 So today I want to talk about, we're going to work sort of through three things. One is to talk about what state capability for policy implementation is and why it might be important. Second, we want to talk about an analytical typology of things you might want to get done that would then ask yourself, be able to ask what kinds of capabilities would I need from the state in order to do that kind of thing. This leads us to a big argument that mostly the things that in development that can get done by the types of state capability that have been built are done. And hence we have to shift our mindset towards the building of capability, towards building capability for doing very different kinds of things. So it's sort of like we've gone on a hike in order to climb a mountain. The first part of the hike, you need the capability to hike. The second part of the hike, when you're going straight up a cliff face, you need an entirely different set of capabilities than the capabilities of hiking. So most of the hiking we've done and we're now facing a set of entirely categorically and analytically different set of challenges for which we need to think entirely differently about how we build the capability to meet those challenges. And then third, we argue that mostly the way in which people have gone about both inside the donor community and in governments themselves have gone out building capability is inimical to building the kinds of capabilities we need. So if you continue to do what you're doing, you are not going to be successful. Most countries are stagnating or going in reverse by most of the measures of whether the capability of their state organizations is getting better, which means if you ask yourself how long will I have to do what I'm doing before I'm successful? The answer is forever, right? I mean, if you're still driving in reverse and your kid says how long till we get there, the answer is like we're not even on the way yet. So the final thing then is we're going to talk a bit about what we think the principles of a new approach are. It's not a method. It's not a technique. It's a set of principles. That's a path forward. So the first point is all of us use the word, all of us in this room use the word policy 12 times a day, right? Talk about policy reform, policy change, good policy, bad policy without too much stopping to think about what a policy means. What a policy is, is it's a mapping. Just like there's a difference between a number and a function and a function as a mapping between a domain and a range, a policy isn't what happens. A policy is the mapping. It's a mapping from states of the world, facts about the world to the appropriate actions by the agent of the state confronting those facts. So a tax policy says, a sales tax of 3% says, a tax agent goes to this nice lady. She decides what her sales are and then is empowered legally to collect 3% of that amount of juridically declared sales as tax. That's a policy, mapping from actions to states of the world, facts about the world to actions by agents, right? Now what is the capability for implementation? The capability for implementation is a feature of organizations, not institutions, organizations to induce their agents to take the appropriate actions. So this is a nice clean example, is in New Delhi there's a policy for getting a driver's license and the policy for getting a driver's license looks just like the policy wherever you live, which is you show up at the place that is authorized to give you the license, you demonstrate that you're of the appropriate characteristics to get a license like you're of the required age, you demonstrate your identity, you demonstrate your residence, and then usually you demonstrate that you can drive. And then an agent of the state authorizes you to drive. He gives you a driver's license, a license to drive. So what really happens, well, they did some experiment from which all of the interesting results come out of the control group, which is often the case, but it's still fun. And they measured people who came to the driver's license bureau to get a driver's license and then tracked what they did. What they found was that of the people that hired a tout, who were of course lounging around in front of the building in order to get their business, only 12% of them took the legally required driving exam. So once you hired a tout, you actually didn't even need to go into the building. They would deliver the license to you. Of course, if you had a little extra money, the tout would come to your house. You never even need to go to the building at all. Of the people that didn't hire a tout, 94% of them had to take the legally required driver exam. Why? Well, it was legally required. If you didn't know to hire a tout, you probably shouldn't be driving in this country, right? Of the 94% that took the driver's license exam, two-thirds of them failed. And then you might think, well, wait a second, if two-thirds of the people who took the exam failed, what about these guys that hired a tout? Well, it was exactly like you would think. Of the people who hired a tout and got a license without having to take a driver's exam, two-thirds of them couldn't drive. Like, not drive badly, couldn't drive. And not, like, drive, not couldn't drive like, oh, Lily-Livered Western researchers were scared of assessing how they drove. They hired people who taught driving in Delhi to assess their driving skills. So people with nerves of steel, because they're already in the business of teaching people to drive in Delhi. These people wouldn't get in the car with two-thirds of the people that had a driver's license, okay? So they couldn't drive at all. So the point is policy reform or policy change or policy, as written in the books, has absolutely nothing to do with the outcomes. The real policy is hire a tout and get a license. And I could hire a consultant and he could determine through traffic patterns using big data that many of the collisions were due to inappropriate left turns. And we could change the policy for a license to include, you know, you must demonstrate that you can make left turns in the driver's exam. And all of that would do jack all because everybody's hiring a tout and getting a license anyway. So it doesn't matter what the content of the de jure policy is if you can't implement it, okay? That's the first point. The second point, and this is like a 45-minute point that I'm going to make in three and a half minutes, the second point is that there's different types of tasks in the world. So when we say capability, we have to differentiate like capability to do what. So think of a sport, right? We were just in Indonesia which is very big on the sport badminton, okay? If I wanted to have the capability to be a badminton athlete, what do I need? I'm asking you. Have you ever seen badminton? Who's seen badminton? You've seen badminton. What do you need to be a badminton athlete? You've never seen badminton? This is a huge sport in Indonesia. There must not be globalized yet. What do you need to be a badminton athlete? Okay, I'll give you a hint. You gotta be quick, right? You never need to run very far because the court's about as big as this table, right? But you need to be very quick and you need to be very fast and you need to have endurance, but you don't need to be big, okay? Now, some of you have seen American football, right? What's the capability you need to be an interior lineman in American football? You need to be bigger and stronger than everybody else. You need to be huge. So if I say I want the capability to be an American football player, you need to eat a lot of food and lift a lot of weights. If I want the capability to be a badminton player, I need to develop agility for which the capability of being big and strong doesn't help me, okay? So what we've created is an analytical classification of tasks by the kind of capability it requires. And let me just sort of walk through what that is. So, and there are basically four questions about the tasks that you have to ask yourself. First is, does capability for accomplishing the objective of my task require a lot of people or a few people? Am I needed to have hundreds of thousands of people? Like a post office? Or do I need 12 people, like a central bank, right? And if it's not transaction intensive, then that requires a different kind of capability than a capability that's transaction intensive. Second question you want to ask yourself is, in order to successfully accomplish my objective, do the agents of the organization need to use discretion or not use discretion? Can I reduce what my agents need to do to a script or not? If it's transaction intensive and I can reduce it to a script based on a known technology, we call that logistics. If I want to build 100,000 schools all over Vietnam, that's logistics, right? Same school, same design, I just pop it down in each place. It might not be the perfect design for each place, but it works, right? If I want to do vaccinations, that's logistics. I grab short people, I give them the vaccination, full stop. Nothing complicated, right? If, on the other hand, I need lots of people and those, by the way, and then if I need to use discretion but it's not transaction intensive, we call this policy making. A few people can sit around and change the policy, make the policy, and some activities are like implementation light. So when the central bank decides to raise the interest rate, it raises the interest rate, and the interest rate's raised. It doesn't need a million agents to raise the interest rate, right? Okay, but there's this thing called implementation intensive service delivery, for instance, that think of primary schooling. I need a lot of teachers. Those teachers need to be in a classroom pursuing an objective that's relevant to what's happening in the classroom. They can rely on a known technology of learning and what's known. And they're trying to provide people with service, right? Now, so the capability of organizations to do implementation intensive stuff is very different than the implementation capability that you would need to do logistics is very different than the implementation capability you need to do policy making. For policy making, you need few numbers of highly trained people. For logistics, you need large numbers of people who are following a script. For implementation intensive, you need large number of people not following a script. You exercising their discretion, right? So part of the issue with the world is that most of the 21st century challenges are challenges of the type for which we will require capability to do implementation intensive things. Why? Well, in part because if it's easy, it's done. So, and policy making has actually gotten mostly a lot better. So the developing world has been fantastically successful actually at doing two kinds of things. It's been fantastically successful at the components of service delivery that involve logistics. So if you look at years of schooling and GDP per capita, what you see is that there has been a massive increase in the years of schooling even in countries that have had very little economic progress. So this is Ghana, whose GDP per capita in 2005 looked a lot like its GDP per capita in 65, but education had quadrupled. Quadrupled, I mean I'm pausing here because you should be stunned by that, right? Zero progress in per capita income quadrupling of the average years of schooling, right? This is Thailand, these guys moved northeast. They got a lot richer and they got a lot more schooling. But Mexico got a little bit richer, got a lot more schooling. France, and we have France up here in part because what's truly astounding about the world we are now in is that Ghana has more years of education of its adult population than France had in 1980. Again, I'm letting that sink in a bit, right? France, a civilization that thinks it leads the world, right? Spoken as an American. You know, France, the average adult, had less years of schooling by a lot than what the typical Ganan does today, okay? So across the board in vaccinations, in child mortality, in fertility, in schooling, the types of things that are logistical are done. Countries like Mali are getting most of their kids vaccinated. What does that tell you? That tells you it's easy to do, right? If Mali can do it, the capability required is low, right? So, but what that means is, but the capability beyond logistics is low, mostly collapses under pressure, and isn't getting fast, isn't getting better very fast. And very fast means quite slow. So what do we mean by capability to do implementation intensive things? Well, let's say we've built a whole bunch of clinics, but to actually accomplish the objective of the clinic with respect to ambulatory curative care where the patient walks in meeting something, you actually need to staff those clinics with people that walk through and exercise a professional knowledge to implement the policy. They've actually got to do a diagnosis and come to a treatment plan for the specific patient. So unlike vaccinations where I just grab a short human being and stick them, each person walks into a clinic with a different problem, okay? Now, when colleagues of mine measured recently in rural India, the capability of the public sector clinics by having patients present at those clinics as if they had certain symptoms and then saw what the doctors actually did, right? So what happens was the typical visit to a public sector clinic lasted only 38% as long as to a private sector clinic. And the average visit lasted 2.4 minutes. So the doctor is basically, what's the matter with you? And you were out the door, right? Only, they did 27, they were only 27% as likely to do simple physical diagnostics like take a pulse. And only 2.6% of these visits resulted in the correct diagnosis of asthma. So the patient presented as if they had the classical symptoms of asthma. And only 2.6% of the visits did the doctor move down the diagnostic treat towards what is not a very difficult to diagnose condition, okay? So what's the capability of these clinics? It's zero. I mean, the clinics there, they might have some stuff there. But the actual capability of the organization to deliver the service is zero, because they don't have the capability to do the implementation intensive stuff. I'm going to skip that slide. So the second thing we looked at is we looked at omnibus kind of rankings of capability, which are things like rule of law, quality of bureaucracy, extent of corruption, which are all measures or intended to be measures of how much the public sector bureaucracy is actually implementing policy versus deviating. And what we see is that if we say, where are countries in terms of absolute level of capability, and where are they in terms of progress? More than half of the countries, well more than half of the countries, have low capability and are moving backwards. So over the 10 years, 98 to 2008, and we've updated the analysis, and it's roughly the same. But basically, most countries' aggregate measures of the capability of the public sector are getting worse, not better. So they're not on a positive trajectory of acquiring the capability they need to do the next generation of things. And even of those that are rising, right, they're rising usually at a very small rate. So on a scale of one to 10, they're going up by less than 0.05 a year. So they're kind of would gain, at the most, it would take them at the, yeah, it takes like 20 years just to gain a single point on this index, at best. So if you calculate at current rates of progress, how long it would take most of these countries to achieve the lowest level of capability of OECD countries, which is Portugal? The answer is 100 years or more. At the current pace of progress, first of all, for half of them at current pace of progress, the time is infinity, right? Because they're going backwards or at zero. And even for those moving forward, almost no countries are less than 100 years at their current pace of progress from achieving Portugal-like levels of state capability. So this is a big problem. We argue that, yeah, you need to warn me about time, because I love to do. I think you're going to get something to warn me about the time. No, no, can you warn me about the time? Yeah, you've got ten minutes. Ten minutes, good, okay, sorry. So we argue that the reason there's such slow progress is that most of, is not that anybody has ever not realized that building state capability was an integral part of development. You read about development in the 50s, you read about it in the 60s, around the 70s, everybody's talking about building capability. No one is ignored the need for capability. It's just, we argue that people have been pursuing mostly a model for building capability that either is just wrong or at best is narrowly adapted to policy and logistics, but is not well adapted to building the capability for implementation intensive things. And it tends to have things like the business as usual approach to building capability, first of all, focuses on human capacity, not organizational capability. So there's a difference between what I could do if I felt like it and what I actually do embedded in the organization I'm working in, right? And most of us in this room feel that difference every day. Cuz we feel, if I just didn't have to put up with this damn bureaucracy, I mean, I could be much better, right? We feel like our capability is being held back, right? So human capacity is typically one of the many constraints to organizational capability, but often not at all the key constraint. So in the example I gave of the doctors in India, they actually demonstrably had the capacity to do much better than they were doing. So if you gave them an examination in principle of what they knew, they knew how to do much better than what they did in practice by a long ways. Second, the business as usual approach to building organizational capability often focuses on trainings that are built around process compliance, not outcomes. We will train you how to fill out expense funds. We will train you how to comply with organizational procedures independent of whether those organizational procedures are in the critical path of accomplishing the organizational objective. They tend to prioritize global solutions over local problems. That is, they frame the local problem as the lack of the global solution. Your problem is, you don't have an MIS system. No, my problem is that people aren't showing up for work. Now you might believe that if I had an MIS system I could make people show up for work, but that's a whole debatable conjecture. But whenever anybody has something to sell, your need looks a lot like their product. And so there's a lot of emphasizing global solutions. Your problem is that you lack my solution, right? The final thing is that there's a lot of emphasis on transplantation over local varieties, which goes along with this. So our contrasting approach, we call PDIA and we invented an acronyms. And we, not an anachronism, we invented, it might well be, but that wasn't our purpose. We invented an acronym knowing that in the donor world you cannot survive without a three or four letter acronym, right? You have to have an acronym. So, and it's about building capability while delivering results. And it focuses on four principles of how you build things, which I'm gonna come back to in a second. Because the problem, one of the major reasons that the business as usual fails, is that it facilitates what we call techniques of successful failure, right? So suppose you were in charge of a bureaucracy and you came to me and you said, I want to be a successful head of this bureaucracy. And I don't really care if anything happens. Can you teach me how to do that? And we're like, yeah, unfortunately, we can. We can teach you how to be a successful failure. Meaning successful at attracting legitimacy and budget and not getting in trouble. But failing at the fundamental purpose of the organization. And a lot of what that is, is that the current systems of capability building facilitate and encourage enormous amounts of isomorphic mimicry. Isomorphic mimicry is an evolutionary strategy where you gain survival value by looking like something dangerous. So this snake is extremely poisonous and this snake is not at all poisonous. Well, then you think, well, why evolutionary has this snake survived? Well, this snake survives by looking enough like that dangerous snake that you leave it alone, right? It's got red, yellow, and black. That's got red, yellow, and black. And so predators are gonna say, gee, I would like to eat me a snake today. But I'm not gonna eat any of these red, yellow, and black snakes. And this guy has just had successful failure. He's failed to produce any venom to be poisonous, but he has succeeded at surviving by mimicking. So a lot of the business as usual stuff encourages developing country organizations to look like functional organizations without actually delving into whether or not they are, right? So how many of you have ever been to a country? No, no, I won't even ask it this way. How would I ask it? How many of you have ever been to a country? Not necessarily your own. You don't have to say anything about your own country. How many of you have ever been to a country where the police were corrupt? Okay? In that country, were the policemen generally wearing uniforms? What's the role of a uniform in a corrupted police force? It's camouflage, right? It's I look like a policeman. I look like I'm enforcing the law independently of what I'm actually doing, right? Okay, so let me now come back to what I think the four principles are and I'm going to go through these very quickly. The first is that we look for local solutions to local problems. Second, you push problem driven positive deviants. That is, you need to create space in which people inside the organization can undertake actions which may or may not be organizationally policy compliant, but which promote the organizational problem designated problem. Based on measurement of the problem and the deviants, you institute ways of learning about what's working. And finally, you plan on scaling, learning through diffusion. That is, your organization is going to get better at doing what it does when the agents decide to adopt better practices, rather than being told to do it this way or that way, right? So, this starts with good locally nominated problems. We got a whole bunch of bad problems, right? Meaning problems that when framed this way will lead you into non capability building things, as opposed to good local problems. Which are problems for which you can actually get the authorizing environment to act. Meaning problems that your politicians really want solved. Not politicians they pretend to want solved, right? So all over the world, politicians will go, yeah, yeah, yeah, we're really about the MDGs. Really? Turnly loose on solving the MDGs. We're not that convinced, right? But they can be convinced that they have certain problems. And once they're convinced they have a problem, they may in fact turn you loose to solve it. But if you come to them with a fake problem, they will give you a fake task to create a fake solution. So you gotta get a problem where the politicians top bureaucrats and inside and outside organizational pressures are aligned with coherence on what the problem really is that they're trying to solve. Once you have a problem, then you have to authorize positive deviation. You have to authorize certain individuals in the organization to act in a way that will lead to solutions to the problem. And this is beyond policy compliance. And I say this because a lot of what's happening is people are claiming they're gonna build organizational capability by enhancing accountability. By which they mean they're gonna hold the individual actors in the organization tighter and tighter responsible for compliance. Now that might work for logistics, but no country has ever beaten teachers into being good teachers. No country has ever beaten, well, maybe not, but anyway. It's not been pretty if it's happened. Maybe Cuba's done it, but I don't know. So partly what you need is you need forces inside the organization to be willing to stick their neck out in order to move in a direction that pulls you forward. And our metaphor for this and we love animals and we love metaphors, cuz that's all you'll really remember other than I kept wandering past the camera, right? Is you can't beat a turtle to move. If you have a corrupt police force, you almost certainly will not get an uncorrupt police force by beating on it exclusively from the outside. You need to have allies inside the police force who are policemen who want to move at your direction. If you have crappy teachers, if you have crappy nurses, you will not get to good nursing by beating on the nurses with a stick. You will get to good nurses when good nurses inside the system pull into your direction, right? So the second thing is you gotta create positive deviance, right? And essentially, this is an autonomy of action in the space of innovation, swap to greater performance accountability rather than process accountability. I just skipped a whole bunch about learning cuz I have like one minute, I think. So I just wanna emphasize, particularly I emphasize this, particularly when I'm in a room of academics, right? Is academics love to have a model of the world in which they're important, right? And what's the model of the world in which academics are important? Well, it's a kinda whisper in the ear of God model of the world, where what the world lacks is knowledge about the right policy. And if I could just go out and do the research and find out what the right policy was, all I have to do is whisper it in the ear of God and he will make it happen by giving dictates and strictures to all those involved and they will change their behavior. Because they've changed the policy, the change in policy will change the behavior. That's just, and that's sort of, if you articulate this as a model of the behavior of agents in the public sector, you'll realize how completely ludicrous and at odds with all possible facts it is as a positive model of why government bureaucracies do what they do and how they get better, right? Except for logistics. There are things that are reducible to logistics and you can do better or worse logistics and you can work on those problems. But the 21st century problems are not this kind of research, right? In the implementation, right? Practices change when people change their practices and hence we need to worry about the combination of incentives, intrinsic motivations and beliefs about the effect of the practices of the people that are engaged. That is, teachers are gonna teach better when they find for themselves techniques that they feel help them teach better. And the organizational incentives and motivations are consistent with that, right? You're not gonna do it by doing some experiment in some corner of the world and then coming and saying adopt this. It just doesn't work that way, right? And you cannot, you cannot juggle without the struggle. Meaning when it's tacit knowledge that needs to be conveyed, the people that are responsible for implementation have to be engaged in the learning process. If they're not engaged in the learning process, they don't actually know how to do it at the end of it, right? Again, unless you can reduce it to logistics and a script. So anything that involves subtle use of judgment and local discretion for successfully accomplishive task is a tacit knowledge like juggling and I can juggle, but I can't transmit to you the skill to juggle by telling you about it, right? It's just not gonna work that way. So, and a lot of what's happening is the research into how to, the research that's trying to find out what better policies are is increasingly finding it doesn't matter what policies are. So when you go into a randomized controlled trial of what would be better policing techniques, what do you find? Well, you find a lot, and this has been increasingly replicated whenever people try into randomized evaluations with real organizations. What you realize is that the senior police supported the reforms and the police station staff basically ceased to carry out the program elements. So we wanted to know if the policemen did it this way versus that way, would it lead to better outcomes? Well, the answer was, we couldn't make policemen do it this way versus that way. They weren't manipulable by the leadership saying do it this way versus that way. Unless they in fact were part of the process of believing that there was a purpose to all this. So of the five experiments they tried, in three of them, the control group and the treatment ended up with exactly the same outcomes. Because they ended up with exactly the same inputs because they couldn't make the experiment happen. Because they were operating on a wrong model of how things were going to happen. So let me just go back, local solutions for local problems. Authorize within the organization the space for problem driven positive deviance. Turn people loose on problems against measurable performance outcomes. Set up a system in which the learning feeds back into the practice. And finally, plan on scaling the learning and the building the capability through your organization by engagement in the learning process rather than by imagining you're going to discover what works, tell everybody else what works. And that will lead to what leads to better organizations. Thank you very much.