gender and dynamics of technology adoption: …...Gender and the dynamics of technology adoption: empirical evidence from a household-level panel data Khushbu Mishraa, 1, Abdoul G.

Gender and the dynamics of technology adoption: empirical evidence from a

household-level panel data

Khushbu Mishraa,1, Abdoul G. Samb, Gracious M. Diiroc, Mario J. Mirandab

aAssistant Professor, Stetson University, 421 N. Woodland Blvd, DeLand, FL 32723, USA bOhio State University cMakerere University

Abstract

Very few empirical studies account for the dynamic nature of the agricultural technology

adoption decision and none of these explores if this dynamic nature depends on the gender of the

decision maker. Using four waves of a household level Ugandan panel data, this is the first

empirical analysis to account for self-learning (one’s own adoption experience) in explaining

current adoption decision in a developing country context, and the first to study the interaction

between self-learning and gender. Technology adoption is defined as the adoption of either hybrid

seed, or inorganic fertilizer, or pesticides. Our results indicate that the dynamic panel data Probit

model is superior to its static counterpart in the sense that self-learning, captured by lagged

technology adoption indicators, is by far the most important determinant of technology adoption.

We also find a weaker impact of self-learning for female-headed households than male-headed

households. Female-headed households face fewer learning opportunities, which produces a

lower self-learning impact in later periods, further exacerbating the gap in technology adoption

among male- and female-headed households.

JEL classifications: O1, O3, Q16

Keywords: technology adoption, panel data, dynamic estimation, Uganda

1 (Corresponding author). The authors are thankful to the participants of Agricultural and Applied Economics Association Annual Meeting and seminars at the Ohio State University and University of Florida and two anonymous referees. The usual disclaimer applies.

0

Gender and the dynamics of technology adoption: empirical evidence from a

household-level panel data

Abstract

Very few empirical studies account for the dynamic nature of the agricultural technology

adoption decision and none of these explores if this dynamic nature depends on the gender of the

decision maker. Using four waves of a household level Ugandan panel data, this is the first

empirical analysis to account for self-learning (one’s own adoption experience) in explaining

current adoption decision in a developing country context, and the first to study the interaction

between self-learning and gender. Technology adoption is defined as adoption of either hybrid

seed, or inorganic fertilizer, or pesticides. Our results indicate that the dynamic panel data Probit

model is superior to its static counterpart in the sense that self-learning, captured by lagged

technology adoption indicators, is by far the most important determinant of technology adoption.

We also find a weaker impact of self-learning for female-headed households than male-headed

households. Female-headed households face fewer learning opportunities, which produces a

lower self-learning impact in later periods, further exacerbating the gap in technology adoption

among male- and female-headed households.

JEL classifications: O1, O3, Q16

Keywords: technology adoption, panel data, dynamic estimation, Uganda

1

1. Introduction

Hunger kills more people every year than AIDS, malaria, and tuberculosis combined

(World Food Programme (WFP), 2015). Among regions most suffering from hunger, Sub-

Saharan Africa (SSA) has the highest prevalence of food insecurity, with 25% of the population

malnourished (WFP, 2015). Agricultural yields (output per acre) have fallen over the last decades

in many SSA economies (Suri, 2011), further exacerbating the food insecurity situation in the

region. There is widespread evidence that adoption of improved agricultural production

technologies can increase yields significantly (Nziguheba et al., 2010)

Despite the apparent advantages of modern agricultural technology, its adoption has been

rather slow in SSA. For example, only 5% of households use modern tractors, 16% use agro-

chemicals, and the use of nutrient application varies from 17% in Niger and Tanzania to virtua lly

non-existent in Uganda (Barrett, 2018; Christiaensen, 2017). Credit constraints, access to

resources, informational barriers, taste preferences, differences in agro-ecological conditions, and

lack of effective commitment devices are among the reasons typically advanced to explain low

rates of technology adoption (Dercon & Christiaensen, 2011).

Moreover, technology adoption rates in the region are lower among female farmers in

comparison to their male counterparts (Fisher & Kandiwa, 2014; Peterman, Behrman, &

Quisumbing, 2014). The reasons of the gender gap in technology adoption have been attributed to

differences between male and female farmers in farm size, asset ownership, social capital, and

access to labor, training, and extension services (Kondylis, Mueller, Sheriff, & Zhu, 2015;

Magnan, Spielman, Gulati, & Lybbert, 2015; Ndiritu, Kassie, & Shiferaw, 2014; Peterman et al.,

2014; Peterman, Quisumbing, Behrman, & Nkonya, 2011; Ragasa, Berhane, Tadesse, &

Taffesse, 2013). This imbalance undermines agricultural efficiency in SSA and by extension food

2

security given that females make up 40% of the agricultural labor force, significantly contributing

to the production process (World Bank, 2019). Leveling the field of access to agricultura l

resources for male and female could increase the total output up to 4%, reducing malnutrition and

lifting hundreds of thousands of farming households out of poverty (World Bank, 2015).

For most of the studies conducted on agricultural technology adoption and gender gaps in

Africa, the dominant approach has been to assume that the technology adoption decision is static

(Abebaw & Haile, 2013; Asfaw & Admassie, 2004; Katungi, 2007; Kondylis et al., 2015; Pamuk,

Bulte, & Adekunle, 2014; Ragasa et al., 2013; Tanellari, Kostandini, Bonabana-Wabbi, &

Murray, 2014). While these models provide important information on determinants of technology

adoption, they disregard the possibility that the adoption decision can be dynamic in that farmers

may base their current adoption decision on their past adoption experiences (Besley & Case, 1994;

Foster & Rosenzweig, 1995; Ma & Shi, 2015). Ignoring the dynamic nature of technology

adoption decision can introduce substantial biases, painting an inaccurate picture of the factors

that affect adoption decisions. For example, using structural theoretical models, Ma and Shi

(2015) find that the dynamic model fits their data better than the static model as the former is

more consistent with farmers’ underlying technology decision process. They also find that

farmers learn more from their own experiences than their neighbors’ experiences in a dynamic

setting. This is because differences in individual farm and farmer characteristics imply that

experiences of one farmer may not apply to others in that true returns from the adoption can only

be learned by farmers’ own experiences. However, their study uses data on US soybean farmers,

hence the findings may not explain the dynamics of technology adoption decisions in the

developing world. In that regard, Dercon and Christiaensen (2011) build a theoretical dynamic

model of how a household’s capacity of protecting itself ex-post from falling consumption affects

3

its ex-ante risk taking from technology adoption decisions. They use data from Ethiopia to test

their model but the empirical estimation does not account for decision making as a dynamic

process. An earlier study by Moser and Barrett (2006) attempt to estimate the dynamics of

smallholder technology adoption and find that learning effects encourage adoption among

farmers. However, their study is based on pseudo-panel data constructed from a cross-sectional

dataset due to lack of a true longitudinal panel dataset. In fact, the lack of household- leve l

longitudinal data has been a major obstacle to a better understanding of the dynamics of

agricultural technological adoption process in developing countries (Moser and Barrett 2006). We

address this gap in the literature by empirically accounting for the dynamic adoption process by

including self-learning for the first time in a developing country context, to the best of our

knowledge, by taking advantage of four waves of a household level national dataset from Uganda.

We define technology adoption as adoption of either hybrid seeds, or inorganic fertilizer, or

pesticides.

Furthermore, the literature has shown a lower rate of adoption among female farmers due

to less favorable environment for technology adoption pertaining to farm and farmer

characteristics (Fisher & Kandiwa, 2014; Ndiritu et al., 2014; Peterman et al., 2014; Quisumbing

et al., 2014; Ragasa et al., 2013; Tanellari et al., 2014). We speculate that the lower rate of

adoption can imply a lower opportunity for learning. This would indicate that female-headed

households have a lower impact from learning than male-headed households. Therefore, we

further investigate if the dynamics of technology adoption decisions differ by the gender of the

household head. To the best of our knowledge, we are the first to do so.

The empirical estimation follows three stages. First, we fit our data using a static Probit

model. Second, we use a Conditional Maximum Likelihood Estimator (CMLE) for dynamic

4

estimations. For the CMLE, we add the lagged dependent variable to account for the dynamic

nature of technology adoption. Third, we investigate the dynamics of technology adoption

disaggregated by gender of the household head. Some interesting results emerge from the

econometric analysis. First, dynamic models are superior than the static models. Second, lagged

technology adoption is the main driver of current adoption, indicating that adoption decisions are

highly determined by farmers’ own experimentation. Third, for the gender-disaggrega ted

analysis, the lagged technology adoption coefficient is significant and positive for female-headed

households but, it is lower than that of male-headed households, indicating a lower self-learning

effect for the female decision-makers. These results have important implications. First, studies

on technology adoption should include self-learning as it is an important determinant of adoption.

Second, female-headed households face fewer learning opportunities earlier, which produces a

lower self-learning impact later. This phenomenon reinforces the gap in technology adoption

among male- and female-headed households.

The remainder of this paper is structured as follows. Section 2 discusses hypotheses and

methodology. Section 3 provides a discussion of the data and descriptive statistics. Section 4

presents the results and discussion. Section 5 concludes.

2. Hypotheses and methodology

There is a widespread research on agricultural technology adoption. Following Griliches

(1957) and Feder, Just, and Zilberman (1985), earlier studies focused on how farmer

characteristics and farmland heterogeneity affect adoption decisions under a static setting. The

more recent literature acknowledges that the adoption decision is an inherently dynamic process

and incorporates learning into adoption models (Baerenklau, 2005; Besley & Case, 1994; Dercon

5

& Christiaensen, 2011; Foster & Rosenzweig, 1995; Ma & Shi, 2015; Moser & Barrett, 2006).

Using myopic (i.e., static) as well as learning (i.e., dynamic) models, Besley and Case (1994)

find that the learning model performs better in predicting the technology diffusion path of the

Green Revolution in India. Using a theoretical model that incorporates learning by doing and

learning spillovers, Foster and Rosenzweig (1995) test their predictions using the same Indian

dataset and find that adoption increases with farmers’ experience with technology. They further

find that farmers’ own experience has a stronger effect on adoption than neighbors’ experience

or formal public information dissemination sources such as government extension agents.

Similarly, Baerenklau (2005) builds an adoption model with a focus on risk preferences,

endogenous learning, and peer-group influence. Testing his model with a group of Wisconsin

dairy farmers, he finds that self-learning is more important than peer-group influence. Moser and

Barrett (2006) model dynamics of farmer technology choice by allowing for farmer

experimentation of technology (learning by doing), exposure to extension educators or other

farmers in the community (learning from others), and one’s education which may affect the rate

of learning. Using data from Madagascar, they find that both self-learning and learning from

others are important determinants of adoption. Lastly, Ma and Shi (2015) construct a dynamic

model to capture forward-looking farmers’ experimental behavior and find that farmers’ own

leaning experience matter more than their neighbors’ as the latter may carry additional noise

either due to information loss during communication or due to inapplicability of the information

given heterogeneous characteristics of farmers and farm plots. Therefore, based on these

theoretical studies, we formulate our first hypothesis as follows:

Hypothesis 1: Technology adoption increases with self-learning.

6

While the above-mentioned theoretical studies are silent on the issue of gender and the

dynamics of technology adoption, we can use their general framework to draw some inferences

about how adoption dynamics may vary between female and male-headed households (FHH and

MHH hereafter). For example, per Moser & Barrett (2006)’s findings, we can infer that FHH are

likely to experience less learning due to lower rates of technology adoption (learning by doing),

lesser exposure to extension educators or other farmers in the community (learning from others),

and lower education. Likewise, per Dercon and Christiaensen (2011), we can conjecture that

lower ex-post coping capacity among FHH in case of a negative shock means they invest in input

technology at a lower rate. In fact, studies have found that FHH have higher barriers to technology

adoption (due to lower income, reduced access to inputs and extension services, and poor quality

of social networks) and heterogeneity in production experiences post adoption (due to smaller

farm size, lower soil quality, and labor) (Kondylis et al., 2015; Magnan et al., 2015; Ndiritu et

al., 2014; Peterman et al., 2014, 2011; Ragasa et al., 2013). The predictions from the theoretica l

dynamic models combined with the findings from static empirical models suggest that FHH have

less learning opportunities due to their own lower experiences in adoption as well as higher

barriers to learn from social and institutional networks than MHH. Therefore, we formulate our

second hypothesis as follows:

Hypothesis 2: The impact of self-learning is weaker for FHH than MHH.

To test these hypotheses, we conduct our econometric analysis in three phases: a static

Probit model, a dynamic Probit CMLE model, and the CMLE model disaggregated by gender of

the household head. As commonly done, we include a lag of the dependent variable as an

additional explanatory variable to account for the possibility of dynamics in the decision to adopt.

7

This lag captures self-learning effect from state dependence, i.e., the impact on technology

adoption in the present period due to adoption in the past period (own experience).

However, incorporating a lagged dependent variable in the context of Probit estimation

with panel data leads to the violation of strict exogeneity which produces inconsistent estimates

(Wooldridge, 2000). For example, suppose our observations start at a date 𝑡 = 0, so that 𝑌𝑖0 is

the first observation of technology adoption 𝑌. For 𝑡 = 1, . . . , 𝑇, we are interested in the dynamic

fixed effects model:

𝑃(𝑌𝑖𝑡 = 1|𝑌𝑖𝑡−1 ,… , 𝑌𝑖0, 𝒙𝑖 , 𝑐𝑖) = Ф(𝒙𝑖𝑡𝜽 + λ𝑌𝑖𝑡−1 + 𝑐𝑖) (1)

where 𝒙𝑖𝑡 = (𝒙𝑖1, … , 𝒙𝑖𝑇) is a vector of contemporaneous explanatory variable, and Ф is the

Probit link function. Equation 1 simply captures the fact that technology adoption behavior

depends on past behavior, farm and household characteristics 𝒙𝑖 , and fixed (unobserved)

effects 𝑐𝑖. The joint distribution of the data stemming from Equation 1 is:

𝑓(𝑌1, 𝑌2, … , 𝑌𝑇 = 1|𝑌0, 𝒙,𝑐; 𝜽, λ) = ∏ 𝑓(𝑌𝑡|𝑌𝑡−1, … , 𝑌1,𝑌0 , 𝒙𝑡 , 𝑐; 𝜽, λ)

𝑇

𝑡=1

(2)

With fixed-T asymptotics, the parameters of primary interest 𝜃 and λ cannot be

consistently estimated due to the unobserved effects 𝑐𝑖 (Wooldridge, 2010). To circumvent this

incidental parameters problem, we need to integrate 𝑐𝑖 out of the likelihood function, which

requires further assumptions about the baseline period observation 𝑌𝑖0. This is known as the init ia l

condition problem. One way to get around it is to assume a density for the ci conditional on some

elements of the covariates 𝒙𝑖𝑡 (as in Chamberlain (1980)) and the initial adoption period 𝑌𝑖0 in

the framework of Wooldridge (2010)’s conditional maximum likelihood estimator (CLME) for

dynamic models. This allows for correlation between the initial condition 𝑌𝑖0 and the unobserved

heterogeneity 𝑐𝑖 in that 𝑐𝑖 = 𝜓 + 𝜉0𝑌𝑖0 + �̅�𝑖𝜉 + 𝜂𝑖 where 𝜂𝑖 ~Normal(0, 𝜎𝜂𝑖) is independent of

8

(𝑌𝑖0, �̅�𝑖) and �̅�𝑖 is the averages of 𝒙𝑖𝑡 , 𝑡 = 1,… , 𝑇. By the assumption of the distribution of 𝑐𝑖, the

latent version of the dynamic Probit model can now be written as:

𝑌𝑖𝑡∗ = 𝜓 + λ𝑌𝑖𝑡−1 + 𝒙𝑖𝑡𝜽 + 𝜉0𝑌𝑖0 + �̅�𝑖𝜉 + 𝜂𝑖 + 𝜀𝑖𝑡 (3)

where 𝑌𝑖𝑡∗ represents the latent net benefit of technology adoption and 𝑌𝑖𝑡 is the indicator variable

of modern technology adoption by household i at time t. In practice, this means that in addition

to contemporaneous explanatory variable 𝒙𝑖𝑡 , we add lagged technology adoption 𝑌𝑖𝑡−1, baseline

technology adoption 𝑌𝑖0, and the averages of explanatory variables �̅�𝑖. Region and time fixed

effect dummies which do not vary across time are omitted from �̅�𝑖.We refer to Equation 3 as

CMLE and use it as our main equation for dynamic analysis. To compare our dynamic CMLE

with a static model, we simply drop the two variables 𝑌𝑖𝑡−1 and 𝑌𝑖0.

In addition to the self-learning component, the literature has also pointed out the

importance of learning externalities from other mechanisms such as from the success of their

neighbors, membership to farmer groups, social networks, formal information dissemination

programs, extension services, and education (Diiro, Ker, & Sam, 2015; Knight, Weir, &

Woldehanna, 2003; Magnan et al., 2015; Moser & Barrett, 2006; Ndiritu et al., 2014; Ragasa et

al., 2013; Schultz, 1963; Smale, Assima, Kergna, Thériault, & Weltzien, 2018). To account for

these learning externalities, we incorporate the household head’s membership to a farmer’s group,

household’s participation in National Agricultural Advisory Services (NAADS) training

program, number of extension visits, and education (proxied by aggregate literacy of the

household members). The NAADS program started in 2001 with the primary objective of

providing information on inputs, production, and market (Kasirye, 2013). Moreover, we add

several controls that have been found to impact technology adoption: credit constraints, proxied

by number of plots cultivated (as in Dercon and Christiaensen (2011)), farm income (as in Moser

9

and Barrett (2006)), and off-farm income (as in Diiro and Sam (2015)), ability to cope with

changes in income proxied by livestock (as in Dercon and Christiaensen (2011)), and land

ownership (proxied by land certificate as in Gao et al. (2018)). We use lags of these variables in

our empirical model to mitigate possible endogeneity bias due to reverse causality. We also

control for soil quality (as in Ma and Shi (2015), labor availability proxied by household size and

number of adult male members in the household (as in Doss and Morris (2000)), and age of the

household head. Studies have found that since taking on risky behavior--adoption of a new

technology in our case--requires astute memory and learning, reduction in cognitive abilities due

to aging is associated with diminished tolerance for risky rewards (Albert & Duffy, 2012; Grubb,

Tymula, Gilaie-Dotan, Glimcher, & Levy, 2016). Furthermore, households may not have access

to risk coping mechanisms post production, which implies that their adoption decisions may be

influenced by risks related to shocks. Therefore, we include weather shocks (i.e., drought, flood,

and landslides), health shocks (i.e., death and illness), and other shocks (i.e., job loss, theft, fire,

violence, crop pests, and livestock pests) in our empirical model. Finally, we also include region

and year fixed effects to account for any local agro-climatic conditions (rainfall and temperature

variation) that vary across geographic areas and other shocks that vary over time but are common

to all regions (Griliches, 1957; Ouma et al., 2002; Smale et al., 2018). While these are also the

factors that are generally heterogeneous over MHH and FHH, we include marital status as an

additional control to account for those females who became household heads in FHH either due

to the death of their spouse, or divorce and separation, or never married. Over 66% of the FHH

are in this category across the years on average.

10

3. Data and descriptive analysis

About 70% of the female and 58% of the male working population are engaged in

agriculture, making an overall contribution of 25% to the Ugandan national GDP in 2017

(Ministry of Agriculture, Animal Industry, and Fisheries (MAAIF), 2019). Recognizing the need

for deeper insight into factors affecting this sector, the government added a detailed agricultura l

module to its Uganda National Household Survey (UNHS) 2005/06. The module collects data on

land, crop area, inputs, outputs, livestock, poultry, and agricultural extension services and

technologies. The UNHS 2005/06 surveyed 7,417 households nationally of which 3,123

households were selected for panel surveys known as Uganda National Panel Survey (UNPS)

(Uganda Bureau of Statistics (UBOS), 2012). The UNPS 2009/10, 2010/11, and 2011/12

successfully retained 83, 82, and 75% of the original sample and replaced the rest with split-off

tracking.1 Our study utilizes roughly 74% of the surveyed households from one wave of the

UNHS 2005/06 and three waves of UNPS 2009/10, 2010/11, and 2011/12 as these households

are engaged in agriculture (UBOS, 2012). After data cleaning, we end up with a total of 8,293

and 7,904 observations across all four years for the first and second cropping seasons,

respectively. Uganda has two cropping seasons, the first season runs from January through June

and the second runs from July through December. Annual crops, predominantly maize, beans,

and cassava, are grown in both seasons since rainfall is available in both, with shorter and more

intense (3 months) versus longer and spread out (4 months), respectively (Orlove, Roncoli,

1 The reasons for attrition cited are migration to unknown locations, natural causes such as death, and

refusal. UBOS generates a 20% sample of the households from each enumeration area selected for the UNPS to adjust the size and composition of sample that maybe impacted by attrition (UBOS 2013). This is a national level study that is not primarily conducted to study technology adoption; hence, we should not be concerned about attrition for the purpose of our work. However, to rule out any possible correlation between technology adoption and attrition, we construct an attrition variable equal to one if the household was not present in all waves and zero otherwise and find no correlation between technology adoption and attrition (see Table A1 in Appendix).

11

Kabugo, & Majugu, 2010). Usually more households cultivate plots in the first than the second

season, for example, 77% versus 72% in 2009/10 (Roberts & Azzarri, 2014). Therefore, we use

the former as our primary analysis and the latter as robustness check.2

In principle, technology adoption could be measured as a continuous variable in terms of

quantities of inputs used but, due to unreliable data on input quantities, we take technology

adoption to be binary throughout the paper as often done in the literature (Suri, 2011). The UNPS

collected data on three kinds of technology adoption: hybrid seed, inorganic fertilizer, and

pesticides. Due to very low rates of adoption of these inputs individually and by gender, we define

technology adoption to be one if the farmer used any of the three sources of technology, and zero

otherwise. Figure 1 presents the dynamics of technology adoption over time. Overall, a total of

25% of Ugandan households use technology over the four waves of data collection. Looking at

adoption over the waves, we note a general pattern that households that adopted (did not adopt)

technology in any particular period are more (less) likely to adopt technology the following period

(with few exceptions). This suggests that the nature of technology adoption is dynamic and should

be modeled accordingly.

[Figure 1 about here]

Table 1 presents tests of equality of means of the variables (reviewed earlier) used in our

empirical analysis between adopters and non-adopters by wave. The differences between the

households that adopt and those that do not adopt generally follow the economic intuition for all

the waves. For example, the means of factors that are associated with learning externalities (total

2 Results from both seasons exhibit the same pattern (see Results section for details).

12

household literacy, number of extension visits, participation in NAADS training, and farmer group

membership) are higher for adopting households. Likewise, the means of variables proxied for

credit access and labor availability are higher for adopters than non-adopters; for example, number

of plots, indicator for land certificate, livestock value, household size, and total adult males (with

an exception off-farm income for 2009/10). Furthermore, technology adopters tend to have

younger household heads (with an exception of 2005/06) and a higher proportion are married.

When it comes to experiencing shocks, we generally do not find any pattern of significant

differences between the adopter and non-adopter households.

[Table 1 about here]

Figure 2 presents the technology adoption proportion of MHH versus FHH; on average,

the adoption rates are 28 and 19%, respectively, across all waves. The adoption rates are generally

increasing for both MHH and FHH (with an exception of UNPS 2009/10) and are higher for MHH

than FHH for all waves. This later pattern can be inferred from the stark difference in initia l

adoption, i.e., first period of available panel data in UNHS 2005/06. Since FHH have lower

adoption rate initially, this translates into lower learning opportunity and consequently lower rate

of adoption in the later periods.

[Figure 2 about here]

In order to delve further into the heterogeneity between MHH and FHH that may have

caused the differences in adoption, we conduct a pairwise mean ttest comparison of variables used

13

in our analysis by wave and gender of the household head. The differences between FHH and

MHH generally follow the economic intuition for all the waves (see Table A2 in Appendix for

details).

4. Results and discussion

To test our hypotheses, our econometric analysis employs three phases. Phase one

employs the static Probit model. This is a correlated random effects model which controls for

individual heterogeneity and time fixed effects. However, this estimation fails to account for

adoption decision as a dynamic process which is problematic given that farmers update their

decision making over time based on their prior experiences. Therefore, phase two employs CMLE

which allows for the dynamic process of technology adoption and unobserved farmer

heterogeneity. To build more robust results, the estimations within each phase control for region

and time effects. Phase three employs the most robust CMLE (with region and time fixed effects)

to investigate the dynamics of technology adoption across MHH and FHH.

Table 2 presents marginal effects of the technology adoption from static Probit and

dynamic CMLE models. Our results provide support for Hypothesis 1 as evidenced by large and

significant coefficient on lagged technology adoption variable. The CMLE model shows that

households that adopt technology in the previous period are more likely to adopt in the following

period by about 18 percentage points per the preferred specification (that controls for time and

region fixed effects). This impact is by far the largest among the other significant marginal effects.

We also find a positive and significant estimate of over 4 percentage points on the baseline

adoption variable. Together, these results underscore the importance of self-learning on

technology adoption decisions of farmers. The pseudo R-squared and AIC/BIC measures of

14

goodness of fit (Posada & Buckley, 2004) for the preferred dynamic model are 0.2056 and

3186/3457 compared to 0.0282 and 3873/4136, respectively for the preferred static model,

providing clear statistical evidence that the dynamic model is a superior way to model adoption

decision. For learning externalities, extension visits, participation in NAADS training, farmer

group membership, and aggregate household literacy, we do not find any statistically significant

impact. These results concur with literature on the importance of learning (Besley & Case, 1994;

Foster & Rosenzweig, 1995; Ma & Shi, 2015) and more specifically with the literature that

emphasizes the importance of learning by doing over learning from others (Baerenklau, 2005;

Conley & Udry, 2010; Munshi, 2004). Among the proxies for credit access, only farm income is

statistically significant. We also note the negative and significant sign on land certificate which

seems puzzling initially. A test of comparison of mean off-farm income for those with and

without land certificate shows that the mean for the former is over three times higher than the

latter. This may indicate that households that own formal certificates are not primarily invested

in agriculture and hence are less likely to adopt agricultural technology. Moreover, age of the

household head is negative and significant, which substantiates previous research that aging

reduces tolerance for risky rewards (Albert & Duffy, 2012; Grubb et al., 2016). Finally, the

coefficients of shock variables are also negative, but statistically insignificant.

[Insert Table 2 about here]

Next, we present gender-disaggregated analysis of the dynamics of technology adoption

decision using the preferred CMLE model (that controls for time and region fixed effects) in

Table 3. Our results provide support for Hypothesis 2 as evidenced by significant but lower

15

coefficient on lagged technology adoption variable for FHH than MHH. These coefficients are

14 and 20 percentage points, respectively. While these findings reiterate the role of self-learning

in technology adoption, the coefficients on lagged technology adoption for FHH is significantly

lower than those for MHH at 5% level. We further match FHH and MHH by propensity score

(PS) method to account for possible differences that could arise due to non-linear effects of one

or more control variables in our model (Rosenbaum & Rubin, 1983). The propensity score is

based on all the observables used as controls in our analysis except for region and time fixed

effects which we include additionally in our PS method (see Mishra and Sam (2016) for

application details). Lagged technology adoption coefficients still remain significantly lower for

FHH than MHH. We speculate these learning differences occur for two reasons. First, FHH adopt

at a lower rate than MHH. The lower adoption creates reduced self-learning opportunity which

then translates into a relatively lower rate of adoption in later periods. Moreover, adoption is a

binary decision in our study whereas in reality there are differences in intensity of adoption (Ali,

Bowen, Deininger, & Duponchel, 2015) which can further translate into learning differences

between FHH and MHH. Although there are no explicit dynamic models of FHH versus MHH

adoption behavior in the literature, studies have found that lower ex-post coping capacity in case

of a bad shock lowers adoption in the current period, therefore lowering their net marginal gain

and consequently causing lower adoption in later periods (Dercon & Christiaensen, 2011; Ma &

Shi, 2015; Suri, 2011). Second, previous work has found that females have lower level of trust

and higher risk perception than males implying lower adoption rates of new technologies among

FHH (Siegrist, Gutscher, & Earle, 2005) and lower associated learning from adoption.

Unfortunately, we cannot test this speculation directly as data limitation neither allows us to

examine the intensity of adoption nor provides us with measures of trust and risk perception

16

pertaining to agricultural households. Overall, the results imply that current gaps in adoption can

propagate later gaps in adoption due to associated gaps in learning opportunities.

[Insert Table 3 about here]

Lastly, we repeat these three estimation phases with data for the second season; results

follow the same pattern as those from the first season (see Table A3 in Appendix). However, we

note that the coefficients are smaller than those in the first season. Given that less households

farm in the second season and have lower rate of adoption than the first season (Roberts &

Azzarri, 2014), the results further bolster our findings that lower current adoption implies a lower

self-learning generating a lower future adoption.

5. Conclusion and policy implications

Several studies in the literature investigate the determinants of technology adoption. Yet,

most of these studies assume the technology adoption decision to be static and therefore do not

account for learning from past experience. While there are theoretical papers that model

technology adoption as a dynamic process, very few empirical studies do so owing, until recently,

to the lack of longitudinal panel dataset from developing countries (Moser & Barrett, 2006).

Besides, none of the few dynamic studies considers whether the dynamic nature of the adoption

decision varies by gender. Therefore, using four waves of household level longitudinal panel data

from Uganda, we empirically investigate the dynamics of technology adoption at pooled and

gender-disaggregated household levels in this paper. By doing so, we make a significant

contribution to the limited literature on empirical estimation of technology adoption decision as a

17

dynamic process and shed light on a yet to be explored topic of adoption dynamics and gender.

Using static and dynamic Probit models, we have three major findings. First, the dynamic

models perform markedly better than the competing static model highlighting the need for

modeling technology adoption decision as a dynamic process. We find that recent experience

(technology adoption) in the previous period is the primary determinant of technology adoption

in the current period. We also find that the positive influence of adoption experience lingers over

time with a positive and significant impact of adoption in the initial/baseline adoption year on all

future period adoption decisions. Unlike self-learning, the variables proxying learning

externalities are not statistically significant. Together, the results imply that learning from self-

experience is the primary determinant of adoption of the technologies considered in this study

and therefore static models can present an incomplete story of adoption decisions. In future, we

hope to see more studies that can take advantage of more recently available longitudinal panel

datasets and present a more accurate empirical evidence on technology adoption decision when

modeled as a dynamic process.

Finally, we find that the positive and significant estimates of lagged technology adoption

are lower for female-headed households than male-headed households. These results imply that

the experience of having adopted in the past carries a lesser weight for future adoption decisions

for female- relative to male-headed households. We speculate that current adoption differences

(both in rate and intensity) cultivate differences in self-learning opportunity which can reinforce

later differences in adoption. Since modern agricultural technologies are accompanied by a higher

yield variance (Miranda & Farrin, 2012; Traxler, Falck-Zepeda, Ortiz-Monasterio R., & Sayre,

1995), lower ex-post risk coping capacity among female farmers may relatively weaken their

commitment to technology adoption (Dercon & Christiaensen, 2011; Suri, 2011). For example,

18

Magnan et al. (2015) find that women have poorer quality of social networks in the form of poorer

households compared to men, indicating a weak social insurance post-adoption. Likewise,

heterogeneity among the decision makers such as trust and risk perception that are associated

with adoption of new technologies could also spur this difference (Cai, Chen, Fang, & Zhou,

2014; Cole et al., 2013). This result has important policy implications. Since learning by doing is

important for continuity in adoption, policies should be in place to provide complimentary

environments (e.g., insurance through groups as in Hill, Hoddinott, and Kumar, (2013)) so that

initial barriers to adoption are broken for female-headed households making learning more

effective for future periods. The reduction in learning gap can encourage stickiness to technology

adoption in the later periods, boost the agricultural productivity, and reduce malnutrition which

is much needed in SSA.

19

References

Abebaw, D., & Haile, M. G. (2013). The impact of cooperatives on agricultural technology

adoption: Empirical evidence from Ethiopia. Food Policy, 38, 82–91.

https://doi.org/10.1016/j.foodpol.2012.10.003

Albert, S. M., & Duffy, J. (2012). Differences in Risk Aversion between Young and Older Adults.

Neuroscience and Neuroeconomics, 2012(1). https://doi.org/10.2147/NAN.S27184

Ali, D., Bowen, D., Deininger, K., & Duponchel, M. (2015). Investigating the Gender Gap in

Agricultural Productivity: Evidence from Uganda. 32.

Asfaw, A., & Admassie, A. (2004). The role of education on the adoption of chemical fertilise r

under different socioeconomic environments in Ethiopia. Agricultural Economics, 30(3),

215–228. https://doi.org/10.1111/j.1574-0862.2004.tb00190.x

Baerenklau, K. A. (2005). Toward an Understanding of Technology Adoption: Risk, Learning,

and Neighborhood Effects. Land Economics, 81(1), 1–19.

https://doi.org/10.3368/le.81.1.1

Barrett, C. B. (2018, July). In the Weeds: Effects of Temperature on Agricultural Input Decisions

in Moderate Climates. Presented at the 2018 International Conference of Agricultura l

Economists, Vancouver, Canada. Retrieved from

https://cslide.ctimeetingtech.com/icae2018/attendee/confcal/session/calendar?q=Christo

pher+B.+Barrett

Besley, T., & Case, A. (1994). Diffusion as a Learning Process: Evidence from HYV Cotton (No.

174). Retrieved from Princeton, Woodrow Wilson School - Development Studies

website: https://ideas.repec.org/p/fth/priwds/174.html

20

Cai, H., Chen, Y., Fang, H., & Zhou, L.-A. (2014). The Effect of Microinsurance on Economic

Activities: Evidence from a Randomized Field Experiment. The Review of Economics and

Statistics, 97(2), 287–300. https://doi.org/10.1162/REST_a_00476

Chamberlain, G. (1980). Analysis of Covariance with Qualitative Data. The Review of Economic

Studies, 47(1), 225–238. https://doi.org/10.2307/2297110

Christiaensen, L. (2017). Agriculture in Africa – Telling myths from facts: A synthesis. Food

Policy, 67, 1–11. https://doi.org/10.1016/j.foodpol.2017.02.002

Cole, S., Giné, X., Tobacman, J., Topalova, P., Townsend, R., & Vickery, J. (2013). Barriers to

household risk management: Evidence from India. American Economic Journal: Applied

Economics, 5(1), 104–135.

Conley, T. G., & Udry, C. R. (2010). Learning about a New Technology: Pineapple in Ghana.

American Economic Review, 100(1), 35–69. https://doi.org/10.1257/aer.100.1.35

Dercon, S., & Christiaensen, L. (2011). Consumption risk, technology adoption and poverty

traps: Evidence from Ethiopia. Journal of Development Economics, 96(2), 159–173.

https://doi.org/10.1016/j.jdeveco.2010.08.003

Diiro, G. M., Ker, A. P., & Sam, A. G. (2015). The role of gender in fertiliser adoption in Uganda.

African Journal of Agricultural and Resource Economics Volume, 10(2), 117–130.

Diiro, G. M., & Sam, A. G. (2015). Agricultural Technology Adoption and Nonfarm Earnings in

Uganda: A Semiparametric Analysis. The Journal of Developing Areas, 49(2), 145–162.

https://doi.org/10.1353/jda.2015.0013

Doss, C. R., & Morris, M. L. (2000). How does gender affect the adoption of agricultura l

innovations? Agricultural Economics, 25(1), 27–39. https://doi.org/10.1111/j.1574-

0862.2001.tb00233.x

21

Feder, G., Just, R. E., & Zilberman, D. (1985). Adoption of Agricultural Innovations in

Developing Countries: A Survey. Economic Development and Cultural Change, 33(2),

255–298.

Fisher, M., & Kandiwa, V. (2014). Can agricultural input subsidies reduce the gender gap in

modern maize adoption? Evidence from Malawi. Food Policy, 45, 101–111.

Foster, A. D., & Rosenzweig, M. R. (1995). Learning by Doing and Learning from Others:

Human Capital and Technical Change in Agriculture. Journal of Political Economy,

103(6), 1176–1209.

Gao, L., Zhang, W., Mei, Y., Sam, A. G., Song, Y., & Jin, S. (2018). Do farmers adopt fewer

conservation practices on rented land? Evidence from straw retention in China. Land Use

Policy, 79, 609–621.

Griliches, Z. (1957). Hybrid Corn: An Exploration in the Economics of Technological Change.

Econometrica, 25(4), 501–522. https://doi.org/10.2307/1905380

Grubb, M. A., Tymula, A., Gilaie-Dotan, S., Glimcher, P. W., & Levy, I. (2016). Neuroanatomy

accounts for age-related changes in risk preferences. Nature Communications, 7, 13822.

https://doi.org/10.1038/ncomms13822

Hill, R. V., Hoddinott, J., & Kumar, N. (2013). Adoption of weather-index insurance: learning

from willingness to pay among a panel of households in rural Ethiopia. Agricultural

Economics, 44(4–5), 385–398. https://doi.org/10.1111/agec.12023

Kasirye, I. (2013). Constraints to Agricultural Technology Adoption in Uganda: Evidence from

the 2005/06-2009/10 Uganda National Panel Survey. Retrieved from

https://ageconsearch.umn.edu/record/159669/

22

Katungi, E. M. (2007). Social capital and technology adoption on small farms : the case of

banana production technology in Uganda (Thesis, University of Pretoria). Retrieved from

https://repository.up.ac.za/handle/2263/24679

Knight, J., Weir, S., & Woldehanna, T. (2003). The role of education in facilitating risk-taking

and innovation in agriculture. The Journal of Development Studies, 39(6), 1–22.

https://doi.org/10.1080/00220380312331293567

Kondylis, F., Mueller, V., Sheriff, G., & Zhu, S. (2015). Policy experiment in Mozambique

highlights importance of gender in dissemination of sustainable land management

techniques. Retrieved from http://agris.fao.org/agr is-

search/search.do?recordID=QB2015100113

Ma, X., & Shi, G. (2015). A dynamic adoption model with Bayesian learning: an application to

U.S. soybean farmers. Agricultural Economics, 46(1), 25–38.

https://doi.org/10.1111/agec.12124

Magnan, N., Spielman, D. J., Gulati, K., & Lybbert, T. J. (2015). Information Networks Among

Women and Men and the Demand for an Agricultural Technology in India (SSRN

Scholarly Paper No. ID 2564538). Retrieved from Social Science Research Network

website: https://papers.ssrn.com/abstract=2564538

Ministry of Agriculture, Animal Industry, and Fisheries (MAAIF). (2019). Agricultural Sector

Potential in Uganda. Retrieved May 23, 2019, from http://agricultureug.org/agricultura l-

sector-potential/

Miranda, M. J., & Farrin, K. (2012). Index Insurance for Developing Countries. Applied

Economic Perspectives & Policy, 34(3), 391–427.

23

Mishra, K., & Sam, A. G. (2016). Does Women’s Land Ownership Promote Their

Empowerment? Empirical Evidence from Nepal. World Development, 78, 360–371.

https://doi.org/10.1016/j.worlddev.2015.10.003

Moser, C. M., & Barrett, C. B. (2006). The complex dynamics of smallholder technology

adoption: the case of SRI in Madagascar. Agricultural Economics, 35(3), 373–388.

https://doi.org/10.1111/j.1574-0862.2006.00169.x

Munshi, K. (2004). Social learning in a heterogeneous population: technology diffus ion in the

Indian Green Revolution. Journal of Development Economics, 73(1), 185–213.

https://doi.org/10.1016/j.jdeveco.2003.03.003

Ndiritu, S. W., Kassie, M., & Shiferaw, B. (2014). Are there systematic gender differences in the

adoption of sustainable agricultural intensification practices? Evidence from Kenya. Food

Policy, 49, 117–127. https://doi.org/10.1016/j.foodpol.2014.06.010

Nziguheba et al. (2010). The African Green Revolution: Results from the Millennium Villages

Project. Advances in Agronomy, 109, 75–115. https://doi.org/10.1016/B978-0-12-

385040-9.00003-7

Orlove, B., Roncoli, C., Kabugo, M., & Majugu, A. (2010). Indigenous climate knowledge in

southern Uganda: the multiple components of a dynamic regional system. Climatic

Change, 100(2), 243–265.

Ouma, J., Murithi, F., Mwangi, W., Verkuijl, H., Gethi, M., & De Groote, H. (2002). Adoption

of seed and fertiliser technologies in Embu District, Kenya. Kenya Agricultural Research

Institute (KARI) and International Maize and Wheat Improvement Center (CIMMYT),

Mexico, DF.

24

Pamuk, H., Bulte, E., & Adekunle, A. A. (2014). Do decentralized innovation systems promote

agricultural technology adoption? Experimental evidence from Africa. Food Policy, 44,

227–236. https://doi.org/10.1016/j.foodpol.2013.09.015

Peterman, A., Behrman, J. A., & Quisumbing, A. R. (2014). A Review of Empirical Evidence on

Gender Differences in Nonland Agricultural Inputs, Technology, and Services in

Developing Countries. In Gender in Agriculture (pp. 145–186).

https://doi.org/10.1007/978-94-017-8616-4_7

Peterman, A., Quisumbing, A., Behrman, J., & Nkonya, E. (2011). Understanding the

Complexities Surrounding Gender Differences in Agricultural Productivity in Nigeria and

Uganda. Journal of Development Studies, 47(10), 1482–1509.

https://doi.org/10.1080/00220388.2010.536222

Posada, D., & Buckley, T. (2004). Model Selection and Model Averaging in Phylogenetics :

Advantages of Akaike Information Criterion and Bayesian Approaches Over Likelihood

Ratio Tests. Retrieved May 12, 2019, from

https://academic.oup.com/sysbio/article/53/5/793/2842928

Quisumbing, A., Meinzen-Dick, R., Raney, T. L., Croppenstedt, A., Behrman, J. A., & Peterman,

A. (2014). Synopsis of Gender in agriculture: Closing the knowledge gap. Retrieved from

http://ebrary.ifpri.org/cdm/ref/collection/p15738coll2/id/128473

Ragasa, C., Berhane, G., Tadesse, F., & Taffesse, A. S. (2013). Gender Differences in Access to

Extension Services and Agricultural Productivity. The Journal of Agricultural Education

and Extension, 19(5), 437–468. https://doi.org/10.1080/1389224X.2013.817343

Roberts, C., & Azzarri, C. (2014). UGANDA AGRICULTURAL SNAPSHOT 2009/10.

Retrieved May 12, 2019, from

25

http://harvestchoice.org/sites/default/files/downloads/publications/WorkingPaper%20-

%20Uganda_3.pdf

Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in

observational studies for causal effects. Biometrika, 70(1), 41–55.

Schultz, T. W. (1963). The economic value of education (Vol. 63). Retrieved from

http://www.tcrecord.org/library/abstract.asp?contentid=2593

Siegrist, M., Gutscher, H., & Earle, T. C. (2005). Perception of risk: the influence of general trust,

and general confidence. Journal of Risk Research, 8(2), 145–156.

https://doi.org/10.1080/1366987032000105315

Smale, M., Assima, A., Kergna, A., Thériault, V., & Weltzien, E. (2018). Farm family effects of

adopting improved and hybrid sorghum seed in the Sudan Savanna of West Africa. Food

Policy, 74, 162–171. https://doi.org/10.1016/j.foodpol.2018.01.001

Suri, T. (2011). Selection and Comparative Advantage in Technology Adoption. Econometrica,

79(1), 159–209. https://doi.org/10.3982/ECTA7749

Tanellari, E., Kostandini, G., Bonabana-Wabbi, J., & Murray, A. (2014). Gender impacts on

adoption of new technologies: the case of improved groundnut varieties in Uganda.

African Journal of Agricultural and Resource Economics Volume, 9(4), 300–308.

Traxler, G., Falck-Zepeda, J., Ortiz-Monasterio R., J. I., & Sayre, K. (1995). Production Risk and

the Evolution of Varietal Technology. American Journal of Agricultural Economics,

77(1), 1–7. https://doi.org/10.2307/1243883

Uganda Bureau of Statistics. (2012). UGANDA NATIONAL PANEL SURVEY 2010/2011.

Uganda Bureau of Statistics Colville Street, Plot 9 P.O. Box 7186, Kampala. Retrieved

26

from http://siteresources.worldbank.org/INTSURAGRI/Resources/7420178-

1294154354661/UNPS_Report_wave2.pdf

Uganda Bureau of Statistics. (2013). UGANDA NATIONAL PANEL SURVEY 2011/2012.

Uganda Bureau of Statistics Colville Street, Plot 9 P.O. Box 7186, Kampala.

Wooldridge, J. M. (2000). A framework for estimating dynamic, unobserved effects panel data

models with possible feedback to future explanatory variables. Economics Letters, 68(3),

245–250.

Wooldridge, J. M. (2010). Econometric Analysis of Cross Section and Panel Data (second edition

edition). Cambridge, Mass: The MIT Press.

World Bank. (2015). The cost of the gender gap in agricultural productivity in Malawi, Tanzania,

and Uganda (English) (No. 100234). Retrieved from UN Women, the joint United

Nations Development Programme–United Nations Environment Programme Poverty-

Environment Initiative, and the World Bank website:

http://documents.worldbank.org/curated/en/847131467987832287/pdf/100234-WP-

PUBLIC-Box393225B-The-Cost-of-the-Gender-Gap-in-Agricultural-Productivity- in-

Malawi-Tanzania-and-Uganda.pdf

World Bank. (2019). Women, Agriculture and Work in Africa [Text/HTML]. Retrieved May 22,

2019, from World Bank website: http://www.worldbank.org/en/programs/africa-myths-

and-facts/publication/women-agriculture-and-work- in-africa

World Food Programme. (2015). Hunger Statistics. Retrieved from

http://www.wfp.org/hunger/stats

27

Tables and Figures

Fig. 1. Dynamics of technology adoption for pooled households

28

Fig. 2. Fraction of households adopting technology by wave and gender

29

Table 1

Mean ttest comparisons of variables by survey year and technology status

2005-06 2009-10 2010-11 2011-12

Variables Not Adopt Adopt Not Adopt Adopt Not

Adopt

Adopt Not

Adopt

Adopt

Number of plots 6.03 8.28*** 5.87 7.17*** 6.40 7.29*** 5.67 6.32***

Land certificate 0.07 0.12*** 0.21 0.24* 0.17 0.24*** 0.21 0.23

Soil quality 1.72 1.61*** 1.55 1.51 1.47 1.44 1.44 1.40*

Farm income/USh 173,797 528,933*** 853,214 1,194,701 637,723 1,313,836*** 930,461 1,877,419

Off-farm income/USh 150,376 213,722 561,199 540595 563,905 1,467,292** 663,489 1,122,478

Livestock value/USh 618,385 976,052** 1,038,583 1,538,971* 321,507 500,594*** 797,579 1,340,530

HH head sex 0.72 .81*** 0.69 .81*** 0.67 0.79*** 0.68 0.75***

HH head age/years 43 45*** 47 46 48 46* 49 48

HH head married 0.78 .84*** 0.73 .84*** 0.74 0.81*** 0.74 0.80***

30

HH size 5.8 6.8*** 6.2 7.0*** 6.84 7.6** 7.48 8.21***

Total adult males 1.3 1.6*** 2.8 3.2*** 2.14 2.42* 1.49 1.64**

Total HH literacy 2.11 2.37** 3.29 4.09*** 3.64 4.40*** 4.25 5.23***

Extension visits 0.15 0.59*** 1.02 2.56*** 0.60 1.68*** 0.75 1.17***

Participation in training 0.07 0.12*** 0.14 0.27*** 0.16 0.25 0.19 0.34***

Farmer group member 0 0 0.11 0.22*** 0.12 0.22 0.15 0.28***

Weather shocks 0.36 0.34 0.19 0.19 0.11 0.14 0.10 0.10

Other shocks 0.09 0.09 0.02 0.03** 0.01 0.01 0.01 0.01

Health shocks 0.09 0.11* 0.08 0.07 0.06 0.06 0.04 0.02**

*** p<0.01, ** p<0.05, * p<0.1 denote the significant differences in means between households that adopt and do not adopt

technology.

31

Table 2

Marginal effects of technology adoption from static and dynamic models

Variables Static Probit Static Probit Dynamic CMLE Dynamic CMLE

Lagged technology 0.2307*** 0.1807***

(0.0138) (0.0141)

Baseline technology 0.0729*** 0.0471***

(0.0158) (0.0158)

Number of plots -0.0015 -0.0019 -0.0051 -0.0045

(0.0042) (0.0041) (0.0042) (0.0042)

Land certificate -0.0296 -0.0320 -0.0560** -0.0600**

(0.0274) (0.0266) (0.0266) (0.0261)

Soil quality -0.0005 -0.0046 0.0063 0.0059

(0.0165) (0.0159) (0.0158) (0.0154)

Ln farm income/USh 0.0031 0.0031 0.0047* 0.0049**

(0.0024) (0.0023) (0.0025) (0.0024)

Ln off-farm income/USh -0.0000 -0.0004 0.0012 0.0008

(0.0029) (0.0028) (0.0031) (0.0030)

Ln livestock value /USh -0.0009 -0.0008 -0.0024 -0.0023

(0.0028) (0.0028) (0.0029) (0.0029)

HH head sex 0.0458 0.0406 0.0792 0.0666

(0.0562) (0.0537) (0.0509) (0.0494)

HH head age/years -0.0039* -0.0048** -0.0038* -0.0047**

32

(0.0024) (0.0023) (0.0021) (0.0020)

HH head married -0.0140 0.0152 -0.0075 0.0219

(0.0613) (0.0581) (0.0552) (0.0536)

HH size 0.0097*** 0.0012 0.0106*** 0.0035

(0.0028) (0.0043) (0.0027) (0.0044)

Total adult males -0.0074 -0.0041 -0.0098 -0.0069

(0.0109) (0.0107) (0.0105) (0.0104)

Total HH literacy 0.0010 -0.0000 0.0041 0.0026

(0.0047) (0.0045) (0.0046) (0.0045)

Extension visits -0.0011 -0.0015 -0.0023 -0.0025

(0.0025) (0.0024) (0.0023) (0.0022)

Participation in training 0.0082 0.0162 0.0115 0.0210

(0.0322) (0.0314) (0.0309) (0.0304)

Farmer group member 0.0352 0.0236 0.0411 0.0286

(0.0359) (0.0351) (0.0353) (0.0348)

Weather shocks 0.0320 -0.0314 0.0345 -0.0204

(0.0497) (0.0483) (0.0470) (0.0464)

Other shocks -0.2775 -0.2666 -0.2829 -0.2387

(0.2076) (0.1999) (0.1910) (0.1862)

Health shocks -0.0277 -0.0234 -0.0130 -0.0138

(0.0584) (0.0563) (0.0567) (0.0552)

Region/year dummy No Yes No Yes

Observations 3,818 3,818 3,334 3,334

33

Number of HHID 2,058 2,058 1,770 1,770

Log likelihood -2003 -1895 -1628 -1549

Pseudo R2 0.0282 0.2056

AIC/BIC 4083/4320 3873/4136 3338/3582 3187/3456

*** p<0.01, ** p<0.05, * p<0.1; standard errors in parentheses.

34

Table 3

Marginal effects of dynamic technology adoption for gender disaggregated households

CMLE CMLE PS PS

Variables MHH FHH MHH FHH

Lagged technology 0.1993*** 0.1398*** 0.2295*** 0.1714***

(0.0168) (0.0259) (0.0215) (0.0309)

Baseline technology 0.0440** 0.0492* 0.0547** 0.0511

(0.0191) (0.0282) (0.0258) (0.0353)

Region/year dummy Yes Yes Yes Yes

Observations 2,310 1,024 2,102 942

Number of HHID 1,246 578 1,078 498

Log likelihood -1119 -419

*** p<0.01, ** p<0.05, * p<0.1; standard errors in parentheses; both CMLE and PS methods account for

the eighteen control variables on income, farm and farmer characteristics, learning externalities, and shocks

as in Table 2; for brevity, we exclude them from the table.

35

Appendix

Table A1

Probit model regression on attrition

Variables Probit

Lagged technology 0.0767

(0.2445)

Number of plots -0.0810**

(0.0361)

Land certificate -0.0624

(0.2553)

Soil quality -0.0126

(0.1580)

Ln farm income/USh -0.0000

(0.0000)

Ln off-farm income/USh 0.0000

(0.0000)

Ln livestock value /USh 0.0000

(0.0000)

Hh head sex 0.2699

(0.2850)

HH head age/years -0.0321***

(0.0089)

HH head married -0.4082

(0.2866)

HH size -0.0865

36

(0.0583)

Total adult males 0.0531

(0.1134)

Total HH literacy 0.0252

(0.0462)

Extension visits -0.0071

(0.0403)

Participation in training -0.1102

(0.3898)

Farmer group member -0.1533

(0.5082)

Weather shocks -0.3438

(0.4494)

Other shocks -0.6270

(1.3371)

Health shocks -0.0158

(0.5292)

Region/year dummy Yes

Observations 8,293

Log likelihood -1603

*** p<0.01, ** p<0.05, * p<0.1; standard

errors in parentheses

37

Table A2

Mean t-test comparison of variables by survey wave and gender of the household head

2005-06 2009-10 2010-11 2011-12

Variables FHH MHH FHH MHH FHH MHH FHH MHH

Number of plots 6.00 6.75*** 5.64 6.42*** 6.16 6.80*** 5.35 6.08***

Land certificate 0.08 0.08 0.22 0.22 0.18 0.20 0.19 0.22*

Soil quality 1.75 1.67*** 1.53 1.54 1.50 1.44** 1.49 1.40***

Farm income/USh 103,853 311,935*** 882,650 964,188 485,900 923,744*** 697,382 1,427,187***

Off-farm income/USh 138,112 174,942 257,158 671,049** 429,535 916,752 679,186 848,894

Livestock value/USh 475,195 783,191* 968,862 1,244,709 246,334 412,233*** 589,022 1,114,614

HH head age/years 48 42*** 52 45*** 51 46*** 52 47***

HH head married 0.40 0.94*** 0.30 .93*** 0.34 0.93*** 0.35 0.94***

HH size 5.4 6.2*** 5.5 6.7*** 6.23 7.35*** 6.96 8.01***

Total adult males 0.9 1.5*** 2.4 3.1*** 1.75 2.41** 1.10 1.72**

Total HH literacy 1.74 2.32*** 2.87 3.74*** 3.40 3.98*** 4.05 4.74***

38

Extension visits 0.25 0.26 0.86 1.64*** 0.53 0.98*** 0.58 1.00***

Participation in training 0.06 0.09* 0.13 0.20*** 0.13 0.20*** 0.18 0.26***

Farmer group member 0 0 0.10 0.15*** 0.10 0.16*** 0.14 0.21***

Weather shocks 0.35 0.36 0.20 0.19** 0.11 0.12* 0.10 0.10

Other shocks 0.09 0.09 0.02 0.03 0.01 0.01 0.007 0.01**

Health shocks 0.12 0.09*** 0.09 0.07* 0.07 0.06 0.04 0.03**

*** p<0.01, ** p<0.05, * p<0.1 denote significant differences in means between female- and male-headed households.

Table A3

Marginal effects of dynamic technology adoption from Probit and CMLE models for second season

Variables Static Probit Dynamic CMLE MHH CMLE FHH CMLE

Lagged technology 0.1346*** 0.1433*** 0.1053***

(0.0150) (0.0177) (0.0290)

Baseline technology 0.0528*** 0.0656*** -0.0009

(0.0151) (0.0178) (0.0306)

Number of plots -0.0059* -0.0098*** -0.0073* -0.0160**

(0.0031) (0.0034) (0.0041) (0.0063)

Land certificate -0.0278 -0.0396* -0.0472 -0.0268

(0.0230) (0.0241) (0.0299) (0.0418)

Soil quality 0.0234 0.0224 0.0283 0.0028

(0.0143) (0.0148) (0.0184) (0.0264)

Ln farm income/USh 0.0004 0.0002 0.0002 0.0010

(0.0021) (0.0024) (0.0030) (0.0041)

Ln off-farm income/USh 0.0029 0.0034 0.0067* -0.0026

(0.0026) (0.0029) (0.0037) (0.0051)

Ln livestock value /USh -0.0003 -0.0014 -0.0073* 0.0051

(0.0026) (0.0029) (0.0038) (0.0047)

HH head sex 0.0037 -0.0091

(0.0485) (0.0486)

HH head age/years -0.0026 -0.0023 0.0003 -0.0045

(0.0021) (0.0020) (0.0045) (0.0036)

HH head married -0.0488 -0.0721 -0.0773 -0.0289

(0.0523) (0.0517) (0.0931) (0.0760)

40

HH size 0.0003 0.0020 0.0025 0.0037

(0.0038) (0.0041) (0.0051) (0.0071)

Total adult males -0.0115 -0.0069 -0.0090 -0.0153

(0.0098) (0.0102) (0.0122) (0.0195)

Total HH literacy 0.0021 0.0041 0.0071 -0.0059

(0.0040) (0.0043) (0.0051) (0.0079)

Extension visits -0.0004 0.0004 0.0011 -0.0078

(0.0021) (0.0021) (0.0023) (0.0083)

Participation in training 0.0138 0.0025 0.0171 -0.0885

(0.0275) (0.0285) (0.0335) (0.0615)

Farmer group member 0.0055 0.0142 0.0061 0.0872

(0.0308) (0.0323) (0.0375) (0.0704)

Weather shocks 0.1546*** 0.1297*** 0.1488*** 0.1082

(0.0415) (0.0427) (0.0514) (0.0805)

Other shocks 0.3143* 0.2822 0.2564 0.3596

(0.1698) (0.1748) (0.2070) (0.3516)

Health shocks -0.0172 -0.0266 -0.0222 -0.0381

(0.0500) (0.0527) (0.0647) (0.0969)

Region/year dummy Yes Yes Yes Yes

Observations 3,385 2,938 2,093 845

Number of HHID 1,835 1,570 1,132 481

Log likelihood -1390 -1150 -1625 -1548

AIC/BIC 2865/3122 2387/2650

*** p<0.01, ** p<0.05, * p<0.1; standard errors in parentheses.