are needed for a good impact evaluation, it is time to go on a journey with me of an actual impact evaluation. This is in our city that I conducted in Karnataka state in India, with Professor Rima Khanna at the Harvard Kennedy School. And the National Health Mission of government of Karnataka, with support from USAID's development innovations venture. The links to the paper, as well as a J-PAL summary and policy insight are included in the resources for this lecture. So what was the challenge, or the problem that policymakers were confronting? Despite high healthcare spending, health outcomes in India are unsatisfactory with very slow improvements over time. When you talk to policymakers, one reason they will often highlight is that health staff absenteeism is widespread, leading to suboptimal care. For example, surveys have shown that healthcare workers in India are absent 43% of the time they should be at work. We found that the numbers were about 40% absenteeism in Karnataka state. While another study found that in Udaipur, in the state of Rajasthan, health centers were closed 56% of the time that they should have been open. What is causing the doctors not to show up for work? And how can we improve their attendance? If we address this, will it improve the quality of care, and therefore the health outcomes? So what are possible contributing factors to the problem? Potentially, there is not enough health staff. Or it could be that the health staff has low intrinsic motivation. So they don't want to show up to work. It could also be that they are discouraged from showing up due to a lack of basic facilities and medicines at the clinics. Or it could be that information about attendance is not properly collected. Or that there is a lack of enforcement against those absent leading to furthermore absenteeism. Depending on what the underlying causes, there are different possible solutions. If you think there is not enough health personnel, you can hire additional staff. If instead you believe intrinsic motivation is the main contributing factor, you may want to remind health staff of the importance of their work. If the lack of basic facilities and medicines may discourage workers from coming to the clinics you could invest in the infrastructure and medicine availability. And finally, if the underlying source is the lack of information on attendance and poor enforcement, you could make attendance data available to the supervisors. To decide which of these options is best suited to our context, we need to pair a good understanding of the problem and knowledge of local conditions with rigorous evidence from around the world. In this case, the government decided to improve attendance data via a tracking system that would collect fingerprints from present staff, and transmit the data in real time using cell phones. It is important to note, that this was a program that the government itself conceived, based on a similar model that had worked very well for government staff working in the state capital of Bangalore, now Bengaluru. They were going to expand it to all primary health centers in the state. And were in the process of securing funding to do so. But before doing that, they invited us to work with them and to hone down the theory of change. And add in an impact evaluation of a pilot rollout to understand if it really works or not, before they scale it up. So kudos to the state government of Karnataka for their foresight, as most government programs in my experience, just go straight from an idea, to a grand announcement, to an ad scale rollout. With consequences that we are all well aware of in terms of bad implementation, and a lack of impact. Next, we collaboratively develop the program's theory of change. That is, the envisioned change process from the identified problem or need. What are the direct input and outputs to the desired intermediate, and to the final outcome. In our example, the government had the theory that the tracking machines will improve attendance and thereby, lead to better care and health outcomes. Let's go through the theory of change step by step. As discussed, the identified need is the low attendance of health staff and poor health outcomes. We have also already talked about the suggested program or input. Namely, the installment of a biometric attendance tracking system in primary health care centers to transmit attendance data in real time. The output of this program is that data is available in real time and is being used. Sometimes it can feel a bit artificial to distinguish between inputs and outputs, but differentiating the two is really important. Imagine the program is not implemented as planned. You will be able to see this because you will not observe the outputs. So it is useful to define those in advance. What are the outcomes of interest? Note, that intermediate outcomes often refer to a change in attitudes or behavior. Our theory was, that the program would lead to improved attendance of staff. And that staff would spend more time with patients. Or more patients would show up to be examined. Finally, we hope that these intermediate outcomes would lead to better standards of care and better health outcomes. Note, that for there to actually be a change in health outcomes, a number of other assumptions have to hold. For example, visits with doctors have to result in better care due to accurate diagnosis. Given the focus of this lecture, we will not go into all those details today.