Internally Valid Econometric Studies
Edited by ITO Seiro
We have tried four interventions: delayed notification, opt out, risk assessment, supportive information. Virtually no one exersized the option of delayed notification, so we used it as a control arm. Opt out resulted in robust negative impacts on uptake among Whites-Others in HCT sample. Risk assessment showed marginally significant positive impacts on Whites-Others in MSP sample. Supportive information increased the uptake of Whites-Others by almost 100% at the margin. We thus find substantial heterogeneity in responses. Generally, all experimental arms were ineffective in increasing the uptake of Africans and Colored who are deemed to be difficult to reach out and are inferred to have a
higher chance of infection. This general ineffectiveness is common among both MSP and HCT samples whose educational background differ significantly. We thus conjecture that factors related to their ethnic background to be the possible deterrants to tests.
We find robust and strong negative association of subjective probability of HIV infection with uptake.
Among the takers of HIV tests, we find a positive correlation between infection and subjective probability.
This indicates that there is a group of individuals who correctly expects their true status as HIV positive, yet rejects the tests. This pauses a threat to the containment of the disease. We recommend to ask subjective probabilities and continue targetting the individuals based on reported probabilities.
concern of being discovered HIV positive reduces a worker’s incentive to take the test, because he might fail to enjoy the full extent of the value of future relationship with colleagues. Since this expected loss might be reinforced by both the strength and size of social relationship, the negative impact of relational value apparently increases with a worker’s tenure and the number of connected workers. Using a worker’s behavioral data drawn from a large enterprise surveyed in South Africa, this paper provides evidence to support these views. As confidentiality is rigorously enforced in the surveyed enterprise, providing all workers with an excuse to test and/or encouraging them to privately test outside the workplace might be effective when introducing a HIV counseling and testing (HCT) program into the workplace of similar kind. Our findings suggest a great difficulty in designing an effective HCT program into a corporate sector, or more generally, any small community in which members strongly connect to each other.
It is also shown that the search methods among the job searchers to be inefficient in terms of time and money which require personal visits to potential employers without dependable prior information. Another popular search method emphasizes the importance of network in finding jobs, which is another manifestation of inefficiency in labor markets.
It is well known that impact evaluation of infrastructure is difficult. This is due to difficulty in randomization and broad spillover of program impacts that effectively wipes out the control group. Our focus on the tertiary canal was strategically determined in light of these
difficulties: First, tertiary canals are partially constructed and we could expect to employ plot-level difference-in-differences (DID) estimator. Ordering of tertiary canal construction was deternmined administratively, started from the closest area to the pump to the furthest. This ordering is expected to be uncorrelated with farmer ability, and provides credible ground for implementing DID estimator. Second, smaller program impacts due to limited capacity of tertiary canals to serve plots simultaneously allow us to find the control group within the irrigation scheme. The availability of control group in the neighborhood of the treated group lends support for credible impact evaluation. We will examine the impact heterogeneity in the coming rounds of additional surveys.