Strengthening the Dairy Value Chain Progress_May 2012

18
Analysis of SDVC Data March 2009 - April 2012 Presented May 31, 2012

description

This presentation uses infographics to illustrate the progress made by the CARE Strengthening the Dairy Value Chain Program. The initiative aims to double the dairy-related incomes of 35,000 smallholder and landless producers in rural Bangladesh. It is supported by the Bill and Melinda Gates Foundation.

Transcript of Strengthening the Dairy Value Chain Progress_May 2012

Page 1: Strengthening the Dairy Value Chain Progress_May 2012

Analysis of SDVC DataMarch 2009 - April 2012

Presented May 31, 2012

Page 2: Strengthening the Dairy Value Chain Progress_May 2012

ARTIFICIAL INSEMINATIONUSAGE RATES OVER TIME

mar-09 jun-09 oct-09 mar-10 jul-10 jan-11 jul-11 apr-12

8.6% 7.1% 7.4% 13.8% 11% 10.3% 8.8%23.3%

PERCENT OF DEWORMING USAGE RATES OVER TIME

mar-09 jun-09 oct-09 mar-10 jul-10 jan-11 jul-11 apr-12

58% 51% 42% 41% 34% 31% 47%79%

wealthy

at least3%

DEWORMING

Use of Artificial Insemination increases for all households. The average household can expect to see at least a 3% increase in household income from milk if they use artificial insemination.

Vaccination of cattle is most beneficial for more wealthy households. In poorer households, the use of vaccination does not seem to increase income. However, in wealthier households, income can increase by about 3% if the cattle are vaccinated.

Deworming of cattle has a very positive effect on household income for all households. The average household can expect an increase in income of between 5 and 10% if they deworm their cattle.

Whether or not a household uses Artificial Insemination is strongly predicted by the availability of the service, the economic status of the household and the skills of the household’s Livestock Health Worker.

Households that have lower then average incomes use CARE Livestock Health Workers or Other Livestock Health Workers or a Government Vet. Households that have higher than average incomes tend to use their own family members to provide vaccinations.

Interestingly, a household with a low knowledge score and a high level of confidence in their LHW and a household with a high knowledge score but a low level of confidence in their LHW are both about equally likely to deworm their cattle. Both are about 15% less likely to deworm their cattle than the high knowledge and high confidence household.

The strongest predictor of whether or not a household chooses to deworm their cattle is their overall knowledge score and the level of confidence they feel in their Livestock Health Worker. A household with a high knowledge score and a high level of confidence in their Livestock Health Worker is 30% more likely to deworm their cattle than a household with a low knowledge score and a low level of confidence in their Livestock Health Worker.

However, for all households – the household income is related to vaccination provider choices.

income increase

at least3%

5% 10%-

VACCINATION

ARTIFICIAL INSEMINATION

Page 3: Strengthening the Dairy Value Chain Progress_May 2012

GROUP LEADER BY GROUP COMPOSITION

Overall, Households within Learning Groups with Female Leaders have incomes that are 3-6% higher.

Learning Groups with Female Leaders do relatively better as the Phase progresses

FARM LEADER GENDERIMPROVES INCOME FROM MILK

phase 1

7%

phase 2

5%

phase 3

2%

phase 4

0%

female leader

percent improvedperformance over male leaders

In Phase 1, the groups with female leaders do 7% betterIn Phase 2, the groups with female leaders do 5% betterIn Phase 3, the groups with female leaders do 2% better than groups with male leadersIn Phase 4, there is no difference in income between female led groups and male led groups.

GROUP COMPOSITIONLEADER GENDER

12%

2%

5%

or

farmleader gender

percentincrease

groupcomposition

Learning Groups with a high percentage of women producers with a female group leader perform the best overall.

Learning Groups with a high percentage of men producers do moderately well regardless of group leader gender.

Learning Groups with a high percentage of women producers and a male group leader perform the least well.

higher income

3% 6%-

GROUP LEADER BY GENDER

GROUP LEADER BY PHASE

Page 4: Strengthening the Dairy Value Chain Progress_May 2012

wealthier district poorer district

5% 7%

better inearning income

-

po

or

po

or

Market Linkage by household economic status and presence of a group selected

MARKETINFORMAL

poor wealthy

MARKETS

MV RD PRAN BRAC

wealthy wealthy

MARKETS

GRAMEENDANOON AKIJ

At MV & RD, the households making the most money are wealthy and do not have their own collectors.On the informal markets, the

poorer households with their own collectors do the best. At PRAN and BRAC, the households

doing the best are wealthy with their own milk collector.

At Akij, it seems to be irrelevant if you have your own collector or not.

At Grameen Dannon, most households do the same – with the very significant exception of the very poor households without their own collector. These households do very poorly at this market.

milk collectorWHAT PERCENT DO HOUSEHOLDS FROMRICH GROUPS DO BETTERTHAN HOUSEHOLDS FROM POOR GROUPS

MV6.19%

RD5.48%

PRAN4.78%

BRAC4.21%

AKIJ3.49%

GRAMEENDANNON

2.05%

INFORMALSECTOR

1.47%

OTHER0.88%

wealthy

GROUP ECONOMIC STATUS LARGER ECONOMIC CONTEXTThe initial economic status and the larger economic environment of a group has a heavy influence on their milk income.

In general, if a group is poor initially, their progress is better if they operate within a wealthy District.

A poor Learning Groups that operate within one of the wealthier Districts do 5-7% better in earning income than equivalent poor Learning Groups that operate within one of the poorer Districts.

And in general, a group that is more wealthy to begin with a operates within a wealthy District does the best overall – a full 10% - 12% better than even an equivalently rich group that operated in a poor District.

MARKET LINKAGE How and where a household sells their milk significantly affects their income.

MARKETINFORMAL

poor wealthy

MARKETS

MV RD PRAN BRAC

5-8%

poor wealthy

MARKETS

GRAMEENDANOON AKIJ

poor wealthy

The poorest households do the same as the wealthier households when selling their milk in an informal market.

However, the rich households do much better than the poor households when they sell their milk to the MV, RD, PRAN, and BRAC markets. When all else is equal, a rich household makes between 5-8% more money than a poor household when selling milk in these markets.

There is a slight advantage to the wealthier households is the Grameen Danoon and Akij markets, but it is much less consistent.

Market Linkage by household economic status

AN

D

10% 12%-

better in earning money

Page 5: Strengthening the Dairy Value Chain Progress_May 2012

PERMISSION TO ATTEND FAR AWAY MEETINGS

mar-09 jun-09 oct-09 mar-10 jul-10 jan-11 jul-11 apr-12

0.58 0.68 0.78 0.88 0.98 1.08 1.17 1.27

2.56 2.48 2.38 2.27 2.17 2.07 1.96 1.86

0.37 0.47 0.57 0.67 0.77 0.86 0.96 1.06

2.37 2.27 2.16 2.06 1.96 1.85 1.75 1.65

high incomelearning group

low incomelearning group

HOUSEHOLDS IN WHICH WOMEN OWN CATTLE

CATTLE SELLING DECISIONS

PERMISSIONS TO ATTEND MEETINGS

Households where women own cattle do about 10% better in earning money than do households where women do not own cattle.

However, this relationship is complex and is changing over time.

Households in which women own cattle and women make the cattle selling decisions are more likely to sell cattle and are more likely to have higher incomes overall.

Whether or not women producers need permission to attend meetings, both within and outside of their village is influenced by whether or not they own cattle, the economic status of their group and time.

Women who own cattle are less likely to need permission to attend meetings far away.

Women in high income learning groups are slightly more likely to need permission to attend meetings.

However, the rates of women needing permission to attend meetings is dropping amongst women who don’t own cattle.

GENDER: GROUP AND HOUSEHOLD

mar-09 jun-09 oct-09 mar-10 jul-10 jan-11 jul-11 apr-12

-0.22 -0.13 -0.04 0.04 0.13 0.22 0.31 0.4

-0.21 -0.11 -0.028 0.06 0.14 0.23 0.33 0.41

-0.45 -0.36 -0.27 -0.18 -0.09 -0.004 0.08 0.17

-0.34 -0.25 -0.17 -0.08 0.01 0.1 0.19 0.28

group has fewhouseholds wherewomen own cattle

group has manyhouseholds wherewomen own cattle

PERMISSION TO ATTEND MEETINGS

mar-09 jun-09 oct-09 mar-10 jul-10 jan-11 jul-11 apr-12

-1.45 -1.25 -1.1 -0.97 -0.8 -0.6 -0.5 -0.45

-1.6 -1.35 -1 -0.75 -0.48 -0.25 0.05 0.4

-1.2 -1 -0.9 -0.8 -0.6 -0.47 -0.3 -0.07

-1.43 -1.05 -0.82 -0.5 -0.3 -0.05 0.4 0.55

high incomelearning group

low incomelearning group

Women who own cattle need less permission to attend meetings.

10%

better in earning money

Page 6: Strengthening the Dairy Value Chain Progress_May 2012

LIVESTOCK HEALTH WORKERS

MILK COLLECTORS

TRAINING BY SEX IS IMPORTANT

SEX BY RECEIVE LOAN IS IMPORTANT

Livestock Health Workers income is influenced by:

• the gender of the worker• the training the worker received• whether or not the worker received a loan.

LIVESTOCK HEALTH WORKERS INCOME

BASIC

ADVANCE

BOTH

LEVEL OF TRAININGFEMALE LHW IMPROVEMENTover MALE LHW with

the same training

33%

22%

17%

LOANFEMALE LHW

35%

24%

IMPROVEMENTover MALE LHW withthe same loan status

Female LHW with loans have a 35% higher increase in income than men.

Female LHW without loans have a 24% higher increase than men.

Milk collectors income is most influenced by the sex of the collector in combination with the market linkage of the collector

Grameen DannonWomen milk collectors who sell here can expect a 30% higher income increase over time than men collectors selling here.

MILK COLLECTORS INCOME

BRAC

AKIJ

INFORMAL

WOMENMILK COLLECTOR

100%

80%

-10%

MARKET INCREASE OVER MEN MC

GRAMEENDANNON 30%

NAMV PRANRD

RD & PRANDo not have enough women selling milk here to discuss.

Informal Women selling here had an income increase that was 10% lower than men (3%)MVVery few women collectors sell milk here. The few that do achieve a much higher income increase than the male milk collectors.

BRAC Women milk collectors who sell here can expect a 100% higher income increase over time than men collectors selling here.

AkijWomen milk collectors who sell here can expect a 80% higher income increase over time than men collectors selling here.

Female LHW with basic training achieve a 33% higher income increase than men.

Female LHW with advanced training achieve a 22% higher income increase than men.

Female LHW with both basic and advanced training achieve a 17% higher income increase than men.

Page 7: Strengthening the Dairy Value Chain Progress_May 2012

FEED SOURCE COMPARED

CA

RB

OH

YD

RA

TES

PRO

TEIN

SV

ITA

MIN

S &

MIN

ERA

LSO

THER

increaseper litre

% of householdsusing this feed

feed source average costper kg in taka

4.4%

45.4%

0.30.19 0.0294.9%

58.6%

BDT %

0.50.05 0.30

BDT %

00.04 0.30

BDT %

0.30.06 0.60

BDT %

0.8 0.605.5%

21.4%

0.04

BDT %

0.8 0.300.03

BDT %

0.126.1% 0.200.17

BDT %

0.13.6% 0.100.08

BDT %

% increase inmonth milk income

RICE BRAN

WHEAT BRAN

PULSE HUSK

BROKEN RICE

OIL CAKE

M. OIL CAKE

VITAMINS& MINERALS

READY FEED

per monthly 10 kg increase

For the best nutrition, cattle need a combination of Carbohydrates, Proteins and Vitamins and Minerals.

CATTLENUTRITION

Carbo

hydra

tes

Miner

als

ProteinsVitamisn

The most cost effective and beneficial forms of carbohydrates seems to be Wheat Bran and Broken Rice.

The most cost effective and beneficial forms of proteins are various forms of Oil Cakes.

Vitamins and minerals are very important for the health and milk production of cattle.

Over time, our farmers have increased their regular use of vitamins and minerals by about 20% overall.

Over time, our farmers have increased their Wheat Bran use from 50% to 75% of all households. And our farmers have held their rates of Broken Rice steady over time. About half of all households use broken rice.

Over time, our farmers have increased their use of various types of oil cakes by about 10% overall.

FEED SOURCEPRICE OVER TIME

RICE BRAN

WHEAT BRAN

PULSE HUSK

BROKEN RICE

OIL CAKE

M. OIL CAKE

VITAMINS& MINERALS

READY FEED

0.25

BDT

0.06

BDT

0.05

BDT

0.09

BDT

0.06

BDT

0.00

BDT

0.23

BDT

0.25

BDT

jun-09

0.20

BDT

0.06

BDT

0.05

BDT

0.07

BDT

0.04

BDT

0.04

BDT

0.03

BDT

0.20

BDT

oct-09

0.19

BDT

0.06

BDT

0.05

BDT

0.05

BDT

0.05

BDT

0.04

BDT

0.02

BDT

0.19

BDT

mar-10

0.21

BDT

0.05

BDT

0.04

BDT

0.04

BDT

0.03

BDT

0.04

BDT

0.57

BDT

0.21

BDT

jul-10

0.16

BDT

0.04

BDT

0.04

BDT

0.04

BDT

0.04

BDT

0.04

BDT

0.03

BDT

0.16

BDT

jan-11

0.16

BDT

0.04

BDT

0.04

BDT

0.05

BDT

0.04

BDT

0.04

BDT

0.31

BDT

0.16

BDT

jul-11

0.15

BDT

0.04

BDT

0.04

BDT

0.05

BDT

0.04

BDT

0.04

BDT

0.02

BDT

0.15

BDT

apr-12

0.19

BDT

0.05

BDT

0.04

BDT

0.06

0.04

BDT

BDT

0.03

BDT

0.19

BDT

0.19

BDT

overallaverage

FEED SOURCEPROPORTIONS

RICE BRAN

WHEAT BRAN

PULSE HUSK

BROKEN RICE

OIL CAKE

M. OIL CAKE

READY FEED

jun-09 oct-09 mar-10 jul-10 jan-11 jul-11 apr-12

18% 57% 31% 10% 8% 9% 7%

4% 10% 9% 3% 2% 2% 3%

4% 7% 4% 1% 0.7% 0.8% 0.9%

0% 0.7% 0.3% 0% 0.1% 0.2% 0.3%

0% 2% 1% 0.3% 0.3% 0.3% 0.4%

0.8% 0.2% 62% 2% 72% 0.2% 0.1%

2.3% 3.6% 3.2% 1.1% 2.3% 1.2% 1.1%

CA

RB

OH

YD

RA

TES

PRO

TEIN

S

VITAMINS& MINERALS

4% 0.6% 0.5% 0.9% 0.4% 0.1% 0.1%

VIT

AM

INS

&M

INER

ALS

OTH

ER

WHEAT BRAN

OIL CAKES

VITAMINS MINERALS

75%

10%

20%

Page 8: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Household Compostion - Entire Dataset

Respondents' GenderCount Percent

1 Women 7290 80.15%2 Men 1805 19.85%Total 9095 100.00%

Count of In-milk Local Breed Cows

Count Percent0 4192 46.09%1 4138 45.50%2 690 7.59%3 75 0.82%Total 9095 100.00%

Count of In-milk Cross Breed Cows

Count Percent0 8093 88.98%1 800 8.80%2 173 1.90%3 29 0.32%Total 9095 100.00%

Overview of Entire Dataset:Household Overview

Count of Total In-Milk Cows in Household

Count Percent0 3302 36.31%1 4734 52.05%2 935 10.28%3 124 1.36%Total 9095 100.00%

Count Percent1 Yes 1202 13.22%2 No 6248 68.70%Total 7450 100.00%

Count of Households that have Cattle Owned by Women

Page 9: Strengthening the Dairy Value Chain Progress_May 2012

Count of Households who Dewormed Cattle

Count Percent1 Yes 3589 39.46%2 No 5382 59.18%Total 8971 100.00%

Overview of Entire Dataset:Vet Practices

Count of Households Who Got AI for Cattle

Count Percent1 Yes 943 13.81%2 No 5884 86.19%Total 6827 100.00%

Type of Treatment Provider, in general

Count Percent

1 CARE LHW6093 66.99%

2 Other LHW1305 14.35%

3 Govt Vet344 3.78%

4 Other people of DLS

107 1.18%

5 Milk Processor Vet

39 0.43%

6 Medicine/Feed Compant Vet

10 0.11%

7 Kabiraj30 0.33%

8 Own Family Member

7 0.08%

9 Others63 0.69%

Total7998 100.00%

Page 10: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Entire Dataset:Financial Practices

Count of Households that Got Loans

Count Percent1 Yes 126 1.39%2 No 8969 98.61%Total 9095 100.00%

Count Percent1 Relatives 7 5.56%2 MFI 87 69.05%3 Commercial Bank 7 5.56%4 Merchent 2 1.60%5 Govt Institution 2 1.59%6 Milk Processing Company2 1.59%7 Milk Trading Association9 7.14%8 Other Association 7 5.56%9 Others 3 2.38%Total 126 100.00%

Source of Loans for Households that Got Them

Count Percent1 Yes 1165 55.19%2 No 946 44.81%Total 2111 100.00%

Count of Households that Engaged in Group Savings

Page 11: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Entire Dataset:Gender Roles

Gender of Person Engaged in Cow Rearing

Count Percent1 Women 5037 55.38%2 Men 892 9.81%3 Both 3166 34.81%Total 9095 100.00%

Gender of Person Engaged with Milk Selling

Count Percent1 Women 2279 25.06%2 Men 2765 30.40%3 Both 871 9.58%Total 5915 100.00%

Count Percent1 Yes 3670 40.35%2 No 3898 42.86%Total 7568 100.00%

Count of Women Who Need Permission to Attend Group Meetings

Count Percent1 Yes 6534 86.34%2 No 1034 13.66%Total 7568 100.00%

Count of Women Who Need Permission to Attend Meetings at a Distance

Gender of Person Engaged in Feed Purchase

Count Percent1 Women 653 7.18%2 Men 6529 71.79%3 Both 1051 11.56%

Page 12: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Entire Dataset:Cattle Productivity

Local Breed Cow Productivity (Daily Litres) Over Time and According to Phase

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12 Phase 1 Phase 2 Phase 3 Phase 4

Cross Breed Cow Productivity (Daily Litres) Over Time and According to Phase

Page 13: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Entire Dataset:Knowledge & Practical Education

0

1

2

3

4

5

6

7

8

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

Total Knowledge Score Over Time and According to Phase

Total Practical Score Over Time and According to Phase

0

2

4

6

8

10

12

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

Page 14: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Entire Dataset:Feed Costs & Milk Income

0.00

200.00

400.00

600.00

800.00

1000.00

1200.00

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

0.00

200.00

400.00

600.00

800.00

1000.00

1200.00

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

0.00

0.50

1.00

1.50

2.00

2.50

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

Monthly Feed Costs per Cow (taka) Over Time and According to Phase

Monthly Income per Cow (taka) Over Time and According to Phase

Ratio of Milk Income to Feed Costs Over Time and According to Phase

Page 15: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Entire Dataset:Cattle Productivity

0%

10%

20%

30%

40%

50%

60%

70%

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

Percent of Women Who Need Permission to Attend Group Mee Over Time and According to Phase

Percent of Women Who Need Permission to Attend Group Mee Over Time and According to Phase

0%

20%

40%

60%

80%

100%

Mar-09 Sep-09 Mar-10 Sep-10 Mar-11 Sep-11 Mar-12

Phase 1 Phase 2 Phase 3 Phase 4

Page 16: Strengthening the Dairy Value Chain Progress_May 2012

Overview of Entire Dataset:Where Milk is Sold

Phase One Groups Over Time

29%  

0%  5%  

5%  12%  

0%  2%  

47%  

1   2   3   4   5   6   7   8  

25%  

0%  3%  

14%  

1%  

15%  7%  

35%  

1   2   3   4   5   6   7   8  

24%  

0%  3%  

13%  

1%  13%  

3%  

43%  

1   2   3   4   5   6   7   8  

20%  

0%  4%  

11%  

0%  10%  

2%  

53%  

1   2   3   4   5   6   7   8  

25%  

0%  

3%  14%  

2%  

16%  5%  

35%  

1   2   3   4   5   6   7   8  

26%  

0%  4%  

10%  

1%  10%  7%  

42%  

1   2   3   4   5   6   7   8  

27%  

0%  3%  9%  

1%  15%  6%  

39%  

1   2   3   4   5   6   7   8  

29%  

0%  6%  

10%  

1%  

11%  

10%  

33%  

1   2   3   4   5   6   7   8  

24%  

0%  6%  

12%  

1%  

17%  

9%  

31%  

1   2   3   4   5   6   7   8  

27%  

0%  

2%  

6%  

2%  15%  21%  

27%  

1   2   3   4   5   6   7   8  

Phase Four Groups Over Time

October 2009

April 2012August 2011

April 2012

June 2009

August 2011

March 2009

August 2010 January 2011

March 2010

1= Percent milk consumed by household 2=Percent milk spoiled 3= Percent milk sold to neighbors 4=Percent milk sold on open markets 5= Percent milk sold to tea shops 6= Percent milk sold to milk collector 7= Percent milk sold to sales point 8= Other

Page 17: Strengthening the Dairy Value Chain Progress_May 2012

Summary of Statistical Models Used

Data Collectiong & Variables

SDVC has collected and analyzed over 350 variables encompassing 863 groups, 45 field facilitators and 2 regions spanning 4 years.

The data has been collected at the household level, the static group level, and the dynamic group level (which changes over time) over eight waves from 2009-2012.

Given this, advanced statistical methods are required to produce accurate results.

Household  ID   Count  of  all  cows  

Milk  product.  

Group  ID   Phase   Region   PPT  Round  1  

PPT  Round  2  

PPT  Round  3  

737   1   .25   10111   1   1   35   47   75  

1601   1   1.6   10111   1   1   35   47   75  

2492   3   4.25   20245   2   1   NA   57   90  

4962   2   2.5   30865   3   2   NA   NA   82  

Household  Level  Data   Sta=c  Group  Level  Data   Dynamic  Group  Level  Data    

Page 18: Strengthening the Dairy Value Chain Progress_May 2012

Summary of Statistical Models Used

Data Analysis Methods Generalized Linear Mixed-Effects models

To accurately analyze the evidence on how SDVC interventions are working, we built statistical models that looked at all the levels simul-taneously and controlled for the context in which the household exists (in this case, we included various group and program level vari-ables).

Most of the trends and effects presented in the findings have controlled for many confounding and mediating variables in addition to the primary variables of interest, including: Geographic variables (upazila, region)Group effect (group number, group contextual variables)Household differences (family size, number of cows, breed of cows).

We used the R software for statistical computing. R is a free software environment that is widely used by statisticians. R is powerful and uses the most up to date algorithms available due to its open source nature. The R packages contain functions for working with the com-plex type of data that is involved in this project. These functions are not established in most other statistical packages.

We primarily used R to build mixed-effects regression models with both fixed and random effects. This we essential for accuracy as this data has both nested effects (such as households within groups within regions) and crossed effects (such as groups within phases within PPT rounds).

Due to the complex nature of the data, all of the models in this analysis were done using generalized linear mixed-effect models. Each of the models in this presentation control for the size of the household cattle herd, the phase of the learning group of the household, the ef-fects of time on the outcomes, and the contextual difference between the household’s results and the group’s results (ie. the within-house-hold trend and the between-household trend.)

Generalized linear mixed-effect models (GLMMs) are a class of models designed for the analysis of clustered and longitudinal data with non-normal dependent variables. In our models we have used a binomial link funtion and a penalized quasi-likelihood methods. All our models include both a random intercept and a random slope. Each model includes fixed effects such as size of herd, time of collection and phase of group. Each model also includes a series of random effects including the learning group and time. This method properly controls for the fact that each group is meaured repeatedly as well as the fact that the data is clustered in several dimensions (ie. phase and geography)

The acceptable significance level for all of our models is alpha = 0.05.