, etc.), psychometric variables such as farmers’ information seeking attitudes and risk attitudes5, production related variables including the quantity of labor, fertilizers, pesticides, and equipment used as well as the total harvest and sales (including units and price). Special attention was given to the collection of detailed agricultural labor input data. This remains challenging and relatively uncommon in household surveys. Labor inputs were collected at the plot level in person-days6 for 5 We follow Läpple & Rensburg (2011) to derive the information attitude and risk aversion measures. These variables are obtained from Principal Component Analysis of a set of items that describes the interest of the producer in information gathering, and the level of risk aversion. The individual items were measured using a five points Likert scale, in which responders specify their level of agreement to a statement going from: (1) Strongly disagree; (2) Disagree; (3) Neither agree nor disagree; (4) Agree; to (5) Strongly agree. For instance, a household with a high risk averse score indicates that the household is more risk averse. 6 To measure the labor input provided by households for each agricultural activity, we use the concept of “person - days”. This information is collected by gender categories: men, women and children (over 5 years old) and for each farming activity. The number of person-days is calculated as follows. For each gender type (men, women, children), 9 each farming activity (land preparation, sowing, weeding, fertilizer application, pesticides application, and harvesting), disaggregated by gender (male, female, and children), both for both household and hired labor. This approach to measuring labor input in agriculture follows the latest recommendations (Sagesaka et al., 2021). Information on the type of equipment used in land preparation (e.g., manual, draught animal, tractor, etc.) was also collected. To have a broader picture of the household living standards, detailed information regarding household assets and food insecurity were asked. Food insecurity is measured using the food insecurity experience scale survey module (FIES-SM) (Cafiero et al., 2018). The household head and/or his/her spouse were asked to answer eight questions about the experience of food insecurity in the past twelve months. These detailed questions about the food experience allow to calculate the food insecurity experience scale score. To proxy for the level of village development and access to household labor for agriculture development, we also include geospatial population density values in the analysis, extracted from NASA Socioeconomic Data and Applications Center (SEDAC). The 2019 available data were merged, with the resolution of one square kilometer per pixel, to each household in our survey data, based on the GPS coordinates. Overall, the household interviews were conducted with the household head or the person responsible for farming activities. Table 1 lists the key outcome variables used in the analysis. Table 1: Definition of outcome variables Variable Definition Yield Rice output per hectare of rice land (kg/ha) Variable cost Cash expenses of seeds, fertiliser, herbicides, land preparation and hired labour per hectare of rice land (FCFA/ha) the respondent is asked how many people worked on the plot during the implementation of each of the farming activity and for each person we then asked how many days the person worked on this plot during the production season. The gender disaggregated amount of person-days for each activity is calculated as the sum of the person-days of all individuals of the same gender working on that activity. 10 Profit (return to land) Total net returns of rice output per hectare of rice land, i.e. Gross income (price times total output/ha) minus variable costs (FCFA/ha) Labor productivity Rice output (kg) per person-days of labour, including both hired and household labor Return to family labor Return to land, over total labor (Gross value output - variables cost, including the cost of hired labor, divided by total labor) Return to total labor Return to land, excluding the cost of hired labor over total labor (Gross value output - variables cost, excluding the cost of hired labor, divided by total labor) Total labor use per ha Sum of hired and household labor used (person-days/ha) Total rice land cultivated Total land cultivated with rice (ha) Degree of Proportion of rice sold over rice produced commercialisation Per capita rice Quantity of rice sold divided by the number of household members commercialisation (kg per capita) Food insecurity FAO household food insecurity experienced scale 2.2 Production packages Rice farmers in the dataset have applied a range of technologies in rice production. Considering the most representative production packages three criteria for categorization were retained: (i) whether land preparation has been done manually, with draught animal or tractor7, (ii) whether chemical fertilizer has been applied and (iii) whether chemical pesticides including fungicides, insecticides, and herbicides were used on the plots. This leads to the differentiation of four production packages (j) described in Figure 1, and to a sample size of 1039 households8. The first group, i.e., the traditional production package (TP), includes farmers who do not apply any yield increasing inputs, and perform all land preparation manually. The second group, i.e., the semi-traditional production package (STP), differs from the latter by the use of chemical fertilizer 7 Three levels of mechanization can be differentiated based on the power sources: human powered, animal powered, and fossil fuel powered (Magezi, et al., 2021; Malabo Montpellier Panel Report, 2018; Mano et al., 2020). Land preparation activities are key areas where animal and machine power could be used to improve agricultural productivity and diversity (Kirui, 2019). 8 The households that are not included in the analyses, were scattered across different technology combinations and much smaller in number. 11 and pesticides in rice production. In the third group, named the semi-modern production package (SMP), draught animal traction is used for land preparation in addition to agro-chemicals, and motorised land preparation (in addition to agro-chemicals) is performed in the fourth production package, i.e., the modern production package (MP). Overall, most farmers have adopted the intermediate production packages: 46 percent have adopted the semi-modern and 40 percent the semi-traditional production package. About one in ten farmers have adopted the traditional production package (possibly because of market constraints). Finally, as expected, only a small fraction of farmers was able to use a tractor for land preparation (4 percent) (in addition to agro- chemicals), suggesting that considerable efforts have to be done to reverse this trend. Figure 1: Distribution of the production packages. 50 46 40 40 Percent (%) 30 20 10 10 4 0 TP STP SMP MP Note: TP stands for traditional production package, STP for Semi-traditional production package, SMP for Semi-modern, and MP for modern production package. See table A1 in the supplementary appendix for further description. 12 3 Empirical Strategy 3.1 The multinomial treatment effect model To investigate the effect of rice intensification on farm performance, farm specialisation, market orientation, and food security, we employ a multinomial treatment effect model (Deb & Trivedi, 2006a, 2006b). It allows for the estimation of the effects of an endogenous multinomial ‘treatment’ variable on binary, count or continuous outcomes, while accounting for selectivity bias. Farmers’ decisions to adopt any of the production packages might be influenced by both observed and unobserved characteristics which could be correlated with the outcomes of interest. To capture the effect of each improved production package, following two stage estimation procedures were employed: (a) the drivers of the production packages adoption are modeled in the first estimation stage under a multinomial logit specification model and (b) their effects on farm outcomes are assessed in a second estimation by Ordinary Least Squares (OLS) with selectivity correction terms. In the first stage, we follow Deb & Trivedi (2006a) and assume that the adoption of any of the four production packages follows a mixed multinomial distribution. Let be a set of binary variables representing the observed improved production packages, such that T = (T1, T2, T3) standing for semi-traditional production package, semi-modern production package, and modern production package respectively. Traditional production package is the baseline choice. The probability of adopting any production package j can be expressed as: ( = | , , ) = ( ′ + l ), = 0,1,2,3. (1) Where denotes exogenous covariates including instruments, and are the associated parameters. l denotes unobserved characteristics affecting both the farmers’ decision to adopt a th production package and the outcome variables and are the selection effects, associated with 13 unobservable characteristics. g is a function that has a mixed multinomial logit structure. The probability of adopting one of the J production packages is thus determined by the following multinomial logit equation: exp( ′ + l ) (2) ( = ⎹ , I ) = 1 + ∑3 ′ , k =1, 2, 3. =1 exp( + l ) where k is one of the improved production packages and ( = ⎹ , I ) is the probability that the ith farmer adopt improved production package k. The second stage of the estimation can be specified as: (∗ ′ ) = + ∑= + ∑= l + µ (3) Where ∗ denotes the latent value underlying the outcome variables of interest such as productivity, profitability, as well as farm specialisation, market orientation indicators, and food security. The vector is the effect of the jth production package on the outcome variable of interest, relative to the base category (j=0). is a vector of exogenous covariates and is the associated coefficients; and µ is the error term. Following Deb & Trivedi (2006b), we specify the joint distribution of selection and outcome variables, conditional on the common unobserved factors, as : ( = , = 1| , , , ) = (′ + ∑3 3 ′ + ∑=0 l ) × g( + l ) =1 (4) To identify the second stage outcome equations, we included four instruments in , and estimate the model using a maximum simulated likelihood approach with 200 Halton sequence-based quasi- random draws. Dealing with Endogeneity As low-land rice farmers decide to adopt any technologies depending on both observed and unobserved factors, the choice of the production package is likely endogenous, which could bias 14 our estimates of the effects of the intensification packages on farm outcomes due to the presence of unobserved heterogeneity. In order to account for this, we use three approaches. First, we use a set of control variables such as education, age, participation in training, household size, dependency ratio, household assets, etc. Second, we also account for unobserved heterogeneity with two sets of attitudinal variables: (1) risk attitudes and (2) information seeking attitudes (Läpple & Rensburg, 2011). Third, we use a set of instruments that are likely to influence the adoption decision, but not the outcome variables to further control for endogeneity of the adoption decision. The set of instruments includes: availability of tractor service, availability of draft animal service, the proportion of households that use agrochemical inputs, excluding the household itself9, and the number of agricultural cooperatives in the village. To check for the validity of the instruments, we examine their relevance and exogeneity. Indeed, our instruments are (jointly) significantly correlated with the adoption decision, which supports the relevance criterion. Specifically, the availability of tractor or draft animal service, captures whether low-land rice farmers in each village have access to modern equipment for land preparation. The presence of the rental markets of these technologies in the village increases the probability of farmers using them, consistent with Diao et al. (2014) and Magezi et al. (2021). The number of agricultural cooperatives within the village also denotes the exposure of the individual farmers to new technologies, and their management at the farm level. Our core hypothesis is that the more households are exposed to the technologies, the more likely they are to adopt these technologies. 9 The proportion of households that used chemical inputs that we used, are inspired by Tabe-Ojong et al. (2021), Wuepper et al. (2018), and Magnan et al. (2015) who emphasize how social learning drives technology adoption in rural area. The instruments also capture social network effects, as farmers located in the same village are likely to be exposed to the barriers and constraints of adoption of agricultural technologies (Sellare et al., 2020). 15 Another condition is the exogeneity of the instruments, which holds that these variables affect our dependent variables only through the production packages. We argue that the adoption of any production technologies in each village may be more pronounced in villages with stronger leadership, and this may have affected farm outcomes of smallholders’ farmers, because farmers living in villages with stronger leadership may have access to improved technologies and appropriate extension services, which have the potential to increase farm performance (yield, labor allocation, market orientation, food security, etc.). It could also be that our set of production packages are promoted in locations with shortage of labor, with high population density, that is with high food needs, because yield increasing and labor saving technologies are promoted to deal with decreasing soil fertility and labor availability concerns. Furthermore, climatic factors such as rainfall can also affect farm performance, particularly yield, and may condition the adoption of a particular farm technology. The presence of all these factors may have affected farm outcomes and thus violate the exogeneity of the instruments. However, we account for these potential variables such as availability of labor, population density, farmers’ heterogeneity in terms risk and information seeking attitudes, and rainfall average in our specifications, to close potential pathways from the instruments to the outcome variables that may violate the exogeneity assumption of the instruments. Therefore, we are confident that the instruments are valid. 4 Findings 4.1 Rice growing households in Côte d’Ivoire The average household head in our data is 46.6 years old, with 17.8 years of experience in rice production, male and has 2.6 year of schooling (Table 2). Surprisingly, households that have adopted the traditional package are more educated than the other types. On average, a household 16 has between four and six members; households adopting the modern and semi-modern production packages tend to have more household members. Approximately half of the household members are either younger than 15, or older than 60. Interestingly, the more intensified the farm is, the more dependent members live in the households, implying that, insufficient household labor within the household can lead to the adoption of modern labor saving technologies in rice production. About one third of the households received training on rice production methods. Households adopting more intensive production methods are those with the highest rate of participation in training on rice production; they also receive more extension visits on their plots. Farmers that have adopted more intensified production package are more likely to have access to credit and to have an off-farm activity. The average household head adopting the semi-modern or modern production package is more likely to look for appropriate production information to improve its farm outcomes than the other groups, but there is no significance difference between the groups. A household’s risk attitudes likely affects its technology adoption decision. As expected, Traditional package (TP) farmers are more risk averse than others. Households that adopt modern production packages are found to have more information seeking behaviors than others. The average land cultivated with rice is 0.98 ha. It becomes increasingly larger as the farm gets more intensified. Walking distance to the rice plot is on average 50 minutes. Households adopting MP own approximately 1.6 times more total livestock units than the average household. They also have more household assets than other types. Villages where tractor or draught animal services are available or with a high number of agricultural cooperatives appear more likely to adopt modern agricultural technologies. Finally, labor is available in most of the villages. 17 Table 2: Descriptive statistics of explanatory variables Variables All TP STP SMP MP P-value Household head Age 46.63 47.50 46.03 47.03 45.63 0.944 Sex (1=male) 0.92 0.88 0.93 0.91 1.00 0.070 Years of education 2.59 5.58 3.03 1.49 2.89 <0.001 Experience in rice farming (years) 17.81 20.19 16.99 17.81 20.05 0.750 Received training (1=Yes) 0.36 0.17 0.44 0.32 0.58 <0.001 Number of training participation 1.17 0.53 1.43 1.07 1.53 0.319 Plot visited by extension agent (1=Yes) 0.45 0.17 0.47 0.46 0.74 <0.001 Number of plot visits by extension agent 1.16 0.39 1.38 1.12 1.45 0.325 Off-farm activity (1=Yes) 0.07 0.04 0.10 0.06 0.11 0.072 Access to credit (1=Yes) 0.32 0.03 0.37 0.35 0.29 <0.001 Information seeking attitudes10 (PCA) -0.03 0.50 -0.11 -0.16 1.01 0.360 Risk attitudes (PCA) -0.09 1.11 -0.33 -0.17 -0.04 <0.001 Household Household size 5.98 4.55 5.87 6.37 6.32 <0.001 Number aged 0-14 2.90 1.95 2.87 3.08 3.53 <0.001 Number aged 15-35 1.52 1.26 1.46 1.67 1.05 0.030 Number aged 36-65 1.33 1.05 1.33 1.38 1.66 0.001 Number aged > 66 0.23 0.29 0.22 0.24 0.08 0.215 Dependency ratio 0.49 0.45 0.50 0.49 0.53 0.054 Land cultivated with rice (ha) 0.98 0.59 0.92 1.10 1.34 <0.001 Total land owned (ha) 9.65 2.92 4.65 15.83 5.84 0.351 Wealth index (PCA score) 0.22 -1.00 0.32 0.38 0.56 <0.001 Total livestock units 1.67 0.04 1.60 2.03 2.76 0.010 Distance to plot (walking minutes) 49.7 46.7 54.7 46.2 48.6 0.121 Population density 287 386 279 274 256 0.357 Labor availability (1=Yes) 0.60 0.50 0.62 0.60 0.63 0.164 Instruments Availability of tractor service (1=Yes) 0.79 0.02 0.80 0.96 0.97 <0.001 Availability of draft animal service 0.23 0.02 0.30 0.22 0.32 <0.001 (1=Yes) Number of agricultural cooperatives 1.78 1.18 1.62 2.05 1.84 <0.001 Proportion of households using chemical 0.80 0.16 0.84 0.89 0.99 <0.001 inputs Region dummies <0.001 Poro 0.37 0.00 0.29 0.54 0.05 Tchologo 0.47 0.02 0.55 0.46 0.95 Tonkpi 0.16 0.98 0.15 0.00 0.00 Observations 1039 109 415 477 38 Notes: Mean values of key variables are shown. P-values of continuous variables are obtained by one way anova test for equality of mean values and P-values of binary variables are obtained by Pearson’s χ2-tests for equal proportions. 18 4.2 Labor allocation, input use, and performance On average, rice farmers spend 98 person-days in rice production per ha and obtain 1938 kg/ha of paddy rice. Adult male and female household labor allocation declines as the production package intensifies (Table 3). The MP records the highest yield, and it is more than twice as high as the yield obtained under the TP. Labor productivity and return to labor increase drastically (8 and 6-fold respectively) as the production packages get modernized, except with the SMP where draught animals are used for land preparation. It increases from 6 to 52 kg paddy rice per person- day per ha, and from 899 FCFA/person day/ha to 5,413 FCFA/person-day/ha. For comparison, the going daily wage rate for agricultural manual laborers is between 1,000 and 2,000 FCFA, depending on the location and task. Clearly, much labor is saved when the production package is mechanized, and the bi-variate findings (not controlling for selectivity bias) suggest that modernization pays. Net profit (FCFA/ha) also increases when moving from TP to MP. Along with increasing labor productivity as the production packages intensify come an increase in the area under rice suggesting some specializations, as well as an increase in the share of rice sold. Overall, the bivariate comparisons suggest that intensification of rice cultivation pays, that it increases the returns to labor and facilitates specialization, a transition to production for the market and higher welfare. Yet, to obtain more confidence in these descriptive findings, multivariate analysis is needed which also controls for endogeneity in the adoption of the packages. Food insecurity also drops progressively suggesting a corresponding increase in welfare. Table 3: Labor and input use, and performance Variables All TP STP SMP SM P-value Labor allocation (person-days/ha) Total household labor (Men) 29.31 70.78 29.75 20.98 10.11 <0.001 Total household labor (Women) 15.01 35.51 13.32 12.54 5.82 <0.001 Total household labor (Children) 5.32 3.09 5.54 5.87 2.39 0.315 Total household labor (person-days/ha) 49.6 109.48 48.6 39.4 18.3 <0.001 19 Variables All TP STP SMP SM P-value Total hired labor (person-days/ha) 48.2 86.0 61.1 29.4 35.6 <0.001 Total labor (HL + HH) 97.9 195.4 109.7 68.8 53.9 <0.001 Total hired labor cost (1000 FCFA) 56.3 89.7 64.3 40.2 75.8 <0.001 Implicit wage (FCFA/day) 1,166 1,047 1,049 1,379 2,142 <0.001 Other inputs Use of improved seed (1=Yes) 0.25 00 0.32 0.23 0.53 <0.001 Total quantity of seed (Kg/ha) 86.0 77.1 89.1 85.7 80.6 0.811 Water source (1=Low-land) 0.68 0.96 0.65 0.66 0.47 <0.001 Quantity of Urea (Kg/ha) 62.9 0.00 75.4 66.9 57.4 <0.001 Quantity of NPK (Kg/ha) 113.1 0.00 127.6 122.7 157.0 <0.001 Total quantity of fertilizers (Kg/ha) 176.0 0.00 203.0 189.6 214.4 <0.001 Use of fungicide (1=Yes) 0.31 0.00 0.36 0.32 0.58 <0.001 Use of herbicide (1=Yes) 0.90 0.00 1.00 1.00 1.00 <0.001 Total quantity of herbicides (L/ha) 8.92 0.00 10.25 9.77 9.20 <0.001 Production cost (1000 FCFA/ha) 142.22 101.27 161.49 132.30 173.79 0.251 Productivity and profitability Yield (kg/ha) 1937.7 1036.7 2081.8 1993.7 2246.6 <0.001 labor productivity (kg/person-days) 30.3 6.5 26.8 37.1 51.9 <0.001 Profit (1000 FCFA/ha) 148.4 54.2 150.8 166.8 163.2 <0.001 Alternative profit (1000 FCFA/ha) 99.71 -52.88 102.69 128.46 143.96 <0.001 Return to family labor (FCFA/person- 7142 1440 8060 7065 14463 <0.001 days) Return to total labor (FCFA/person- 3146 899 2766 3810 5413 <0.001 days) Farm specialisation, market orientation and food security Share of cash crop area (%) 33.35 54.83 29.63 33.01 16.70 <0.001 Rice plot size (ha) 0.98 0.59 0.92 1.10 1.34 <0.001 Rice production per ha (kg/per capita) 356.24 162.12 360.58 379.76 570.29 <0.001 Proportion of rice commercialized 0.26 0.08 0.30 0.25 0.50 <0.001 Rice sold per capita (Kg/capita) 137.40 24.67 151.20 127.48 434.52 <0.001 Food insecurity 4.28 7.24 4.08 3.93 2.42 <0.001 Observations 1039 109 415 477 38 Notes: Mean values are shown. P-values of continuous variables are obtained by one way anova test for equality of mean values and P-values of binary variables are obtained by Pearson’s χ2-tests for equal proportions. All labor related variables are expressed in person-days/ha. Implicit wage is obtained as ratio of total expenses on hired labor divided by average person-days hired. Turning to the labor input in more detail, Table 4 presents a deep dive in the evolution of labor inputs by age, sex, and activity along the four production packages. Notably, as the production package intensifies, households tend to use less household labor in most farming activities. Specifically, much labor is spent in the traditional production package in land preparation, weeding and harvesting, compared to the other improved production packages. This 20 difference can be attributed to the use of labor-saving technologies in the intensified production packages. Male and female household labor allocated on the plots tend to reduce. However, a slight increase is observed in the amount of household labor spent by children, except for modern production packages. The amount of hired labor applied on the plots also decreases as production processes intensify, but at a slower pace than household labor. Table 4: Labor input by agricultural activity All TP STP SMP MP P-value 1. Land preparation Use of household labor (1=Yes) 0.85 0.93 0.86 0.86 0.47 <0.001 Numbers of Person-days/ha (Men) 8.95 25.38 9.04 5.74 1.24 <0.001 Numbers of Person-days/ha (Women) 2.85 5.39 2.59 2.72 0.13 0.001 Numbers of Person-days/ha (Children) 1.47 0.82 1.48 1.72 0.08 0.407 Total Person-days (Household) 13.28 31.58 13.11 10.18 1.45 <0.001 Use of hired labor (1=Yes) 0.65 0.86 0.71 0.54 0.82 <0.001 Person-days/ha of hired labor 12.88 32.04 19.51 3.65 1.44 <0.001 Person-days/ha of total labor 26.16 63.63 32.63 13.82 2.89 <0.001 2. Sowing and transplanting Use of household labor (1=Yes) 0.91 0.94 0.87 0.94 0.89 0.002 Numbers of Person-days/ha (Men) 4.91 14.45 4.81 3.10 1.32 <0.001 Numbers of Person-days/ha (Women) 3.68 9.00 3.46 2.84 1.28 <0.001 Numbers of Person-days/ha (Children) 1.34 0.83 1.46 1.42 0.59 0.729 Total Person-days (Household) 9.93 24.28 9.74 7.36 3.19 <0.001 Use of hired labor (1=Yes) 0.56 0.78 0.67 0.42 0.50 <0.001 Person-days/ha of hired labor 12.72 19.71 17.83 6.85 10.62 <0.001 Person-days/ha of total labor 22.65 43.98 27.57 14.21 13.81 <0.001 3. Fertilizer: Urea NPK Use of Urea/NPK (1=Yes) 0.90 0.00 1.00 1.00 1.00 <0.001 Use of household labor (1=Yes) 0.86 0.00 0.95 0.96 0.97 <0.001 Numbers of person-days/ha (Men) 2.55 0.00 3.34 2.50 1.91 <0.001 Numbers of person-days/ha (Women) 0.67 0.00 0.72 0.80 0.43 0.001 Numbers of person-days/ha (Children) 0.31 0.00 0.41 0.29 0.28 0.397 Total person-days (Household) 3.53 0.00 4.47 3.59 2.62 <0.001 Use of hired labor (1=Yes) 0.12 0.00 0.13 0.13 0.13 0.001 Person-days/ha of hired labor 0.55 0.00 0.71 0.55 0.28 0.510 Person-days/ha of total labor 4.07 0.00 5.18 4.13 2.90 <0.001 4. Weeding Use of household labor (1=Yes) 0.95 0.99 0.95 0.94 0.92 0.123 Numbers of person-days/ha (Men) 3.97 9.06 4.39 2.60 1.86 <0.001 Numbers of person-days/ha (Women) 1.19 4.80 1.28 0.37 0.32 <0.001 Numbers of person-days/ha (Children) 0.28 0.00 0.38 0.28 0.00 0.802 Total person-days/ha (Household) 5.44 13.86 6.05 3.25 2.17 <0.001 Use of hired labor (1=Yes) 0.18 0.20 0.18 0.17 0.18 0.886 Person-days/ha of hired labor 1.82 6.63 1.72 0.93 0.36 <0.001 21 All TP STP SMP MP P-value Person-days/ha of total labor 7.27 20.49 7.77 4.19 2.54 <0.001 5. Harvesting Use of household labor (1=Yes) 0.91 0.95 0.88 0.93 0.89 0.033 Numbers of person-days/ha (Men) 8.93 21.90 8.17 7.04 3.79 <0.001 Numbers of person-days/ha (Women) 6.62 16.31 5.26 5.82 3.66 <0.001 Numbers of person-days/ha (Children) 1.92 1.44 1.81 2.16 1.44 0.264 Total person-days/ha (Household) 17.47 39.65 15.24 15.02 8.89 <0.001 Use of hired labor (1=Yes) 0.87 0.83 0.91 0.84 0.92 0.010 Person-days/ha of hired labor 20.25 27.61 21.35 17.40 22.90 <0.001 6. Labor allocation Total household labor (Men) 29.31 70.78 29.75 20.98 10.11 <0.001 Total household labor (Women) 15.01 35.51 13.32 12.54 5.82 <0.001 Total household labor (Children) 5.32 3.09 5.54 5.87 2.39 0.315 Total household labor 49.65 109.38 48.62 39.39 18.32 <0.001 Total hired labor 48.23 85.98 61.12 29.39 35.61 <0.001 Total labor (HL + HH) 97.87 195.36 109.74 68.78 53.93 <0.001 Observations 1039 109 415 477 38 Notes: Mean values are shown with standard deviations in parentheses. P-values of continuous variables are obtained by one way anova test for equality of mean values and P-values of binary variables are obtained by Pearson’s χ2-tests for equal proportions. All labor data are expressed in person-days/ha. 4.3 Econometric results 4.3.1 Factors explaining the adoption of modern farm technologies. Table 5 presents the estimated results from the first stage of the multinomial endogenous treatment effects model. Following insights regarding the correlates of the adoption of different technology packages emerge. First, those with access to credit and wealthier households are more likely to adopt improved farm technologies in rice production, highlighting the continuing importance of access to financial markets and the ability to take risks. Second, contact with extension services and the number of plot visits by extension agents is also positively correlated with adoption of intensified production packages. Third, there appears less adoption of intensification on sole-managed plots (see Haider et al., 2018), possibly because the benefits stemming from managing those plots individually contributes less to the household’s overall 22 welfare. Table 5: Factors affecting the adoption of the technological packages (1) (2) (3) VARIABLES Semi-traditional Semi-modern Modern package package package Age of the household head 0.02 0.02 0.00 (0.03) (0.03) (0.04) Year of education 0.08 -0.02 0.12 (0.06) (0.06) (0.08) Household size -0.12 -0.05 -0.11 (0.14) (0.15) (0.15) Dependency Ratio 0.72 0.61 1.38 (1.43) (1.48) (1.65) Plot management type (1=Individual) -2.03*** -1.98*** -2.46*** (0.59) (0.63) (0.88) Experience in rice production -0.03 -0.03 0.02 (0.03) (0.03) (0.03) Number of training participation 0.43** 0.40* 0.32 (0.21) (0.22) (0.25) Plot visited by extension agent (1=Yes) 1.96** 2.28** 3.07*** (0.91) (0.92) (1.09) Off-farm activity (1=Yes) 1.29 0.85 1.20 (1.36) (1.38) (1.44) Access to credit (1=Yes) 3.42*** 3.19*** 2.46*** (0.85) (0.87) (0.95) Attitudes towards information 0.05 0.04 0.35 (0.15) (0.16) (0.23) Risk Attitudes -0.34 -0.26 -0.37 (0.24) (0.25) (0.32) ln(Total Plotsize) 0.24 0.32 0.61 (0.56) (0.56) (0.61) Wealth index 0.83*** 0.95*** 0.70*** (0.18) (0.19) (0.21) ln(Population density) -0.09 -0.27 -0.71* (0.39) (0.40) (0.43) ln(Distance to plot) -0.58* -0.76** -0.58 (0.35) (0.36) (0.40) Labor availability (1=Yes) -0.24 -0.40 -0.73 (0.48) (0.58) (0.64) Availability of tractor service (1=Yes) 3.99*** 5.64*** 3.68*** (0.90) (1.21) (1.19) Availability of draught animal service 2.39*** 1.86** 2.16** (1=Yes) (0.88) (0.95) (1.06) Number of agricultural cooperative 0.23 0.63* 1.04** (0.34) (0.35) (0.43) Proportion of households using chemical 5.68*** 6.99*** 37.64*** inputs (expressed as a fraction 0-1) (1.77) (2.00) (6.42) ln(Rainfall average) -2.18 4.76* -22.88** 23 (1) (2) (3) VARIABLES Semi-traditional Semi-modern Modern package package package (2.34) (2.77) (11.59) Constant -7.73 13.44 -109.72*** (7.48) (9.20) (41.25) Observations 1,039 1,039 1,039 Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1 At times, the different correlates also affect the intensification practices differently. Plots located far away from the dwelling, for example, are less likely to be ploughed with a draught animal, while distance to the plot does not matter for the adoption of the modern package (with mechanized ploughing). Similarly, while the use of draught animal service seems appropriate for largely humid areas, the use of tractor services is more suitable for less humid areas. In areas with limited mechanized services, farmers tend to rely on manual work. Concerning the instruments, the results show that the availability of draught animals, tractor services and chemical fertilizers and pesticides, are also driving factors of the adoption of more intensified production packages. Consistent with Kirui (2019), the results show that the existence of a rental market is a key driver for the adoption of machinery. Moreover, households located in a village with a high number of agricultural cooperatives, are more likely to adopt the technological packages with mechanization. Integrating cooperatives effects in the adoption and diffusion of new technologies has already been documented as important in the empirical literature (Sellare et al., 2020). 4.3.2 The effect of intensification on farm performance and structural transformation This sub-section summarises the effects of the intensification packages on farm performance, specialisation, market orientation and food security of rice growing households in Côte d’Ivoire. Full results are presented in tables A2 to A9 in the supplementary appendix. 24 Effects of rice intensification on productivity and profitability Compared to the traditional package with mean yield of 1037 kg/ha, yields increase progressively by 641, 794 and 1334 kg/ha going from the semi-traditional, to the semi-modern and the modern production packages respectively (Table 6), or an increase by 62, 77, and 129 percent. Other studies have also reported an increase in yields when draught animals and tractors are used (Mano et al. (2020) for the use of tractors in land preparation in Côte d’Ivoire and Kirui (2019) using pooled data from 11 countries in Africa). As for profit (return to land), there is only a positive and significant effect of the semi- modern package. Specifically, rice profit increases by 80.7 thousand FCFA/ha, which corresponds to a 48 percent increase for the SMP (compared to the traditional package). Consistent with Mano et al. (2020), modern production packages involving tractor use in land preparation is not significantly associated with a greater return per ha. The increase in yields (and gross revenue) does not suffice to compensate for the greater cost of production (higher input use) and especially the greater use of hired labor. These results are also consistent with Adu-Baffour et al. (2019), who demonstrate that the demand for hired labor in mechanized production systems, increases due to land expansion and due to a shift from family labor, including that of children, to hired labor. As shown in the next column, the real benefit from the modern package lies in the greater return to labor (i.e. family labor). As family labor input in rice production comes down substantially, it also frees up time for other activities (on or off the farm). Turning to labor productivity, the ultimate variable of interest in the context of structural transformation, this increases drastically as production systems intensify, by 22, and 32 kg of paddy rice per person-days, in the semi-modern and modern production packages respectively, implying a three (367 percent) to five (533 percent) fold increase, respectively, when compared to 25 the mean labor productivity of about 6 kg of paddy rice per person-days of the traditional production package. Magezi et al. (2021) also report substantial labor productivity gains with different types of tractors used in land preparation (between 12 and 14 kg per person days of labor use). The higher gain obtained here likely follows from the additional effect of using chemical fertilizers and pesticides as part of the advanced production technology. Similar patterns are observed when looking at the return to family labor. It increases by 3 3330 and 7370 FCFA per person-days in the semi-modern and modern production packages, respectively. Compared to the mean values of 1440 FCFA of the traditional packages, this corresponds to more than doubling (231 percent increase) and quintupling (512 percent) the return to family labor in the respective intensified production packages. Increasing patterns are also observed when looking at the return to total labor. It increases by 2090 and 3180 FCFA per person- days in the semi-modern and modern production packages, respectively. Compared to the mean values of 899 FCFA of the traditional packages, this corresponds to more than doubling (232 percent increase) and tripling (354 percent) the return to family labor in the respective intensified production packages. Despite the limited effect obtained for the semi-traditional package, our results confirm that modern green revolution type technologies in rice production in Côte d’Ivoire used in combination result in higher land and especially much higher labor productivity and that they also substantially increases returns to labor. Put differently, low adoption of these packages should not be primarily attributed to their inability to generate sufficient returns, which appear on average sufficiently large to compensate for the greater risk involved (they are more than twice as profitable). Other factors must be at work to explain relatively limited adoption. 26 Table 6: Effects of rice intensification on farm performance Variables Yield (kg/ha) Profit (1000 Labor Productivity Return to family Return to total FCFA/ha) (kg/person-days) labor labor Semi-traditional 641.45*** -25.20 4.49 -0.18 0.12 package (176.11) (40.32) (2.86) (1.93) (0.36) Semi-modern 793.69*** 80.70*** 21.64*** 3.33* 2.09*** package (178.67) (26.50) (3.19) (1.79) (0.41) Modern package 1333.61*** 52.66 31.88*** 7.37** 3.18*** (240.62) (42.41) (6.96) (3.48) (1.03) Control variables Yes Yes Yes Yes Yes Note: Control variables include age of the household head, year of education, household size, dependency ratio, plot management type (1=Individual), experience in rice production, number of training participation, plot visited by extension agent (1=Yes), off-farm activity (1=Yes), access to credit (1=Yes), attitudes towards information, risk attitudes, ln(total plot size), wealth index, ln(population density), ln(distance to plot), and ln(rainfall average). Yield is the rice output per hectare of rice land (kg/ha); profit refers to the total net returns of rice output per hectare of rice land; labor productivity indicates rice output (kg) per person-days, return to family labor captures the return to land in 1000 FCFA per person-days of family labor and return to total labor captures the return to land in 1000 FCFA per person-days of total labor. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 27 Table 7: Effects of rice intensification on labor allocation in rice production by age and type Variables Total labor use Household labor Men labor Women labor Children labor Hired labor Semi-traditional -30.61*** -22.44** -16.51** -17.50*** 3.83 2.22 package (13.39) (10.37) (7.27) (6.02) (2.36) (6.46) Semi-modern -104.45*** -67.39*** -46.06*** -17.83*** 6.70*** -47.27*** package (13.95) (11.55) (7.22) (5.25) (1.99) (5.68) Modern package -69.52*** -62.33*** -43.87*** -29.00*** 3.16 -18.73*** (14.86) (10.20) (7.43) (6.01) (2.34) (7.13) Control variables Yes Yes Yes Yes Yes Yes Note: Control variables include age of the household head, year of education, household size, dependency ratio, plot management type (1=Individual), experience in rice production, number of training participation, plot visited by extension agent (1=Yes), off-farm activity (1=Yes), access to credit (1=Yes), attitudes towards information, risk attitudes, ln(total plot size), wealth index, ln(population density), ln(distance to plot), and ln(rainfall average). All labor outcomes variables are expressed in person-days. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 28 Table 8: Effect of rice intensification on labor allocation in rice production by activity. Variables Total labor use Household labor Men labor Women labor Children labor Hired labor Land preparation Semi-traditional -15.12**( 5.98) -5.85(3.83) -8.33**(3.51) -2.29(2.31) 0.63(0.83) -4.48(3.56) Semi-modern -44.14***(6.29) -22.79***(4.17) -20.06***(3.87) -0.44(2.13) 1.67**(0.72) -25.93***(3.16) Modern package -40.49***(6.35) -20.46***(4.32) -19.12***(3.83) -4.22*(2.36) -0.53(0.69) -21.43***(3.66) Sowing Semi-traditional -1.92(5.65) -4.15(3.33) -3.47(2.24) -3.60(2.22) 1.92**(0.94) 5.21(3.36) Semi-modern -26.93***(5.03) -16.71***(3.86) -9.48***(2.07) -6.80***(2.19) 1.41**(0.66) -13.86***(2.76) Modern package -12.41**(5.96) -11.23***(3.43) -8.11***(2.02) -5.40***( 2.04) 0.76(0.65) -0.19(3.50) Fertilizer application Semi-traditional 6.69***(0.62) 6.25***(0.13) 4.05***(0.64) 0.78***(0.18) 0.71***(0.23) 1.26***(0.37) Semi-modern 5.03***(0.56) 4.25***(0.14) 1.70***(0.26) 1.42***(0.22) 0.45***(0.14) 1.01**(0.42) Modern package 4.59***(0.76) 1.42***(0.28) 2.61***(0.37) 1.01***0.28) 0.77***(0.27) 1.20**(0.49) Weeding Semi-traditional -5.94**(2.69) -3.43**(1.33) -1.36(0.97) -2.54(1.57) 0.35*(0.21) -1.67(1.52) Semi-modern -14.37***(3.56) -10.34***(1.67) -5.05***(0.91) -5.11***( 1.92) 0.39**(0.21) -4.83**(2.10) Modern package -10.09***(3.15) -7.89***(1.56) -4.32***(1.08) -4.18**(1.96) 0.35(0.31) -2.58(1.64) Harvesting Semi-traditional -11.98**(4.94) -14.32***(1.92) -6.22***(2.34) -5.92(4.38) 0.55(0.66) 3.354( 2.68) Semi-modern -24.28***(4.95) -31.40***(2.26) -14.72***(2.28) -9.91***(2.78) 2.33***(0.77) -3.90(2.91) Modern package -18.14**(7.61) -31.21***(4.79) -13.51***(2.48) -11.25*(6.19) 1.50(0.99) 3.44(4.64) Note: Control variables include age of the household head, year of education, household size, dependency ratio, plot management type (1=Individual), experience in rice production, number of training participation, plot visited by extension agent (1=Yes), off-farm activity (1=Yes), access to credit (1=Yes), attitudes towards information, risk attitudes, ln(total plot size), wealth index, ln(population density), ln(distance to plot), and ln(rainfall average). All labor outcomes variables are expressed in person-days. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 29 Effects of rice intensification on labor allocation By raising labor productivity rice intensification also frees up labor for other activities, a core corollary of the structural transformation. This begs the question how labor allocation within rice production changes. Does it become less self-employed and more waged (less family and more hired wage labor use) and do the changes differ by gender and age. To begin, Table 7 confirms that, in the aggregate, total labor use in rice production declines, as rice production intensifies, by between 31 (STP) and 104 (SMP) person-days per ha compared to the traditional package. This comes along with a shift from family towards waged labor. While family labor input (per ha) exceeds hired labor use in the traditional package (109 vs 86 days per ha) (Table 3), this changes as rice production intensifies with hired labor input exceeding family labor input in the STP and SM systems. The multivariate regression findings confirm this shift, with the decline in household labor substantially larger than the decline in hired labor (Table 7). Across technology packages male household labor input in rice production is on average about twice as large as female labor input, suggesting a proportionate decline as rice production intensifies (Table 3). This is broadly confirmed in the multivariate analysis (Table 7), with the exception of the STP package where both male and female household labor input declined with a similar amount (by 16-17 person-days/ha). Female labor input declined also somewhat more rapidly in the modern package. Importantly, child labor input does not decrease with rice intensification. In fact, it even increases under the semi-modern package (by 6.7 person-days). In developing countries, it is common for children (boys) to do the land preparation when using draught animals. The technology does not require a lot of physical strength, and boys can easily perform the task. That said, in the semi-modern cultivation system children are slightly more involved across activities (land preparation, sowing, fertilizer application, weeding, harvesting) (Table 8). The greater involvement in land preparation of boys when using draught animals only partially explains the 30 greater use of child labor under the SMT package. Different technologies affect labor use in rice production at different stages. Mechanization typically begins with land preparation because this is where the labor savings are largest, followed by harvesting. Consistently, when looking at the effects on labor use by activity, Table 8 shows how the majority of the labor savings in the semi-modern and modern package follow from labor savings during land preparation (by about 20 person days of family labor (virtually all male) and another 20-25 days of hired labor), followed by labor savings in harvesting (for both men and women). Except for land preparation, hired labor does not really decline for any of the other activities. There the labor savings are fully absorbed by family members. Finally, except for fertilizer application, where labor use increases—there is no fertilizer use in the traditional system by definition, labor use declines across all activities. Effect of rice intensification on land expansion, market orientation and welfare Along with intensification (and greater labor productivity) usually also comes specialization and greater market orientation. This is confirmed in the study area. The adoption of labor-saving technologies significantly increases the total land cultivated with rice (Table 9). This increase could be explained by the use of draught animal and tractor for land preparation to deal with labor shortage or high cost of manual land preparation. Rice area increases by 0.48 and 0.54 ha in the semi-modern and modern production packages respectively. Compared to the mean value of 0.54 ha for the control group, these effects correspond to a 81 and 91 percent increase in total land cultivated with rice. The likelihood of selling some of the rice produced increases by 47, 27 and 78 percent in the STP, SMP and MP respectively. For the modern production package, the share of rice sold further increased by 39 percentage points. As auto-consumption needs often also affect farmers’ 31 market orientation, the adoption of the improved production packages on the amount of rice sold per capita was also analyzed. It increases by 80, 94 and 374 kg of rice sold per capita for the STP, SMP, and MP respectively. To address the question whether greater rice orientation following intensification also results in greater welfare, household income and or consumption data would be needed. In the absence of such information, Table 9 looks at household food insecurity instead. Consistent with the notion of reduced poverty (and increased welfare), food insecurity decreases significantly by 2.22, and 3.17 points on the FAO household food insecurity experience scale in the semi-modern and modern production packages respectively. 32 Table 9: Effect of production packages on agricultural extensification, commercialisation and food security Variables Area under rice Decision to Share of rice sold Rice sold per capita Food insecurity cultivation (ha) commercialise (0-1) (kg) (1=Yes) Semi-traditional 0.09 0.47*** 0.28 79.92*** -0.08 package (0.09) (0.03) (0.20) (25.94) (0.78) Semi-modern package 0.48*** 0.27*** 0.07 93.62*** -2.22*** (0.12) (0.04) (0.09) (27.66) (0.48) Modern package 0.54*** 0.78*** 0.39** 374.41*** -3.17** (0.16) (0.04) (0.16) (87.90) (1.41) Control variables Yes Yes Yes Yes Yes Note: Control variables include age of the household head, year of education, household size, dependency ratio, plot management type (1=Individual), experience in rice production, number of training participation, plot visited by extension agent (1=Yes), off-farm activity (1=Yes), access to credit (1=Yes), attitudes towards information, risk attitudes, ln(total plot size), wealth index, ln(population density), ln(distance to plot), and ln(rainfall average). Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 33 5 Concluding remarks Green Revolution type technologies such as agro-chemicals and mechanization are being promoted in Africa to raise smallholder labor productivity, specialization and greater market orientation. This in turn fosters Africa’s structural transformation through the productive release of agricultural labor (Beegle & Christiaensen, 2019). Yet, input adoption and mechanization in Sub-Saharan Africa remain limited (Sheahan & Barrett, 2017) and studies have started to question whether the adoption of intensified agricultural production methods is indeed profitable (Burke et al., 2017; Liverpool-Tasie et al., 2017). When adopted, inputs and mechanization are furthermore rarely applied in combination, thereby foregoing possible synergetic effects, possibly one of the reasons for their limited adoption. Using cross-sectional data collected from rice farmers in Côte d’Ivoire and instrumental variable techniques to control for endogeneity in the adoption of different technology packages, this paper investigates to what extent rice intensification affects farm performance, specialisation, market orientation and food security of smallholder rice growing households. Rice intensification is measured, through four production packages, compiled through different combinations of available labor saving technologies such as using tractor or draught animals for land preparation, chemical fertilizer for soil fertility, and herbicides for weeding. Using available information about labor inputs across activities, by age, gender, and type (family/hired), the implication for employment (number of jobs, type of employment (self-versus wage-), and worker characteristics (adult/child, male/female) is further reviewed in detail, including the agricultural activities through which the changes occurred. Results indicate significant positive effects of the adoption of more intensive production packages on land and labor productivity, as well as the return to labor, which increases more than 34 two to threefold with the adoption of semi-modern and modern production packages respectively. Labor productivity gains and monetary returns to labor are especially large when modern input use (fertilizer and pesticides) is combined with the use of draught animals, and especially tractor services, for land preparation, highlighting the benefits from the combined use of different technologies (including mechanization) to make a substantial difference. Importantly, these results do not carry through to the returns to land (net profit per ha), where statistically significant gains are only found for the semi-modern production package. For the modern production package, yield gains are not large enough to compensate for the larger input costs and especially the greater reliance on hired labor. The results further confirm that rice intensification increases specialization (larger shares of the farm land allocated to rice production) (at least for the semi-modern and modern packages where inputs and mechanization are combined) and greater market orientation. This in turn comes along with greater food security and the productive release of labor for other crops and off-farm activities. Between 70 and 100 days less person days are used per ha rice cultivated (for the modern and semi-modern production packages), while yields more than double compared to the traditional package. The largest labor reduction follows from reduced use of labor in land preparation, followed by harvesting. It is largest among family members, but equally felt among men and women. Reduction in hired labor is less and largely confined to land preparation. As the proportional decline in family labor exceeds the decline in hired labor, an increase in the share of wage employment in rice production ensues (while the overall number of jobs in on farm rice production decreases). Child labor use does not decline, and in semi-modern production systems (with animal traction), it actually increases, in land preparation, but also across the other activities. 35 The empirical specification of the intensification packages studied in the regression equations cannot exclude that additional factors also matter, such as the associated use of improved seeds, water control or soil fertility. These effects are controlled for to the extent that the instruments are able to isolate these practices from the ones studied (agrochemicals, mechanization). There are no obvious reasons to believe that they do not, but it cannot be fully excluded. Given the cross- sectional nature of the data, and despite the inclusion of a broad set of controls including farmers’ attitudes (risk, information) to reflect farmer heterogeneity as much as possible, and the use of instruments, caution remains warranted in interpreting the coefficients as causal. Analysis using panel data or randomized control trials complemented with more complete information on the adopted packages and input quantities would be needed to establish this more firmly. To conclude, the findings support the general notion that agricultural intensification can be profitable and that helping farmers intensify their production through the adoption of modern agricultural technologies such as tractors and chemical inputs amongst others, contributes to raising the returns to their labor, enhances specialization and commercialization, raises household income, and helps productively release agricultural labor for other crops and off-farm activities. It points to alleviation of the barriers in accessing these modern technologies as good policy entry points for supporting agricultural development and rural transformation. Second, the findings underscore the importance of looking at returns to labor (instead of returns to land). Labor productivity is the more relevant metric in instigating structural transformation, despite the somewhat misguided focus of ministries and agricultural development agencies as well as the literature on yields (Cock et al., 2022). While this partly follows from the challenges related to measuring labor input in agriculture (Sagesaka et al., 2021), the focus of policymakers and the profession on yields also reflects the sectorial (focused on aggregate food 36 production) as opposed to the farmer and economic perspective (focused on individual welfare and broader structural transformation). The greater availability of agricultural labor input data makes it possible to start shifting this balance. 37 References Adu-Baffour, F., Daum, T., & Birner, R. (2019). Can small farms benefit from big companies’ initiatives to promote mechanization in Africa? A case study from Zambia. Food Policy, 84, 133–145. https://doi.org/10.1016/J.FOODPOL.2019.03.007 Athukorala, W., Lee, B. L., Wilson, C., Fujii, H., & Managi, S. (2023). Measuring the impact of pesticide exposure on farmers’ health and farm productivity. Economic Analysis and Policy, 77, 851–862. https://doi.org/10.1016/J.EAP.2022.12.007 Barrett, C. B., Christian, P., & Shiferaw, B. A. (2017). The structural transformation of African agriculture and rural spaces: introduction to a special section†. Agricultural Economics, 48(S1), 5– 10. https://doi.org/10.1111/AGEC.12382 Beegle, K., & Christiaensen, L. (2019). Accelerating Poverty Reduction in Africa. World Bank Publications, Washington D.C. https://doi.org/10.1596/978-1-4648-1232-3 Burke, W. J., Jayne, T. S., & Black, J. R. (2017). Factors explaining the low and variable profitability of fertilizer application to maize in Zambia. Agricultural Economics, 48(1), 115–126. https://doi.org/10.1111/AGEC.12299 Cafiero, C., Viviani, S., & Nord, M. (2018). Food security measurement in a global context: The food insecurity experience scale. Measurement, 116, 146–152. https://doi.org/10.1016/J.MEASUREMENT.2017.10.065 Cock, J., Prager, S., Meinke, H., & Echeverria, R. (2022). Labour productivity: The forgotten yield gap. Agricultural Systems, 201, 103452. https://doi.org/10.1016/J.AGSY.2022.103452 Danso-Abbeam, G., & Baiyegunhi, L. J. S. (2018). Does fertiliser use improve household welfare? Evidence from Ghana’s cocoa industry. Development in Practice, 29(2), 170–182. https://doi.org/10.1080/09614524.2018.1526887 de Janvry, A., & Sadoulet, E. (2020). Using agriculture for development: Supply- and demand-side approaches. World Development, 133, 105003. https://doi.org/10.1016/J.WORLDDEV.2020.105003 Deb, P., & Trivedi, P. K. (2006a). Maximum simulated likelihood estimation of a negative binomial regression model with multinomial endogenous treatment. Stata Journal, 6(2), 246–255. https://doi.org/10.1177/1536867X0600600206 Deb, P., & Trivedi, P. K. (2006b). Specification and simulated likelihood estimation of a non‐normal 38 treatment‐outcome model with selection: Application to health care utilization. The Econometrics Journal, 9(2), 307–331. https://doi.org/10.1111/J.1368-423X.2006.00187.X Diao, X., Cossar, F., Houssou, N., & Kolavalli, S. (2014). Mechanization in Ghana: Emerging demand, and the search for alternative supply models. Food Policy, 48, 168–181. https://doi.org/10.1016/J.FOODPOL.2014.05.013 Emmanuel, D., Owusu-Sekyere, E., Owusu, V., & Jordaan, H. (2016). Impact of agricultural extension service on adoption of chemical fertilizer: Implications for rice productivity and development in Ghana. NJAS - Wageningen Journal of Life Sciences, 79, 41–49. https://doi.org/10.1016/J.NJAS.2016.10.002 Ghosh, B. K. (2010). Determinants of farm mechanisation in modern agriculture: A case study of Burdwan Districts of West Bengal. International Journal of Agricultural Research, 5(12), 1107– 1115. https://doi.org/10.3923/IJAR.2010.1107.1115 Haider, H., Smale, M., & Theriault, V. (2018). Intensification and intrahousehold decisions: Fertilizer adoption in Burkina Faso. World Development, 105, 310–320. https://doi.org/10.1016/J.WORLDDEV.2017.11.012 Hörner, D., & Wollni, M. (2022). Does integrated soil fertility management increase returns to land and labor? Agricultural Economics, 53(3), 337–355. https://doi.org/10.1111/AGEC.12699 Kahan, D., Bymolt, R., & Zaal, F. (2017). Thinking outside the plot: Insights on small-scale mechanisation from case studies in East Africa. The Journal of Development Studies, 54(11), 1939– 1954. https://doi.org/10.1080/00220388.2017.1329525 Kassie, M., Teklewold, H., Marenya, P., Jaleta, M., & Erenstein, O. (2015). Production risks and food security under alternative technology choices in Malawi: Application of a multinomial endogenous switching regression. Journal of Agricultural Economics, 66(3), 640–659. https://doi.org/10.1111/1477-9552.12099 Keil, A., D’Souza, A., & McDonald, A. (2016). Growing the service economy for sustainable wheat intensification in the Eastern Indo-Gangetic Plains: lessons from custom hiring services for zero- tillage. Food Security, 8(5), 1011–1028. https://doi.org/10.1007/S12571-016-0611-9 Kirui, O. (2019). The agricultural mechanization in Africa: Micro-level analysis of state drivers and effects. SSRN Electronic Journal. https://doi.org/10.2139/SSRN.3368103 Komatsu, S., Saito, K., & Sakurai, T. (2022). Changes in production, yields, and the cropped area of 39 lowland rice over the last 20 years and factors affecting their variations in Côte d′Ivoire. Field Crops Research, 277, 108424. https://doi.org/10.1016/J.FCR.2021.108424 Läpple, D., & Rensburg, T. Van. (2011). Adoption of organic farming: Are there differences between early and late adoption? Ecological Economics, 70(7), 1406–1414. https://doi.org/10.1016/j.ecolecon.2011.03.002 Liverpool-Tasie, L. S. O., Omonona, B. T., Sanou, A., & Ogunleye, W. O. (2017). Is increasing inorganic fertilizer use for maize production in SSA a profitable proposition? Evidence from Nigeria. Food Policy, 67, 41–51. https://doi.org/10.1016/j.foodpol.2016.09.011 Magezi, F. Yuko, N. Takeshi, S. (2021). Can smallholder farmers benefit from mechanization in Sub- Saharan Africa? Evidence from rice farming in Tanzania. 2021 Conference, August 17-31, 2021, Virtual 315066, International Association of Agricultural Economists. https://econpapers.repec.org/paper/agsaaea21/313949.htm Magnan, N., Spielman, D. J., Lybbert, T. J., & Gulati, K. (2015). Leveling with friends: Social networks and Indian farmers’ demand for a technology with heterogeneous benefits. Journal of Development Economics, 116, 223–251. https://doi.org/10.1016/J.JDEVECO.2015.05.003 Malabo, M. P. R. (2018). Mechanized: Transforming Africa’s agriculture value chains. https://www.ifpri.org/publication/mechanized-transforming-africas-agriculture-value-chains Mano, Y., Takahashi, K., & Otsuka, K. (2020). Mechanization in land preparation and agricultural intensification: The case of rice farming in the Cote d’Ivoire. Agricultural Economics, 51(6), 899– 908. https://doi.org/10.1111/AGEC.12599 Martey, E., Kuwornu, J. K. M., & Adjebeng-Danquah, J. (2019). Estimating the effect of mineral fertilizer use on Land productivity and income: Evidence from Ghana. Land Use Policy, 85, 463– 475. https://doi.org/10.1016/J.LANDUSEPOL.2019.04.027 Mason, N. M., & Smale, M. (2013). Impacts of subsidized hybrid seed on indicators of economic well- being among smallholder maize growers in Zambia. Agricultural Economics, 44(6), 659–670. https://doi.org/10.1111/AGEC.12080 Mellor, J. W. (2017). Agricultural Development and Economic Transformation. Agricultural Development and Economic Transformation. https://doi.org/10.1007/978-3-319-65259-7 Minten, B., Koru, B., & Stifel, D. (2013). The last mile(s) in modern input distribution: Pricing, profitability, and adoption. Agricultural Economics (United Kingdom), 44(6), 629–646. 40 https://doi.org/10.1111/AGEC.12078 Mottaleb, K. A., Rahut, D. B., Ali, A., Gérard, B., & Erenstein, O. (2017). Enhancing Smallholder Access to Agricultural Machinery Services: Lessons from Bangladesh. Journal of Development Studies, 53(9), 1502–1517. https://doi.org/10.1080/00220388.2016.1257116 Niang, A., Becker, M., Ewert, F., Dieng, I., Gaiser, T., Tanaka, A., Senthilkumar, K., Rodenburg, J., Johnson, J. M., Akakpo, C., Segda, Z., Gbakatchetche, H., Jaiteh, F., Bam, R. K., Dogbe, W., Keita, S., Kamissoko, N., Mossi, I. M., Bakare, O. S., … Saito, K. (2017). Variability and determinants of yields in rice production systems of West Africa. Field Crops Research, 207, 1–12. https://doi.org/10.1016/j.fcr.2017.02.014 Oparinde, L. O. (2021). Fish farmers’ welfare and climate change adaptation strategies in southwest, Nigeria: Application of multinomial endogenous switching regression model. Aquaculture Economics & Management, 25(4), 450–471. https://doi.org/10.1080/13657305.2021.1893863 Park, T., Paudel, K., & Sene, S. (2018). Sales impacts of direct marketing choices: treatment effects with multinomial selectivity. European Review of Agricultural Economics, 45(3), 433–453. https://doi.org/10.1093/ERAE/JBX038 Paudel, G. P., KC, D. B., Rahut, D. B., Justice, S. E., & McDonald, A. J. (2019). Scale-appropriate mechanization impacts on productivity among smallholders: Evidence from rice systems in the mid- hills of Nepal. Land Use Policy, 85, 104–113. https://doi.org/10.1016/J.LANDUSEPOL.2019.03.030 Pingali, P. (2007). Chapter 54 Agricultural Mechanization: Adoption Patterns and Economic Impact. Handbook of Agricultural Economics, 3, 2779–2805. https://doi.org/10.1016/S1574-0072(06)03054- 4 Sagesaka, A., Palacios-Lopez, A., & Amankwah, A. (2021). LSMS GUIDEBOOK: A guidebook for designing household surveys measuring agricultural labor. World Bank Group. https://www.worldbank.org/lsms Sellare, J., Meemken, E. M., Kouamé, C., & Qaim, M. (2020). Do sustainability standards benefit smallholder farmers also when accounting for cooperative effects? Evidence from Côte d’Ivoire. American Journal of Agricultural Economics, 102(2), 681–695. https://doi.org/10.1002/ajae.12015 Sheahan, M., & Barrett, C. B. (2017). Ten striking facts about agricultural input use in Sub-Saharan Africa. Food Policy, 67, 12–25. https://doi.org/10.1016/J.FOODPOL.2016.09.010 41 Sheahan, M., Barrett, C. B., & Goldvale, C. (2017). Human health and pesticide use in Sub-Saharan Africa. Agricultural Economics, 48(S1), 27–41. https://doi.org/10.1111/agec.12384 Sheahan, M., Black, R., & Jayne, T. S. (2013). Are Kenyan farmers under-utilizing fertilizer? Implications for input intensification strategies and research. Food Policy, 41, 39–52. https://doi.org/10.1016/J.FOODPOL.2013.04.008 Suri, T., & Udry, C. (2022). Agricultural Technology in Africa. Journal of Economic Perspectives, 36(1), 33–56. https://doi.org/10.1257/JEP.36.1.33 Tabe-Ojong, M. P. J., Mausch, K., Woldeyohanes, T. B., & Heckelei, T. (2021). Three hurdles towards commercialisation: integrating subsistence chickpea producers in the market economy. European Review of Agricultural Economics, 00(00), 1–28. https://doi.org/10.1093/ERAE/JBAB023 Takeshima, H., Nin-Pratt, A., & Diao, X. (2013). Mechanization and agricultural technology evolution, agricultural intensification in Sub-Saharan Africa: Typology of agricultural mechanization in Nigeria. American Journal of Agricultural Economics, 95(5), 1230–1236. https://doi.org/10.1093/AJAE/AAT045 Tanaka, A., Diagne, M., & Saito, K. (2015). Causes of yield stagnation in irrigated lowland rice systems in the Senegal River Valley: Application of dichotomous decision tree analysis. Field Crops Research, 176, 99–107. https://doi.org/10.1016/J.FCR.2015.02.020 van den Ban, A. (2011). Increasing labour productivity in agriculture and its implications. Journal of Agricultural Education and Extension, 17(5), 401–409. https://doi.org/10.1080/1389224X.2011.596414 Wuepper, D., Yesigat Ayenew, H., & Sauer, J. (2018). Social capital, income diversification and climate change adaptation: Panel data evidence from rural Ethiopia. Journal of Agricultural Economics, 69(2), 458–475. https://doi.org/10.1111/1477-9552.12237 42 Appendix Table A1: Description of technologies packages Technologies Land Preparation Chemical Chemical Number of Frequency (%) packages fertilizer Pesticides observations Traditional Manual No No 109 10,49 package Semi-traditional Manual No No 415 39,94 package Semi-modern Draught animal Yes Yes 477 45,91 package Modern package Tractor Yes Yes 38 3,66 Total 1039 100 43 Table A2: Effects of rice intensification on farm performance (full results) Yield (kg/ha) Profit (1000 FCFA/ha) Labor Productivity Return to family Return to total labor labor (kg/person-days) (FCFA/person- (FCFA/person-days) days) VARIABLES Semi-traditional package 641.45*** -25.20 4.49 -0.18 0.12 (176.11) (40.32) (2.86) (1.93) (0.36) Semi-modern package 793.69*** 80.71*** 21.64*** 3.33* 2.09*** (178.67) (26.50) (3.19) (1.79) (0.41) Modern package 1,333.61*** 52.66 31.88*** 7.37** 3.18*** (240.62) (42.42) (6.96) (3.48) (1.03) Age of the household head -4.41 -0.56 -0.13 -0.03 -0.02 (3.25) (0.49) (0.09) (0.04) (0.01) Year of education 8.41 0.09 -0.01 0.27 0.01 (9.28) (1.71) (0.21) (0.17) (0.03) Household size 2.62 2.18 -0.55* -0.56*** -0.04 (20.57) (2.90) (0.33) (0.20) (0.04) Dependency Ratio -266.41 -50.97* 0.19 1.61 -0.07 (191.89) (28.74) (3.73) (1.72) (0.50) Plot management type 44.12 -3.52 -2.26 -1.49 -0.32 (103.88) (16.00) (2.57) (1.59) (0.32) Experience in rice production 5.41 0.68 0.04 0.00 0.01 (4.17) (0.63) (0.11) (0.04) (0.01) Number of training participation -9.10 -1.18 0.15 0.29 0.02 (22.98) (3.48) (0.69) (0.40) (0.10) Plot visit by extension agent 179.53 9.60 1.24 0.54 0.15 (114.08) (16.46) (2.34) (1.10) (0.30) Off-farm activity 25.46 21.90 3.65 0.63 0.47 (160.38) (22.13) (2.53) (1.42) (0.35) Access to credit -209.49** -27.42* -5.18** -2.06* -0.78*** (101.25) (15.13) (2.12) (1.08) (0.27) Attitudes towards information 10.22 1.08 0.68 0.86*** 0.08 44 Yield (kg/ha) Profit (1000 FCFA/ha) Labor Productivity Return to family Return to total labor labor (kg/person-days) (FCFA/person- (FCFA/person-days) days) VARIABLES (21.84) (3.58) (0.56) (0.24) (0.07) Risk Attitudes -19.95 -3.82 -1.68** -1.02* -0.20* (40.48) (5.94) (0.82) (0.53) (0.10) ln(Total Plotsize) -86.41 -4.01 2.75** 1.35** 0.28* (63.86) (10.18) (1.36) (0.61) (0.17) Wealth index (PCA) 74.75** 8.36 1.30* 0.63** 0.14 (34.80) (5.24) (0.70) (0.29) (0.09) ln(Population density) 98.51 11.34 -0.67 -0.00 0.03 (65.99) (9.88) (1.43) (0.67) (0.18) ln(Distance to plot) 24.11 1.22 -0.15 0.84 0.01 (59.90) (7.64) (1.30) (0.64) (0.16) Labor availability (1=Yes) -50.28 3.77 1.08 -0.12 0.08 (128.56) (20.26) (2.64) (1.44) (0.34) ln(Rainfall average) -2,599.41*** -508.46*** -64.26*** -33.02*** -7.32*** (812.06) (136.71) (14.10) (7.83) (1.75) Constant -7,617.93*** -1,591.75*** -185.16*** -104.58*** 1.08*** (2,682.70) (441.46) (45.42) (24.41) (0.08) lnsigma 6.90*** 4.94*** 3.12*** 2.54*** 1.34*** (0.08) (0.14) (0.07) (0.12) (0.29) lambda_Semi traditional package 208.87** 80.56** 10.05*** 3.27*** -0.04 (97.83) (36.74) (2.24) (1.13) (0.54) lambda_Semi modern package -95.49 -58.60* -0.50 -2.49*** 0.07 (153.49) (30.28) (4.41) (0.95) (0.37) lambda_Modern package -669.04*** -44.05*** 0.70 -1.68*** -21.56*** (154.44) (16.06) (2.60) (0.63) (5.67) Observations 1,039 1,039 1,039 1,039 1,039 Note: Yield is the rice output per hectare of rice land (kg/ha); profit refers to the total net returns of rice output per hectare of rice land; labor productivity indicates rice output (kg) per person-days; return to family labor captures the return to land in1000 FCFA per person-days of family labor and return to total labor captures the return to land in 1000 FCFA per person-days of total labor. Clustered robust standard errors at village level in parentheses.*** p<0.01, ** p<0.05, * p<0.1. 45 Table A3: Effects of rice intensification on labor allocation in rice production by age and type (full results) VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor Semi-traditional package -30.61** -22.44** -16.51** -17.50*** 3.83 2.22 (13.39) (10.37) (7.27) (6.02) (2.36) (6.46) Semi-modern package -104.45*** -67.39*** -46.06*** -17.83*** 6.70*** -47.27*** (13.95) (11.55) (7.22) (5.25) (1.99) (5.68) Modern package -69.53*** -62.33*** -43.87*** -29.01*** 3.17 -18.73*** (14.86) (10.20) (7.43) (6.01) (2.34) (7.13) Age of the household head 0.07 0.31 0.14 0.09 0.02 -0.19 (0.21) (0.21) (0.11) (0.08) (0.04) (0.13) Year of education 0.63 -0.31 0.23 -0.27 0.08 0.50 (0.83) (0.49) (0.35) (0.21) (0.11) (0.43) Household size 2.89*** 4.33*** 2.23*** 0.80*** 1.05*** -1.17** (0.86) (0.59) (0.38) (0.21) (0.21) (0.54) Dependency Ratio -7.12 -25.69*** -24.52*** -2.44 2.68 17.60*** (8.88) (7.18) (4.70) (2.96) (1.85) (5.98) Plot management type 2.22 -0.25 -2.14 -1.06 1.67* 3.36 (5.36) (4.00) (2.81) (1.76) (0.88) (3.84) Experience in rice production 0.00 -0.06 -0.18* 0.10 0.06 0.03 (0.22) (0.17) (0.10) (0.06) (0.04) (0.13) Number of training participation 0.84 -0.34 -0.44 0.09 0.57* 0.51 (1.08) (0.81) (0.51) (0.37) (0.29) (0.84) Plot visit by extension agent 13.50** 4.92 5.58** 0.82 -2.54*** 10.72*** (5.27) (3.77) (2.64) (1.36) (0.97) (3.59) Off-farm activity (1=Yes) -16.50 -10.21 -3.59 -1.25 -2.94** -9.18 (10.08) (6.36) (3.87) (2.22) (1.37) (5.72) Access to credit (1=Yes) 4.18 5.02* -0.42 4.23*** 2.24* -3.09 (4.84) (2.81) (1.74) (1.40) (1.19) (3.17) Attitudes towards information -4.12*** -3.96*** -2.17*** -1.14*** -0.43** -0.12 (1.54) (0.90) (0.60) (0.30) (0.21) (1.05) Risk Attitudes 5.25*** 4.38*** 1.98*** 0.94** 1.09*** 1.32 (1.74) (1.22) (0.77) (0.44) (0.30) (1.04) ln(Total Plotsize) -10.44*** -5.13** -1.79 -2.83*** -0.96** -4.60*** (2.82) (2.10) (1.24) (0.76) (0.43) (1.44) Wealth index (PCA) 0.74 -1.08 -1.08 -0.17 -0.41 2.11 (1.89) (1.38) (0.87) (0.50) (0.34) (1.47) 46 VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor ln(Population density) 4.31* 1.04 0.57 0.52 0.76* 2.38 (2.47) (1.91) (1.19) (0.66) (0.40) (1.49) ln(Distance to plot) -0.71 -1.82 0.05 -0.95 -0.64 0.44 (2.31) (1.71) (1.20) (0.64) (0.63) (1.51) Labor availability (1=Yes) -6.12 -4.21 -2.93 0.11 -1.27 -1.74 (5.59) (4.09) (2.62) (1.58) (1.25) (3.24) ln(Rainfall average) 237.57*** 152.36*** 87.74*** 44.68*** 16.31** 86.53*** (49.42) (35.32) (25.94) (15.24) (7.79) (22.05) Constant 921.26*** 580.64*** 346.33*** 174.57*** 44.34* 349.85*** (159.91) (119.25) (86.58) (49.08) (25.47) (71.13) lnsigma 3.80*** 2.96*** 2.76*** 2.70*** 2.38*** 3.53*** (0.07) (0.12) (0.10) (0.08) (0.09) (0.13) lambda_Semi traditional package -13.51*** -24.45*** -15.30*** -0.82 2.43 -9.07** (4.42) (3.66) (2.27) (1.92) (1.65) (3.86) lambda_Semi modern package 39.99*** 27.49*** 16.60*** -2.65 -2.11 22.96*** (4.90) (3.69) (2.27) (2.12) (1.31) (3.11) lambda_Modern package -17.60*** 3.88 3.18** 9.24*** 0.80 -9.97*** (3.14) (5.70) (1.59) (1.77) (1.49) (1.79) 1,039 1,039 1,039 1,039 1,039 1,039 Note: All labor related outcomes are expressed in person-days/ha. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 47 Table A4: Effect of rice intensification on labor allocation in rice production for land preparation (full results). VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor Semi-traditional package -15.12** -5.85 -8.33** -2.29 0.63 -4.48 (5.98) (3.83) (3.51) (2.31) (0.83) (3.56) Semi-modern package -44.14*** -22.79*** -20.06*** -0.44 1.67** -25.93*** (6.29) (4.17) (3.87) (2.13) (0.72) (3.16) Modern package -40.49*** -20.46*** -19.12*** -4.22* -0.53 -21.43*** (6.35) (4.32) (3.83) (2.36) (0.69) (3.66) Age of the household head -0.01 0.06 0.01 0.02 0.00 -0.05 (0.09) (0.07) (0.05) (0.03) (0.02) (0.05) Year of education 0.25 -0.16 0.06 -0.05 0.05 0.16 (0.34) (0.22) (0.15) (0.07) (0.04) (0.22) Household size 1.01*** 1.29*** 0.69*** 0.17** 0.32*** -0.17 (0.38) (0.20) (0.16) (0.08) (0.07) (0.27) Dependency Ratio 1.72 -7.04** -6.30*** -0.44 1.01 7.50*** (3.59) (2.75) (2.00) (1.04) (0.81) (2.54) Plot management type -1.34 -0.49 -1.48 0.34 0.50 -0.77 (2.56) (1.63) (1.23) (0.55) (0.36) (1.92) Experience in rice production -0.03 0.03 -0.04 0.04** 0.02 -0.06 (0.09) (0.07) (0.04) (0.02) (0.02) (0.06) Number of training participation -0.16 -0.61* -0.37* -0.02 0.01 0.17 (0.42) (0.33) (0.21) (0.14) (0.07) (0.38) Plot visit by extension agent 4.92** 1.33 1.99* -0.72 -0.58 4.70*** (1.93) (1.53) (1.09) (0.52) (0.37) (1.72) Off-farm activity (1=Yes) -6.64* -3.84** -2.13 0.12 -1.03** -3.75* (3.59) (1.92) (1.52) (0.78) (0.44) (2.28) Access to credit (1=Yes) 3.05 2.51* 0.31 1.79*** 1.17*** -0.63 (2.06) (1.31) (0.75) (0.61) (0.42) (1.44) Attitudes towards information -1.37** -1.08*** -0.89*** -0.15 -0.07 -0.19 (0.66) (0.39) (0.29) (0.11) (0.08) (0.53) Risk Attitudes 2.18*** 1.11** 0.65** 0.24 0.24** 1.09** (0.73) (0.49) (0.32) (0.15) (0.10) (0.51) ln(Total Plotsize) -3.49*** -1.93** -1.04** -0.85*** -0.27 -1.32** (1.00) (0.87) (0.45) (0.25) (0.18) (0.55) Wealth index (PCA) 0.41 -0.02 -0.17 -0.01 -0.13 0.68 (0.84) (0.55) (0.36) (0.19) (0.12) (0.71) 48 VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor ln(Population density) 1.15 -0.38 -0.27 -0.14 0.20 1.29* (1.03) (0.70) (0.43) (0.24) (0.16) (0.76) ln(Distance to plot) 1.30 -0.33 0.30 -0.03 -0.10 1.06* (0.91) (0.99) (0.58) (0.21) (0.26) (0.63) Labor availability (1=Yes) -2.42 -0.67 -0.69 0.73 -0.28 -2.13* (2.17) (1.67) (1.09) (0.67) (0.51) (1.21) ln(Rainfall average) 65.19*** 43.19*** 28.90** 12.29** 2.62 20.85** (20.00) (16.18) (11.75) (5.01) (2.73) (9.95) Constant 258.15*** 169.04*** 120.09*** 43.46*** 5.46 88.46*** (65.87) (53.85) (38.66) (15.69) (9.02) (31.93) lnsigma 3.12*** 1.83*** 2.04*** 1.85*** 1.40*** 2.93*** (0.09) (0.24) (0.07) (0.09) (0.11) (0.13) lambda_Semi traditional package -4.64*** -10.34*** -6.34*** 0.51 0.58** -2.83*** (1.18) (2.08) (1.20) (0.81) (0.24) (0.83) lambda_Semi modern package 12.03*** 10.80*** 6.59*** -2.50*** -0.83*** 7.27*** (1.82) (2.46) (1.14) (0.65) (0.26) (1.16) lambda_Modern package -5.44*** -1.98 0.39 0.86* 0.58*** -3.21*** (1.13) (1.42) (0.72) (0.51) (0.17) (0.72) Observations 1,039 1,039 1,039 1,039 1,039 1,039 Note: All labor related outcomes are expressed in person-days/ha. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 49 Table A5: Effect of rice intensification on labor allocation in rice production forsowing (full results) VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor Semi-traditional package -1.92 -4.15 -3.47 -3.60 1.92** 5.22 (5.65) (3.33) (2.24) (2.22) (0.94) (3.36) Semi-modern package -26.93*** -16.71*** -9.48*** -6.80*** 1.41** -13.86*** (5.03) (3.86) (2.07) (2.19) (0.67) (2.76) Modern package -12.41** -11.23*** -8.11*** -5.40*** 0.76 -0.19 (5.96) (3.43) (2.02) (2.04) (0.65) (3.50) Age of the household head 0.06 0.09 0.03 0.05** 0.01 -0.04 (0.08) (0.07) (0.04) (0.02) (0.01) (0.06) Year of education -0.02 -0.14 0.02 -0.11 -0.01 0.02 (0.25) (0.16) (0.10) (0.07) (0.04) (0.16) Household size 0.43* 1.01*** 0.47*** 0.22*** 0.28*** -0.47** (0.26) (0.17) (0.09) (0.08) (0.07) (0.19) Dependency Ratio -0.13 -4.27** -3.73*** -0.79 0.70 3.72 (3.57) (2.16) (1.15) (1.10) (0.62) (2.42) Plot management type 4.54** 1.51 -0.47 1.11** 0.63** 3.07* (1.86) (1.13) (0.66) (0.54) (0.25) (1.59) Experience in rice production -0.06 -0.06 -0.05 -0.01 0.00 0.01 (0.08) (0.06) (0.03) (0.02) (0.01) (0.05) Number of training participation 0.81* 0.18 0.09 -0.06 0.28** 0.41 (0.43) (0.26) (0.13) (0.11) (0.14) (0.35) Plot visit by extension agent 3.02* 0.71 0.64 0.55 -0.76** 2.70** (1.69) (1.20) (0.62) (0.48) (0.36) (1.35) Off-farm activity (1=Yes) -5.43* -2.25 0.08 -0.80 -0.94** -4.10* (3.00) (1.54) (1.03) (0.66) (0.46) (2.28) Access to credit (1=Yes) -1.09 0.64 0.09 0.58 0.06 -2.23* (1.69) (0.89) (0.52) (0.44) (0.27) (1.32) Attitudes towards information -0.82* -0.87*** -0.46*** -0.17* -0.12 -0.09 (0.48) (0.29) (0.15) (0.09) (0.08) (0.38) Risk Attitudes 1.57** 1.28*** 0.54*** 0.32*** 0.31*** 0.45 (0.66) (0.36) (0.20) (0.12) (0.10) (0.51) ln(Total Plotsize) -3.00*** -1.07* -0.39 -0.50** -0.20 -1.78*** (0.74) (0.62) (0.31) (0.23) (0.13) (0.53) Wealth index (PCA) 0.28 -0.24 -0.36 0.03 -0.10 0.82 (0.66) (0.40) (0.27) (0.16) (0.13) (0.51) ln(Population density) 2.45** 0.88 0.61* 0.25 0.14 1.34* 50 VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor (0.97) (0.59) (0.36) (0.19) (0.12) (0.75) ln(Distance to plot) -0.22 -0.91 -0.48 -0.15 -0.26 0.59 (1.01) (0.56) (0.35) (0.29) (0.24) (0.68) Labor availability (1=Yes) -3.96* -2.08 -1.41 -0.52 -0.17 -1.86 (2.25) (1.31) (0.86) (0.52) (0.33) (1.61) ln(Rainfall average) 34.74** 34.72*** 23.47*** 6.47 6.39** -0.95 (16.07) (13.07) (8.79) (5.32) (2.74) (10.15) Constant 137.60*** 128.12*** 87.37*** 26.28 18.70** 8.30 (51.15) (43.78) (29.30) (16.81) (9.02) (32.15) lnsigma 2.70*** 1.81*** 1.76*** 1.61*** 1.36*** 2.44*** (0.08) (0.19) (0.08) (0.09) (0.11) (0.10) lambda_Semi traditional package -5.67*** -5.98*** -2.92*** -0.66 -0.15 -5.95*** (2.13) (1.14) (0.74) (0.70) (0.47) (2.12) lambda_Semi modern package 13.61*** 9.13*** 3.72*** 2.99*** 0.36 8.14*** (1.78) (1.62) (0.75) (0.55) (0.56) (1.92) lambda_Modern package -5.31 -1.15 0.91*** -0.23 0.95*** -7.39*** (3.36) (1.26) (0.35) (0.15) (0.31) (1.63) Observations 1,039 1,039 1,039 1,039 1,039 1,039 Note: All labor related outcomes are expressed in person-days/ha. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 51 Table A6: Effect of rice intensification on labor allocation in rice production for fertilizer application (full results) VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor Semi-traditional package 6.69*** 6.25*** 4.05*** 0.78*** 0.71*** 1.26*** (0.62) (0.13) (0.64) (0.18) (0.23) (0.37) Semi-modern package 5.03*** 4.25*** 1.70*** 1.42*** 0.45*** 1.01** (0.56) (0.14) (0.26) (0.22) (0.14) (0.42) Modern package 4.59*** 1.42*** 2.61*** 1.02*** 0.78*** 1.20** (0.76) (0.28) (0.37) (0.28) (0.27) (0.49) Age of the household head 0.03* 0.03*** 0.00 0.01* 0.00 0.01 (0.01) (0.00) (0.01) (0.01) (0.00) (0.01) Year of education 0.03 -0.01 0.00 -0.02* 0.00 0.02 (0.04) (0.01) (0.03) (0.01) (0.01) (0.03) Household size 0.05 0.10*** 0.15*** -0.06** 0.04** -0.05* (0.06) (0.02) (0.04) (0.03) (0.01) (0.03) Dependency Ratio -0.32 -1.36*** -1.65*** 0.61** 0.02 0.61* (0.60) (0.17) (0.40) (0.26) (0.16) (0.37) Plot management type 0.31 0.15 0.26 0.04 0.03 0.04 (0.29) (0.18) (0.20) (0.11) (0.08) (0.19) Experience in rice production 0.02 -0.01*** -0.00 0.00 0.00 0.02 (0.02) (0.00) (0.01) (0.01) (0.00) (0.01) Number of training participation 0.01 0.02 -0.03 0.02 0.03 -0.01 (0.09) (0.03) (0.04) (0.04) (0.03) (0.06) Plot visit by extension agent 0.29 0.13 0.31 -0.09 -0.05 0.13 (0.41) (0.14) (0.21) (0.16) (0.08) (0.31) Off-farm activity (1=Yes) -0.66 0.02 -0.39 0.22 -0.22* -0.26 (0.42) (0.15) (0.34) (0.23) (0.13) (0.18) Access to credit (1=Yes) 0.51 -0.00 -0.08 0.26** 0.13 0.19 (0.32) (0.14) (0.19) (0.12) (0.11) (0.26) Attitudes towards information -0.06 -0.04 0.02 -0.01 -0.05*** -0.04 (0.08) (0.04) (0.05) (0.04) (0.02) (0.05) Risk Attitudes -0.05 0.03 -0.13* -0.04 0.06** 0.06 (0.11) (0.04) (0.07) (0.04) (0.03) (0.08) ln(Total Plotsize) -0.43** -0.30*** -0.11 -0.21*** -0.08* -0.03 (0.20) (0.04) (0.09) (0.07) (0.04) (0.14) Wealth index (PCA) -0.13* -0.09* 0.04 -0.05 -0.01 -0.06 52 VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor (0.07) (0.05) (0.07) (0.04) (0.04) (0.05) ln(Population density) 0.24* 0.09* 0.08 0.08 0.05 0.03 (0.13) (0.05) (0.10) (0.07) (0.03) (0.07) ln(Distance to plot) -0.23 -0.10 -0.08 0.08 -0.14* -0.15 (0.24) (0.07) (0.10) (0.06) (0.08) (0.17) Labor availability (1=Yes) 0.00 -0.30*** -0.14 0.02 -0.22** 0.31** (0.25) (0.11) (0.16) (0.12) (0.10) (0.16) ln(Rainfall average) 7.56*** 5.99*** 1.14 2.55*** 1.47** 2.71** (1.82) (0.81) (1.24) (0.82) (0.73) (1.12) Constant 21.98*** 18.24*** 3.65 7.20** 4.62* 7.81** (6.11) (2.60) (3.94) (2.80) (2.43) (3.72) lnsigma 1.37*** -1.21*** 0.24 0.31*** 0.03 0.93*** (0.10) (0.15) (0.50) (0.06) (0.10) (0.18) lambda_Semi traditional package -0.40* -1.58*** -0.40 0.26* 0.16 0.13 (0.24) (0.04) (0.55) (0.15) (0.10) (0.08) lambda_Semi modern package 0.70*** 0.25*** 1.97*** -0.50*** 0.37*** 0.31*** (0.25) (0.06) (0.56) (0.11) (0.10) (0.11) lambda_Modern package 0.02 2.53*** -0.23 -0.19* 0.08 -0.31** (0.40) (0.05) (0.39) (0.11) (0.05) (0.13) 1,039 1,039 1,039 1,039 1,039 1,039 Note: All labor related outcomes are expressed in person-days/ha. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 53 Table A7: Effect of rice intensification on labor allocation in rice production for weeding or herbicides application (full results) VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor Semi-traditional package -5.95** -3.43*** -1.36 -2.54 0.35* -1.67 (2.69) (1.33) (0.97) (1.57) (0.21) (1.52) Semi-modern package -14.37*** -10.34*** -5.05*** -5.11*** 0.39* -4.83** (3.56) (1.68) (0.91) (1.92) (0.21) (2.10) Modern package -10.09*** -7.89*** -4.32*** -4.18** 0.35 -2.58 (3.15) (1.56) (1.08) (1.96) (0.31) (1.64) Age of the household head 0.01 0.01 0.01 -0.01 0.00 -0.00 (0.03) (0.02) (0.01) (0.01) (0.00) (0.02) Year of education 0.07 -0.02 0.06 -0.05 -0.01 0.06 (0.13) (0.08) (0.06) (0.04) (0.01) (0.08) Household size 0.24 0.33*** 0.17*** 0.09* 0.05** -0.05 (0.15) (0.10) (0.05) (0.05) (0.02) (0.09) Dependency Ratio -0.49 -2.12** -2.43*** -0.11 0.43*** 1.56** (1.27) (1.08) (0.78) (0.48) (0.13) (0.74) Plot management type -2.17 -1.17 0.11 -1.31** 0.06 -1.01* (1.32) (0.93) (0.54) (0.60) (0.11) (0.57) Experience in rice production -0.00 -0.02 -0.03* 0.01 -0.00 0.02 (0.03) (0.02) (0.01) (0.02) (0.00) (0.02) Number of training participation 0.08 0.15 0.06 0.06 0.04 -0.09 (0.21) (0.11) (0.06) (0.06) (0.03) (0.12) Plot visit by extension agent 1.30 0.28 0.03 0.36 -0.14 1.10 (1.03) (0.53) (0.44) (0.37) (0.09) (0.67) Off-farm activity (1=Yes) -3.08** -2.47*** -1.15** -0.88 -0.31*** -0.84 (1.35) (0.87) (0.52) (0.54) (0.11) (0.73) Access to credit (1=Yes) 0.80 0.44 -0.04 0.33 0.19* 0.26 (0.70) (0.34) (0.26) (0.20) (0.11) (0.47) Attitudes towards information -0.92*** -0.38** -0.00 -0.31*** -0.05** -0.53** (0.35) (0.16) (0.08) (0.11) (0.02) (0.22) Risk Attitudes 0.04 -0.06 -0.04 -0.06 0.05 0.11 (0.24) (0.17) (0.11) (0.08) (0.03) (0.14) ln(Total Plotsize) -0.38 -0.25 -0.26** -0.04 0.01 -0.06 (0.36) (0.22) (0.13) (0.11) (0.04) (0.21) Wealth index (PCA) 0.13 -0.05 -0.06 0.03 -0.04 0.21 (0.30) (0.22) (0.13) (0.15) (0.07) (0.16) 54 VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor ln(Population density) 0.12 0.14 -0.05 0.14 0.07* -0.07 (0.49) (0.27) (0.15) (0.18) (0.04) (0.27) ln(Distance to plot) -0.44 -0.11 0.17 -0.20 -0.04 -0.41 (0.41) (0.25) (0.15) (0.16) (0.06) (0.26) Labor availability (1=Yes) -1.23 -0.68 -0.41 -0.19 -0.13 -0.54 (1.03) (0.55) (0.26) (0.42) (0.10) (0.58) ln(Rainfall average) 35.62*** 20.32*** 14.53*** 4.57 1.31 15.26*** (9.62) (3.99) (2.31) (3.37) (0.88) (5.74) Constant 136.33*** 79.36*** 55.11*** 20.66* 3.59 57.17*** (33.31) (13.75) (7.92) (10.89) (2.78) (20.07) lnsigma 2.25*** 1.50*** 1.32*** 1.22*** 0.32** 1.83*** (0.10) (0.09) (0.10) (0.11) (0.14) (0.12) lambda_Semi traditional package -3.32*** -2.25** -1.47*** -1.32*** 0.35*** -1.64** (1.21) (0.91) (0.45) (0.44) (0.12) (0.64) lambda_Semi modern package 4.23*** 4.09*** 1.62*** 1.13*** 0.06 2.01*** (1.08) (0.86) (0.18) (0.35) (0.07) (0.63) lambda_Modern package -1.72*** 0.47 0.28 0.20 -0.00 -1.05** (0.62) (0.70) (0.26) (0.14) (0.14) (0.45) Observations 1,039 1,039 1,039 1,039 1,039 1,039 Note: All labor related outcomes are expressed in person-days/ha. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 55 Table A8: Effect of rice intensification on labor allocation in rice production for harvesting (full results) VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor Semi-traditional package -11.98** -14.32*** -6.22*** -5.92 0.55 3.35 (4.94) (1.92) (2.34) (4.38) (0.66) (2.68) Semi-modern package -24.28*** -31.41*** -14.72*** -9.91*** 2.34*** -3.90 (4.95) (2.26) (2.28) (2.78) (0.77) (2.91) Modern package -18.14** -31.21*** -13.51*** -11.25* 1.50 3.44 (7.61) (4.79) (2.48) (6.19) (0.99) (4.64) Age of the household head -0.02 0.17*** 0.09** 0.02 -0.00 -0.12** (0.08) (0.06) (0.04) (0.03) (0.01) (0.06) Year of education 0.24 -0.14 0.06 -0.18 0.03 0.22 (0.28) (0.13) (0.10) (0.16) (0.05) (0.20) Household size 1.14*** 1.73*** 0.76*** 0.49*** 0.38*** -0.45** (0.31) (0.15) (0.13) (0.14) (0.10) (0.22) Dependency Ratio -8.10** -13.51*** -10.69*** -2.59 0.49 4.15 (3.70) (2.13) (1.89) (1.60) (0.73) (2.96) Plot management type 1.32 -2.41** 0.03 -0.93 0.46 1.88 (2.05) (1.14) (0.85) (0.81) (0.39) (1.97) Experience in rice production 0.07 -0.02 -0.06* 0.05* 0.03** 0.05 (0.08) (0.03) (0.04) (0.03) (0.01) (0.06) Number of training participation 0.06 -0.53* -0.32* -0.07 0.20 0.05 (0.39) (0.32) (0.17) (0.21) (0.13) (0.36) Plot visit by extension agent 4.23** 4.85*** 2.80*** 0.91 -0.97** 2.04 (1.85) (1.04) (1.04) (0.69) (0.43) (1.33) Off-farm activity (1=Yes) -1.25 -3.15 -0.50 -0.84 -0.53 -0.04 (3.41) (2.50) (1.16) (1.38) (0.59) (2.41) Access to credit (1=Yes) 0.55 1.64 -0.68 0.98 0.66 -0.84 (1.88) (1.37) (0.85) (0.74) (0.51) (1.24) Attitudes towards information -0.82* -1.62*** -0.88*** -0.60*** -0.14 0.82** (0.46) (0.19) (0.23) (0.20) (0.09) (0.36) Risk Attitudes 1.57** 1.58*** 1.05*** 0.63*** 0.45*** -0.45 (0.62) (0.32) (0.30) (0.22) (0.15) (0.43) ln(Total Plotsize) -3.11*** -0.27 0.13 -1.12*** -0.39** -1.39** (1.09) (0.35) (0.73) (0.34) (0.16) (0.68) Wealth index (PCA) -0.14 -0.38 -0.45 0.11 -0.10 0.41 56 VARIABLES Total labor Household labor Men labor Women labor Children labor Hired labor (0.75) (0.30) (0.34) (0.41) (0.11) (0.61) ln(Population density) 0.13 -0.19 -0.09 -0.00 0.27 -0.23 (1.00) (0.65) (0.53) (0.39) (0.17) (0.67) ln(Distance to plot) -1.41* -1.76*** 0.15 -0.90*** -0.12 -0.74 (0.77) (0.50) (0.40) (0.35) (0.15) (0.70) Labor availability (1=Yes) 1.40 -0.88 -0.52 0.03 -0.48 2.24* (1.94) (0.75) (0.96) (0.71) (0.46) (1.21) ln(Rainfall average) 95.76*** 56.39*** 20.70*** 22.27*** 4.53* 50.02*** (18.89) (4.26) (7.07) (7.42) (2.65) (10.84) Constant 373.31*** 223.65*** 84.23*** 89.69*** 12.29 193.41*** (60.80) (12.99) (23.93) (24.77) (8.79) (33.79) lnsigma 3.10*** 0.76** 1.86*** 1.85*** 1.51*** 2.90*** (0.09) (0.30) (0.09) (0.16) (0.09) (0.06) lambda_Semi traditional package -2.27 -8.14*** -4.49*** -3.58 1.29*** -0.64 (2.64) (0.56) (0.92) (5.94) (0.30) (1.13) lambda_Semi modern package 10.60*** 11.67*** 6.39*** 2.40 -1.12*** 6.09*** (1.93) (0.43) (0.98) (2.61) (0.29) (1.05) lambda_Modern package 5.08*** 8.40*** 1.21** 3.39 -0.08 3.25*** (0.88) (0.56) (0.50) (8.19) (0.22) (0.75) Observations 1,039 1,039 1,039 1,039 1,039 1,039 Note: All labor related outcomes are expressed in person-days/ha. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 57 Table A9: Full results of the effects of the intensification packages on agricultural extensification, market orientation and food security (full results) Area under Decision to Share of rice Rice sold per Food rice cultivation commercialise sold capita (kg) insecurity (ha) VARIABLES (1=Yes) Semi-traditional package 0.09 0.48*** 0.28 79.92*** -0.08 (0.09) (0.03) (0.20) (25.94) (0.78) Semi-modern package 0.48*** 0.27*** 0.07 93.62*** -2.22*** (0.12) (0.04) (0.09) (27.66) (0.48) Modern package 0.54*** 0.78*** 0.39** 374.41*** -3.17** (0.16) (0.04) (0.16) (87.90) (1.41) Age of the household head -0.00 -0.00 0.00 0.36 -0.00 (0.00) (0.00) (0.00) (1.03) (0.01) Year of education -0.01* -0.00* -0.00 5.69 -0.06 (0.01) (0.00) (0.00) (3.55) (0.04) Household size 0.01 0.00*** -0.00 -19.40*** -0.03 (0.01) (0.00) (0.01) (4.65) (0.07) Dependency Ratio -0.33*** -0.11*** 0.02 -143.41** 0.73 (0.12) (0.02) (0.07) (56.30) (0.80) Plot management type -0.21** 0.00 -0.03 -55.14* 0.07 (0.09) (0.01) (0.12) (33.32) (0.45) Experience in rice production 0.01 -0.00*** -0.00 0.84 0.01 (0.00) (0.00) (0.00) (1.23) (0.01) Number of training participation -0.01 -0.00 0.00 -0.24 -0.02 (0.01) (0.00) (0.01) (5.08) (0.06) Plot visit by extension agent -0.04 0.10*** 0.07** 15.39 -1.07*** (0.06) (0.02) (0.03) (27.95) (0.26) Off-farm activity (1=Yes) 0.18 0.01 -0.01 89.91 -0.41 (0.12) (0.01) (0.06) (67.81) (0.45) Access to credit (1=Yes) 0.06 0.02*** 0.02 2.10 1.10*** (0.09) (0.00) (0.09) (22.09) (0.30) Attitudes towards information 0.02 0.02*** 0.01 9.00 -0.32*** (0.02) (0.00) (0.01) (6.12) (0.08) Risk Attitudes -0.05** -0.01*** 0.01 -3.72 0.31* (0.02) (0.00) (0.03) (7.36) (0.16) 58 Area under Decision to Share of rice Rice sold per Food rice cultivation commercialise sold capita (kg) insecurity (ha) VARIABLES (1=Yes) ln(Total Plotsize) 0.23*** 0.01*** 0.00 18.40 -0.04 (0.05) (0.00) (0.02) (12.56) (0.15) Wealth index (PCA) 0.02 -0.01* -0.00 0.62 0.07 (0.02) (0.00) (0.02) (7.68) (0.08) ln(Population density) -0.00 -0.03*** -0.03 -12.17 -0.20 (0.04) (0.00) (0.06) (17.01) (0.17) ln(Distance to plot) 0.12*** -0.01** 0.00 21.95** -0.18 (0.04) (0.00) (0.02) (10.59) (0.15) Labor availability (1=Yes) 0.07 -0.10*** -0.09*** -48.74* -0.48 (0.07) (0.02) (0.03) (26.32) (0.33) ln(Rainfall average) 0.21 -0.07 -0.13 -441.85** 11.85*** (0.35) (0.10) (0.24) (181.18) (3.01) Constant 0.97 0.29 -0.18 -1,261.74** 46.79*** (1.11) (0.34) (0.95) (611.15) (10.00) lnsigma -0.34*** -3.93*** -2.33 5.56*** -0.26 (0.08) (0.22) (2.79) (0.13) (0.57) lambda_Semi traditional package 0.11** -0.23*** -0.19 29.93 -2.29*** (0.04) (0.00) (0.30) (22.02) (0.21) lambda_Semi modern package -0.31*** 0.11*** 0.05 0.86 0.72*** (0.10) (0.00) (0.10) (12.93) (0.27) lambda_Modern package 0.01 -0.40*** -0.16 -35.06** 1.66*** (0.03) (0.00) (0.18) (15.22) (0.48) 1,039 1,039 1,039 1,039 1,039 Note: Food insecurity is the household food insecurity experienced scale developed by FAO. Clustered robust standard errors at village level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Most Recent Jobs Working Papers: 72. Cost-Effectiveness of Jobs Projects in Conflict and Forced Displacement Contexts—Annexes. (2022) Virginia Barberis, Laura Brouwer, Jan Von Der Goltz, Timothy Hobden, Mira Saidi, Kirsten Schuettler and Karin Seyfert. 72. Cost-Effectiveness of Jobs Projects in Conflict and Forced Displacement Contexts. (2022) Virginia Barberis, Laura Brouwer, Jan Von Der Goltz, Timothy Hobden, Mira Saidi, Kirsten Schuettler and Karin Seyfert. 71. Towards a Just Coal Transition Labor Market Challenges and People’s Perspectives from Wielkopolska. (2022) Luc Christiaensen, Céline Ferré, Maddalena Honorati, Tomasz Janusz Gajderowicz and Sylwia Michalina Wrona. 70. Towards a Just Coal Transition Labor Market Challenges and People’s Perspectives from Silesia. (2022) Luc Christiaensen, Céline Ferré, Tomasz Gajderowicz and Sylwia Wrona. 69. Towards a Just Coal Transition Labor Market Challenges and People’s Perspectives from Lower Silesia. (2022) Luc Christiaensen, Céline Ferré, Tomasz Gajderowicz and Sylwia Wrona. 68. A Tale of Two Countries: Labor Market Profiles of Youth in Urban and Rural Cameroon. (2022) Ioana Botea and Mitja Del Bono 67. Improving Smallholders’ Jobs Through Agribusiness Linkages : Findings of the Mozambique Agricultural Aggregator Pilot (MAAP). (2022) Michael Baxter,Christopher Brian Delgado, Jose Romero and David Ian Walker. 66. Rural Employment in Africa: Trends and Challenges. (2022) Luc Christiaensen and Miet Maertens 65. Understanding and Predicting Job Losses Due to Covid-19: Empirical Evidence from Middle-Income Countries. (2021) Maho Hatayama, Yiruo Li, and Theresa Osborne 64. Revisiting Labor Market Regulations in The Middle East and North Africa. (2021) Maho Hatayama. Click here for full Jobs Paper Series Address: 1776 G St, NW, Washington, DC 20006 Website: http://www.worldbank.org/en/topic/jobsanddevelopment Twitter: @WBG_Jobs Blog: https://blogs.worldbank.org/jobs/