Relaxing Constraints for Family Farmers: Providing Capital and Information in Malawi
Current version: May, 2016 Very preliminary draft, please do not circulate Kate Ambler, Alan de Brauw and Susan Godlonton Abstract This paper evaluates the short-term impacts of a combined agricultural extension and cash transfer program in Malawi on agricultural productivity and household wellbeing. Using a randomized controlled trial we alleviate capital and information constraints among approximately 1,200 smallholder farmers. The capital and information treatments were cross randomized. In the capital treatment, farmer groups were randomly allocated to a control group that received no transfers, a cash group that received a series of three framed cash transfers at planting, mid-season, and harvest, and an input group that received transfers of equivalent value but of which approximately 50% were given in-kind. In the information treatment, farmers received standard NASFAM extension services (primarily including group trainings run by lead farmers from the community) or intensive extension services in which they receive at least three one-on-one visits from professional extension workers in addition to group trainings. The cross randomization allows us to explicitly study the complementarities of the capital and information treatments. We find strong evidence that both the framed cash transfer and input packages led to large increases in the overall value of agricultural production, as well as production yields for groundnuts. Evidence also strongly suggests that farmers are increasing resources to inputs, specifically pesticides and the use of casual labor that leads directly to these increases. Furthermore, intensive extension complements the transfers particularly for crops that farmers have less experience growing.
Key words: Agriculture, Extension, Cash transfer, graduation model, Malawi
1
1. Introduction Despite tremendous progress in reducing extreme poverty in the last two decades, more than 800 million people still live in extreme poverty (UN, 2016). Understanding effective policy interventions to reduce poverty is thus critical. Using a cluster cross-cutting randomized controlled trial we examine the independent effects of common components of multi-faceted graduation programs as well as the existence and extent of complementarities of these components. We focus on two key components: asset transfers (cash and input packages) and agricultural extension advice that are design to relax capital and information constraints respectively. A total of 120 existing farmer groups of approximately 10 farmers per group were randomly allocated to treatment groups. In the capital treatment, farmer groups were randomly allocated to a control group that received no transfers, a cash group that received a series of three framed cash transfers at planting ($36), mid-season ($22), and harvest ($26), and an input group that received transfers of equivalent value of which approximately 50% were given in-kind (seed, hoes, inoculant, storage sacks). In the information treatment, farmers received standard NASFAM extension services (primarily including group trainings and demonstration plots run by lead farmers) or intensive extension services in which they receive at least three one-on-one visits from professional extension workers in addition to group trainings. A farm management plan is developed as part of these visits. The cross randomization allows us to explicitly study the complementarities of the capital and information treatments. Our results show that both cash and in-kind transfers lead to increases in production while there is no evidence of a short-run impact of intensified extension services relative to group extension support. Specifically, we observe the transfers lead to shifts away from tobacco towards groundnuts and soy production. We observe no impact on yields attributable to the intensive extension treatment but do observe an increase in groundnut yields for the capital transfer treatment. Groundnut yields increase by approximately 30 percent for those receiving the asset transfer. Relatedly, we observe large increases in the overall value of production. Overall, we find a statistically significant increase in overall production value only for the cash intervention, representing an increase of 21 percent. We observe a smaller (but still positive) increase in overall production value for the input group, but it is not statistically significant at a reasonable threshold.
2
We explore the mechanisms driving these production increases by examining changes to the use of and expenditures on agricultural inputs. Transfers increased use of both pesticides (57 and 37 percent increase for cash and input farmers respectively) and ganyu (66.7 and 60 percent increase for cash and input farmers respectively), but does not change fertilizer use. Similarly, we observe increased expenditures on ganyu and total input expenditures for both transfer interventions. The estimated impacts are large, representing approximately a 30 percent increase in total input expenditures. We observe no increase in fertilizer use on either the extensive (adoption) or intensive (expenditure) margins. This is not terribly surprising given the high penetration of the Fertilizer Input Subsidy Program (FISP). Lastly, we find important complementarities between intensive extension and capital transfer interventions in input use. We observe that intensive extension and the capital transfers are not complementary for ganyu use for groundnuts, but strongly complementary for ganyu use for soy. We observe a similar pattern of results for ganyu expenditures. A larger proportion of farmers were unfamiliar with soy at baseline, implying a larger scope for the intensive extension to be effective when capital constraints are simultaneously alleviated. Our findings contribute to several important literatures. First, these results add to the burgeoning cash transfer literature. Cash transfers have been widely and rigorously studied particularly in the context of health and education starting with Progressa in the late 90s (e.g. Gertler, 1999). This literature generally finds positive impacts but less is known about the potential of cash transfers to improve productivity. We provide strong evidence in support of asset transfers as part of an integrated agricultural support program. This is consistent with recent work on multi-dimensional BRAC style graduation models (Banerjee et al. 2015). Second, we provide further evidence that framing of cash transfers may be sufficient to generate large impacts in the sector of interest consistent with similar findings in Morocco in education (Duflo et al. 2014). Third we contribute to the important literature on the effectiveness of agricultural extension. Millions of dollars are spent annually on agricultural programs utilizing extension workers. In the last three decades, the World Bank alone has spent close to $4 billion supporting such programs. Two reviewsβBirkahaeuser et al. (1991) and Evenson (2001) β document that most existing studies found significant positive impacts, with highly variable returns. However, very little rigorous evidence exists on the impact of extension services on agricultural productivity and poverty reduction; nor the conditions under which different models are effective. We provide rigorous short term evidence comparing intensive and group extension, as well as contribute to understanding the conditions under which more expensive individualized extension might be effective.
3
The paper proceeds as follows. Section 2 describes the institutional setting and the experimental design. Section 3 presents the data and discusses the measurement of key outcomes, while Section 4 outlines the empirical analysis. We present the main results in Section 5 and Section 6 discusses the mechanisms underpinning the main findings. Section 7 concludes. 2. NASFAM Program and Experimental Design 2.1. Background Malawi is one of the worldβs poorest countries, ranking 173rd out of 188 countries on the Human Development Index in 2015 (HDI, 2015). Given that most of the population resides in rural areas, unsurprisingly rural poverty is high at 50.7 percent (World Bank, 2010). Agriculture is the backbone of the Malawi economy contributing approximately 40 percent to GDP and 80 percent of exports (GoM, 2015). The agricultural sector is primarily smallholder farmers cultivating maize, cassava and sweet potatoes. Few smallholder farmers are engaged in farming high value cash crops. The National Smallholder Farmers Association of Malawi (NASFAM) an organization of national scale that provides farmers with both commercial and social services. Operational since 1998 it currently serves approximately 165,000 farmers in 13 districts spanning all regions and is steadily growing its farmer base. Farmers self-organize into clubs of up to 15 members and are required to pay a membership fee of 3,000 MWK per club. NASFAM provides an input starter package of high yield variety seeds of a cash crop suited to the particular region, for example groundnuts and soya the focus crops for this project. At the end of the agricultural season farmers are expected to repay 1.5 to 2 times the amount of seed provided. These seed loans are provided to farmer clubs for the first two years after registering with NASFAM. Farmers also receive group based extension support and extension support primarily through the lead farmer approach that is commonplace in Malawi. 2.2. Experimental Design The experiment was conducted in Ntchisi and Dowa districts in central Malawi between August 2014 and September 2015. These districts were selected in consultation with the implementation partner, NASFAM. This region, like much of the rest of the country is characterized by high rates of rural poverty, with the majority of households engaged in subsistence agriculture. We selected the 120 newest farmer clubs in the region as identified by NASFAM totaling approximately 1200 smallholder farmers. The selection rule excluded farmer clubs that exclusively grow
4
tobacco. Tobacco is the primary cash crop and accounts for more than half of export earnings. Due to the reliance on tobacco many government policies and NGOs seek to broaden the export base by encouraging adoption of other cash crops. For our purposes two project focus crops were selected: groundnuts and soya. The experimental design for phase one is illustrated in Figure 1. We implement a cross-cutting clustered randomized design. The capital intervention has three study arms: cash only, an input-cash package and control. The information intervention offers either intensive extension support or standard group based and on-demand extension support. Next, we provide more specific information about each of these interventions. 2.2.1. Framed Agriculture Cash Transfer NASFAM implemented the Capital transfer according to the timeline presented in Figure 1. The total value of the capital transfers is approximately equivalent to $84. The value of the first transfer is $36, the second is $22 and the final transfer is $26. The value of each transfer is set to equalize the cost of inputs at each time point. For individuals assigned to the cash group they received the Malawi Kwacha equivalent of the amounts reported. For individuals assigned to the input group they received seed, inoculant (if soya seed was provided) and hoes to the value of $36 in the first transfer; they received cash for ganyu (day-labor, the relevant input mid-season) for the second transfer and sacks, strings and cash for transportation and ganyu to total the value of the third cash transfer. 2.2.2. Intensive Extension: Business Management Tool To support the intensive extension component 15 additional agricultural field officers (AFOs) were hired. The AFOs provided standard NASFAM support (on-demand extension support) to farmers in the control group as well as the intensive extension. The additional extension workers reduced the extension worker to farmer ratio dramatically enabling the provision of one on one consultation to farmers in the intensive treatment group. The extension workers were expected to visit farmers on an individual basis three times during the agricultural season. During the first visit at the beginning of the season, AFOs assisted farmers to develop a farm management plan. The farm management plans cover a wide variety of topics including: land allocation across crops; livestock production income: Breeding; by-product revenue; crop production and livestock production expenses; and the appropriate timing of activities. During the follow-up visits, AFOs were
5
expected to check in with farmers to assess their progress relative to their management plan. In practice, AFOs did not refer much to the farm management plan but instead focused primarily on providing specific agricultural expertise related to farmer specific issues. 3. Data 3.1 Data Sources We utilize several data sources in the analysis combining both survey data and administrative data sources. Baseline Household Survey. A baseline survey was administered in August-September 2014. It includes modules on socioeconomic characteristics, extensive information pertaining agricultural production and practices, livestock, time use and employment related information by household member, as well as remittances, credit and savings modules. Many of the modules were developed closely following the Integrated Household Survey to enable comparisons to nationally representative data. Midline Household Survey 1 (sub-sample). A midline survey was conducted during the lean season for a subsample of households in March 2015. The subsample comprises all club chairs as well as two randomly selected club members per club. This survey includes modules on food security, time use and employment related information, as well as projected agricultural output for the season. Midline Household Survey 2. A second midline survey was conducted in August-September 2015 for the entire sample. This survey included similar information to that collected during the Baseline survey as well as additional modules pertaining to household enterprises, program implementation monitoring, and an extensive consumption module (analogous to that used in the Integrated Household Survey). Administrative Data. We utilize several administrative data sources collected by NASFAM. This includes activity log data that provides coarse self-reported information pertaining to each visit conducted by the extension officer; farm management plans that provide a useful benchmark for comparing how household agricultural intentions compare to final outcomes; seed storage data documenting how much seed each individual farmer stored with NASFAM at the end of the agricultural season; and finally marketing data of crops sold directly to NASFAM.
6
3.2 Sample Characteristics Table 2 Panel A presents summary statistics of the NASFAM member in the household using the baseline survey data. The NASFAM members are disproportionately female (63%), most are married (82%) and on average they are 41 years old. The NASFAM members have little education, 70% have either never been to school or only completed less than primary schooling. On average, the NASFAM respondents spent the majority of their time engaged in agricultural activities in the past week, approximately 13 hours on average. In addition, respondents spent an average of 5.3 hours running their own business, and 2.6 hours engaged in ganyu labor. In the past 12 months, almost half engaged in some ganyu work. Those who did engage in casual labor worked approximately 26 days in the past 12 months, at an average daily wage of $2.50.1 In the past 12 months, only 14% were ever engaged in salaried work indicating that these smallholder farmers depend primarily on agriculture to support the household and any agricultural productivity enhancing intervention has the potential to improve their welfare substantially. Table 2 Panel B presents summary statistics regarding the household of the NASFAM respondent. In 47% of households the NASFAM respondent is not the identified household head. The characteristics of the household head also likely impact the decisions taken regarding agricultural activities. On average, only 14% of household heads are women, most are married (84%) and they are slightly older than the NASFAM respondent with an average age of 44 years. The household heads also have limited formal education: 14% have no schooling, 48% have some primary schooling, 21% have completed primary schooling, the remainder (17%) have completed more than primary schooling. Households engage in numerous agricultural activities growing on average 4.5 distinct crops in the last agricultural season. Most land used for agriculture is owned by the households. On average they own about four acres, and in the last agricultural season rented or borrowed on average less than one additional acre. Households earned a total (across crops) of only $300 per annum. While the amount of income from agriculture is low, it is the primary source of income for these households.
1
The exchange rate used is 389.453, the average exchange rate during the data collection period.
7
4. Empirical Approach 4.1. Estimation Strategy Due to the experimental design of the proposed study, the analyses avoid the limitations of crosssectional or longitudinal studies that may be affected by omitted variable bias. The randomized design of the study permits a relatively straightforward analysis. First, we will model the analysis of the effects of receiving one mode of extension services relative to the control group. To do this we will estimate the following equation: πππ = πΌ + π½1 πΈπ₯π‘πππ ππππ + πΎπππ + πππ , where Yij indicates agricultural productivity of individual i in farmer group j, Extensioni indicates if the respondent was allocated to the treatment group receiving intensive agriculture extension services, and Xij is a vector of baseline covariates (e.g., baseline agricultural productivity, input use, and other background characteristics). Second, we are also interested in the differential effects of the modality of the production transfer. For example, we will estimate the effect of intensive extension (increased information); cash transfers and extension; or input packages and extension relative to the control group on agricultural productivity as follows: πππ = πΌ + π½1 πΆππ βπ + π½2 πΌπππ’π‘π π + πΎπππ + πππ , where: Cashi indicates if the respondent was allocated to the treatment group receiving a production cash transfer, Inputsi indicates if the respondent was allocated to the treatment group receiving an equivalently valued input package. In this case, we are not only interested in the direction and size of π½2 and π½3 , but also the relationship between π½2 πππ π½3 . Standard errors are clustered at the farmer group level. A third set of analyses enables us to examine the degree of complementarity/substitutability between the capital and information interventions. Specifically we estimate the following equation: πππ = πΌ + π½1 πΈπ₯π‘πππ ππππ + π½2 πΆππ βπ + π½3 πΌπππ’π‘π π + π½4 (πΆππ β π πΈπ₯π‘πππ πππ)π +π½5 (πΌπππ’π‘π π πΈπ₯π‘πππ πππ)π + πΎπππ + πππ ,
8
In this case π½4 and π½5 enable us to test the existence of interactions when alleviating the capital and information constraints simultaneously. If π½4 , π½5 > 0 then the intensive extension support and the asset transfers are complementary. If π½4 , π½5 < 0 then the intensive extension support and the asset transfers are substitutes. The size of the coefficients speak to the magnitude of the complementarity/ substitutability of alleviating these constraints. 4.2. Balance To check that the randomization βworkedβ we examine balance across the treatment groups. To do this, we compare a host of baseline indicators across our two cross cutting interventions. In general, our sample is well balanced, there exist very few indicators (certainly no more than would be expected by chance) that differ systematically between the different treatment groups. The balancing test for the information intervention is presented in Table 1 while the balancing tests corresponding to the capital intervention are presented in Table 2. Table 1 Column 1 presents the group mean among respondents assigned to receive intensive extension services and develop the plan with the assigned extension officer. Column 2 presents the group mean among respondents assigned to receive standard NASFAM agricultural services. Column 3 presents the p-value associated with the test that the group means are the same. With few exceptions we observe that there are no statistically significant differences. Individuals designated to receive intensive agricultural services are slightly more likely to be married, and reside in slightly larger households. Next, we turn to examine the balancing statistics across the different transfer intervention treatment arms. Table 2 Column 1 presents the group mean among respondents assigned to the control group; Column 2 presents the group mean among respondents assigned to receive inputs; Column 3 presents the group mean among those designated to receive cash. Column 4 through 7 present the pvalue associated with the test of whether the group means (of those indicated) are the same. As with the planning intervention we observe very few cases in which the groups seem to be different. Individuals in the input group (20.8%) are more likely to have no education relative to the control group (14.9%); and slightly less likely to have some secondary education than the control group. These differences are statistically significant at the 10 percent level. Individuals look very similar across groups with respect to agricultural inputs, and crop production. One exception is the case of pumpkins. Production of pumpkins in the cash group is much higher than in the other groups at baseline. This is likely to be due to chance and not of particular concern.
9
5. Results 5.1 Extensive Margin: Crop adoption First, we examine whether the treatments impact the decision to adopt different crops. Table 3 presents these results. Panel A shows that there is no evidence intensive extension services alone impacts the decision to grow the key crops. This is not particularly surprising given that even in the control group adoption rates of all key crops is relatively high, ranging from 45 percent for Tobacco to 99.5 percent for Maize. Panel B shows that this is a modest increase in adoption of groundnuts attributable to the cash and input groups. Although adoption of groundnuts is already high, 71 percent among farmers receiving no capital transfer, farmers were 5.9 and 7.1 percentage points more likely to adopt groundnuts if they received the cash or input treatment respectively. The cash transfer also increased the adoption of soy relative to the input group, although the increase relative to the control group is not statistically significant. These results make sense because a larger share of farmers in the input treatment group received groundnut seed rather than soy seed from NASFAM, whereas farmers in the cash treatment were not limited in their decision making. Given that most farmers who are members of NASFAM within this area already grown Groundnuts or Soy, it is not terribly surprising that we do not observe large impacts on the adoption of new crops. Next, we turn to the intensive margin, specifically how much land farmers allocated to each of these crops.
5.2. Intensive Margin: Land allocation Table 4 presents the impacts of the interventions on land allocation. The standard unit of area used in Malawi is acres rather than hectares. Plot sizes are relatively small. We observe no evidence that the intensive extension changed the allocation of land across different crops. Panel A presents these results. We observe coefficients that are small in magnitude that are not statistically significant. Panel B demonstrates that the capital transfers do impact land allocation across crops. Specifically, we find an increase in land allocated to groundnuts (0.17 acres) and soy (0.13 acres) for those receiving the cash transfer; and an increase in land allocated to groundnuts (0.25 acres) for farmers receiving the input intervention. The increase in land allocated to groundnuts is larger and statistically significant for the input intervention relative to the cash intervention. This aligns with the greater disbursement of groundnut rather than soy seed. The estimated impacts are meaningful. For the cash intervention, the coefficients represent a 22 and 21 percent increase in land allocated to
10
groundnuts and soy (relative to the control) respectively. For the input intervention, the increase in land allocated to groundnuts is a 32 percent increase in acreage. We find suggestive evidence that the land allocated to tobacco declines, but the impact is only significant for the input treatment arm.
5.3. Productivity: Total production (in kgs); yields; value of production and sales (in Kwacha, MWK) Table 5 and 6 examine the impacts on productivity. Table 5 presents total production as measured in kilograms for the key crops. Once again we find evidence that intensive extension alone changed production for the key crops (Panel A). The capital transfers did result in increased production for groundnuts, and reduced production of tobacco. The effect sizes are similar for the cash and input interventions. We observe an increase of approximately 36 to 55 percent increase in groundnut production (53 to 80 kilogram increase in overall production); and a reduction of approximately 26 to 28 percent reduction in tobacco production (42 to 46 kilogram decrease in net production). We do not observe a statistically significant increase in soy production for either the cash or input intervention relative to the control. However, we do observe an increase in soy production in the cash treatment group relative to the input group, where the cash group produces on average 30 kilograms more soy. Table 6 presents the impact of the interventions on yields for the key crops are not surprising. We observe no impact on yields attributable to the intensive extension treatment (Table 6 Panel A) but do observe an increase in groundnut yields for the capital transfer treatment (Table 6 Panel B). The estimated coefficients represent a 25 to 31 percent increase in groundnut yields. Tobacco yields decline for the input intervention by approximately 29 percent. Next, we examine the value of production. We observe increases in production for some crops and decreases for others so it is important to examine what happens to aggregate sales of crop production. Furthermore, farmers may increase their production without necessarily increasing the value of their production if for example they produce lower quality crops, or conditional on quality obtain lower prices. Therefore, it is important to examine the value of the production produced as well as the value of the portion of production sold. These results are presented in Tables 7 and 8 respectively. Table 7 shows that the intensive extension alone has no impact on the value of the key crops all the aggregate crop production value (Column 7). Table 8 demonstrates large impacts on the value of groundnut production, consistent with the previous results. The estimated coefficients suggest the value of groundnut production increased by approximately 18,000 MWK or 42 percent. We observe suggestive evidence of an increase in maize production, but it is only statistically significant for the cash
11
intervention. While the treatment impact on the value of tobacco production is negative, suggesting a reduction in the value of tobacco it is not statistically significant. Overall, we find a statistically significant increase in overall production value only for the cash intervention, representing an increase of 21 percent. We observe a smaller (but still positive) increase in overall production value for the input group, but it is not statistically significant at a reasonable threshold. We find similar results for the value of crop sales presented in Table 8, which is only a fraction of total production. In sum, these results indicate that transfers lead to increases in production while there is no evidence of short-run impact of extension services. Specifically, we observe the transfers lead to shifts away from tobacco towards groundnuts and soy consistent with farmer expectations measured during the midline survey conducted mid-season. There is some evidence to suggest that cash is more effective than inputs in the short-run, but further analysis is still needed. Next, we turn to examine the mechanisms that may have facilitated these increases. 6. Discussion 6.1. Mechanisms: Input adoption and usage Table 9 examines the impact of the treatments on input use and expenditures by input type and overall. The intensive extension increased pesticide use by 5.57 percentage points (or 22 percent) as shown in Panel A. On the intensive margin, ganyu (day-labor) and overall input expenditures increased for the intensive extension group by 35 and 12 percent respectively. We observe no increase in fertilizer use on either the extensive (adoption, column 1) or intensive (expenditure, column 2) margins. However, it is worth highlighting that fertilizer is quite high, 84.1 percent report using any fertilizer in the control group. Table 9 Panel B shows the impact of the capital transfers on overall input use and expenditures. Transfers increased use of both pesticides (57 and 37 percent increase for cash and input farmers respectively) and ganyu (66.7 and 60 percent increase for cash and input farmers respectively), but does not change fertilizer use. Similarly, we observe increased expenditures on ganyu and total input expenditures for both transfer interventions. The estimated impacts are larger in magnitude for the cash treatment group relative to the input group but the difference is not statistically significant. The estimated impacts are large, representing approximately a 30 percent increase in total input expenditures. Furthermore, Table 9 Panel C demonstrates important complementarities between intensive extension and capital transfer interventions. We observe no impact on input use for farmers who only
12
received extension services, but we find the largest increases when capital transfers are combined with extension services. Tables 10 through 13 consider each input type separately and present the impacts of the treatments on input use by crop. Table 10 presents the results for fertilizer, Table 11 pesticide, Table 12 irrigation and Table 13 ganyu (day-labor). This set of tables provides important insight into input use by crop that is obscured in the average usage and expenditure results.
Fertilizer usage by crop In general, the treatment impacts on fertilizer usage by crop mimic the aggregated results, no change to fertilizer use attributable to either the intensive extension or capital transfers (Table 10 Panel A and B). However, we do observe an increase in fertilizer usage for soy for only the cash intervention (relative to both the control and the input intervention). Furthermore, this effect is driven by farmers who receive both cash and the intensive extension reflecting the complementarities between these two interventions (Table 10 Panel C). A larger proportion of farmers were unfamiliar with soy at baseline, implying a larger scope for intensive extension to be effective. Combining the framed but unrestricted cash transfer with intensive extension appears most effective in this case.
Pesticide usage by crop In general, the treatment impacts on fertilizer usage by crop mimic the aggregated results, and increase in pesticide use attributable to both the intensive extension or capital transfers (Table 11 Panel A and B). The estimated effects are driven by increases in pesticide use on soy (capital and extension) and maize (capital). Panel C shows strong complementarities between the capital transfer and intensive extension service interventions, i.e. the largest increases in pesticide use for soy are driven by farmers who receive both a transfer and intensive extension.
Irrigation usage by crop We observe limited effects of irrigation use across both interventions (Table 12 Panels A and B). This is not particularly surprising given that irrigation adoption in the area in general is very low. Although we do observe an increase in irrigation adoption for soy both for the intensive extension and cash treatments we are hesitant to over interpret these results as they are driven by a handful of farmers
13
adopting irrigation. Nonetheless, these results are consistent with the other input decisions, an increase in input usage particularly for soy.
Ganyu usage and expenditure by crop Table 13 presents the impact on ganyu (day-labor) usage. We observe strong evidence that both interventions increased ganyu use and expenditures on ganyu. Intensive extension services increased ganyu use for maize, soy and tobacco (Panel A). The transfers dramatically increased use of ganyu for maize, groundnuts, and soy (Panel B). Panel C shows some really fascinating results. We observe that intensive extension and the capital transfers are not complementary for ganyu use for groundnuts, but strongly complementary for ganyu use for soy (Panel C). We observe a similar pattern of results for ganyu expenditures (Table 14).
6.2. Mechanisms: Other productivity enhancing investments? Another mechanism through which our results may be realized is the acquisition of relevant productivity enhancing assets. Table 15 presents initial results examining assets and expenditures. There is no clear evidence intensive extension increased assets or total expenditures (Panel A). The transfers increase the value of agricultural assets and total assets (Panel B). The impacts on assets are strongest when cash transfers are combined with intensive extension (Panel C). There is also suggestive evidence that treatments may have slightly increased expenditures but the standard errors are large. We will conduct more robustness analysis to further examine this. These results suggest that farmers are investing more in agriculture both in terms of input use (pesticide and ganyu in particular) and agricultural asset purchases (at least within the cash group). Figure 1 presents summary information on what farmers said they spent the cash on. The most commonly reported categories are seeds for the November transfer and ganyu labor for the February and March transfers. Other common uses reported are food and tools. Tools include primarily hoes and sacks, with a few sickles and one ox-cart. This is broadly consistent with our results with the exception of a noticeable increase in pesticide expenditure in the self-reported data.
7. Conclusion Using a cluster cross-cutting randomized controlled trial we examine the direct impacts of a graduation model that combines cash or in-kind (input packages) transfers with intensive
14
agricultural extension support in rural Malawi. In the capital treatment, farmer groups were randomly allocated to a control group that received no transfers, a cash group that received a series of three framed cash transfers at planting ($36), mid-season ($22), and harvest ($26), and an input group that received transfers of equivalent value of which approximately 50% were given in-kind (seed, hoes, inoculant, storage sacks). In the information treatment, farmers received standard NASFAM extension services (primarily including group trainings and demonstration plots run by lead farmers) or intensive extension services in which they receive at least three one-on-one visits from professional extension workers in addition to group trainings. We find strong evidence that both the framed cash transfer and input packages led to large increases in the overall value of agricultural production, as well as production yields for groundnuts. Evidence also strongly suggests that farmers are increasing resources to inputs, specifically pesticides and the use of casual labor that leads directly to these increases. In addition, farmers invest in both agricultural specific assets particularly among those receiving the cash transfers. Furthermore, intensive extension complements the transfers particularly for crops that farmers have less experience growing.
15
Figure 1: Timeline of Activities
Interventions (Year 1) Baseline Survey
β’ Transfer + AFO visit #1 β’ Transfer + AFO visit #2
Interventions (Year 1) Midline survey 1 (Subsample)
β’ Transfer + AFO visit #3
Interventions (Year 2) Midline survey 2
β’ Business Farm Tool β’ Conservation
Midline survey 3
Figure 2: Self-reported use of transfers 14000 Inputs Food
12000
Ganyu Health
10000
Household items Land Fees 8000
Livestock Milling
6000
Savings Education
4000
Seeds Shelling Tools
2000
Transport Other
0 November
February
April
16
Table 1 Panel A: Balance Information Intervention
NASFAM member characteristics Female Married Age Education None Some primary Completed primary Some secondary Completed secondary Some post-secondary
Intensive (Mean) (1)
Standard (Mean) (2)
Intensive = Standard (pvalue) (3)
0.660 0.840 40.81
0.605 0.799 40.77
0.309 0.0891 0.968
0.164 0.553 0.154 0.0760 0.0512 0.00177
0.178 0.523 0.163 0.0874 0.0469 0.00162
0.596 0.341 0.709 0.536 0.733 0.950
Table 1 Panel B: Balance Information Intervention
Household Composition Household size Number of female adults (16-59 years) Number of male adults (16-59 years) Number of female adults (60+ years) Number of male adults (60+ years) Number of children (under 16 years) Agricultural Production: Number of crops grown Total value of sales (in USD) Area of land owned (in acres) Area of land rented (in acres) Area of land borrowed (in acres)
Intensive (Mean) (1)
Standard (Mean) (2)
Intensive = Standard (p-value) (3)
5.743 1.326 1.312 0.125 0.138 2.801
5.385 1.238 1.238 0.142 0.121 2.636
0.0141 0.102 0.265 0.470 0.476 0.118
4.560 277.7 4.178 0.557 0.0592
4.545 318.1 4.110 0.510 0.0891
0.914 0.281 0.742 0.516 0.379
17
Table 1 Panel C: Balance Information Intervention
Production (in kgs): Maize Cassava Sweet Potatoes Irish Potatoes Groundnuts Common beans Soy Ground beans Pigeon peas Tobacco Pumpkins Other vegetables
Intensive (Mean)
Standard (Mean)
(1)
(2)
Intensive = Standard (p-value) (3)
1138.000 4.938 37.680 9.824 213.600 23.660 144.300 5.787 2.187 163.100 178.800 8.236
1175.000 10.620 42.570 19.080 245.500 22.890 152.400 4.588 2.304 182.300 195.400 9.383
0.707 0.379 0.584 0.234 0.399 0.909 0.751 0.504 0.951 0.509 0.532 0.726
18
Table 2 Panel A: Balance Transfer Intervention
NASFAM member characteristics Female Married Age Education None Some primary Completed primary Some secondary Completed secondary Some post-secondary Household Composition Household size Number of female adults (16-59 years) Number of male adults (16-59 years) Number of female adults (60+ years) Number of male adults (60+ years) % of adult household members (16-59) who are female Number of children (under 16 years) Agricultural Production: Number of crops grown Total value of sales (in USD) Area of land owned (in ha) Area of land rented (in ha) Area of land borrowed (in ha)
Control = Cash (5)
Cash = Inputs (6)
Control = Inputs = Cash (7)
Control (1)
Inputs (2)
Cash (3)
Control = Inputs (4)
0.638 0.811 40.500
0.598 0.826 41.230
0.655 0.818 40.660
0.551 0.623 0.565
0.792 0.823 0.898
0.400 0.752 0.644
0.694 0.883 0.830
0.149 0.520 0.184 0.097 0.050 0.000
0.208 0.544 0.141 0.060 0.044 0.003
0.158 0.548 0.151 0.088 0.053 0.003
0.055 0.486 0.192 0.090 0.709 0.317
0.785 0.479 0.277 0.696 0.848 0.316
0.104 0.934 0.743 0.210 0.586 0.980
0.104 0.697 0.382 0.202 0.853 0.366
5.458 1.261 1.239 0.132 0.129
5.575 1.256 1.259 0.158 0.130
5.635 1.323 1.323 0.113 0.128
0.550 0.937 0.810 0.371 0.995
0.332 0.381 0.262 0.429 0.949
0.697 0.295 0.436 0.123 0.943
0.622 0.549 0.502 0.298 0.997
0.517 2.667
0.533 2.738
0.521 2.740
0.368 0.592
0.806 0.588
0.520 0.989
0.657 0.833
4.535 282.700 4.088 0.544 0.060
4.440 304.400 4.076 0.521 0.083
4.678 309.800 4.262 0.532 0.082
0.540 0.635 0.960 0.802 0.551
0.393 0.551 0.512 0.896 0.581
0.201 0.911 0.434 0.896 0.975
0.439 0.812 0.707 0.968 0.776
19
Control (1) Production (in kgs): Maize Cassava Sweet Potatoes Irish Potatoes Groundnuts Common beans Soy Ground beans Pigeon peas Tobacco Pumpkins Tomato Other vegetables
1151.000 6.953 28.970 14.340 229.000 23.300 140.700 4.679 1.649 179.400 171.200 7.052 5.142
Table 2 Panel B: Balance Transfer Intervention Control= Inputs Cash Inputs (2) (3) (4)
1086.000 2.591 41.420 10.510 207.400 22.100 148.200 5.210 2.184 170.100 149.200 7.591 10.620
1233.000 14.000 50.400 18.990 253.600 24.330 156.700 5.595 2.913 169.800 240.800 6.662 10.830
0.572 0.241 0.157 0.650 0.607 0.893 0.813 0.796 0.819 0.796 0.423 0.923 0.168
Control= Cash (5)
Cash= Inputs (6)
Control=Inputs =Cash (7)
0.483 0.459 0.062 0.640 0.619 0.891 0.597 0.687 0.546 0.793 0.043 0.896 0.092
0.217 0.230 0.447 0.433 0.335 0.795 0.800 0.859 0.774 0.993 0.007 0.869 0.964
0.464 0.306 0.124 0.732 0.621 0.966 0.869 0.920 0.833 0.957 0.024 0.983 0.135
20
Table 3: Crop Adoption
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Maize
Groundnuts
Soy
Groundnuts & Soy
Tobacco
Other crops
(1) -0.00248 (0.00428) 0.0777 (0.0898)
(2) -0.0173 (0.0287) 0.179*** (0.0514)
(3) 0.0415 (0.0316) 0.214*** (0.0318)
(4) 0.00205 (0.00743) 0.0406 (0.0385)
(5) 0.0182 (0.0313) 0.468*** (0.0381)
(6) -0.0371 (0.0328) -0.0294 (0.0367)
1,145 0.058 0.995
1,145 0.189 0.842
1,145 0.215 0.700
1,145 0.037 0.988
1,145 0.363 0.450
1,145 0.092 0.492
-0.0119** (0.00468) -0.00683 (0.00433) 0.0786 (0.0893)
0.0592* (0.0332) 0.0718* (0.0369) 0.178*** (0.0517)
0.0459 (0.0366) -0.0229 (0.0408) 0.211*** (0.0315)
0.0286*** (0.00779) 0.0145 (0.0101) 0.0422 (0.0382)
0.0143 (0.0368) -0.0363 (0.0386) 0.467*** (0.0381)
-0.00448 (0.0388) 0.0301 (0.0383) -0.0289 (0.0368)
1,145 0.061 1
1,145 0.194 0.795
1,145 0.216 0.716
1,145 0.046 0.976
1,145 0.364 0.465
1,145 0.091 0.478
0.321
0.713
0.0591
0.0795
0.171
0.392
0.0143* (0.00780) -0.00159 (0.00524) 0.00490 (0.00801) -0.0227 (0.0137) -0.0262** (0.0132) 0.0792 (0.0896)
0.0482 (0.0545) 0.0886 (0.0581) 0.157*** (0.0503) -0.0602 (0.0909) -0.176** (0.0853) 0.181*** (0.0511)
-0.0463 (0.0675) -0.0149 (0.0597) -0.0917 (0.0606) 0.122 (0.102) 0.146 (0.113) 0.210*** (0.0311)
0.00904 (0.0133) 0.0257** (0.0106) 0.0340*** (0.0103) 0.00521 (0.0191) -0.0393** (0.0185) 0.0437 (0.0382)
-0.0680 (0.0688) -0.0168 (0.0617) -0.136* (0.0691) 0.0677 (0.105) 0.208* (0.107) 0.464*** (0.0380)
-0.0926 (0.0634) -0.0658 (0.0619) 0.0297 (0.0640) 0.145 (0.105) 0.0237 (0.0990) -0.0246 (0.0366)
Observations 1,145 1,145 1,145 1,145 1,145 R-squared 0.063 0.199 0.220 0.050 0.368 Control mean 1 0.805 0.681 0.969 0.487 p-value for equality of coefficients 0.482 0.153 0.142 0.230 0.0665 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.710 0.385 0.348 0.0236 0.699 Input*Intensive) Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
1,145 0.095 0.509 0.173 0.618
21
Table 4: Land Allocation by Crop (Acres)
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
(6)
Total Land Cultivated (Crop Level) (7)
Total Land (Household Level) (8)
-0.0110 (0.0389) 0.366*** (0.0470)
-0.126 (0.0779) 0.0513* (0.0300)
-0.158 (0.133) 0.316*** (0.0425)
-0.119 (0.120) 0.266*** (0.0394)
1,145 0.237 1.562
1,144 0.334 0.445
1,142 0.097 0.650
1,141 0.296 4.375
1,145 0.233 3.963
0.133** (0.0673) 0.0255 (0.0597) 0.251*** (0.0479)
0.300*** (0.0701) 0.282*** (0.0644) 0.364*** (0.0386)
-0.0395 (0.0473) -0.127*** (0.0483) 0.368*** (0.0469)
-0.142 (0.0922) -0.119 (0.104) 0.0535* (0.0307)
0.0537 (0.174) -0.00742 (0.154) 0.317*** (0.0431)
0.357*** (0.134) 0.231 (0.154) 0.264*** (0.0398)
1,145 0.366 0.774
1,145 0.212 0.627
1,145 0.255 1.400
1,144 0.339 0.481
1,142 0.097 0.715
1,141 0.295 4.331
1,145 0.236 3.789
0.546
0.163
0.0710
0.766
0.0216
0.795
0.661
0.373
-0.178 (0.109) -0.178* (0.0983) -0.119 (0.103) 0.271* (0.159) 0.228 (0.174) 0.404*** (0.0442)
-0.0474 (0.0899) 0.201** (0.0858) 0.281*** (0.0845) -0.0422 (0.152) -0.0496 (0.147) 0.384*** (0.0407)
-0.180 (0.110) 0.00658 (0.104) -0.165* (0.0903) 0.273* (0.162) 0.412** (0.174) 0.255*** (0.0478)
-0.248** (0.112) 0.191* (0.103) 0.111 (0.0953) 0.261 (0.167) 0.387** (0.182) 0.366*** (0.0382)
-0.115 (0.0785) -0.105 (0.0770) -0.218*** (0.0795) 0.148 (0.114) 0.203* (0.119) 0.365*** (0.0470)
-0.477*** (0.145) -0.474*** (0.125) -0.256* (0.153) 0.762*** (0.206) 0.381 (0.241) 0.0542* (0.0307)
-1.016*** (0.239) -0.599** (0.258) -0.490** (0.240) 1.494*** (0.372) 1.176*** (0.373) 0.314*** (0.0426)
-0.958*** (0.221) -0.188 (0.186) -0.293 (0.242) 1.267*** (0.331) 1.242*** (0.375) 0.262*** (0.0395)
Maize
Groundnuts
Soy
Groundnuts + Soy
Tobacco
Other Crops
(1)
(2)
(3)
(4)
(5)
-0.0257 (0.0469) 0.405*** (0.0438)
-0.0523 (0.0489) 0.383*** (0.0410)
0.0494 (0.0545) 0.253*** (0.0483)
-0.00699 (0.0628) 0.364*** (0.0393)
1,145 0.329 1.712
1,145 0.349 0.921
1,145 0.207 0.641
-0.0580 (0.0614) -0.0237 (0.0581) 0.407*** (0.0442)
0.170*** (0.0524) 0.250*** (0.0570) 0.384*** (0.0407)
1,145 0.329 1.735
Observations 1,145 1,145 1,145 1,145 1,144 1,142 1,141 1,145 R-squared 0.331 0.369 0.219 0.258 0.341 0.106 0.304 0.244 Control mean 1.840 0.803 0.582 1.385 0.526 0.862 4.613 3.962 p-value for equality of coefficients 0.549 0.376 0.0950 0.453 0.0802 0.0678 0.624 0.649 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.848 0.395 0.708 0.669 0.362 0.254 0.393 0.499 Input*Intensive) Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
22
Table 5: Crop Production (Kgs)β
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
-51.38 (67.94) 0.356*** (0.0456)
-9.328 (15.46) 0.136*** (0.0353)
-2.495 (15.56) 0.315*** (0.0513)
1,139 0.245 992.6
1,141 0.288 192.4
58.36 (85.23) 100.4 (81.75) 0.357*** (0.0458)
Tobacco
Other Crops
All Crops
(5)
(6)
(7)
-11.11 (24.70) 0.199*** (0.0414)
8.191 (16.52) 0.533*** (0.0541)
-46.61** (19.91) 0.0765** (0.0311)
-62.07 (99.35) 0.414*** (0.0434)
1,143 0.265 146.3
1,131 0.273 349.2
1,128 0.417 127.5
1,075 0.106 135.2
1,051 0.337 1635
52.90*** (17.46) 80.54*** (19.98) 0.139*** (0.0355)
22.06 (18.56) -7.960 (18.04) 0.312*** (0.0510)
67.55** (28.11) 66.11** (30.23) 0.199*** (0.0421)
-45.64** (19.49) -42.35** (19.53) 0.531*** (0.0538)
16.44 (28.32) 16.10 (22.48) 0.0775** (0.0313)
47.26 (123.4) 139.8 (123.4) 0.416*** (0.0445)
1,139 0.245 996.1
1,141 0.302 145.3
1,143 0.268 142.8
1,131 0.280 304.6
1,128 0.422 162.7
1,075 0.102 106.6
1,051 0.338 1601
0.594
0.160
0.0797
0.962
0.827
0.989
0.466
23
Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
-156.8 (141.8) -5.013 (124.6) 44.66 (127.1) 160.8 (200.1) 143.3 (218.6) 0.356*** (0.0458)
-7.952 (33.40) 52.04 (33.94) 99.78*** (32.98) 6.262 (57.27) -35.27 (55.97) 0.139*** (0.0357)
-56.59* (30.95) -31.37 (31.06) -23.59 (27.20) 116.8** (48.31) 44.94 (47.18) 0.314*** (0.0510)
-51.47 (44.89) 19.42 (43.05) 78.92* (45.36) 108.9 (74.82) -11.87 (74.27) 0.200*** (0.0429)
-21.02 (35.66) -70.51** (28.26) -70.08** (28.46) 50.41 (44.41) 59.41 (47.24) 0.530*** (0.0535)
-90.52* (46.17) 6.747 (60.01) -22.09 (36.83) 39.90 (86.25) 91.80 (68.89) 0.0734** (0.0314)
-252.7 (203.1) -101.2 (205.6) 48.18 (193.8) 345.9 (340.2) 233.4 (324.1) 0.414*** (0.0446)
Observations 1,139 1,141 1,143 1,131 1,128 1,075 1,051 R-squared 0.246 0.303 0.276 0.283 0.423 0.108 0.339 Control mean 978.6 142.1 133.9 291.4 154.6 130.1 1562 p-value for equality of coefficients 0.715 0.197 0.806 0.306 0.989 0.547 0.522 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.812 0.808 0.00940 0.154 0.651 0.423 0.841 Input*Intensive) β Observations greater than the 99% percentile winsorized Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
24
Table 6: Yield of Major Crops (Kg/Acre) Maize Groundnuts (1) (2) Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Soy (3)
Tobacco (4)
-24.47 (47.61) 0.265*** (0.0546)
-22.33 (13.84) 0.0432 (0.0272)
-15.58 (21.11) 0.0271 (0.0738)
43.77 (35.78) 0.549** (0.241)
1,139 0.162 634
1,141 0.183 186.8
1,143 0.079 198.4
1,127 0.148 138.8
10.83 (53.46) 52.53 (61.83) 0.267*** (0.0539)
37.22** (17.69) 46.97** (18.52) 0.0430 (0.0269)
4.414 (25.02) 15.69 (30.79) 0.0245 (0.0750)
-36.36 (26.83) -57.63** (25.79) 0.545** (0.240)
1,139 0.162 633.7
1,141 0.187 149.3
1,143 0.079 186.2
1,127 0.148 197.8
0.474
0.577
0.726
0.345
7.488 (94.89) 22.93 (95.09) 98.78 (89.58) -20.13 (152.2) -91.47 (126.5) 0.268*** (0.0536)
11.05 (32.31) 54.25 (33.44) 94.18*** (28.77) -30.44 (54.94) -94.07* (51.98) 0.0429 (0.0275)
-67.33 (43.75) -77.87* (41.78) 39.41 (52.29) 180.0** (71.83) -27.12 (78.20) 0.0370 (0.0761)
96.09 (101.2) 0.0128 (51.33) -41.18 (47.34) -95.35 (110.6) -52.10 (120.4) 0.546** (0.240)
Observations 1,139 1,141 1,143 1,127 R-squared 0.163 0.193 0.086 0.150 Control mean 571.9 145.5 176.3 149.9 p-value for equality of coefficients 0.525 0.200 0.0951 0.361 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.954 0.308 0.0146 0.946 Input*Intensive) Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1 25
Maize (1) Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input
Table 7: Value of Production (Kwacha)β Groundnuts Groundnuts Soy + Soy (2) (3) (4)
Tobacco
Other Crops
All Crops
(5)
(6)
(7)
-5,435 (9,416) 0.577*** (0.0963)
-455.0 (4,276) 0.275*** (0.0571)
495.8 (3,119) 0.292*** (0.0651)
-278.4 (6,847) 0.377*** (0.0729)
4,679 (7,808) 0.569*** (0.0578)
-3,088** (1,502) 0.0942*** (0.0343)
8,991 (24,153) 0.770*** (0.0698)
1,139 0.207 97805
1,141 0.364 53614
1,143 0.199 27066
1,139 0.261 86573
1,128 0.364 65291
1,062 0.122 9356
1,038 0.312 278270
19,212* (11,238) 14,047 (10,989) 0.580*** (0.0956)
17,503*** (4,948) 18,404*** (4,817) 0.277*** (0.0577)
5,256 (3,665) -290.3 (3,646) 0.287*** (0.0649)
22,901*** (8,674) 14,300* (7,671) 0.376*** (0.0742)
-7,795 (10,347) -14,261 (8,612) 0.568*** (0.0577)
54.22 (1,988) -67.53 (1,667) 0.0957*** (0.0349)
55,970* (29,424) 31,010 (27,980) 0.770*** (0.0716)
1,139 0.209 93872
1,141 0.374 43635
1,143 0.202 25578
1,139 0.267 75999
1,128 0.365 76209
1,062 0.118 8408
1,038 0.315 267906
0.633
0.870
0.118
0.323
0.471
0.947
0.356
26
Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
-24,485 (19,045) 12,862 (15,837) -1,771 (15,188) 17,479 (26,313) 35,898 (29,461) 0.574*** (0.0962)
-17,188** (8,452) 10,915 (8,386) 4,740 (7,621) 16,087 (15,766) 30,305** (14,754) 0.276*** (0.0577)
-7,970 (5,950) -2,763 (6,541) -3,022 (5,540) 17,336* (10,028) 7,382 (9,712) 0.287*** (0.0651)
-25,850** (11,566) 8,907 (13,713) -3,400 (12,573) 32,297 (20,528) 40,350* (20,971) 0.375*** (0.0745)
-4,134 (16,104) -11,387 (13,355) -26,722* (15,925) 6,583 (20,837) 25,254 (26,318) 0.567*** (0.0580)
-6,030* (3,247) -593.4 (4,054) -2,877 (2,661) 2,605 (5,795) 6,627 (4,818) 0.0930*** (0.0346)
-11,919 (52,368) 64,139 (40,976) -3,257 (40,047) -16,514 (70,402) 68,584 (75,501) 0.767*** (0.0720)
Observations 1,139 1,141 1,143 1,139 1,128 1,062 1,038 R-squared 0.211 0.377 0.206 0.270 0.366 0.124 0.316 Control mean 92662 42951 24282 73183 76014 10215 250346 p-value for equality of coefficients 0.413 0.565 0.966 0.461 0.350 0.467 0.184 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.849 0.356 0.0521 0.746 0.850 0.461 0.709 Input*Intensive) β Observations greater than the 99% percentile winsorized Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
27
Maize (1) Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Table 8: Value of Crop Sales (Kwacha)β Groundnuts Groundnuts Soy + Soy (2) (3) (4)
Tobacco
Other Crops
All Crops
(5)
(6)
(7)
671.7 (1,314) 0.419*** (0.119)
1,768 (2,644) 0.326*** (0.0508)
973.3 (1,965) 0.255*** (0.0664)
3,042 (4,018) 0.361*** (0.0692)
-558.5 (5,960) 0.469*** (0.0505)
-1,254* (658.4) 0.246*** (0.0662)
-997.1 (7,781) 0.544*** (0.0552)
1,132 0.141 7293
1,119 0.437 21320
1,136 0.222 16971
1,110 0.349 39075
1,090 0.341 54607
1,142 0.135 3243
1,105 0.387 108273
4,346*** (1,603) 1,672 (1,359) 0.421*** (0.118)
9,704*** (3,040) 9,334*** (2,903) 0.322*** (0.0511)
2,012 (2,357) -1,463 (2,212) 0.253*** (0.0665)
12,712*** (4,664) 8,236* (4,417) 0.356*** (0.0697)
-4,863 (8,113) -7,542 (7,124) 0.469*** (0.0502)
95.67 (888.4) 50.46 (753.8) 0.248*** (0.0665)
15,227 (9,691) 3,546 (9,722) 0.545*** (0.0556)
1,132 0.147 6909
1,119 0.445 16614
1,136 0.224 17415
1,110 0.355 35096
1,090 0.341 59989
1,142 0.131 2562
1,105 0.389 107072
0.0878
0.909
0.0918
0.345
0.695
0.955
0.152
-3,313 (2,167) 1,854 (2,175) -847.8 (2,083) 5,445 (3,564) 5,679 (3,967) 0.422*** (0.118)
-7,991 (5,197) 3,515 (4,734) 3,441 (4,595) 13,586 (8,884) 13,396 (9,040) 0.322*** (0.0514)
-3,098 (4,318) -2,911 (3,941) -1,934 (3,433) 10,262 (6,499) 1,808 (6,541) 0.253*** (0.0671)
-9,805 (7,937) 1,606 (7,007) 3,436 (6,860) 23,616* (12,403) 11,932 (13,200) 0.357*** (0.0698)
-21,306 (12,911) -15,043 (10,650) -28,031** (13,386) 23,120 (16,504) 44,736** (21,676) 0.467*** (0.0503)
-2,635 (1,611) -320.1 (1,946) -1,113 (1,286) 1,451 (2,892) 2,841 (2,175) 0.245*** (0.0664)
-33,337** (16,247) -1,968 (13,978) -21,345 (14,688) 39,826* (21,225) 56,565** (24,176) 0.543*** (0.0556)
Observations 1,132 1,119 1,136 1,110 1,090 1,142 1,105 R-squared 0.149 0.448 0.228 0.359 0.344 0.137 0.391 Control mean 6884 15877 15635 31517 61837 3118 106508 p-value for equality of coefficients 0.217 0.988 0.780 0.815 0.307 0.598 0.188 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.474 0.959 0.0315 0.176 0.532 0.557 0.855 Input*Intensive) β Observations greater than the 99% percentile winsorized Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
28
Table 9: Input Use & Expenditures (Kwacha) Fertilizer Used Fertilizer Expenditure Used Pesticides Used Irrigation (Kwacha) (1) (2) (3) (4) Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Used Ganyu Labourβ‘
Ganyu Labour Expenditureβ‘
Total Input Expenditure
(5)
(6)
(7)
-0.00259 (0.0179) 0.295*** (0.0401)
1,509 (2,116) 0.275*** (0.0345)
0.0557* (0.0331) 0.185*** (0.0485)
-0.0261 (0.0269) 0.153** (0.0647)
0.0595 (0.0408) -
4,869*** (1,839) -
6,492** (2,955) 0.338*** (0.0419)
1,145 0.231 0.841
1,124 0.307 28546
1,145 0.125 0.252
1,143 0.077 0.291
1,144 0.123 0.703
1,143 0.107 13935
1,145 0.340 52428
0.0137 (0.0187) -0.00298 (0.0249) 0.294*** (0.0402)
2,087 (2,783) -1,622 (2,154) 0.275*** (0.0348)
0.118*** (0.0355) 0.0748** (0.0367) 0.186*** (0.0488)
-0.0108 (0.0330) -0.0257 (0.0303) 0.155** (0.0652)
0.337*** (0.0305) 0.303*** (0.0306) -
6,665*** (2,186) 6,520*** (2,087) -
14,936*** (3,681) 10,051*** (2,900) 0.338*** (0.0421)
1,145 0.231 0.849
1,124 0.308 28269
1,145 0.132 0.205
1,143 0.077 0.292
1,144 0.216 0.505
1,143 0.112 12107
1,145 0.347 47727
0.486
0.131
0.280
0.659
0.278
0.946
0.169
-0.121*** (0.0420) -0.0621* (0.0330) -0.0912* (0.0512) 0.170*** (0.0620) 0.199** (0.0780) 0.292*** (0.0402)
-2,854 (4,561) -2,585 (3,474) -2,922 (4,168) 9,652 (5,905) 3,366 (7,406) 0.275*** (0.0347)
-0.148*** (0.0562) 0.0246 (0.0579) -0.119** (0.0580) 0.199** (0.0856) 0.409*** (0.0965) 0.190*** (0.0497)
-0.0540 (0.0518) -0.0125 (0.0549) -0.0607 (0.0484) 0.0151 (0.0824) 0.0789 (0.0847) 0.155** (0.0651)
-0.0604 (0.0514) 0.313*** (0.0440) 0.204*** (0.0495) 0.0529 (0.0779) 0.204** (0.0804) -
-2,969 (3,534) 696.4 (2,714) 1,452 (3,293) 11,704** (5,140) 10,694* (5,748) -
-6,551 (5,429) 4,487 (5,125) 3,450 (5,367) 21,141*** (8,058) 14,641 (9,149) 0.338*** (0.0419)
Observations 1,145 1,124 1,145 1,143 1,144 1,143 1,145 R-squared 0.237 0.310 0.148 0.078 0.221 0.120 0.351 Control mean 0.858 28560 0.221 0.327 0.496 10917 47070 p-value for equality of coefficients 0.441 0.932 0.00984 0.359 0.0425 0.809 0.850 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.997 0.173 0.207 0.722 0.360 0.949 0.234 Input*Intensive) β‘ Baseline value unavailable for dependent variable (questions were added to midline survey), not included in specification Notes: 1,3-5 Linear Probability Model, 2,6-7 Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
29
Table 10a: Fertilizer Usage by Crop
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
-0.00679 (0.0211) 0.271*** (0.0352)
0.00153 (0.00151) -0.00297 (0.00508)
0.0126 (0.0110) 0.00726 (0.0326)
1,145 0.268 0.785
1,145 0.043 0
-0.0313 (0.0233) -0.0153 (0.0283) 0.271*** (0.0350)
Tobacco
Other Crops
(5)
(6)
0.0140 (0.0111) -0.00347 (0.0268)
0.00902 (0.0292) 0.439*** (0.0383)
-0.00652 (0.0193) -0.00782 (0.0232)
1,144 0.095 0.0151
1,144 0.092 0.0151
1,145 0.320 0.396
1,145 0.085 0.128
0.000477 (0.00147) 0.00346 (0.00259) -0.00388 (0.00562)
0.0281** (0.0127) 0.00353 (0.0115) 0.00976 (0.0329)
0.0284** (0.0128) 0.00700 (0.0119) -0.00206 (0.0270)
0.0248 (0.0346) -0.0188 (0.0352) 0.439*** (0.0380)
-0.0186 (0.0239) -0.00519 (0.0225) -0.00702 (0.0232)
1,145 0.269 0.805
1,145 0.044 0
1,144 0.099 0.0135
1,144 0.095 0.0135
1,145 0.321 0.408
1,145 0.085 0.132
0.576
0.173
0.0351
0.0724
0.158
0.581
-0.0284 (0.0434) -0.0265 (0.0390) -0.0581 (0.0590) -0.00585 (0.0691) 0.0883 (0.0857) 0.268*** (0.0347) 1,145 0.270 0.801
0.00116 (0.00175) 0.00242 (0.00213) 0.000357 (0.00167) -0.00443 (0.00315) 0.00563 (0.00460) -0.00396 (0.00576) 1,145 0.047 0
-0.0122 (0.0172) 0.00288 (0.0148) -0.00333 (0.0204) 0.0510** (0.0230) 0.0171 (0.0279) 0.00864 (0.0331) 1,144 0.102 0.0221
-0.0112 (0.0173) 0.00510 (0.0151) -0.00316 (0.0204) 0.0468** (0.0233) 0.0232 (0.0285) -0.00237 (0.0271) 1,144 0.098 0.0221
-0.0959 (0.0680) -0.0204 (0.0579) -0.125** (0.0619) 0.102 (0.0950) 0.227** (0.0994) 0.436*** (0.0380) 1,145 0.325 0.438
-0.0671 (0.0416) -0.0879** (0.0411) -0.0185 (0.0364) 0.151** (0.0690) 0.0435 (0.0598) -0.00422 (0.0233) 1,145 0.090 0.155
Observations R-squared Control mean p-value for equality of coefficients 0.468 0.202 0.705 0.617 0.0652 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.124 0.161 0.0337 0.113 0.679 Input*Intensive) Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.0892 0.266
30
Table 10b: Fertilizer Usage by Crop (If Grown)
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
-0.00310 (0.0215) 0.281*** (0.0366)
0.00181 (0.00181) -0.00446 (0.00784)
0.0100 (0.0148) 0.0287 (0.0367)
1,127 0.270 0.789
890 0.058 0
-0.0198 (0.0236) -0.0146 (0.0292) 0.281*** (0.0365)
Tobacco
Other Crops
(5)
(6)
0.0150 (0.0103) 0.00182 (0.0261)
-0.0291 (0.0245) 0.144 (0.105)
0.00873 (0.0426) -0.0542 (0.0626)
633 0.086 0.0216
1,092 0.084 0.0153
437 0.223 0.881
413 0.217 0.259
0.000506 (0.00183) 0.00419 (0.00312) -0.00550 (0.00841)
0.0256 (0.0164) -0.0146 (0.0150) 0.0351 (0.0375)
0.0239* (0.0121) 0.000446 (0.0109) 0.00345 (0.0264)
0.0170 (0.0317) 0.00520 (0.0348) 0.150 (0.106)
0.0188 (0.0450) -0.0444 (0.0550) -0.0511 (0.0633)
1,127 0.270 0.805
890 0.060 0
633 0.093 0.0189
1,092 0.087 0.0139
437 0.222 0.878
413 0.219 0.277
0.861
0.177
0.0143
0.0383
0.713
0.240
-0.0440 (0.0445) -0.0302 (0.0394) -0.0682 (0.0594) 0.0271 (0.0717) 0.113 (0.0866) 0.278*** (0.0361) 1,127 0.272 0.801
0.00152 (0.00224) 0.00300 (0.00265) 0.000501 (0.00209) -0.00569 (0.00408) 0.00644 (0.00530) -0.00558 (0.00864) 890 0.063 0
-0.0309 (0.0254) -0.0141 (0.0213) -0.0282 (0.0297) 0.0817** (0.0340) 0.0335 (0.0443) 0.0336 (0.0373) 633 0.098 0.0325
-0.00982 (0.0162) 0.00114 (0.0139) -0.0110 (0.0197) 0.0449** (0.0221) 0.0255 (0.0276) 0.00332 (0.0265) 1,092 0.090 0.0228
-0.135** (0.0624) -0.0609 (0.0443) -0.0426 (0.0699) 0.190** (0.0930) 0.145 (0.110) 0.152 (0.105) 437 0.229 0.900
-0.0687 (0.0756) -0.0850 (0.0729) 0.00102 (0.0868) 0.253* (0.132) -0.0313 (0.126) -0.0408 (0.0640) 413 0.229 0.304
Observations R-squared Control mean p-value for equality of coefficients 0.403 0.189 0.479 0.340 0.756 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.254 0.161 0.0222 0.106 0.547 Input*Intensive) Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.383 0.00464
31
Table 11a: Pesticide Usage by Crop
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
0.00233 (0.00945) 0.279** (0.119)
-0.00530 (0.00581) 0.00308 (0.00842)
0.0909*** (0.0315) 0.139 (0.159)
1,145 0.105 0.0386
1,145 0.085 0.0101
0.0388*** (0.0111) 0.0468*** (0.0123) 0.285** (0.117)
Tobacco
Other Crops
(5)
(6)
0.0854*** (0.0321) 0.0194 (0.0957)
-0.0110 (0.0189) 0.263*** (0.0479)
-0.00951 (0.0141) 0.0629 (0.0464)
1,144 0.158 0.108
1,144 0.150 0.116
1,145 0.163 0.0839
1,145 0.079 0.0805
0.00770 (0.00612) 0.00128 (0.00620) 0.00424 (0.00922)
0.124*** (0.0309) 0.0833** (0.0386) 0.203 (0.151)
0.128*** (0.0317) 0.0845** (0.0382) 0.0624 (0.0972)
-0.0138 (0.0235) -0.0370* (0.0215) 0.265*** (0.0476)
0.00899 (0.0186) 0.00676 (0.0157) 0.0641 (0.0461)
1,145 0.114 0.0189
1,145 0.085 0.00270
1,144 0.163 0.0649
1,144 0.157 0.0676
1,145 0.165 0.0973
1,145 0.079 0.0730
0.509
0.397
0.286
0.255
0.236
0.901
-0.0492** (0.0198) 0.0171 (0.0219) 0.00223 (0.0232) 0.0509 (0.0337) 0.0975** (0.0373) 0.290** (0.116) 1,145 0.118 0.0265
-0.00448 (0.0104) 0.0108 (0.00809) 0.00212 (0.0143) -0.00458 (0.0188) -0.000815 (0.0213) 0.00532 (0.00925) 1,145 0.086 0.00442
-0.0762 (0.0495) 0.0319 (0.0432) -0.0689 (0.0463) 0.179** (0.0696) 0.312*** (0.0826) 0.154 (0.144) 1,144 0.185 0.0708
-0.0756 (0.0488) 0.0433 (0.0436) -0.0633 (0.0454) 0.166** (0.0695) 0.304*** (0.0817) 0.0326 (0.0919) 1,144 0.176 0.0752
-0.0208 (0.0377) 0.00726 (0.0335) -0.0772** (0.0352) -0.0388 (0.0488) 0.0805 (0.0569) 0.267*** (0.0481) 1,145 0.169 0.0885
-0.0876*** (0.0302) -0.0465 (0.0340) -0.0421 (0.0263) 0.126** (0.0523) 0.116** (0.0454) 0.0630 (0.0466) 1,145 0.084 0.0973
Observations R-squared Control mean p-value for equality of coefficients 0.454 0.405 0.0292 0.0232 0.0201 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.112 0.526 0.552 0.575 0.213 Input*Intensive) Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.896 0.786
32
Table 11b: Pesticide Usage by Crop (If Grown)
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
0.00314 (0.00959) 0.277** (0.119)
-0.00816 (0.00754) 0.00449 (0.0153)
0.112*** (0.0387) 0.242 (0.193)
1,127 0.107 0.0388
890 0.099 0.0120
0.0401*** (0.0111) 0.0479*** (0.0126) 0.283** (0.116)
Tobacco
Other Crops
(5)
(6)
0.0875*** (0.0326) 0.0473 (0.106)
-0.0103 (0.0432) 0.311*** (0.0661)
-0.00893 (0.0366) 0.0796 (0.0866)
633 0.190 0.154
1,092 0.154 0.117
437 0.263 0.187
413 0.185 0.164
0.00836 (0.00767) 7.82e-05 (0.00747) 0.00487 (0.0167)
0.152*** (0.0378) 0.107** (0.0483) 0.317* (0.178)
0.122*** (0.0325) 0.0814** (0.0399) 0.0873 (0.107)
0.00760 (0.0569) -0.0655 (0.0518) 0.317*** (0.0667)
0.000580 (0.0423) -0.0275 (0.0449) 0.0812 (0.0868)
1,127 0.116 0.0189
890 0.099 0.00340
633 0.196 0.0906
1,092 0.159 0.0693
437 0.268 0.209
413 0.185 0.153
0.530
0.383
0.366
0.321
0.199
0.520
-0.0507** (0.0200) 0.0176 (0.0219) 0.00110 (0.0236) 0.0532 (0.0341) 0.102*** (0.0379) 0.287** (0.115) 1,127 0.121 0.0265
-0.00672 (0.0110) 0.0130 (0.00916) 0.000590 (0.0165) -0.00719 (0.0206) 0.000357 (0.0235) 0.00812 (0.0163) 890 0.100 0.00549
-0.0896 (0.0609) 0.0432 (0.0492) -0.0784 (0.0587) 0.216*** (0.0761) 0.386*** (0.102) 0.250 (0.165) 633 0.222 0.104
-0.0821* (0.0493) 0.0377 (0.0425) -0.0826* (0.0456) 0.166** (0.0676) 0.337*** (0.0857) 0.0610 (0.100) 1,092 0.182 0.0776
0.0663 (0.0779) 0.104 (0.0916) -0.0792 (0.0875) -0.201 (0.134) -0.0137 (0.128) 0.321*** (0.0669) 437 0.273 0.182
-0.183*** (0.0653) -0.120** (0.0588) -0.105 (0.0685) 0.313*** (0.106) 0.228** (0.104) 0.0742 (0.0873) 413 0.197 0.191
Observations R-squared Control mean p-value for equality of coefficients 0.411 0.333 0.0479 0.00826 0.118 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.108 0.614 0.445 0.413 0.955 Input*Intensive) Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.840 0.341
33
Table 12a: Irrigation Usage by Crop
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
-0.0147 (0.0193) 0.218** (0.0971)
0.00368 (0.00230) -0.00578 (0.00487)
0.00857** (0.00328) -0.0128** (0.00538)
1,145 0.077 0.121
1,145 0.204 0
0.0315 (0.0233) 0.0100 (0.0208) 0.227** (0.0971)
Tobacco
Other Crops
(5)
(6)
0.0123*** (0.00386) -0.00889* (0.00451)
0.0125 (0.0221) 0.0624 (0.139)
-0.0227 (0.0147) 0.0248 (0.0481)
1,143 0.040 0
1,143 0.084 0
1,145 0.084 0.128
1,145 0.073 0.107
-0.000840 (0.00130) 0.00136 (0.00120) -0.00507 (0.00482)
0.00869** (0.00388) 0.00688 (0.00503) -0.0124** (0.00521)
0.00783* (0.00406) 0.00823 (0.00520) -0.00579 (0.00434)
-0.0101 (0.0252) -0.0299 (0.0268) 0.0680 (0.139)
-0.0261 (0.0203) 0.0153 (0.0192) 0.0222 (0.0481)
1,145 0.078 0.105
1,145 0.203 0
1,143 0.040 0
1,143 0.081 0
1,145 0.085 0.143
1,145 0.075 0.103
0.387
0.155
0.714
0.939
0.471
0.0314
0.0286 (0.0399) 0.0716* (0.0399) 0.0402 (0.0374) -0.0817 (0.0661) -0.0663 (0.0710) 0.227** (0.0974) 1,145 0.080 0.111
0.00121 (0.00228) -0.00227 (0.00315) -0.00205 (0.00234) 0.00196 (0.00423) 0.00641 (0.00455) -0.00796 (0.00526) 1,145 0.205 0
-0.00117 (0.00620) 0.000978 (0.00481) 0.000131 (0.00592) 0.0144 (0.00992) 0.0136 (0.0136) -0.0157** (0.00666) 1,143 0.043 0
4.33e-05 (0.00652) -0.00129 (0.00538) -0.00195 (0.00627) 0.0163 (0.0103) 0.0201 (0.0140) -0.0115** (0.00495) 1,143 0.086 0
-0.0159 (0.0413) -0.00377 (0.0396) -0.0880** (0.0379) -0.0145 (0.0604) 0.115* (0.0688) 0.0650 (0.138) 1,145 0.088 0.159
-0.0575 (0.0378) -0.0694* (0.0390) 0.0160 (0.0333) 0.100 (0.0633) 0.0131 (0.0578) 0.0229 (0.0476) 1,145 0.079 0.119
Observations R-squared Control mean p-value for equality of coefficients 0.413 0.928 0.811 0.874 0.0234 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.667 0.0776 0.862 0.742 0.287 Input*Intensive) Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.0140 0.946
34
Table 12b: Irrigation Usage by Crop (If Grown)
Panel A: Extension Treatment Intensive extension treatment Baseline value of dependent variable Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Baseline value of dependent variable Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Baseline value of dependent variable
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
-0.0144 (0.0197) 0.216** (0.0971)
0.00444 (0.00279) -0.00667 (0.00480)
0.00945** (0.00400) -0.0173** (0.00756)
1,127 0.076 0.121
890 0.251 0
0.0325 (0.0236) 0.0114 (0.0213) 0.225** (0.0971)
Tobacco
Other Crops
(5)
(6)
0.0126*** (0.00398) -0.00910* (0.00500)
0.0225 (0.0477) 0.0380 (0.194)
0.0256 (0.0379) 0.0994 (0.0929)
632 0.088 0
1,091 0.087 0
437 0.153 0.284
413 0.186 0.218
-0.00140 (0.00145) 0.00109 (0.00129) -0.00569 (0.00459)
0.0117** (0.00532) 0.00367 (0.00565) -0.0190** (0.00828)
0.00790* (0.00423) 0.00855 (0.00547) -0.00596 (0.00479)
-0.0223 (0.0573) -0.0711 (0.0610) 0.0433 (0.196)
-0.0179 (0.0473) -0.0236 (0.0437) 0.105 (0.0955)
1,127 0.078 0.105
890 0.250 0
632 0.089 0
1,091 0.084 0
437 0.155 0.308
413 0.186 0.215
0.408
0.203
0.229
0.906
0.450
0.901
0.0284 (0.0404) 0.0722* (0.0402) 0.0416 (0.0382) -0.0810 (0.0672) -0.0662 (0.0722) 0.225** (0.0974) 1,127 0.080 0.111
0.000581 (0.00242) -0.00408 (0.00341) -0.00314 (0.00273) 0.00456 (0.00451) 0.00802 (0.00520) -0.00859* (0.00514) 890 0.252 0
0.00295 (0.00687) 0.00549 (0.00601) 0.000626 (0.00985) 0.0106 (0.0113) 0.00604 (0.0152) -0.0226** (0.0101) 632 0.091 0
1.23e-05 (0.00669) -0.00163 (0.00555) -0.00164 (0.00647) 0.0171 (0.0106) 0.0203 (0.0145) -0.0118** (0.00536) 1,091 0.089 0
-0.0133 (0.0870) 0.00239 (0.0929) -0.201** (0.0804) -0.0438 (0.145) 0.232 (0.145) 0.0273 (0.194) 437 0.163 0.327
-0.0986 (0.0875) -0.128* (0.0762) -0.0585 (0.0696) 0.265* (0.139) 0.120 (0.123) 0.0871 (0.0964) 413 0.194 0.235
Observations R-squared Control mean p-value for equality of coefficients 0.434 0.739 0.383 0.998 0.0360 Cash = Input p-value for equality of coefficients (Cash + Cash*Intensive) = (Input + 0.678 0.0885 0.466 0.750 0.454 Input*Intensive) Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.373 0.238
35
Table 13a: Ganyu Usage by Cropβ‘ Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
0.0776** (0.0337)
0.0261 (0.0444)
0.0983** (0.0495)
1,144 0.121 0.362
1,143 0.242 0.471
0.134*** (0.0351) 0.142*** (0.0420)
Tobacco
Other Crops
(5)
(6)
0.0788 (0.0482)
0.0698*** (0.0253)
-0.00390 (0.00910)
1,145 0.162 0.294
1,143 0.112 0.629
1,145 0.095 0.106
1,145 0.054 0.0336
0.291*** (0.0448) 0.301*** (0.0435)
0.260*** (0.0471) 0.168*** (0.0512)
0.427*** (0.0347) 0.379*** (0.0353)
0.0333 (0.0335) 0.00530 (0.0281)
0.0119 (0.0123) 0.0156 (0.00954)
1,144 0.131 0.308
1,143 0.306 0.289
1,145 0.196 0.197
1,143 0.242 0.389
1,145 0.089 0.122
1,145 0.055 0.0243
0.828
0.840
0.0790
0.163
0.338
0.734
-0.0222 (0.0639) 0.0441 (0.0476) 0.100 (0.0632) 0.173* (0.0974) 0.0905 (0.101)
0.0580 (0.0726) 0.376*** (0.0757) 0.309*** (0.0711) -0.178 (0.133) -0.0318 (0.118)
-0.240*** (0.0734) 0.0245 (0.0702) -0.0483 (0.0700) 0.497*** (0.110) 0.480*** (0.125)
-0.100* (0.0571) 0.353*** (0.0450) 0.262*** (0.0577) 0.157* (0.0830) 0.251*** (0.0908)
0.00365 (0.0474) -0.0250 (0.0409) -0.0440 (0.0492) 0.105 (0.0688) 0.0972 (0.0763)
-0.0346** (0.0158) -0.0131 (0.0181) 0.00227 (0.0166) 0.0565** (0.0271) 0.0344 (0.0275)
1,144 0.137 0.305
1,143 0.309 0.270
1,145 0.225 0.212
1,143 0.247 0.394
1,145 0.098 0.106
1,145 0.057 0.0398
0.399
0.363
0.272
0.0902
0.657
0.297
p-value for equality of coefficients 0.651 0.260 0.218 0.960 0.631 (Cash + Cash*Intensive) = (Input + Input*Intensive) β‘ Baseline value unavailable for dependent variable (questions were added to midline survey), not included in specification Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.752
Panel A: Extension Treatment Intensive extension treatment Observations R-squared Control mean VARIABLES Panel B: Capital Treatment Cash treatment Input treatment Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Observations R-squared Control mean p-value for equality of coefficients Cash = Input
36
Table 13b: Ganyu Usage by Crop (If Grown)β‘ Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
0.0800** (0.0342)
0.0313 (0.0445)
0.111* (0.0575)
1,136 0.122 0.364
940 0.203 0.559
0.139*** (0.0353) 0.145*** (0.0423)
Tobacco
Other Crops
(5)
(6)
0.0782 (0.0484)
0.130*** (0.0454)
-0.00385 (0.0192)
845 0.152 0.420
1,129 0.113 0.636
533 0.138 0.235
554 0.100 0.0683
0.301*** (0.0435) 0.306*** (0.0423)
0.305*** (0.0483) 0.258*** (0.0576)
0.416*** (0.0352) 0.380*** (0.0364)
0.0956 (0.0588) 0.0663 (0.0569)
0.0298 (0.0251) 0.0293 (0.0193)
1,136 0.132 0.308
940 0.270 0.364
845 0.201 0.275
1,129 0.239 0.399
533 0.129 0.262
554 0.102 0.0508
0.865
0.916
0.393
0.300
0.626
0.984
-0.0274 (0.0637) 0.0439 (0.0478) 0.0982 (0.0639) 0.185* (0.0973) 0.101 (0.102)
0.0143 (0.0685) 0.366*** (0.0689) 0.258*** (0.0693) -0.137 (0.120) 0.0821 (0.107)
-0.289*** (0.0824) 0.0339 (0.0738) -0.00982 (0.0724) 0.575*** (0.115) 0.581*** (0.135)
-0.108* (0.0573) 0.340*** (0.0477) 0.249*** (0.0582) 0.163* (0.0842) 0.283*** (0.0894)
0.00492 (0.0849) -0.0276 (0.0911) -0.00385 (0.102) 0.221 (0.158) 0.144 (0.154)
-0.0856*** (0.0301) -0.0258 (0.0322) 0.000157 (0.0322) 0.137** (0.0562) 0.0897* (0.0506)
1,136 0.138 0.305
940 0.274 0.335
845 0.234 0.312
1,129 0.246 0.406
533 0.145 0.218
554 0.108 0.0783
0.419
0.140
0.509
0.0882
0.812
0.408
p-value for equality of coefficients 0.627 0.0877 0.610 0.599 0.549 (Cash + Cash*Intensive) = (Input + Input*Intensive) β‘ Baseline value unavailable for dependent variable (questions were added to midline survey), not included in specification Notes: Linear probability model, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
0.635
Panel A: Extension Treatment Intensive extension treatment Observations R-squared Control mean VARIABLES Panel B: Capital Treatment Cash treatment Input treatment Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs Observations R-squared Control mean p-value for equality of coefficients Cash = Input
37
Table 14a: Ganyu Expenditure by Crop (Kwacha)β‘
Panel A: Extension Treatment Intensive extension treatment Observations R-squared Control mean VARIABLES Panel B: Capital Treatment Cash treatment Input treatment Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
1,129* (655.1)
748.2 (704.0)
400.6 (666.1)
1,143 0.102 4218
1,143 0.168 5130
1,525* (871.3) 1,472** (683.4)
Tobacco
Other Crops
(5)
(6)
1,149 (1,082)
2,706*** (1,006)
-16.41 (139.6)
1,145 0.075 2857
1,143 0.103 7985
1,144 0.076 1419
1,145 0.037 313.1
3,804*** (720.3) 4,605*** (815.3)
2,764*** (825.4) 1,509** (716.3)
6,568*** (1,180) 6,119*** (1,110)
-1,143 (1,053) -1,326 (1,097)
-183.8 (191.2) 372.0 (259.0)
1,143 0.103 3899
1,143 0.201 2880
1,145 0.091 1719
1,143 0.139 4599
1,144 0.069 3328
1,145 0.041 275.4
0.949
0.335
0.105
0.690
0.838
0.0188
178.9 (1,177) -158.7 (1,029) 1,965 (1,249) 3,227 (1,951) -902.0 (2,018)
-1,386 (1,330) 3,132*** (1,173) 2,789** (1,241) 1,478 (2,301) 3,829 (2,432)
-2,350** (1,009) 906.3 (1,377) -62.28 (996.5) 4,063** (1,731) 3,623** (1,668)
-3,739** (1,686) 4,036** (1,821) 2,734* (1,639) 5,549* (2,896) 7,452*** (2,777)
1,535 (1,975) -2,260* (1,157) -2,960* (1,631) 1,492 (2,171) 2,875 (2,748)
-758.8* (413.6) -767.5* (394.3) -91.72 (276.5) 1,286** (625.5) 1,085 (676.8)
Observations 1,143 1,143 1,145 1,143 1,144 R-squared 0.109 0.204 0.097 0.144 0.080 Control mean 3698 2846 1964 4810 1952 p-value for equality of coefficients 0.0788 0.780 0.362 0.397 0.669 Cash = Input p-value for equality of coefficients 0.246 0.143 0.217 0.751 0.642 (Cash + Cash*Intensive) = (Input + Input*Intensive) β‘ Baseline value unavailable for dependent variable (questions were added to midline survey), not included in specification Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
1,145 0.044 450.9 0.00575 0.183
38
Table 14b: Ganyu Expenditure by Crop (Kwacha) (If Grown)β‘
Panel A: Extension Treatment Intensive extension treatment Observations R-squared Control mean VARIABLES Panel B: Capital Treatment Cash treatment Input treatment Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs
Maize
Groundnuts
Soy
(1)
(2)
(3)
Groundnuts + Soy (4)
1,150* (660.5)
925.3 (825.1)
452.2 (796.9)
1,135 0.102 4239
940 0.153 6093
1,562* (877.6) 1,498** (683.4)
Tobacco
Other Crops
(5)
(6)
1,129 (1,137)
5,415*** (1,620)
-106.0 (285.1)
845 0.081 4083
1,129 0.105 8080
532 0.154 3156
554 0.067 636.9
3,970*** (840.6) 4,919*** (960.6)
3,413*** (927.9) 2,382*** (906.3)
6,460*** (1,232) 6,195*** (1,188)
-1,139 (1,600) -2,228 (1,927)
-174.6 (348.5) 757.3 (502.4)
1,135 0.103 3899
940 0.185 3624
845 0.101 2400
1,129 0.140 4713
532 0.140 7159
554 0.072 575.7
0.939
0.303
0.248
0.817
0.552
0.0196
135.6 (1,176) -176.5 (1,030) 1,962 (1,251) 3,351* (1,958) -842.1 (2,026)
-2,072 (1,423) 2,864** (1,264) 2,379* (1,355) 2,416 (2,310) 5,399** (2,529)
-2,959** (1,332) 1,139 (1,730) 333.8 (1,301) 4,984** (2,198) 4,571** (2,025)
-3,929** (1,782) 3,851** (1,876) 2,557 (1,710) 5,724* (2,956) 8,025*** (2,821)
4,232 (3,138) -2,789 (2,133) -5,310* (3,112) 1,371 (4,418) 4,348 (4,921)
-1,311* (756.1) -984.0 (625.2) 241.6 (522.4) 2,003* (1,092) 1,490 (991.6)
Observations 1,135 940 845 1,129 532 R-squared 0.109 0.190 0.107 0.146 0.158 Control mean 3698 3534 2882 4964 4011 p-value for equality of coefficients 0.0768 0.714 0.503 0.401 0.478 Cash = Input p-value for equality of coefficients 0.239 0.110 0.360 0.609 0.868 (Cash + Cash*Intensive) = (Input + Input*Intensive) β‘ Baseline value unavailable for dependent variable (questions were added to midline survey), not included in specification Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
554 0.075 886.1 0.0119 0.216
39
Table 15: Assets and Expenditures (Kwacha)β β‘ Food Nonfood Expenditure Expenditure (1) (2) Panel A: Extension Treatment Intensive extension treatment Observations R-squared Control mean Panel B: Capital Treatment Cash treatment Input treatment Observations R-squared Control mean p-value for equality of coefficients Cash = Input Panel C: Treatment interactions Intensive extension treatment Cash treatment Input treatment Intensive * Cash Intensive * Inputs
Total Expenditure (3)
Agricultural Assets Value (4)
Total Assets Value (5)
5,976 (20,407)
41,691** (19,579)
54,475 (38,513)
-252.8 (3,935)
6,242 (5,414)
1,145 0.127 568476
1,145 0.134 270317
1,145 0.154 850101
1,144 0.081 19351
1,145 0.120 55464
26,161 (25,014) 54,272** (25,442)
41,392 (25,049) 31,137 (21,751)
60,098 (48,776) 70,678 (46,574)
10,048** (5,056) 5,863 (3,873)
16,773** (7,208) 10,644** (5,315)
1,145 0.130 555052
1,145 0.133 277534
1,145 0.155 850985
1,144 0.086 14195
1,145 0.125 51665
0.305
0.660
0.822
0.413
0.378
-46,949 (43,165) -5,050 (39,156) 15,208 (50,064) 68,968 (65,257) 86,857 (84,078)
-3,073 (38,219) 9,128 (32,368) -4,102 (35,642) 58,427 (56,271) 70,163 (59,080)
-7,851 (79,091) 19,916 (65,974) 18,801 (83,372) 73,001 (113,968) 103,627 (134,753)
-6,133 (5,718) 4,874 (7,601) 5,146 (7,119) 11,635 (11,780) 2,958 (11,925)
-13,346 (8,372) 172.3 (10,237) 1,727 (9,552) 34,661** (15,726) 20,874 (16,173)
1,145 1,145 1,145 1,144 Observations 0.131 0.137 0.156 0.087 R-squared 550158 258017 816329 14888 Control mean p-value for equality of coefficients 0.660 0.733 0.989 0.971 Cash = Input p-value for equality of coefficients 0.430 0.976 0.749 0.397 (Cash + Cash*Intensive) = (Input + Input*Intensive) β Observations greater than the 99% percentile winsorized β‘ Baseline value unavailable for dependent variable (questions were added to midline survey), not included in specification Notes: Ordinary Least Squares Regression, including baseline controls for NASFAM member's household size, gender, polygamy education level; Robust standard errors clustered by farmer association reported in parenthesis; *** p<0.01, ** p<0.05, * p<0.1
1,145 0.128 51657 0.882 0.383
40
