respondents. Risk preferences were based on the point at which respondents
switched from the safe to the risky option and were measured as the total number
of safe options chosen.
9 According to the expected payouts, a risk neutral person
would switch to the risky option under the fifth choice, and if the risky option were
chosen before this fifth choice, the person choosing would be deemed to be on the
risk preferring side, and if choosing after the fifth option, would be deemed risk
averse to varying degrees. One of the choices was randomly selected via the toss of
a ten-sided die, to decide an actual payout. Unlike the MPL, in which risk
preferences are inferred by non-hypothetical payout options, the self-assessment
scale allows respondents to identify themselves by the level of risk they are willing
to take on a scale of 0–10.
10 For both the MPL and the self-assessment scale, higher
numbers represent higher degrees of risk aversion.
The results show that, on average, respondents were risk averse according to
both elicitation methods. Using the MPL method, 10 % of respondents were
identified as preferring to take risks, 15 % were risk neutral and the remaining
75 % could be seen as risk averse. Using the self-assessment scale, 25 % chose a
value of less than five, 33 % chose exactly five and the remaining respondents chose
a value greater than five – though five should not be interpreted as indicating risk
neutrality. A Pearson Correlation Test carried out of the two methods indicated that
they were not significantly correlated.
We now turn to the question of if and how risk preference differs depending on
respondent characteristics. Table 5.4 shows the mean risk preferences of the
respondents, differentiated using two measures of poverty: absolute poverty based
on daily average expenditure per capita in 2010, and relative poverty based on a
poverty index constructed with data from 2007. According to the absolute poverty
measure, respondents living in poor households were, on average, significantly more
risk averse when compared to those living in non-poor households, to a 5 % level.
The relative poverty measure revealed a similar interpretation, that is respondents in
the poorest tercile were, on average, more risk averse when compared to those in the
wealthiest tercile, though the difference in means was statistically significant when
using the MPL method only (to a 1 % level). Furthermore, results from the Pearson
Correlation Test indicate that our two wealth variables (daily per capita expenditure,
and the 2007 wealth index) were significantly and negatively correlated with both risk
preference measures, providing further evidence that higher levels of wealth are
associated with lower degrees of risk aversion.
Further investigation of respondent characteristics indicates that respondents who
never completed their formal education were significantly more risk averse, and that
women were more risk averse than men (p < 0.01). Likewise, differences in risk
aversion were observed between male- and female-headed households, whereby the
second group was on average more risk averse than the first (p < 0.05).
9 For more details on the methodology, please see Holt and Laury (2002).
10 This method is based on a German Socio-Economic Panel Study and has been widely used to
assess risk preferences (cf., Caliendo et al. 2009).
5 Linkages Between Agriculture, Poverty and Natural Resource Use. . .
189
switched from the safe to the risky option and were measured as the total number
of safe options chosen.
9 According to the expected payouts, a risk neutral person
would switch to the risky option under the fifth choice, and if the risky option were
chosen before this fifth choice, the person choosing would be deemed to be on the
risk preferring side, and if choosing after the fifth option, would be deemed risk
averse to varying degrees. One of the choices was randomly selected via the toss of
a ten-sided die, to decide an actual payout. Unlike the MPL, in which risk
preferences are inferred by non-hypothetical payout options, the self-assessment
scale allows respondents to identify themselves by the level of risk they are willing
to take on a scale of 0–10.
10 For both the MPL and the self-assessment scale, higher
numbers represent higher degrees of risk aversion.
The results show that, on average, respondents were risk averse according to
both elicitation methods. Using the MPL method, 10 % of respondents were
identified as preferring to take risks, 15 % were risk neutral and the remaining
75 % could be seen as risk averse. Using the self-assessment scale, 25 % chose a
value of less than five, 33 % chose exactly five and the remaining respondents chose
a value greater than five – though five should not be interpreted as indicating risk
neutrality. A Pearson Correlation Test carried out of the two methods indicated that
they were not significantly correlated.
We now turn to the question of if and how risk preference differs depending on
respondent characteristics. Table 5.4 shows the mean risk preferences of the
respondents, differentiated using two measures of poverty: absolute poverty based
on daily average expenditure per capita in 2010, and relative poverty based on a
poverty index constructed with data from 2007. According to the absolute poverty
measure, respondents living in poor households were, on average, significantly more
risk averse when compared to those living in non-poor households, to a 5 % level.
The relative poverty measure revealed a similar interpretation, that is respondents in
the poorest tercile were, on average, more risk averse when compared to those in the
wealthiest tercile, though the difference in means was statistically significant when
using the MPL method only (to a 1 % level). Furthermore, results from the Pearson
Correlation Test indicate that our two wealth variables (daily per capita expenditure,
and the 2007 wealth index) were significantly and negatively correlated with both risk
preference measures, providing further evidence that higher levels of wealth are
associated with lower degrees of risk aversion.
Further investigation of respondent characteristics indicates that respondents who
never completed their formal education were significantly more risk averse, and that
women were more risk averse than men (p < 0.01). Likewise, differences in risk
aversion were observed between male- and female-headed households, whereby the
second group was on average more risk averse than the first (p < 0.05).
9 For more details on the methodology, please see Holt and Laury (2002).
10 This method is based on a German Socio-Economic Panel Study and has been widely used to
assess risk preferences (cf., Caliendo et al. 2009).
5 Linkages Between Agriculture, Poverty and Natural Resource Use. . .
189