Hope and Its Distribution in Rural Tanzania

Abstract

Recent research at the intersection of psychology and economics sheds light on the influence of hope on economic decisions. A body of that work concentrates on the economics of hope in developing country contexts. We identify two notable gaps: lack of attention to the measurement of hope as a latent psychological construct, and consequently, the lack of description and characterization of hope as a variable that can be measured and targeted. This study addresses these gaps by assessing the effectiveness of a novel hope measurement instrument, utilizing a large primary dataset collected in rural Tanzania. We estimate hope distributions across over 5,000 individuals and conditionally within subgroups defined by gender, region, recent shock, age, food security, income source, and religiosity. A positively-worded question about faith had the greatest information content among all questions, negatively worded questions were more effective in distinguishing people with relatively high hope. Employing generalized structural equation models, we observe significant variations in hope across sub-groups. Correcting for measurement distortions, we find significant heterogeneity in hope distributions across individuals and subgroups. The presence of an income-earning household member and religiosity yield the most pronounced shifts in hope distributions.

Keywords:

• Africa
• item response theory
• psychometrics
• measurement distortions
• graded response model

Acknowledgements

This research was materially supported by World Vision International and World Vision Tanzania as part of an assessment of their Empowered World Vision development model. The project, survey, and Data Transfer Agreement were approved by the Tanzania National Health Research Ethics Review Committee and the President’s Office Regional Administration and Local Government in January 2020. The authors acknowledge the Ifakara Health Institute (IHI) and all offices of World Vision (WV) for their roles in collecting and processing the original data for the baseline survey. The overall assessment and baseline survey were conceived and developed by staff from WV, IHI, and the University of Alberta, including Seamus Anderson, Rhonda Breitkreuz, Jane Chege, Charles Festo, Amy Kaler, Gilbert Kamanga, Vincent Kasuga, Katherine Kraft, Raphael Makoye, Philip Makutsa, Nessarian Mollel, Sally Mtenga, Daniel Muvengi, Elisaria Nassari, John Parkins, Josephine Shabani, and Monica Shandal. Anthony Scioli generously provided guidance to the hope assessment. Many helpful comments were provided by the dozens of World Vision staff and donors who participated in the annual virtual Research Update Meetings in 2021, 2022, and 2023.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 A substantial body of empirical and theoretical literature has traditionally operated under the assumption that household poverty traps arise solely from external constraints beyond the control of the household, such as limited access to credit or inadequate nutrition. These constraints, when coupled with non-linear production technologies, generate thresholds in welfare that the household is unable to cross and, effectively, land the household in a poverty trap (e.g Barrett, Carter, & Chavas, 2019). The emerging perspective of the economics of hope suggests that self-imposed thresholds in welfare, leading to persistent poverty traps, may be attributed to a shortage of hope, aspirations, or a limited reserve of other psychological capital within households.

2 The GRM specification is similar to an Ordered Logit model except that in our case the explanatory variable is latent (Skrondal & Rabe-Hesketh, 2004; Statacorp, 2013).

3 We maintain the assumptions of the Scioli scale including uni-dimensionality of hope, and conditional independence which postulates that the response variables are independent given the latent variables.

4 The theory conceptualizes hope as a goal-oriented cognitive process involving both the planning (pathways) and motivation (agency) to reach goals. Pathways refer to different ways to achieve one’s goals. Agency is the belief that one can create change to achieve goals.

5 Ifakara Health Institute obtained ethical clearance for the survey and data transfer to the University of Alberta from the IHI institutional Research Ethics Committee (IHI/IRB/No:02-2020) and the Tanzania National Institute for Medical Research Research Ethics Committee (NIMR/HQ/R.8a/Vol.IX/3344), dated 28 October 2020. The survey team pays courtesy to regional and district authorities prior to beginning data collection activities. With the help of the World Vision team, all permission letters were obtained well in advance and presented to village leaders during data collection activities. Before interviews, RA’s informed all study participants about the study objectives, risks, and benefits of participating and its procedures. Only respondents who provided written consent were interviewed.

6 Risks could be more or less idiosyncratic (specific to individual respondents or households) or co-variate across respondents in the same region (4 regions in our sample), ward (8 wards in our sample), or village (33 villages in our sample). The survey question and enumerator guidance were worded in a way that might encourage respondents to focus on widespread risks like ‘crop failure, floods, livestock diseases etc’. To evaluate the extent of covariation, we used anova tests to determine if there were statistically significant differences in prevalence of negative shock between regions, wards and villages. The results of that analysis suggest that while there are statistically significant differences between regions, wards and villages, the primary source of variance is idiosyncratic. Only a small portion (4.4-5.6%) of the total variance is due to the differences between regions, wards or villages. We thus conclude that while there are some common shocks affecting people who live in the same locality, most of the variability in reporting experience with shocks is idiosyncratic. That underlies our empirical strategy for treating exposure to negative shocks as a grouping variable in the multi-group model.

7 Information statistics are based on the Fisher information function which measures the information that an observable random variable X carries about an unknown parameter $\theta$ (which the likelihood of X depends upon) and typically measured as the negative of the Hessian of the log-likelihood of X.

8 The appendix provides a detailed analysis of how the hope survey questions probabilistically map and encompass a distribution of hope levels of individuals in our rural economy.

9 We evaluate the fit of the GRM by comparing a constrained model which restricts all parameters across item to be the same and an unconstrained model that assumes that each item has a unique set of parameters. Using AIC and BIC, and LR tests, the unconstrained model was supported, justifying the appropriateness of our specification.

10 In the upcoming distributional analysis, the redundancy of the two questions was confirmed by excluding them from the analysis. The results are qualitatively same across the two specifications. These sensitivity analysis results are available upon request.

11 A table of detailed results of the Kolmogorov–Smirnov tests is available upon request.

12 Results of gender analysis of religiosity effects are available upon request.

13 A table of detailed results of the Kolmogorov–Smirnov tests is available upon request.

14 That is, when given the FAITH question, respondents with: low hope levels (below approximately −3) are most likely to select ‘Strongly Disagree’; medium hope levels (between approximately −3 and -1.5) are most likely to select ‘Disagree’; and with slightly higher hope levels (between -1.5 and 1) will select ‘Agree’; and those with higher hope (above 1) are most likely to answer ‘Strongly Agree’ (bottom panel).