1. We should embark on things that we can do (i.e. don’t take on things that we don’t have the ability to do); and
2. We should walk before we run (i.e. do simple things, achieve victory, then move on to a more complex problem).

These simple and obvious ideas have interesting implications for how we think about doing public policy and public administration in a place like India, where there is a crisis of State capacity, where numerous policy initiatives break down in the implementation.

Premature load bearing

Pritchett, Woolcock and Andrews, 2010 talk about `premature load bearing’ as a source of implementation shortfall. Their metaphor is a bridge that is built to a certain limited capability. If you run a truck on the bridge which has a weight that is too high, the bridge comes crashing down. In this example, the concept of `load’ and `load bearing capacity’ is quite clear.

Example. We commission a government facility that registers land transactions. This is inadequately sized. The staff collapses under the crushing pressure of a large number of transactions. A black market develops where some people pay bribes to get ahead. The staff is partly super-busy fire-fighting, and have no time to think about fixing the broken system. The staff is also happy to receive a steady flow of bribes and lacks incentive to fix the broken system. The load was too high, the capacity was inadequate, and the system collapsed.

What is load in public administration?

With a bridge, the load is clear: the mass of the vehicles that run on the bridge. How do we think about load in the public policy context? What are easy problems vs. what are difficult problems? Pritchett and Woolcock, 2003 suggest two dimensions of load: high transactions and high discretion. The example above (land titling office that collapsed) is a simple example where the question was of transaction processing capability. That is an easy one to comprehend. But we should look beyond this simple engineering perspective of counting the number of transactions. A public administration problem is more difficult when front-line officials have more discretion.

There is a valuable third dimension to thinking about load, which is to think about the stakes. What is at stake? What could a corrupt official stand to gain? It is easier to run a system where the stakes are low. As an example, the personal gains to a school teacher from being absent half the time at school are relatively small. While it’s hard to make school teachers show up to work and to teach well (it is a transaction-intensive discretion-intensive service), this is not that hard, as the stakes are low. But the gains to a bad tax official can be 1000 times larger than the gains to a bad school teacher.

Thinking about the incentives of the civil servant takes us to thinking about load that comes from the magnitude of the principal-agent problem between the objectives of the organisation and the objectives of the individuals that man it. The load that is placed upon a system is the extent to which the objectives of the individuals diverge from the objectives of the organisation.

Example: Parking enforcement

Every economics student is exposed to the problem of enforcing self-service parking meters. Cars are supposed to pay Rs.10 for parking. Suppose we do not police compliance pervasively. Suppose there is only a 0.01% probability of getting caught. Let’s set the fine at Rs.1000. Risk averse persons will then prefer to pay the fee for sure (i.e. pay Rs.10) instead of taking the risk of losing Rs.1000.

The attractive thing about this fable is that it shows us the path to a small traffic police force. Instead of having a large number of policemen watching all cars, we can get by with limited enforcement. We can have a small government and yet get the job done. At first blush, you would think things are always easier in public administration with a small police force, which calls for a smaller number of transactions.

The story changes significantly when we worry about the principal-agent problem between the police department and the front-line enforcer of the fine. Do we have the ability to create checks and balances where the police will actually collect a fine of Rs.1000? Or will this collapse in pervasive corruption?

The magnitude of the fine is the load that the system is placed under. When the fine is small, e.g. Rs.100, the policeman has the choice of taking a bribe of Rs.50 and catering to his personal interest, or insisting on the fine of Rs.100. This is a small divergence between self-interest and the objectives of the organisation. But when the fine is Rs.1000 or Rs.10,000, the gap between the two enlarges. The system is placed under greater load.

Suppose the user charge is u, the probability of getting caught is p1and the fine is F. In the textbook, we try to make p1 small, so as to have a small police force, and compensate by ratcheting up F=u/p1. However, large values of F are a large load upon the system. We have to ask ourselves whether we have designed a public administration mechanism that is able to deal with a large F.

What is load bearing capacity?

We are asked to build a bridge that will be strong enough to take 10 main battle tanks weighing 60 tonnes each. Now we must pull together an elaborate array of design features in the bridge, so that it is able to cope with this load.

In similar fashion, once we know about the number of transactions, the extent of discretion of front-line officials, and about the stakes, we have a characterisation of the load. What are the elements that shape load bearing capacity?

The simplest question is transaction processing capacity. If there are 100 land market transactions a day, we must build a facility that will have commensurate capacity. The second dimension is discretion: if systems can be designed which reduce discretion, this will increase the load-bearing capacity. In some situations, IT systems can remove discretion and thus remove one dimension of the load.

The most important element of load bearing capacity is to think about the stakes, i.e. the maximisation of the official. Public administration is about establishing processes so that the organisation achieves its goals even though individuals have divergent personal interests. The task of management is to reshape incentives so that the narrow self-interest of employees gets aligned with the objectives of the organisation. The quality of the processes, and the checks and balances that have been designed, determine the load-bearing capacity.

Let’s go back to the parking fine problem. What shapes the thinking of the policeman? There is a probability p2 that he gets caught if he asks for a bribe. From his point of view, if p2 is high and F is low, then it’s safer to just enforce the fine. As the fine gets bigger, he is more tempted to ask for a bribe. Under good conditions of public administration, p2 is high. If a cop in London asks for a bribe, there’s a good chance that he gets caught. Our job in public administration is to build the checks and balances so that p2 goes up. The load-bearing capacity of the system is reflected in p2.

There is a tension between making a problem easier by reducing the number of transactions vs. making a problem easier by reducing the stakes. E.g. when a parking fine goes from Rs.1000 to Rs.100, the number of fines (i.e. the number of transactions and citizen-interfaces) has to go up by 10 times.

Consequences of premature load bearing

When an organisation is asked to deal with load that goes beyond its load-bearing capacity, what results is a rout: `a collapse of organisational coherence and integrity’. While lip service to the goals of the organisation continues, on the ground, there is an every day reality of a large divergence between the behaviour of individuals in the organisation versus the objectives of the organisation.

Once an organisation collapses in this fashion, it shifts into a low level equilibrium of pervasive rent collection. All that goes on is the abuse of coercive power of the State in favour of laziness and corruption by the persons manning and wielding the instruments of power. These rents can often get ingrained into a new political arrangement, and political incentives for preservation of the status quo. Pritchett et. al. thus encourage us to watch out for premature load bearing, particularly because it can lead to sustained and persistent implementation shortfall, and create an incumbent set of players who are the biggest obstacles to fixing things.

We see this with entrenched bureaucracies in many areas in India. Premature load bearing led to an organisational rout, and in the wreckage we now have incumbents that man the State machinery who are zealously defending the corruption and laziness.

Walk before you can run

We should build the parking enforcement system through the following constructive strategy:

1. First, we would setup a parallel and independent measurement system to obtain data about the extent to which cars are not paying the user charge and the perception in the eyes of citizens that when they are caught, they pay a bribe instead of the fine.
2. Next, we would build a large police force (i.e. high p1) and a low F. We would put down all kinds of monitoring and checks and balances, in order to overcome the principal-agent problem. We would design accountability mechanisms to create pressure on the leadership and at every level of the police force, so as to get p2 up.
3. We would fight with implementation shortfall until the survey evidence shows us that the offenders are paying the fine and not the bribe.
4. Only then would we announce We know how to run this system for a certain F and p1.
5. Only then would we take the next step, of reducing the police force by 25% and increasing the fine to 1.25F. This will be assisted by behavioural changes among the police and citizens, who would have gotten more ingrained in good behaviour, where misbehaviour is more likely to set off alarms.
6. We would do this, one step at a time, pushing up the fine by 25% at each step, and continuously watching the survey evidence to look at the incidence of bribe-taking. At a step where the survey evidence shows that bribe-taking has gone up unacceptably, we would stop and undertake deeper reforms of the public administration mechanisms so as to push up p2 until the extent of bribe-taking goes back to an acceptable level.

The four hardest problems in State building

The stakes are sky high in four areas:

• The criminal justice system,
• The judiciary,
• Tax collection, and
• Infrastructure + financial regulation.

In these four areas, the personal gains that staffers in government can get, in return for sacrificing the objectives of the organisation, are thousands of times larger than their wage income. They all involve a large number of transactions, and there is inescapable discretion in the hands of front-line officials. Here, creating the public administration machinery to make civil servants behave correctly is the hardest. These four areas are the most challenging problems in State building.

This has two implications:

1. Particularly in these areas, we should learn to walk before we can run. At first, policy pathways should involve low load. In order to do this, we should push towards low transaction intensity, low discretion and low stakes.
2. These four problems should take up the highest priority as the big hairy audacious goals of State building. The top management has to prioritise time and resources for these big four problems.

Example: Punishments in the criminal justice system

Every now and then, we have outrage in India about crime. After a great deal of hand wringing, the outcome too often is: Let’s increase the punishment.

The criminal justice system (laws, police, prosecution, prisons, courts) is one of the hardest problems in public administration. The policeman and the public prosecutor are able to talk with the accused and threaten that if the law is enforced, a certain punishment will flow, and ask for a bribe in exchange for not enforcing the case. The bigger the punishment, the bigger the divergence between personal benefits and the objectives of the organisation.

Ratcheting up punishments in response to failures of enforcement is thus precisely wrong. The criminal justice system is failing when given low load (e.g. 2 years imprisonment for rape). If the load is increased (e.g. 4 years imprisonment for rape) then this places a higher load upon a broken system. This will result in an inferior criminal justice system.

First we have to make the criminal justice system work with low punishments. There is a lot to be done, in building the criminal justice system, with reforms of laws, lawyers, police, public prosecutors and prisons. We should keep punishments low, and make this work in terms of processes, delays, arbitrariness, etc.

Only after that can we consider the tradeoffs in higher punishments. This consideration trumps all others. You may have a strong moral belief that a certain crime deserves a certain punishment. You may be able to demonstrate that the minimum level of punishment required to achieve deterrence is quite high. You may see mainstream practices in developed countries and think that gives a ballpark estimate for what a certain punishment should be. All these considerations are irrelevant. The maximal punishment that should be used is the one that we are able to pull off, in terms of the load bearing capacity of the criminal justice system. Only after we have established a high load bearing capacity can we bring other considerations into play, and potentially ratchet up to larger punishments.

There are other good arguments in favour of low punishments. One is Occam’s razor of public policy: we should desire the lowest punishment which gets the job of deterrence done. James Scott has a meta-principle: Prefer to do things that you can undo if you discover you were wrong. He has a footnote on this saying this is a good reason to not have a death penalty. A certain fraction of people that we convict are always Type 1 errors (innocent but convicted); the harm that we impose is lower when punishments are lower.

Example: Design of the Goods and Services Tax in India

Satya Poddar and I have an article on `walk before you can run’ applied to the design of the Indian GST.

Example: Armed forces

An article from October 2016.

Example: bankruptcy reform

An article from June 2017.

Conclusion

Most management is about principal-agent problems. Most public administration is about the principal-agent problem between citizen and State employees. To be a public policy thinker, we have to ask three questions. What’s the market failure? What’s the minimal intervention that can address this market failure? Do we have the ability to build this intervention, in the light of public choice theory? The third test is a big barrier in India. Many things that sound reasonable and are done by many countries are not feasible in India. We in India have yet to learn about establishing accountability mechanisms through which we obtain high performance agencies. We are at the early stages of this journey.

As Kaushik Basu says, there is libertarianism of choice and there is libertarianism of necessity. All too often, in India, the right answer is to do less in government, out of respect for limited State capacity. I think of `do less’ as two dimensions.

The first is to not do certain things at all e.g. it’s impossible for India to build unemployment insurance. This will permit scarce resources (money, management capacity) to be focused on a few core issues. As an example, Jeff Hammer emphasises that in the field of health, we should use our limited State capacity to emphasise public health.

The second is to get started with low loads. In an environment of pervasive implementation shortfall, we should first score victory with policy pathways that place low load upon public agencies. We should ask for high performance organisations which are able to deliver results when given easy problems : low transaction intensity, low discretion and low stakes. Only after we know how to walk should we consider running.

There is very limited capacity in terms of the top management which will build and run these systems. We have to focus on the few highest priorities that are worth pursuing. The big four are: criminal justice system, judiciary, tax collection and infrastructure + financial regulation. They should be the top priority in State building, and have the highest claim on scarce top management time and resources. In each area of these four areas, our strategy should be:

1. Build an overall strategy;
2. Create independent measurement so as to track the performance of the system e.g. survey-based measurement of the incidence of corruption in tax administration;
3. Rescale the objectives of policy to minimise the load : i.e. favour pathways that involve low number of transactions, low discretion for officials and low stakes.
4. Design for load-bearing capacity. This is about adequate sizing for the required number of transactions, setting up processes which reduce discretion, and creating checks and balances to get accountability. This involves many design elements, as has been done for macro/finance by FSLRC: clarity of objective, minimising conflicts of interest, limited powers and discretion, precision of rules, procedural and transactional transparency, accountability mechanisms including the rule of law.
5. Achieve demonstrable success on low load problems;
6. Consider increased load only after success has been achieved at low load.

How does this link up to the debate about small steps versus grand schemes, the tension between  incrementalism and transformative change? We should do transformative change like the GST, but we should start at a low load GST — low rate, flat rate, comprehensive base. We should first learn how to build the load-bearing capacity for this low-load GST, before contemplating a higher rate or multiple rates or a Balkanised base.

Acknowledgements

I am grateful to John Echeverri-Gent, Lant Pritchett and Vijay Kelkar for stimulating conversations on these issues.

References

Capability traps? The mechanism of persistent implementation failure. Lant Pritchett, Michael Woolcock, Matt Andrews. Working Paper, 2010.
Has a section on premature load bearing.

Solutions when the solution is the problem: Arraying the disarray in development. Lant Pritchett, Michael Woolcock. World Development, 2003.
Introduces the 2×2 scheme for classifying problems as hard when there is discretion and transaction intensity.

Improving governance using large IT systems. Ajay Shah. In S. Narayan, editor, Documenting reforms: Case studies from India, Macmillan India, 2006.
IT systems can be used to remove discretion and thus make some problems easier.