Experimental Design Details
The experiment is attached to the existing interview for the performance-based financing (PBF) study. As such, the experiment will use the existing sample and recruitment procedures. In the larger study, in each facility three patients seeking antenatal or under five curative health services will be randomly sampled for direct observation and exit interviews for the main evaluation. The health workers providing services to these patients will then be automatically selected for a health worker interview to which the experiment is attached.
The larger PBF study tests two financing modalities against a control condition. The three study arms are implemented at the state level. States in the project areas are assigned to either a performance based financing (PBF) setup or to a decentralised facility financing (DFF) setup. Three states have been chosen as candidate comparison states based on similarity to one of the three project states. Similarity is defined as the Euclidian distance between the project state and potential comparison states, based on 13 socio-economic indicators measuring the state of the health care infrastructure in the state.
Participants in each arm are paid in mobile phone airtime. The information-only group receives 1,750 Naira (about USD 5.80), the rewards group receives a base pay of 1,000 Naira (and can gain up to a total of 2,500 Naira), and the penalties group receives a base pay of 2,500 Naira (and can lose 1,500 so that the minimum payout is 1,000 Naira). The minimum and maximum possible amounts for the rewards and penalty groups are identical at 1,000 and 2,500 Naira, respectively.
In secondary analyses we will include clinic and worker covariates, including interaction terms.
At the clinic level, covariates include:
• Level (binary: primary or secondary)
• Tercile of the facility’s catchment area (categorical: small; medium; large. Calculated separately for the primary and secondary facilities)
At the worker level, covariates include:
• Gender (binary)
• Age (continuous)
• Education level (categorical: primary or less; secondary or less; more than secondary)
• Qualification (categorical: medical professional [doctor, nurse, nurse midwife; lower-level medical professional [midwife; public health nurse; community health officer; community health extension worker; junior community health extension worker]; other [e.g., pharmacist, lab technician]);
• Received training in labor/delivery (categorical: less than 1 year ago; more than 1 year ago; never)
We will approach potential further analyses using a split-sample approach. Specifically, we will create a random sample of 30% of clinics (stratified by state and the facilities’ level) to conduct exploratory work. Tests with an absolute t-statistic of 1.2 in the exploratory sample will also be implemented in the full sample, but as one-sided tests (using the sign indicated in the tests on the exploratory sample).