https://orcid.org/0000-0002-2158-9921
![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
This study investigates the willingness to pay (WTP) for index-based drought insurance coupled with agricultural loans by product design and gender, using a contingent valuation method. The empirical estimations reveal several important findings. First, females have lower mean WTP than male farmers for each product design. Second, females do not express preference differences between insured loan product designs while males do. In fact, males prefer insured loans that make payouts directly to them and prefer insured loans without design risk.
Acknowledgements
It would not have been possible without our local research partners, Northern Association of Rural Banks, Ghana Agricultural Insurance Programme, and field survey team leaders: Dr. Francis Mulangu, Dr. Francis Kemeze, and Sigrinath Sherif. We are grateful to Abdoul G. Sam Mario J. Miranda, and Joyce P. Chen from the Ohio State University, Jerrod Penn at Louisiana State University, seminar participants at University of Central Florida, the Agricultural and Applied Economics Association, the Southern Economics Association, and the American Economic Association CSWEP, and two anonymous reviewers and the Editorial team at the Journal of Development Studies for their helpful feedback.
Notes
1 Design risk refers to the component of basis risk that results from a disconnect between the rainfall index and the actual rainfall experienced by the insured.
2 See Croson and Gneezy (2009) for a detailed discussion of gender preference differences as well as an exploration of the nuances in findings that are presented in broad generalizations here for brevity.
3 See Supplementary materials, Section S2, for their exact descriptions used in the surveys.
4 The description of the product to respondents made it clear that the rainfall was measured on the respondent’s farm, and therefore, the product did not include design risk. The concept of design risk was not explicitly discussed. For details on the product descriptions, see Supplementary materials.
5 WTP for Product 2 could be higher than Product 1 if a farmer has a demand for commitment products, i.e., a sophisticated hyperbolic discounter per Ashraf et al. (Citation2006), or if transaction costs are high that would make it costly to use insurance payouts to repay the loan. For a further discussion of this distinction, see Gallenstein et al. (Citation2019).
6 Risk aversion is measured using a self-reported risk aversion level based on a 5-point Likert scale. Our survey questions were designed based on Hardeweg, Menkhoff, and Waibel (Citation2013) who experimentally validate this risk aversion elicitation method in Thailand.
7 The coefficient on land holdings is insignificant in model 1 and 2, results not shown for conciseness.
8 The coefficients on unconditional remittances and poultry ownership variables are statistically not significant and positive, respectively.
9 The coefficient on age is positive and the coefficient on age squared is negative. These results are not reported for conciseness. These results indicate the relationship between the dependent variable and age for males. The coefficients presented in the table for the interaction terms with females move in the opposite direction. To consider the relationship between the dependent variable and age for females, we must sum the coefficient on age and age*female as well as sum the coefficient on age-squared and age-square*female. When doing so, we find that there is no non-linear effect for females as the resulting terms are not statistically significant.
10 It is important to note that our gender variable indicates female farmers, not female headed households. In Table S1 in Supplementary materials, we show that roughly 36% of females are household heads, compared to 83% of males in the sample. Moreover, we show evidence that females who are the head of their household have higher WTP for insured loans than females who are not (Supplementary materials, Table S5).
11 In Table S2 in Supplementary materials, we report mean t-test comparisons, following the structure of Table S1 (Supplementary materials), and show that the matched sample has far stronger balance as compared to Table S1 (Supplementary materials). Specifically, out of 25 variables, only two have statistically significant differences at baseline.
12 During the calibration of the bid values, we assumed that the loan was repaid within one year of issuing the loan and calculated the average loan repayment amount accordingly (GH₵ 485). Later, we found that loans are usually due within ten months implying that the loan repayment amount for a market-viable insured loan (GH₵ 468) is slightly smaller than the original estimate (GH₵ 485).
13 The delta method is a procedure for estimating non-linear transformation of random variables, like coefficient estimates from a regression. When implementing the Delta Method, we used a slightly different parametric model which leads to some different magnitudes than . The difference in the parametric model were that we did not include the interaction terms with gender and we excluded the district fixed effects.