In an interview with a student at Fudan University, China, Dilip Mookherjee (Member of the I4I Editorial Board) discusses Chinese growth; India-China comparison; and research in development economics.
Q: You mentioned that your first visit to China was 30 years ago in 1988, when people were wearing blue uniforms and caps with a red star in the front. For the one-week stay in China this time, as compared to your 1988 visit, what impresses you the most?
Dilip Mookherjee (DM): Well, obviously, it is an entirely different country. I think China has made a quantum leap from an undeveloped society to an almost developed society. You can see the signs in the physical infrastructure; in how people dress and how they move around with confidence; and in how educational institutions are organised and how close they are to the frontier (in terms of research, the quality of faculty and students). In all respects, so much change in 30 years is an amazing achievement.
Q: The 30 years during 1988-2018 is actually a period when China experience growth so rapid that it surprised the rest of the world. Yet the formal institutions in China remains rather underdeveloped as compared to the advanced nations. This is considered as the “China Miracle”. Many people are trying to find the key to this miracle. What is your take on the growth model of China?
DM: Yes, it is unique in many ways. Property rights is one example. It is not a full-fledged capitalist model; though it is capitalist in some aspects. The judicial systems are different. The political systems are different. I think there are advantages and disadvantages. There is a lot of literatures on institutions these days and I think most of it would benefit by taking a closer look at China.
Acemoglu and Robinson have a book on why nations fail. It is a theory of institutions and their importance for growth and development. But I do not think their theory applies very well to the Chinese context. In some of the research with my co-authors, I am studying the role of some other unique institutions such as social and community networks, which compensate for the lack of formal institutions in China. The research shows that informal institutions have been very successful in overcoming some of those inadequacies, for instance, with regard to entrepreneurship and entry into private industry. There is a lot of learning and cooperation among private entrepreneurs from clans within a common hometown. That is one way in which China has overcome the problem.
Q: As two of the most populous countries in the world and with such close geographical connections, India and China are often mentioned as a pair for comparison. Yet India experienced modest development in the past decades. What do you think lies behind this difference?
DM: I disagree somewhat with your assessment of relative growth performance of the two countries. India has achieved almost similar growth compared to China in the past few years. The modern period of India’s growth is the early 1990s. I think China may have started a little bit earlier, with the reforms led by Deng Xiaoping. So, China may have been ahead by about 10 years. Also, there was comprehensive agricultural development in the first phase, between 1980 and 1984 – a significant land reform and the household responsibility system. But in the later decades – the 1990s and the first decade of the 21st century, the growth of China may have been a lit bit above India, but not a lot.
Nevertheless, there is a very wide difference in the pattern of the economy and the nature of the development process. India never accomplished land reform and the levels of education and health are much lower relative to China. I think one of the achievements of communism in China was the creation of equality of access to assets as well as to health and education. It created a very good base for broad-based development.
There are a large number of poor, malnourished people with poor education in India. But both are very large countries. India has democracy and it is a heterogeneous country. That is another key difference. China is heterogeneous in some ways but India is much more heterogeneous – ethnically, linguistically, in terms of the religious divide between Hindus, Muslims and some other religions, and so on. As it is a much more divided country, it has to proceed on the basis of consensus. In democratic politics, everybody has to agree and hence, things take a bit more time to get going. Despite that, I think the growth India has achieved, the fact that it can come so close to China’s growth rate, is really incredible.
But growth is not everything. Development is much more important than growth – it is about human development, and the composition and distribution of growth. Both societies are experiencing high inequality. There are differences in terms of the regional patterns, environmental pollution, health outcomes. In those aspects, especially on the environmental side now and on human development, China is definitely doing better than India.
Yes, the distinctions between growth and development are worth noting. Now we will turn to the academic side.
Theory and empirics
Q: In your previous work, especially research in development economics, you maintain a good balance between theory and empirics. What do you think of the relationship between theory and empirics in research?
DM: Yes, I think it is very important to have a balance between theory and empirical work. Traditionally, in development economics work until mid-1980s, there was too much theory. There was a lot of armchair theorising, without much connection to empirical research. Starting from the 1990s, there was a new wave of empirical work, particularly involving field experiments. And things went to the other extreme of doing empirical work without thinking about what to infer from what we are seeing, and what it tells us about the underlying mechanism. There was too great an obsession with policy – about what policies work and what policies do not – before trying to understand the economy, how it functions, and why various policies do and do not work.
My perspective is that there are no universal laws in economics, unlike physics. It is more alike to evolutionary biology. There are some general game theoretic principles – like the Darwinian principles of selection in biology. But the way they play out varies from context to context and so every ecosystem is different. They all follow Darwinian principles but the local dynamics are very different. Similarly in economics, we really have to treat each region, each industry individually and try to understand it before we try to decide on what kind of policies will work well. So scientific understanding must precede policy research. For scientific understanding, we need both theories and tests of theories. For that, both theory and empirics are very important.
Q: So in this case, do you think that even if they are equal in principle, we need to go for empirical study first and then infer the theory from it?
DM: Yes. I once spoke at a conference at Cornell University in 2004 about the balance between theory and empirics in development economics. The proceedings of the conference panel were published in an Indian journal called Economics and Political Weekly in 2005. There, I reviewed what historians of science tell us about how scientific knowledge usually proceeds. Even in the natural sciences, it starts from some very basic descriptive evidence. That gives rise to some theories. Then those theories have to be tested and you need more sophisticated empirical works. And then there is a constant back and forth between theory and evidence. Until you understand the mechanisms which requires a balance between theory and empirics, you will never learn how to improve economic policy. I think it is similar in medical research. If you study the history of cancer research, you can see that substantial gains have been achieved in the treatment of cancer only after there have been sufficient progress in genetics that enables people to understand the source of cancer. Before then, people just tried different kinds of medicines but never really knew what was going to work. Every time a new medicine gave people hope and later disappointed them. In economics, we have seen many similar contexts where people think a certain theory is going to work in a particular context. The advisors suggest some policies which the government try out with great enthusiasm. Later it comes out that it does not work and people are disappointed. This is mainly because of a lack of sufficient scientific understanding.
Q: And that requires a lot of patience.
DM: A lot of patience, yes.
Q: A paper by Duflo in 2010 on the research agenda for development economics, suggests that future work on development should (a) revitalise applied theory to address limits exposed in earlier theory by recent empirical work, (b) expand empirical research, and (c) expand both theory and empirical research on aggregate consequences of micro distortions. Now eight years later, from your point of view, has the theory been revitalised and has empirical research been expanded?
DM: Yes, we are seeing a lot of econometric analysis, which is taking more theories and subjecting them to sophisticated tests. So I think there has been a lot of progress.
Q: Seems that we did a good job in the past eight years. So for the future, what do you think will be the frontiers of research in development economics?
DM: There are a lot of new avenues to be explored. One which I find very interesting, is to go beyond government and market as two polar alternatives and treat community networks as a key intermediate institution. There are so many different kinds of networks – social networks, economic networks, political networks. In the last few years, we are seeing an explosion of research on networks, mostly theoretical. We are also just beginning to see some empirical research in development on networks.
Political economy was an exciting new area about 15 years ago when theories were being developed. Now there is a lot of theoretical and empirical research on political economy. Economists expanded the scope of their methodology to include politics. I think now we are expanding it to include sociology. The boundaries of the discipline are becoming wider. I see many new researchers trying to understand social relationships and how they matter for economic outcomes.
There is the related question of how to design development institutions. There is a wide range of choice – purely government-based, purely market-based, and all kinds of combinations in between. I think we are becoming more and more interested in different ways of decentralising the implementation of policies. There is a choice between markets, local governments, NGOs, and local communities as implementing agencies. What is the right approach in any given context? How do we actually make these work, e.g. how do we evaluate the relative competence and accountability of different ways of delegating authority? That is a big challenge. I and many others have been working on these kinds of questions in the last few years. I expect that this kind of research is going to grow.
Q: Do you think that if we want to push that kind of research a little bit forward, cooperation with NGOs and the UN will be necessary? I have learned the theory of development and self-rule in Germany when I was an exchange student there. They mentioned that there is a possibility to have this kind of self-ruling development…
DM: Yes, absolutely, there is a whole range of alternatives. When you go outside of the purely private, profit-oriented institutions and government, you can rely on communities. You can also rely on non-profit, non-government firms. That is another alternative.
Q: Yes, but people are still trying to find a way to carry out the whole mechanism.
DM: There is a huge range of alternatives, irrespective of the sector you talk about. Consider for instance provision of education: you can have public schools, private schools, schools run by NGOs, or by the parents of the students (like a cooperative). Or you can have a school run by a local community leader, or by the local government. There are so many different kinds of alternatives. And once you set it up, let’s say, a school run by NGO, there are so many ways that you can set up the contractual relationships between the State and the NGO. Different ways of financing, different ways of evaluating the performance of the school. There is a lot of very detailed work needed on how to design these contractual relationships.
Reduced form and structural approach
Q: Thanks for sharing your view on the relationship between theory and empirics. Now for the empirical part, randomised controlled trials (RCTs) have been considered as the first resort to establish causal relationships. But as you are well aware, the experimental approach has also been criticised by some economists for a variety of reasons, including weak external validity, ethical considerations, compliance issues, and so on.
What are the key advantages of the experimental approach over other methods like observational studies? Or is it complementary to those?
DM: It is complementary. RCTs are a very useful and important tool. But it is by no means the best tool available. By its very nature, it can overcome some econometric problems of identifying causal relationships in a particular setting. But there are lots of context-specific features of the interventions being considered as well as the of the settings in which they are carried out. We rarely learn which of those are important. So efforts are being made to replicate these experiments in different countries and settings because of concerns of weak external validity. These are very expensive. They require lots of research funds, logistical preparations, and cooperation with local institutions, households, and firms. It is organisationally and financially difficult to set up an experiment in multiple settings. Some people are trying these meta-experiments – but it is impossible for most researchers to do that, to work on that kind of scale.
These are some inherent limitations of RCTs, which can sometimes be overcome in what we call natural experiments. These are based on policies that have actually been tried by governments on a large scale, e.g. a whole province, or a whole country. It can generate a lot of information at low cost. Of course, these studies use observational data and one then has to confront the difficult issue of establishing causal relationships – but there are a number of techniques for that. So these can be more cost-effective and involve fewer concerns regarding external validity compared with RCTs, at the cost of a certain lack of precision or internal validity. These are some of the trade-offs between these two different kinds of methods. We should obviously carry out both.
Q: Another question about the empirical work. There are so many reduced form studies using simple RCT, difference-in-differences (DID), regression discontinuity design (RDD), etc., these days. People are getting enough of these and a trend towards structural approaches seems to be on the way. Structural estimations enable us to see more clearly the channels and counterfactuals. So in your opinion, what are the comparative advantages and disadvantages of these two approaches? Do you think structural estimation will become a must-do in future empirical works?
DM: I think they are complementary. The traditional reduced-form approach is very important for testing between alternative theories because you are never sure in a certain context which theory is going to be the best approximation of reality. So, in any context, we have to test between alternative mechanisms. And for that, reduced forms are much better because when you do structural estimation, you have to commit to a particular model. You really have to be sure that that model captures the true mechanism. A lot of research in macroeconomics and industrial organisation (IO) is overwhelmingly structural, I do not think they are sufficiently careful to check the validity of their model.
For me, the ideal kind of research approach is: first, spend a lot of time and try to find out which is the right model using a combination of institutional detail followed by theory and then testing its reduced-form predictions. Usually when I start an empirical project, I go into the field, talk to people in the local area and get their perspectives. Sometimes economists are too used to running with an off the shelf dataset without understanding the local context. Then I develop a theoretical model capturing the key instiutional details, and derive its predictions. These are used to design household and firm questionnaires, generating data that enable the predictions to be tested. Or find some existing data which can be used for this purpose. Often the first model I try does not work, but understanding how it fails indicated what is wrong with it, and how it needs to be modified. This generates a new model which I again test, and the iterative procedure continues until I have a model that seems to fit the data reasonably well. Only then would I feel confident enough to go structural with that model, estimate or calibrate its paramters. With such a model you can understand what would happen if there were some alternative policies; how things would change if some parameters are changed; and do some welfare analysis. This is where structural analysis is very useful, as reduced-form methods will never tell you anything about welfare implications of policies. So you need a combination of both methods.
Way of thinking
Q: The starting point of your research is straightforward, but the results you bring to us are somewhat unexpected. For instance, the existence of middleman may actually be a good thing in the case of credit constraint and building roads to connect farmers to local markets may not necessarily do them any good. You are always thinking deeper. Is there a common way of thinking behind these research studies?
DM: Yes, you always have to keep an open mind and that the most important thing. That means you should put aside prejudice and preconceptions as much as you can. You have to question everything. Be prepared for surprises. And when surprises come, that tells you have really learnt something new. Otherwise, if the research just justifies your preconceptions, you will have learned nothing. So I think to learn is to be open and to be willing to find all kinds of surprises.
Q: So when you find something counterintuitive, how do you respond?
DM: Quite often it is the case that you start with a certain theory, a model, and then you expect to observe a set of findings if the model is correct. Quite often you do not get what you have expected. Then you have to go back and say, “Well, the model is not the right model! I have learned something.” You have to modify the model and find something that you missed. Ask yourself, “What is it that I missed?” This is the constant back and forth. You keep searching and groping until you get a model that explains all the facts you can see.
Q: You have done a broad range of work, from development to political economics, contract and organisation theory, IO, game theory. Is there a common thread among these topics that you have worked on?
DM: There is a common thread to many of my research projects, which is about how to decentralise, to informed intermediaries or middlemen or local experts. In terms of organisation theory, you can think of a middleman as a manager of a group of households or firms whose development a policymaker may want to promote. The broad question is who those intermediaries should be, how should responsibilities be delegated to them, and how they should be incentivised to generate accountability? Should they be government bureaucrats, private people, NGOs or key community agents? Essentially, you view society as a hierarchical organisation involving policymakers, managers, intermediaries, and common citizens at the ground level. Implementation is all about getting the right intermediaries, and the best way to motivate them. That is the common thread between most of my work. In the context of organisation theory I have been interested mainly in modelling delegation and contracting within hierarchies. In political economy and industrial organisation of developing countries I have been trying to understand how local governments and intermediaries behave. And finally the question of designing the mechanism – who is the right middleman and how to contract with the middleman?
Q: What motivates and give coherence to this wide range of work? Is it an occasional shift in personal interest and this coincidentally leads to the whole range of work, or is it the case that certain questions lead to one work after another?
DM: I think it is a bit of both. I did not anticipate that my work on political economy and on middlemen would come together, or plan it that way. It was subconscious. I now see that these are related issues, and am beginning to perceive the nature of the common thread.
Q: Yes, one of my teachers said that a research study does not lead to an end. It leads to another set of questions.
DM: Yes, indeed. Two decades ago I was trying to understand the impact of land reforms in West Bengal, the state where I come from. I started trying to understand how the ownership of land and the nature of land reforms carried out in the 1980s, affected productivity and inequality. But when I talked with people in the late 1990s on the ground; they said this was not where the action was any longer. They believed the ‘action’ was somewhere else, access to credit and marketing for poor farmers. Thanks partly to the land reforms, and partly to subdivisio of land via population growth, breakup of households and market transactions, the small farmers were no longer facing a problem of security of land rights. In West Bengal around the late 1990s, it was really a problem of how to get loans to cultivate cash crops, and how to sell cash crops at a remunerative price. So that is why I started to be interested in middlemen because farmers were selling to middlemen rather than to wholesale markets directly. They were borrowing mainly from these middlemen to finance their working capital needs. Therefore, I started looking at marketing relationships between farmers and local traders. One thing led to another. I switched from research on land to research on credit and marketing.
Micro and macro
Q: Most of your work has been at the microeconomic level. How do you integrate your findings with macroeconomic considerations that affect development – aggregate growth, international trade, and fiscal and monetary policy?
DM: Ideally, one should aggregate up in some way. But in development, there is a very wide diversity of contexts so it is very hard to aggregate. One should think hard about the aggregate implications. I do this using theoretical models of globalisation or financial development. It is harder to do empirically. My ongoing work on Chinese growth, for instance, is an effort to test an aggregate model of individual networks incorporating their heterogeneity. We are looking at the aggregate implications of having community networks; contrast them with existing theories of growth. So I think it is possible to relate micro with macro but it is challenging. Not just at the theoretical level, but especially at the empirical level. The kind of data you need to be able to aggregate up is quite demanding.
Research ideas and advice
Q: Your research covers a wide range of topics, including inequality, globalisation, corruption, etc. So where do you get the research ideas?
DM: I think I get the research questions and ideas from talking to people, reading newspapers, thinking about what is happening in the world today, what is really the core problem of development. Problems of policy implementation have been important, in the context of which decentralisation has become a big issue. We can see a lot of policy initiatives that are about corruption and decentralisation in one way or another. Globalisation is another, questions about its effects on inequality and development are widespread; I have been working on theoretical models that can help think about these in a systematic way. I also spent many years working on deforestation, on problems of sustainable development in the Himalayas, covering northern India and Nepal. We interviewed a lot of households, communities and tried to measure the economic damage of deforestation, to understand why it was happening, how it interacted with economic growth, and the effects of community-based forestry schemes. So most of my research has been driven by the problems of our generation.
Q: The last question: can you give some advice for the Ph.D. candidates?
DM: Well, be open-minded. That is the first important thing. Do not be wedded to any particular methodology. Look for a problem, a really important problem. Gauge that not just by what the professor thinks, or what your fellow students think, but also by talking to people in regular life. Second, it should not have received a lot of research attention already. You should study something new, or pursue a new angle to a classic problem. So there are two criteria: (1) it should be important. It should matter for people’ lives and for development. (2) It should not be studied already by many people. You should not want to do the 51st paper on some topic. You should do something new and original instead. Those are the most important things in finding problems. Then go acquire whatever tool that is going to be the most useful. Do not be fixated upfront with either structural or reduced-form approaches. Go and see what data is available. Talk to people on the ground and then get a sense of what the right model is. Always look for scientific understanding first. You have to understand what is happening before you evaluate or suggest some new policies. That is what I would recommend.
The original transcript is available here: https://mp.weixin.qq.com/s/hqez2tEwHGag3EZrjv3vlA