Determinants of variety choice Determinants of variety choice among potato growers in the among potato growers in the

Bolivian highlandsBolivian highlands

Presentation at thePresentation at the2008 SANREM CRSP Annual Meeting2008 SANREM CRSP Annual Meeting

Los Baños, PhilippinesLos Baños, PhilippinesMay 26-28, 2008May 26-28, 2008• Michael Castelhano, Virginia TechMichael Castelhano, Virginia Tech

• Jeffrey Alwang, Virginia TechJeffrey Alwang, Virginia Tech• Ruben Botello, PROINPA, BoliviaRuben Botello, PROINPA, Bolivia• Nick Kuminoff, Virginia TechNick Kuminoff, Virginia Tech

BackgroundBackground• Potato producers in the Bolivian highlands adopt Potato producers in the Bolivian highlands adopt

many varieties of potatoesmany varieties of potatoes• Variety selection depends on attributes of the Variety selection depends on attributes of the

potato, including yield, tastes, disease and pest potato, including yield, tastes, disease and pest resistance, and resistance to frost and droughtresistance, and resistance to frost and drought

• Variety choice also depends on producer attributes, Variety choice also depends on producer attributes, such as education, household labor availability, and such as education, household labor availability, and access to information; access to and performance of access to information; access to and performance of markets may be critically important markets may be critically important

• When markets are not available, farmers would be When markets are not available, farmers would be more likely to select varieties for taste and storage more likely to select varieties for taste and storage characteristics; remote farmers may plant highly characteristics; remote farmers may plant highly diversified variety profiles due to individual tastes diversified variety profiles due to individual tastes and preference and exposure to different risks and preference and exposure to different risks (market and others)(market and others)

• Variety development should include considerations Variety development should include considerations of farmer demand for attributesof farmer demand for attributes

ObjectivesObjectives

• To understand the role of variety-specific risk To understand the role of variety-specific risk characteristics in potato variety selection characteristics in potato variety selection decisions made by farmers in Bolivian highlands decisions made by farmers in Bolivian highlands – Describe risk-related and other characteristics of Describe risk-related and other characteristics of

alternative varietiesalternative varieties– Analyze farmer household characteristics, risk Analyze farmer household characteristics, risk

concerns and demand for potato attributesconcerns and demand for potato attributes– Analyze the determinants of “demand” for potato Analyze the determinants of “demand” for potato

attributesattributes

• Provide information for potato researchers Provide information for potato researchers relative to farmer preferences for different relative to farmer preferences for different variety attributesvariety attributes

MethodsMethods

• Review of literature of variety studies: Review of literature of variety studies: • Dalton (West Africa-rice); DeFalco, Chavas Dalton (West Africa-rice); DeFalco, Chavas

and Smale (Ethiopia-wheat); Smale (various-and Smale (Ethiopia-wheat); Smale (various-Malawi & Ethiopia); a number of studies by Malawi & Ethiopia); a number of studies by Renkow (Honduras-maize)Renkow (Honduras-maize)

• Shively and friends (adoption of soil Shively and friends (adoption of soil conservation-Philippines)conservation-Philippines)

• Edmeades & Smale (plantain in Uganda)Edmeades & Smale (plantain in Uganda)

• Major points:Major points:• Risk attributes may be a strong deterrent to Risk attributes may be a strong deterrent to

adoption of variety and techniquesadoption of variety and techniques• Distance to market has been found to affect Distance to market has been found to affect

the risk/return tradeoffsthe risk/return tradeoffs

MethodsMethods• Interviews with scientists and extension Interviews with scientists and extension

personnel (November 2008):personnel (November 2008):• Identify varieties and their attributesIdentify varieties and their attributes• Describe current research program: main constraints Describe current research program: main constraints

being addressed through potato researchbeing addressed through potato research

• Survey of farmers (November 2008): Survey of farmers (November 2008): • Survey instrument: initial design at VT; review and Survey instrument: initial design at VT; review and

revisions in consultation with Fundacion PROINPArevisions in consultation with Fundacion PROINPA• Based on previous SANREM work (baseline surveys) in Based on previous SANREM work (baseline surveys) in

Bolivia and Ecuador, but modified to account specifically Bolivia and Ecuador, but modified to account specifically for variety characteristics and exposure to riskfor variety characteristics and exposure to risk

• Implementation: three villages in Cochabamba Implementation: three villages in Cochabamba Department; 145 total interviewsDepartment; 145 total interviews

• Research (data collection) was designed to Research (data collection) was designed to increase variation in climatic and market increase variation in climatic and market conditionsconditions

Household Survey Household Survey LocationsLocations

• Toralapa (20): Toralapa (20): – High altitude, dry, frost proneHigh altitude, dry, frost prone– Good road access (major paved road nearby goes right to Good road access (major paved road nearby goes right to

Cochabamba city). Cochabamba city). – Problem with Toralapa is that a single variety predominates due to Problem with Toralapa is that a single variety predominates due to

market integration and preferences for uniform varietal market integration and preferences for uniform varietal characteristicscharacteristics

• Colomi (68): Colomi (68): – More moisture (typically) than ToralapaMore moisture (typically) than Toralapa– Much more diversity in potato plantingsMuch more diversity in potato plantings– Less frost (more blight)Less frost (more blight)– Good paved road access, many cobbled roads in smaller villages; Good paved road access, many cobbled roads in smaller villages;

major road to Cochabamba is nearbymajor road to Cochabamba is nearby• Morochata (57): Morochata (57):

– More moisture that Torolapa; warmer by reputation (no weather More moisture that Torolapa; warmer by reputation (no weather station)station)

– Home to PROINPA-supported breeding programHome to PROINPA-supported breeding program– Relatively poor road conditions- necessary to travel on steeply Relatively poor road conditions- necessary to travel on steeply

sloped areas to get around; washouts frequent on many roads; to sloped areas to get around; washouts frequent on many roads; to reach Cochabamba its necessary to cross Andean ridge; fairly reach Cochabamba its necessary to cross Andean ridge; fairly remoteremote

ColomiColomi

Road to Morochata

Potato Market in Colomi

Methods: Theory and Methods: Theory and Econometric ModelEconometric Model

• Basic modeling focus: households make decision Basic modeling focus: households make decision about how much to plant of each of variety about how much to plant of each of variety encountered (choice set). Implicit assumption is encountered (choice set). Implicit assumption is that household could plausibly plant each variety that household could plausibly plant each variety (technology set is available to all)(technology set is available to all)

• Expected utility maximization subject to variety Expected utility maximization subject to variety availability constraints (risk averse, safety first availability constraints (risk averse, safety first framework) and stochastic production and pricesframework) and stochastic production and prices

• Decision is modeled as a function of variety Decision is modeled as a function of variety characteristics and household characteristicscharacteristics and household characteristics

• We observe many zero values for the dependent We observe many zero values for the dependent variablevariable

• Many farmers plant zero acreage of many varietiesMany farmers plant zero acreage of many varieties• No farmer plants every varietyNo farmer plants every variety

Econometric ModelEconometric Model• Tobit model used to account for truncation in the dependent Tobit model used to account for truncation in the dependent

variable:variable:

• E(V|X; Y>=0) = XB +E(UE(V|X; Y>=0) = XB +E(Uvv|U|Uyy>X>XyyBByy))• where:where:• V= area planted to a single variety (0<=V<=4)V= area planted to a single variety (0<=V<=4)• Y= Unobserved latent variable representing demand for Y= Unobserved latent variable representing demand for

varietyvariety• X= variety and household characteristicsX= variety and household characteristics• B= vector of parameters B= vector of parameters • Each variety characteristic will have one coefficient Each variety characteristic will have one coefficient • Each household characteristic will have eleven coefficients (one Each household characteristic will have eleven coefficients (one

for each variety)for each variety)• Desired results of the project: Desired results of the project:

– How do variety characteristics impact potato planting decisions? How do variety characteristics impact potato planting decisions? – Coefficients of variety characteristics will quantify farmers revealed Coefficients of variety characteristics will quantify farmers revealed

preferences for different characteristicspreferences for different characteristics

Results: Potato VarietiesResults: Potato Varieties• During interviews with PROINPA researchers and During interviews with PROINPA researchers and

field visits, we identified a total of 11 varieties (8 field visits, we identified a total of 11 varieties (8 or more observations of variety adoption)or more observations of variety adoption)

• Blight tolerance is the most common ‘improved’ Blight tolerance is the most common ‘improved’ characteristic; others include shorter growing characteristic; others include shorter growing season, differential water toleranceseason, differential water tolerance

• Very little insect resistance (possibly related to Very little insect resistance (possibly related to ease of obtaining pesticide in region) in any of the ease of obtaining pesticide in region) in any of the varietiesvarieties

• Varieties studied cover about 90% of potato Varieties studied cover about 90% of potato planting area in the three study communities planting area in the three study communities (Colomi 86%; Morochata 93%; Toralapa 98%)(Colomi 86%; Morochata 93%; Toralapa 98%)

• All 11 varieties are present in Colomi; 5 in All 11 varieties are present in Colomi; 5 in Toralapa; 8 in MorochataToralapa; 8 in Morochata

Results: Potato VarietiesResults: Potato Varieties• 53 names recorded for varieties in 145 surveys53 names recorded for varieties in 145 surveys• Waych’a by far the most common (124 obs.)Waych’a by far the most common (124 obs.)

– Average area planted-0.4 Ha.Average area planted-0.4 Ha.– Moderately tolerant/resistant to nematodes, frost, drought and Moderately tolerant/resistant to nematodes, frost, drought and

hailhail– Performs well in poor soil; fairly high yieldingPerforms well in poor soil; fairly high yielding– Heavily marketed in cities (restaurants, intermediaries, etc.)Heavily marketed in cities (restaurants, intermediaries, etc.)– Often used as a benchmark for comparison (most familiar) of new Often used as a benchmark for comparison (most familiar) of new

varietiesvarieties• HH (also called Runa Toralapa) (52 obs. on average 0.24 Ha.)HH (also called Runa Toralapa) (52 obs. on average 0.24 Ha.)

– Blight resistant (most common blight resistant cultivar in survey)Blight resistant (most common blight resistant cultivar in survey)– High yields (comparable to Waych’a)High yields (comparable to Waych’a)– Price was generally lower than Waych’a in 06-07Price was generally lower than Waych’a in 06-07

• -Pinta Boca: occurs often (50 obs.), but in small plantings (on -Pinta Boca: occurs often (50 obs.), but in small plantings (on average .09 Ha.)average .09 Ha.)– Lower yielding, less resistance to pests, disease, climateLower yielding, less resistance to pests, disease, climate– Traditional to serve to guests, friends/family and on Easter Traditional to serve to guests, friends/family and on Easter

HolidayHoliday– Larger portion of crop planned for home consumption (average Larger portion of crop planned for home consumption (average

40% of crop is home-consumed) than other common varieties 40% of crop is home-consumed) than other common varieties

Results: Variety Results: Variety AttributesAttributes

• Most plantings and varieties are planned for multiple Most plantings and varieties are planned for multiple end uses (consumption and sale)end uses (consumption and sale)

• Farmers generally do not differentiate between varieties Farmers generally do not differentiate between varieties in terms of their input requirements (including agro-in terms of their input requirements (including agro-chemicals)chemicals)– Example 1: one question asked about fertilizer use, and MANY Example 1: one question asked about fertilizer use, and MANY

farmers were only able to provide information on total quantity farmers were only able to provide information on total quantity of fertilizer purchased and stated that it was spread evenly of fertilizer purchased and stated that it was spread evenly throughout the farmthroughout the farm

– Example 2: most farmers stated that they sprayed all the Example 2: most farmers stated that they sprayed all the varieties evenly for pests/disease; even resistant varieties were varieties evenly for pests/disease; even resistant varieties were reportedly sprayed for late blight (for those farmers who reportedly sprayed for late blight (for those farmers who sprayed for late blight)sprayed for late blight)

• Risk characteristics are categorical: highly susceptible, Risk characteristics are categorical: highly susceptible, moderately susceptible, moderately tolerant, highly moderately susceptible, moderately tolerant, highly tolerant. (moderately susceptible is base category)tolerant. (moderately susceptible is base category)– Information on risk tolerance comes from PROINPA personnel Information on risk tolerance comes from PROINPA personnel

and publications (particularly Julio Gabriel)and publications (particularly Julio Gabriel)

ColomiColomiMorochaMorochatata TorolapaTorolapa

MeaMeann StdStd

MeaMeann StdStd

MeaMeann StdStd

Household CharacteristicsHousehold Characteristics

Household SizeHousehold Size 5.85.8 2.42.4 5.45.4 2.52.5 5.45.4 2.22.2

Head's AgeHead's Age 42.942.9 14.714.7 43.543.5 12.912.9 39.439.4 11.611.6

Head Primary Head Primary EducationEducation 0.80.8 0.40.4 0.90.9 0.40.4 0.70.7 0.50.5

Head Secondary Head Secondary EducationEducation 0.060.06 0.240.24 0.050.05 0.230.23 0.050.05 0.20.2

Members Members Participating in Participating in Potato ActivitiesPotato Activities 3.23.2 1.71.7 3.13.1 1.81.8 33 1.51.5

Members Members Participating in Participating in Non-farm Income Non-farm Income GenerationGeneration 0.50.5 1.11.1 0.20.2 0.70.7 0.10.1 0.30.3

Results: Summary Statistics

Results: Summary Results: Summary StatisticsStatistics

ColomiColomi MorochataMorochata TorolapaTorolapaMeanMean StdStd MeanMean StdStd MeanMean StdStd

Area Planted to Area Planted to PotatoPotato

0.650.65 0.40.4 0.520.52 0.640.64 1.171.17 1.181.18

Pct Certified SeedPct Certified Seed 0.310.31 0.470.47 0.320.32 0.470.47 0.30.3 0.470.47

Number Varieties Number Varieties PlantedPlanted

5.225.22 2.272.27 2.072.07 1.151.15 1.751.75 1.021.02

Pct Receiving Pct Receiving CreditCredit

0.220.22 0.420.42 0.050.05 0.230.23 0.20.2 0.410.41

Pct Farmer Pct Farmer OrganizationsOrganizations

0.560.56 0.50.5 0.950.95 0.230.23 0.90.9 0.30.3

Pct Visiting Pct Visiting Market WeeklyMarket Weekly

0.790.79 0.40.4 0.230.23 0.420.42 0.850.85 0.220.22

Pct Hiring LaborPct Hiring Labor 0.540.54 0.50.5 0.320.32 0.470.47 0.650.65 0.370.37

Results: Summary Results: Summary StatisticsStatistics

• Household size and age very similar across regionsHousehold size and age very similar across regions• Landholdings differ, but total numbers of household members Landholdings differ, but total numbers of household members

participating in potato production is similarparticipating in potato production is similar

• Most participation in off-farm activities is found in Most participation in off-farm activities is found in Colomi (access to transportation and nearby labor Colomi (access to transportation and nearby labor markets)markets)

• Large difference in number of varieties plantedLarge difference in number of varieties planted• Differential access to credit, possibly related to access Differential access to credit, possibly related to access

to markets. Eg.: very little credit availability in to markets. Eg.: very little credit availability in MorochataMorochata

• Strength of farmer organizations varies: most isolated Strength of farmer organizations varies: most isolated farmers are more likely to participate in organizationsfarmers are more likely to participate in organizations

• Farmers outside of Toralapa visit the market less Farmers outside of Toralapa visit the market less frequentlyfrequently

• More agricultural market and labor market More agricultural market and labor market participation in Toralapa participation in Toralapa

Preliminary Results: Model Preliminary Results: Model Estimation Estimation

• Not yet ready for prime timeNot yet ready for prime time• Yield attributes do not appear to be as important Yield attributes do not appear to be as important

determinants of variety selection as are disease resistance determinants of variety selection as are disease resistance and taste: risk management is an important consideration and taste: risk management is an important consideration when choosing potato varietieswhen choosing potato varieties

• Yield is more important in Toralapa and least important in ColomiYield is more important in Toralapa and least important in Colomi• Quantitative tradeoffs between different attributes are measured: Quantitative tradeoffs between different attributes are measured:

substitutability is increased in towns that are closest to markets; substitutability is increased in towns that are closest to markets; distance also affect substitutability within towns (more distance also affect substitutability within towns (more remote=>less substitutability)remote=>less substitutability)

• Substitutability between attributes varies by location: less Substitutability between attributes varies by location: less substitution in remote areassubstitution in remote areas

• No statistically significant relationship between landholding No statistically significant relationship between landholding size and degree of diversity in variety adoption. size and degree of diversity in variety adoption.

• Household labor availability is significant; tradeoffs between Household labor availability is significant; tradeoffs between labor and timing of harvest labor and timing of harvest

• Farmer education (weakly) associated with adoption of Farmer education (weakly) associated with adoption of fewer varietiesfewer varieties

• Access to markets associated with fewer varieties being Access to markets associated with fewer varieties being adopted: the most remote households are the most adopted: the most remote households are the most diversifieddiversified

Summary and Summary and ConclusionsConclusions

• We have quantitative information about We have quantitative information about tradeoffs Bolivian highland potato farmers tradeoffs Bolivian highland potato farmers make between attributes like yield, taste make between attributes like yield, taste and riskand risk

• These tradeoffs depend on farmer location, These tradeoffs depend on farmer location, land farmed and education, with location land farmed and education, with location being the most important determinantbeing the most important determinant

• Information can be used to guide variety Information can be used to guide variety development; there is a clear need to have development; there is a clear need to have a number of varieties with different a number of varieties with different attributesattributes

• Information from this study can be Information from this study can be incorporated into an outreach program incorporated into an outreach program