Public Policy for Realists

Reading Time: 20 minutes

Thanks to Sam Enright, Fergus McCullough, Nihal Sahu and Judah for their comments on this

Making public policy is hard. For even the most simple policy change, such as an increase in the goods and services tax, there are multiple factors to consider (will people cut back spending? How much revenue will it raise? Is it worth it?) and multiple stakeholders to pacify (unions, businesses, specific interest groups, voter anxieties). 

In the end, even after decisions are made and a consensus is reached, someone has to do the hard work of writing laws and rules in a way that conveys the intent of the decision makers. When that is done, there needs to be a bureaucracy that converts these words on paper into action on the ground. 

All this is difficult enough in a normal country. Now imagine your country is large and populous, with over a billion people across 1.2m square miles. They speak different languages, have different customs, and have immense variation in incomes and industries. And to top all of this, your government doesn’t have the ability to implement everything you wanted. 

Welcome to India.

How should public policy be done in this context? Ajay Shah and Vijay Kelkar give their answer in their book In Service of the Republic. The two men are distinguished members of the Indian policy world. Kelkar has been secretary of the Prime Minister’s Economic Advisory Council, Finance Secretary, Chairman of the Finance Commission, head of the GST task force among many other positions; Shah is an academic at Jindal Global University, who runs the excellent Leap Blog with Kelkar. Shah’s research focuses on finance, especially that of equity market regulation, international finance and pension reform. 

The Problem with Public Policy

Policy making starts with data. But in India there is little data, and what exists is not detailed enough. Take the example of air pollution in Delhi. This problem is common knowledge – one has to only look outside the window –  but beyond basic data (like pollution concentration levels), we really don’t know what is going on. We don’t have granular data on air quality monitoring, on its causes, on the number of people affected by it, and so on. 

Another good example is Indian inflation data. Former Reserve Bank of India governor Duvvuri Subbarao made the point in his book that in many cases inflation data was so unreliable that RBI decisions were reversed after better data was collected. How are policymakers supposed to make good decisions with such bad data?

Or to quote Y.V Reddy, another former central banker: 

Everywhere around the world, the future is uncertain; in India, even the past is uncertain. 

Part of the reason why is that statistical data isn’t very well funded in India. The Indian Ministry of Statistics received only Rs. 710 crore (~$87mn) of funding in 2018-19 and Rs. 742 crore ($320 mn adjusted for PPP) for 2019-20. In the US, comparable institutions (like the Bureau of Labor Statistics) produce data more frequently, and are substantially better funded. Richer countries just have much more money to fund statistical agencies, and get better data. In the US, the BEA which produces GDP figures receives $141mn in funding, the BLS which produces CPI statistics and unemployment (far more frequently than India) got $1.7 billion in 2022. 

But  having statistics like the unemployment rate and inflation rate isn’t enough to make decisions about policy – that’s just the first step. Policymakers also need ideas about how to interpret the data, and understanding about causation in the data. For this, you need economists, sociologists, public health researchers, and political scientists. However, that base of knowledge production simply does not exist in India. When government committees look to academia for answers, they don’t get them. And this means that decisions are taken without the required knowledge to build a strong evidence base. Kelkar and Shah quote their personal experience on this:

Most committees and commissions that we have been associated with have been hard-pressed to find the data and research required to make recommendations based on sound evidence. In the absence of sound evidence, it is difficult to take sound decisions. The process of policymaking is always about decision-making under uncertainty, but in the absence of proper data sources and sound research, the uncertainty increases considerably.

For a moment let’s assume that these issues are solved. Let’s assume that policymakers have good enough data, and good enough knowledge about the world that they can make informed decisions. The next step that comes is actually implementing the decisions: raising money for them and executing the policy on the ground. These two steps are fraught with more potential problems. 

More money, more problems

For instance, policymakers might decide to impose a tax. One might think that the only cost of a tax is the value of the tax itself. However, the cost of a tax is not the value of the tax, but the welfare loss it imposes on the rest of society. Income taxes reduce the incentive to work, sales taxes reduce consumption, corporate taxes reduce investment, wages and corporate profits. It’s not only the changes in incentives for the overall economy. Taxes also change the allocation of goods. Parathas (a type of Indian flatbread) have a GST rate (India’s sales tax) of 18%, while rotis (another similar flatbread) have a GST rate of 5%. This misallocates resources in the economy by biassing toward rotis (even if ever so slightly), and these misallocations add up across the economy. 

High levels of taxation also encourage fraud, which ends up being an implicit tax on honest firms and a tax break on those who cheat. Government actions to solve this – like raids and litigation – can take up the mindspace of businesspeople, and reduce the motivation to run their businesses. Finally, at the margin, public spending is financed out of deficits which will end up (assuming interest rates are constant) being paid for through taxation years in the future. Deficits aren’t free from welfare loss either. 

When savings are scarce, governments compete with the private sector for them. In that case higher government investments can lead to lower private sector ones leading to private sector investments being “crowded out”. In India this is worse because deficits are financed by financial repression (i.e. financial institutions have to lend some percentage of their assets to the government, plausibly to keep financing costs low). Shah and Kelkar estimate (admittedly with little justification) that the total cost of each rupee of government financing is thrice the tax raised  – this means that government projects should only be done if the gains to society are expected to exceed three times their cost. Suddenly the policymakers’ tax, which made sense in theory, is at risk of being damaging in practice.

And then finally, even when a decision is made with poor information, poor knowledge, and financed with a high cost to society, the final constraint is state capacity. Some countries can do complex things like run very large vaccination programs and conduct the world’s largest elections in a free and fair manner. But other countries also have very high levels of corruption in basic state services, high levels of absenteeism among state employees providing crucial services all of which leads to bad outcomes. In fact, India is the country in all of these examples, illustrating the wide variety of capacities the Indian state has. Policymakers have to take this into account when formulating policy. There is no point in passing legislation they cannot enforce, writing ordinances nobody follows, or setting up regulations doomed to fail. 

Purvanchal Expressway. From @NidhiiTweets_

Can we solve it?

Shah and Kelkar have some interesting solutions to the problem. Their philosophical approach is liberal: they presume that markets should be left unencumbered by government intervention, except where to establish property rights and enforce contracts. In general they assume that freedom works well and coercion by the state should be minimised. 

But this doesn’t make them axiomatic libertarians. In this they are much like Gregory Mankiw, who wrote in Principles of Economics that governments can sometimes improve market outcomes, but one should consider government failure as well while considering market failure. As one might expect of a liberal, they are against paternalism, and have a high standard for redistribution.

Shah and Kelkar propose a public policy system. The first part of the system is to have a systematised process of data collection that provides policymakers with the data that they need. In their view of the world, Indian policymakers would have data comparable to that found in developed countries. India would have monthly detailed inflation reports, detailed GDP figures that are released rapidly, data specific to some industries or sectors and so on. 

Step 2 is to have a process where there is descriptive and causal research. Data is good, but finding information from the data is better. Shah and Kelkar have an insightful critique of the academic sphere in this regard:

This work should be primarily grounded in the Indian locale. Academic researchers are too often swayed by the curiosity of journal editors and referees in a different continent. This hampers the choice of questions to pursue and the quality of research design through which those questions are sought to be answered.

In their ideal world, there would be research on empirical topics in India. Questions like “how much does banning cars on certain days help with pollution?” would be easier to answer, and so academics and civil servants writing technical reports for the government would have an easier time finding evidence to recommend better policies. This of course would need several changes in the Indian policy environment. At first, it would require large increases in funding to social science departments in Indian universities so they could supply the talent that could perform such analysis, and the development of more think tanks and research institutes that performed these research. 

After you gather all the data and evidence, the next part is to create policy proposals. This isn’t always a straight line from having data to creating policy - you need a model of how the world works as well! Take this example from the book:

In the Indian experience, we set about creating a new private industry of electricity generation, without having solved the problems of distribution or of fuel. Very large amounts of capital are blocked in new-age generation plants, which are stranded with the lack of fuel and/or the lack of financially sound buyers for their electricity.

With the benefit of hindsight, we see a simple failure of sequencing in the reforms of this sector. Problems in distribution and the energy sector should have been addressed before or at least in conjunction with generation reforms.

If Indian policymakers had good research about the Indian electricity industry and an understanding of the bottlenecks in the industry, the policy proposals given by academics and civil servants to governments might have been better! 

After good policy proposals are made, agreed by the public, they should be implemented. But, implementing public policy in a country where the government isn’t very effective (like India) is hard to do. Policymakers should consider small solutions instead of big ones, and when they want large changes, they should try to scale it up slowly. One good example of this is the Insolvency and Bankruptcy Code. When it was enacted in 2016, there was little state capacity in the legal system to implement the law. Neither were there private players who knew how to litigate that subfield of the law.

But over time, the state built capacity to implement the IBC with institutions like the National Company Law Tribunal (NCLT) and the Insolvency and Bankruptcy Board of India (IBBI). Private law firms also started to specialise in this, leading to a market for their services where firms looking to file for bankruptcy could hire lawyers. Their main point is that in India, good public policy takes time. 

Getting data takes time, as does building a body of evidence. Creating policy proposals takes time too. 

The authors’ final point about their proposed system is that rushing things does more bad than good. Politicians want quick solutions, but those solutions aren’t always well developed. In fact many times they aren’t well thought through and cause more problems than they solve. 

Will it work?

Where I agree

There are some aspects of the book’s agenda that are unambiguously good. The first is the emphasis on data collection, and research to build evidence. Indian policymakers deal with much less data than is necessary to make a decision. Indian inflation numbers are noisy, and frequently revised. The most reliable unemployment numbers are provided by a private firm. Government numbers on the job market only capture the organised market (which is a minority of total employment), and GDP numbers have been called into question for the same reason. 

Economic journalists and investors have recognised the issue, and many investors are creating their own indicators combining multiple private sector datasets. But this is no substitute for a government statistics program that provides accurate and timely data on at least the very basic indicators. 

The other part I am unambiguously in agreement with is that policymakers should make policies in relation to existing state capacity. India isn’t a rich country, and the Indian state from the central government to municipal bodies do not have the capability to implement the same policies that those in rich countries do. The Indian state collects much less in taxes than developed countries (India’s was 17%, much lower than the OECD average of 33%) and consequently has less money to spend than they do. 

What should Indian policymakers do in this context? The book (albeit indirectly) recommends a two-pronged strategy. For the current period when state capacity is limited, policymakers should be more discerning when considering additional regulation and spending. There is little point in making laws the government doesn’t have the capacity to enforce or passing spending bills the government will waste on corruption. In the words of Kaushik Basu, Indian policymakers have to engage in ‘libertarianism of necessity’. But along with that, they should start building institutions that are able to enforce the laws they’d ideally want to have. 

One concept the authors should have discussed more in the book is the idea of ‘premature imitation’. Indian policymakers are in some ways more connected to the West than they are to their own people. Parts of the Indian elite speak English at a native speaker’s level and they are educated abroad in the UK or the US. (This is true for the authors of the book too). To quote Rajagopalan and Tabarrok (2019) which introduced the topic:

Among Indian elites, education abroad continues to be common, as do working and even living abroad. Since 1980, for example, only two of the eleven governors of the Reserve Bank of India were fully educated in India, whereas the remainder had British or U.S. graduate degrees. Many had extensive experience living and working abroad, including Raghuram Rajan at the University of Chicago and I. G. Patel, who later became the director of the London School of Economics. Similarly, since 1980, 100 percent of India’s chief economic advisers have had foreign graduate degrees.

This means that Indian policymakers are almost inevitably disconnected with the public they govern leading to them being unable to grasp the reality of state capacity in the country. One example Rajagopalan and Tabarrok cite is that of parking requirements in Indian cities:

Consider, for example, parking requirements. Parking requirements may seem like a sensible and natural requirement to the policy elite, but most people in India do not own cars. They use either public transportation or two-wheelers. Land set aside for parking in affordable-housing complexes is wasted, but, more importantly, if there were [sic] less space for parking, more units could be built, spreading the cost of the land more widely and reducing rents. Further reductions in rent are possible by reducing road widths, which can be done while still satisfying requirements for emergency vehicles. Reduced road width also increases pedestrian safety. In a careful and informed analysis, Bimal Patel, Sweta Byahut, and Brijesh Bhatha (2018) show that lifting a handful of the costliest requirements would reduce housing costs by 34 percent and increase supply by as much as 75 percent without any reduction in safety or quality of building materials.

Where I disagree

There are two specific problems the book’s proposed solutions have. The first is that the authors under emphasised the value of speed in making public policy.

The authors claim that there is a cult of speed. They say the politicians want solutions immediately and that can’t be solved because it takes time to think through solutions, and implement them. But even with this being true, they ignore the significant opportunity cost of taking things slowly. Every year 3% of Indian children under five die, most from preventable diseases. Every year about 5 million Indians enter the job market, and the majority of them quit looking for jobs because those are so hard to find. By 2030, about an eighth of India’s population will be above 60, a higher share than those under age 10. 

When politicians want quick solutions, they have a point. Every year that we delay solutions leads to more joblessness, more child deaths from preventable diseases, and a higher chance that India will not be able to provide for all those in its country. Policymakers should operate with a high degree of urgency. But the book does not emphasise this. It only explains the costs of speed, while ignoring the benefits. Many times it is not possible to think through problems the solutions haven’t been tried yet. Spending more time in committee drafts doesn’t solve this problem. It is also true that hastily written solutions can pose costs larger than the benefits they provide. But that doesn’t mean we should accept the time it takes for these to happen as set in stone! It should be a priority to speed up this process, so that we don’t miss out on helping the current generation of people. 

We must try to weaken the tradeoff between time and the quality of solutions. What are some alternatives? The solutions are found in the book itself. To the extent possible, policymakers should leverage foreign regulatory capabilities, learning from foreign solutions and adapting them to local circumstances. The authors probably have better solutions to this problem than I do. But my main point is that speed is not necessarily a vice, and being slow is not a virtue. 

Second, while Kelkar and Shah say they want to avoid an administrative state where there is rule by officials (instead of the rule of law), their solutions will lead to that very outcome. 

The authors admire the US, where the regulatory agencies make rules under authority given to them by the legislature and enforcement actions can be litigated in the courts:

…we in India look at the US Securities and Exchange Commission (SEC) and think of it as roughly analogous to SEBI. We often fail to see the capability of the US judicial system, through which ‘Administrative Law Judges’ (ALJs) hear cases brought by the US SEC. Hearings at the SEC are conducted by judges, not by SEC employees, and the separation of powers at the SEC is protected. This is made possible by the invisible infrastructure of a well-functioning legal system. We in India have an inferior arrangement, where the SEBI Act does not enshrine the separation of powers. SEBI employees who also perform legislative and executive functions, and lack judicial independence, are performing the judicial function. The lack of this invisible infrastructure—separation of powers and judicial independence—results in lower capability at SEBI when compared with the US SEC.

I agree with their assessment of the situation. It is deeply damaging to the rule of law that SEBI is a lawmaker, enforcer and the judge. But it is also deeply damaging to the rule of law for the same actor to both write the rules and to enforce them. Legislators often write the most broad rules possible, and let regulatory agencies decide the details. 

I’m sceptical of this because even in the best designed systems bureaucrats writing the rules have little incentive to get them right. When Parliament delegates the power to make rules about drugs to the CDSCO, or rules about the stock market to SEBI, the bureaucrats make the decisions. In some ways this is inevitable because market conditions change and so rules cannot be decided much in advance. In addition civil servants have the technical ability to make rules which Parliament lacks. However, a lack of proper incentives means that bureaucrats can build their own fiefdoms in the government and face no consequences for poor regulations. In the book’s own example, when the RBI makes poor regulations about the bond market, it faces no consequences for it. They quote Manish Sabarwal in saying that the RBI’s bond exchange, the Negotiated Dealing System (NDS), ‘has hostages, not customers’. The authors’ solution would give more power to specialised agencies, without feedback mechanisms needed to ensure their rules are good.

Take the recent example of RBI rules reducing the access companies have to consumer credit card data. This is a complex economic issue that affects a large number of stakeholders. The questions that underlie the policy about privacy, consumer protection and data sharing are a political question. It is for elected representatives to decide, not unelected civil servants who will face no consequences for their decisions. 

In other liberal democracies like the United States, the rule-making state faces the same problem. The FDA in the United States was extremely slow over approving COVID-19 test kits, and even stopped a study in the early days of COVID which was studying the spread of the disease. Alex Tabarrok makes a convincing case that FDA risk aversion cost thousands of lives during the pandemic. What consequences will they face from this? Unlike legislators who face periodic votes from their citizens, FDA bureaucrats face no consequences for their decisions. It is true that this is a matter of risk preference, and there are multiple reasonable solutions here. But what is not true is that this decision should be made by bureaucrats who face no costs of being extremely risk averse.

A common thread between the RBI and the FDA (and the Nuclear Regulatory Commission in the US) is that bureaucrats have little incentive to innovate. When things go well, they don’t get the limelight. When things go poorly, politicians blame them.

Improvements to the system

What solutions are there to this problem of bureaucratic risk aversion? 

One is to have an anti-regulator: an agency whose job it is to be adversarial and argue against new regulations. If you buy my thesis that regulators are structurally risk averse because of their asymmetric incentives, then one way to correct that is by having an agency that argues against new regulations. Being adversarial isn’t new in government. We use it in the legal system as a means of getting to the truth and protecting the defendant’s innocence. To quote the blogger Rohit Krishnan:

One of the biggest issues we deal with is the fact that laws and regulations are passed whenever something acute happens. There's an epidemic of false advertising, FDA passes laws. There's a financial crisis, SEC passes more laws. There's thalidomide poisoning … you get the picture.

The anti regulator is different from the Law Commission in the sense that its primary job is to be adversarial and recommend reductions in rules by agencies and the government. One good case study is the Sunset Commission in the state of Texas, USA which has:

eliminated 29 types of licenses that provided little or no benefit to the public and were held by 160,000 businesses and individuals, such as polygraph examiners and combative sports event coordinators. Sunset also streamlined the regulation of about 345,000 barbers, cosmetologists, and their businesses to reduce both the department’s workload and regulatory burdens on the licensees.

My next proposal is to have automatic repeal built into government laws and government agencies. Too often, government rules are outdated: Indian information technology rules are under the Information Technology Act 2000, wireless communications under the Telegraph Act 1885 (no, that’s not a typo) and procedures for criminal cases under the CPC 1973. All of these have been amended over time, but it is important to check the base value of them frequently. Stasis often sets into government bureaucracies, and it is important to have a structural force towards making government smaller. The Sunset rules in Texas allow for an agency to be abolished 12 years after it is passed unless the legislature re-established it again. Several US states have sunset laws that require the government to question again if government agencies and laws are still required. 

Conclusion 

In the end, what do I make of the book’s arguments? Despite my criticisms above I think that if the book’s policy proposals are implemented, policymaking would be much better. The government shouldn’t be short of data. We shouldn’t be making public policy based on guesses and intuition instead of research. We shouldn’t be having inefficient taxes that stifle industry, and we shouldn’t be making large complex decisions on the whims of politicians. 

In the authors’ ideal world, none of this would happen and my objections would be relatively minor when compared to the benefits of following their advice. 

But one lesson of the book is to work with the resources you have, not the ones you wish you had. So in the meantime we’re going to have to deal with more poor public policy. The only hope is that more people read In Service of the Republic and push for the changes that make the world a better place.