In the last blog, Gaby reported that we have great news to report about the impact of our Agriculture intervention.  From the 2012 Nuru farmer harvest, this is our latest & greatest:

Summary

In September 2012, the Nuru Kenya M&E field team collected data on the average maize yield (90kg bags per acre) for Nuru farmers in the 2012 long rain season (LR).   For farmers new to Nuru in 2012 LR, baseline data was collected from their 2011 LR yield and follow-up data from their 2012 LR yield.   Additionally, yield data was collected from returning Nuru farmers in order to report on 2012 LR yields of all Nuru farmers as a whole.

On average, farmers new to Nuru in 2012 LR experienced 123% yield increase from 2011 LR yields.  New farmers averaged 12.9 bags/acre in 2012 LR compared to 5.8 bags/acre average in 2011 LR.

Returning Nuru farmers averaged 11.5 bags/acre in 2012 LR.   The combined group of new and returning Nuru farmers overall averaged 12.5 bags/acre in 2012 LR.   Data is not available for returning farmers from 2010 and 2011.

Methodology

In 2012 LR, 2,781 farmers participated in the Nuru Agriculture program to plant maize in Kuria, Kenya.  Of these farmers, 1,883 (68%) were new to Nuru in the 2012 LR season and 898 (32%) were returning Nuru farmers.

Method of collection:

• 8 enumerators supervised by two Nuru Field Managers and accompanied by Agriculture Field Officers
• Farmers self-report the number of 90 kg bags of maize they harvested per Nuru acre
• The farmer’s field is measured by a pacing technique so the area is confirmed and validated to acreage

The survey collected the following information:

• Maize yield
• Seed variety
• Age
• Education
• Gender
• GPS points

Sample:

• The goal was to analyze the data in two overlapping groups:
  • New Nuru farmers
  • Combined new  and returning Nuru farmers
  • We targeted sampling 10% more farmers than the required statistically significant sample size to account for field challenges, such as a farmer not being home.  Details on the sampling frame can be provided upon request.
  • To collect a statistically significant sample of the 1,883 new Nuru farmers, we collected 2011 LR data from 498 new Nuru farmers and 2012 LR yields from 514 new Nuru farmers.  We used a repeated cross-sectional methodology, meaning that different farmers were sampled and surveyed on 2011 LR yields and 2012 LR yields.  Of these, 134 farmers were included in both 2011 and 2012 samples.
  • To collect a statistically significant sample of all 2,781 Nuru farmers, we collected 2012 LR yields from an additional 182 returning farmers, for a total of 696 sampled farmers (25% of Nuru farmers).
  • Sampling is geographically stratified, but does not stratify by age, gender, or education level of farmers.

• Enumerator training: Thursday, August 30^th.
• Data collection: Monday, September 3^rd – Friday, September 21, 2012.
• Data entry: The Kenyan M&E team entered data as forms come in from the field.
• Analysis:  The Kenyan team analyzed sample data with descriptive statistics, with support from U.S. staff for testing statistical significance and making inferences to population of Nuru farmers.
• Reporting: The M&E reported to the Ag team by mid-October, with an external facing report prepared by end of October.

Results

Sample data indicates that 62% of Nuru farmers in 2012 LR are male.  The average age is 40.2 years.  Approximately half of our farmers did not complete primary school.  Details by sublocation are shown below.

Was there a significant increase in 90 kg bags of maize per acre for new Nuru farmers in 2012 LR? The average baseline yield is 5.81 bags/acre in 2011 LR, with a standard deviation of 4.9 per acre.   The average yield is 12.9 bags/acre in 2012 LR, with a standard deviation of 5.5 per acre.   The sample is sufficient to infer that the sample average yield reflects the yield of the entire population of new Nuru farmers within +/- 1 bag/acre margin of error. [*]

Nyangiti and Nyamaranya have the smallest percentage yield increases and the highest baseline values.   Though no testing was done to determine the cause of this, we postulate that this may be due to their location along roads, which encourages spillover knowledge.

Histograms showing yield distribution are below.   In 2011 LR, more than half (58%) of farmers had less than 5 bags/acre.  In 2012 LR, fewer than 5% had less than 5 bags/acre.

Did returning farmers experience yields similar to new Nuru farmers?  In other words, did the knowledge gained from becoming a Nuru farmer in a prior year result in ongoing high yields in subsequent years? Average yields for returning Nuru farmers was 11.5 bags/acre in 2012 LR, which is 1.47 bags/acre lower than the average yield of new Nuru farmers. However, the geographic concentration of the returning farmers may be the reason for lower average yields than the new Nuru farmers. We do not have the 2010 LR or 2011 LR yields for the returning Nuru farmers. This limits what analysis and conclusions we can make about the ongoing program impact. We aim to measure the ongoing impact beginning with 2013 LR data.

Visualization

The maps below depict farmer yields reported for the baseline (2011 LR) and follow-up Nuru harvest (2012 LR).  Since a random sample of new farmers was taken at the baseline and follow-up, the farmers’ data points appear in different geographic areas in each map.

How do the yields of different seed varieties compare? Seven seed input varieties were available for 2012 LR.  Series 5 seeds (513, 515, 516) are intended for shorter duration maturity and are expected to have lower yields.  Nearly half of farmers used series 5 seeds. Series 6 seeds (614, 624, 625, 628) take longer to grow to maturity and are expected to have higher yields. Farmers may also choose seed variety based on their soil type. Approximately 6% of our sample farmers used a combination of seeds.  We have excluded these farmers from the averages.

2012 LR yields were highest for series 624 seed type (13.8 bags per acre), which 39% of farmers used.  Series 513 yields averages 11 bags/acre, and was the second most common type planted by farmers (30% of farmers).

Conclusions and Next Steps

New Nuru farmers significantly increased yields from the 2011 to 2012 LR seasons.  As a note, Nuru has been operating in the area for several years, and spillovers may have occurred such that the 123% increase is an understatement of the true Nuru impact.

There are several ongoing analysis projects (October-December 2012):

1. Determine final loan repayment
2. Calculate expected increases in maize profits using a newly developed profit model

Future data collection for 2013 LR should aim to address the limitations found this report,  specifically:

1. Include a baseline (2011 LR) and 2 years of follow-up yields for returning farmers (2012 LR and 2013 LR), to analyze the ongoing impact of the program
2. Include a comparison group of non-Nuru farmers in a similar geographic area to measure impact within a season.