Choice of Explanatory Variables

Our selection of explanatory variables is based on an extensive literature review but is limited by data availability. As such, some factors that we have not accounted for in this analysis could alter its results. However, the explanatory variables cover a broad spectrum of factors that re-searchers widely accept as influencing adaptation to climate change. The explanatory variables are summarized in Table 5.1, and they include

• household characteristics such as the age, education and gender of the household head, as well as the household size;

• monetary aspects such as farm income, livestock ownership, household assets and poverty status;

• physical characteristics such as farm size, well ownership and the use of extension ser-vices;

• climate-change awareness, as measured by perceptions regarding a decrease in rainy days and increases in soil erosion, temperature and water depletion.

In addition, we use region fixed effects to control for institutional and climatic differences, as well as for any other unobservable differences among the three watersheds.

Age of household head can be used as a proxy for farming experience. The literature shows mixed evidence regarding the sign and size of age’s effect on the use of adaptation strategies.

Several studies show at least a partially significant negative relationship between age and adap-tation (Deressa et al. 2009; Hassan and Nhemachena 2008; Jiri et al. 2017), but others indicate that age positively influences adaptation to climate change (Hassan and Nhemachena 2008;

Maddison 2007; Nhemachena et al. 2014). Some scholars even detect no significant relation-ship at all (Di Falco et al. 2012; Esham and Garforth 2013). Hence, one could argue that older (and thus more experienced) farmers are more likely than younger ones to adapt their farming behaviors to changes in climatic conditions; however, older farmers might also be more risk-averse and less flexible than younger farmers. Based on these findings, we conclude that the relationship between age and adaptation strategies is nonlinear. Thus, we add the square of the household head’s age to the regression. Nevertheless, we expect the sign to be inconclusive.

Education of householdhead is hypothesized to be positively correlated with the application of new agricultural technologies and the adaptation to climate change. Better-educated farmers are expected to have better knowledge and more information about climate change. Thus, they should be able to use their knowledge to react to perceived changes by applying improved farming methods. We find this to be the conventional wisdom in the literature; for example, Abid et al. (2015), Anley et al. (2007), Bryan et al. (2013), Deressa et al. (2009), Dolisca et al.

(2006), and Hassan and Nhemachena (2008)all find a positive and significant relationship, thus supporting this argument.

Gender of household headis hypothesized to influence decisions regarding adaptation to cli-mate change. Nhemachena et al. (2014) argue that, as compared to male-headed households, female-headed ones are more likely to use adaptation methods, as women are more involved than men in agricultural work and thus are more experienced in farm management. In addi-tion, scholars such asBayard et al. (2007) and Dolisca et al. (2006) provide supporting evidence for this argument because they find a positive relationship between female-headed households and adaptation. Nevertheless, there is also evidence that male-headed households are more likely than female-headed ones to get information about new technologies and farming prac-tices (Deressa et al. 2009), which can positively influence the adaptation process (Hassan and Nhemachena 2008; Hisali et al. 2011). Other evidence is mixed, as researchers such as Bryan et al. (2013), Di Falco et al. (2012), and Ndamani and Watanabe (2016) find that the gender of the household head has no influence on adaptation to climate change.

Household sizeis hypothesized to impact the adaptation process. As Deressa et al. (2009) and Gbetibouo (2009) argue, this issue can be seen from two angles. On the one hand, household size (as a proxy for labor endowment) should be positively related to adaptation because a larger household means more available workers and, thus, higher adaptation capability Jiri et al. (2017). On the other hand, larger families might require more off-farm activities in order to secure more income and decrease their consumption pressure. However, the empirical findings support the former reasoning.

Farm income,livestock ownership,household assetsandpoverty statusrepresent various aspects of wealth. Although farm income provides information about whether a household is solely dependent on farming to pay for the family members’ living, livestock and household assets can be seen as accumulated wealth. It is hypothesized that a higher level of wealth and financial wellbeing is positively related to use of adaptation strategies (Jiri et al. 2017; Knowler and Bradshaw 2007). In addition, we use the poverty status of a household to cover a wide range of wealth aspects that we cannot explicitly control for otherwise.

Farm size measures the total land size that a household holds in hectares and can be seen as a proxy for wealth Abid et al. (2015). In addition, Alauddin and Sarker (2014), Chengappa et al. (2017), and Gbetibouo (2009) argue that households with larger farm sizes are more likely

TABLE5.1: Description of the variables

Explanatory Variable Mean Standard

Deviation

Description Expected

Sign

Age of household head 54.7163 13.0701 Continuous (+) / (-)

Age of household head squared 3164.6977 1462.8538 Continuous (+) / (-)

Education of household head 0.8698 0.5568 Dummy: 2 if high school, graduate or post-graduate degree; 1 if primary, middle or sec-ondary school; 0 if no schooling

(+)

Gender of household head 0.7953 0.4034 Dummy: 1 if male, 0 otherwise (+) / (-)

Household size 4.0000 1.5004 Continuous (+)

Farm income (as only source) 0.0837 0.2770 Dummy: 1 if yes, 0 otherwise (+)

Household assets 6.3488 1.5323 Continuous (+)

Livestock ownership 0.4884 0.4999 Dummy: 1 if owned, 0 otherwise (+)

Household poverty status 0.5767 0.4941 Dummy: 1 if above poverty line, 0 otherwise (+)

Farm size in hectares 0.2858 0.4536 Continuous (+)

Well ownership 0.0651 0.2467 Dummy: 1 if owned, 0 otherwise (-)

Extension services 0.0930 0.2905 Dummy: 1 if received, 0 otherwise (+)

Rainy days 0.4419 0.4978 Dummy: 1 if medium, high or very high; 0 if

negligible or low

(+) / (-)

Soil erosion 0.5023 0.5012 Dummy: 1 if moderate or high, 0 if

unde-tectable

(+) / (-) Temperature rise 0.8556 0.3521 Dummy: 1 if medium, high or very high; 0 if

negligible or low

(+) / (-)

Water depletion 0.9442 0.2301 Dummy: 1 if seasonal, considerable or

al-most complete; 0 if irregular or not problem-atic

(+) / (-)

to adopt inventions earlier compared with smaller farms, as adaptation processes typically involve large transaction and information costs. We follow this argument and hypothesize a positive relationship between adaptation to climate change and farm size.

Well ownershipallows for controlling for an adequate supply of water. If a household does not have enough water for the irrigation of crops, we hypothesize that households are more likely to engage in an adaptation process.

Extension services influence the adaptation decision. They provide assistance and information about climate change, which is required to make an adaptation decision (Deressa et al. 2009). Various studies have found a positive relationship between extension services and households’ adoption behavior (Deressa et al. 2009; Gbetibouo 2009; Maddison 2007;

Nhemachena et al. 2014).

Climate-change awareness is an important factor in determining adaptation strategies. We explicitly control for this using the variables rainy days, soil erosion, temperature rise and water depletion. We measure the perceived existence of a substantial decrease in rainy days, the presence of soil erosion, the possible depletion of water and a considerable rise in temperature over the past few years. Especially, soil fertility has been found to be positively correlated to the decision of using soil-conservation methods (Di Falco et al. 2012; Gbetibouo 2009). In addition, awareness of changes in temperature and precipitation are important to adaptation decisions and the success of those decisions (Deressa et al. 2009; Maddison 2007; Nhemachena et al. 2014).

We expect that a farmer who is aware of a change in the climatic condition will increase the adoption of specific adaptation strategies that will help him in his situation. At the same time, however, farmers will eliminate the strategies that do not work specifically for their situations, are too expensive or address other risks. Thus, we expect the sign to be inconclusive.