Schedule - Parallel Session 3 - Field Data

An unpredictable climate and the increasing likelihood of extreme weather events in Ghana means that farmers are subject to significant issues associated with risk and uncertainty. To understand how this environment influences farmers’ preferences, our study examines risk preferences using (Cumulative) Prospect Theory (PT) so that we can examine how respondents compare prospects in the gain domain as opposed to those in the loss domain. An feature of this study is that we explicitly assess if the restrictions on the power parameters for value functions used by Tversky and Kanheman (1992) are supported by our data. That is, the concavity and convexity of the value functions are exactly reversed in the loss and gain domains. We allow for heterogeneity amongst respondents in the way in which we econometrically model our data following Balcombe and Fraser (2015). Specifically, we model heterogeneity by employing a Hierachical Bayesian (HB) approach that has only been employed in a limited number of studies within the risk literature to date (e.g., Nilsson, Rieskampa and Waenmakers 2011). This research adds to the literature on risk preferences by authors such as Tanaka et al (2010), Liu (2013), Liu and Huang (2013), de Brauw and Eozenou (2014), Liebenehm and Waibel (2014). In general we find broadly support for “Weak” PT over Expected Utility (EU), and a model allowing farmer heterogeneity is supported relative to a representative agent model. As PT suggests, an “Inverse S” probability weighting is supported with respect to the “largest gain” within a prospect, but farmers’ seem almost globally optimistic concerning the “largest loss”. Farmers in our analysis show little tendency towards weighting losses more heavily than gains, within the range of the prospects considered. Our results indicate that models that allow for individual parameter heterogeneity are warranted. Also PT models cannot be usefully restricted to EU models and, the restrictions frequently employed in the literature on the power parameters for value functions, is not universally supported. This is an important result as it has implications for parameter interpretation. Specifically, as we discuss, this means that different parameter values interact and require careful interpretation. We pay particular attention to whether, `loss aversion’ or `loss seeking’ behaviour can be identified. Thus, in contrast with what appears to be the majority finding in the literature, we find is mixed evidence with regard to loss aversion, with many individuals being largely `loss neutral’ within a relevant payoff range. However, we do not make this conclusion simply with regard to the “loss aversion” parameter, as it has been labelled in the literature. Finally, we ask whether the estimates of individuals PT parameters can be used to explain a hypothetical decision to adopt a drought resistant technology. We find that the PT has some capacity to do so, but not in the way that we expected.

In the last decades birth order effects have been widely discussed in the psychological literature (e.g. Schooler, 1972; Sulloway, 2007) and in recent years also economists got interested in this topic (e.g. Björklund & Jaentti, 2012) as fertility rates and higher order birth rates decreased in industrialized countries. Looking at birth order effects, first parents’ behavior is important. Studies by Yeung et al. (2001) and Price (2008) show, that parent spend different amounts of time with their children favoring e.g. first born children. Additionally Hertwig et al. (2002) show that although parents tread their children equally at any given point in time, this yields an unequal distribution of investment or time if they have three or more children. Secondly, the child’s behavior matter. Sulloway (1995, 1996) assumes that siblings tend to maximize their differences to avoid direct competition and make it more to difficult for parents to compare their offspring’s. Both behaviors result in possible birth order effects. Many psychological studies concentrate on birth order differences in personality, whereas economists are more interested in differences in economic preferences, e.g. risk taking behavior. While the birth order effect on personality is unclear (see Rohrer et al, 2015 for a recent study), many studies find robust results for differences in risk taking (e.g. Sulloway & Zweigenhaft, 2010; Lampi & Nordblom, 2013). The key finding of these studies is that later born or middle born children are less risk averse. One problem of all these studies is that they use survey questions or look only at field behavior. Our study is therefore the first which investigates the influence of birth order, number of siblings and sibling’s sex composition on risk preferences in the school context with pupils aged on average around 16, who are still close to the family context, and with experiments which were incentivized, so that each subject was paid according to her or his choices. The field study was conducted between March and May 2015 in schools in Schleswig-Holstein, Germany with a total of 525 pupils all in the 10th grade. The experiments were conducted during school hour which helped us to minimize dropouts and self-selection. We elicited risk preferences through the choice task by (Eckel & Grossman, 2002) which is easy to understand and has been shown to yield good results over alls social-economic groups. We find no effects for the number of siblings. However, we find that only boys are more risk averse than all other boys with siblings. Looking at birth order effects we also find that middle born boys are more risk seeking, or in general, that second born boys are more risk seeking than first born boys. A single-sex environment seems to play only a role for boys in this context with boys having brother being more risk averse than all other boys with siblings. For girls we find no significant results.