Policy as Experiment

Policy-makers often claim that they understand the problem and know the solution. But policy-making is often like moving into uncharted territory. Policy-makers may have inadequate understanding of the problem they are trying to address and even less of an idea of whether the proposed solution would work. 

They face incomplete information, rapidly changing conditions, and heterogeneous responses across firms and sectors. This is particularly true in periods of turbulence.

For Albert Hirschman, one of the most insightful thinkers in economics, this is not a huge problem. Many years ago, in his 1958 classic The Strategy of Economic Development, Hirschman argued that development, and policy-making, is a lot about experimentation.

Policy-makers don’t start with perfect foresight, but with imperfect information and limited understanding. They learn as they start implementing their policies.

Thus, in the Hirschmanian world, what is important is sequential discovery. Policy-makers frequently discover constraints only after implementation begins. As policy interacts with practice, unanticipated problems crop up. But so do solutions, especially when a government is willing and equipped to learn.

Hirschman is not the only one who thought that way. There is a whole body of literature on evolution economics that analyzes how both firms and governments learn through adaptation under uncertainty.

In brief, excessive insistence on perfect design or waiting for complete knowledge can lead to policy paralysis. The more realistic approach may be to act, observe carefully, learn from outcomes, and progressively improve policy design.

Let us illustrate with a concrete example.

Bangladesh has recently announced a substantial stimulus package aimed at supporting enterprises facing severe financial stress. Consisting of subsidized loans and refinancing facilities, the package comes in the light of widespread concern about industrial stagnation, job losses and factory closures.

The hope is that the stimulus package will help turn things around -- factories will reopen, capacity utilization will increase, industrial growth will resume, jobs will be created or at least saved, and all this will help contain inflation and revive growth.

These are all noble objectives. But does the government know exactly what it is doing and what impact this policy might have in practice? The package seems to be based on a particular diagnosis of the problem, i.e., the economy is in "balance-sheet recession," with firms facing intense balance sheet stress.

To put it in simple terms, firms are finding it difficult to service their existing debts and thus reluctant to take on new loans to revive production. Some firms may be over-leveraged, having over-borrowed in good times.

Some may have been prudent in the past, but now tight economic conditions (such as poor domestic demand, import restrictions, energy shortages, exchange rate uncertainty, and broader macro-economic instability) have reduced their sales and earnings. So, they too find it difficult to service debts.

Businesses have become conservative, increasingly focused on preserving cash and servicing existing obligations rather than undertaking new investment. But many businesses are inherently good, and a little liquidity support may help them cross the hump. That is what the stimulus is intended to do.

It is possible that the diagnosis is correct. But even if it is, there could be legitimate concerns about how the support program will play out. Policy-makers may not have a clear idea about the characteristics of the transmission mechanisms (i.e. how a policy intervention impacts the behavior of economic actors and thereby transmits its effect through the economy).

They may have a vague idea of the implementation capacity of the relevant agents, such as the banks, and can not fully predict the behavioral responses of the intended beneficiaries, i.e. the enterprises.

In other words, the stimulus package will be implemented under considerable uncertainty. But that is what makes it interesting. This is not merely a macro-economic intervention. It is also potentially an important opportunity for the Bangladeshi government to learn something valuable, i.e. how to do adaptive policy -- policy-making under uncertainty.

So, what exactly is this? And why is this valuable?

As with other policies or programs, many things will be revealed as the implementation of the stimulus program proceeds. We will see different kinds of responses at the firm level. Some firms will use the funds as intended and revive production.

Some may genuinely try to put the funds to good use but will be unable to do so because other factors, such as energy shortages, are preventing them from reviving or expanding production. But there may also be firms who borrow the funds but don’t deploy these to revive production, diverting these to other uses instead.

The government may discover that some firms primarily need working capital while others need debt restructuring. It may find that export-oriented SMEs respond differently from large conglomerates. It may discover that firms integrated into diversified supply chains recover more rapidly than those dependent on narrow markets.

There may be variations among the banks too. Some banks may demonstrate strong capacity to identify viable firms. But others may simply refinance politically connected borrowers.

How well the government notices these variations will depend on how good their feedback mechanisms are. With good feedback mechanisms, the government will get insights that will help it to better craft the next round of policies. The goal would not merely be disbursement of funds but systematic generation of knowledge.

Which firms recover quickly after support? Which sectors respond most dynamically? Which financing instruments work best? Which banks allocate funds more effectively? Which firms resume exports, retain workers, or restore profitability? Which enterprises continue to deteriorate despite assistance?

In other words, the proposed stimulus package may be viewed not as a one-time finalized solution but as the beginning of an adaptive policy process. This is true of other policies or programs that are being rolled out or contemplated.

Take for instance the Finance Minister’s proclamations about initiating a substantial deregulation program. The intentions are good -- after all regulatory hassles and uncertainty are frequently mentioned by businesses as huge problems. But there are numerous examples from Bangladesh and abroad of how well-intentioned deregulation programs are derailed.

Bureaucrats who lose power or the opportunity to take bribes due to deregulation may take actions to dilute the programs. We must also remember that regulations are there for a purpose -- ensure workplace safety, protect workers’ rights, prevent harm to the environment, safeguard consumer welfare, etc. Deregulation runs the risk of diluting these protections.

Here too, the government will be moving into uncharted territory. Thus, rather than trying to engineer a perfect deregulation program from the start, the government could design a program based on some knowledge of global good practices adapted to local conditions, and start implementing it. But then it should monitor the implementation experience carefully and take corrective actions where necessary.

Adaptive policy-making does not happen automatically. Governments do not learn merely because they implement programs. Learning requires institutions, systems, incentives, and deliberate effort. Without these, experimentation can degenerate into ad hoc improvisation.

First, and foremost, we need a sound monitoring and feedback system. Collecting relevant data in timely fashion is key. Policy-makers need timely, and fairly granular, information on how firms are performing, what is happening to jobs, whether loans are being repaid and capacity utilization increasing.

This will require a "whole-of-government" approach, since different parts of government collect data on different aspects of the economy. Non-governmental players, such as commercial banks, industry associations and individual businesses will have to join the effort. In fact, willingness to provide the required data could be a condition of support.  Data collection should not be treated as a secondary bureaucratic exercise.

The government may also encourage some variation in implementation. Consider the stimulus package. The same financing structure and eligibility criteria may not work for all sectors or enterprise types. Some variation may be required; but one should also ensure that there isn’t scope for undue discretionary decisions.

In other words, we may need diversity in policy implementation with adequate discipline. That way, diversity itself can become a source of learning. If every intervention is identical, policymakers learn less about what works.

We must also recognize that some approaches may not work. Policy makers are often reluctant to accept failure. Political incentives are such that certainty and decisiveness are rewarded. But adaptive policymaking requires some degree of political tolerance for partial failure, course correction, and experimentation.

Adaptation is not new to Bangladesh. Many successful Bangladeshi enterprises did not emerge from detailed master plans. They evolved through experimentation, improvization, adaptation to shocks, and gradual capability building. Firms learned how to survive amidst abysmal infrastructure, policy and regulatory uncertainty, financing constraints, and volatile markets.

The challenge now is whether the state itself can become more adaptive and evolutionary in its policymaking. The most successful developmental states were not necessarily those that possessed perfect knowledge from the outset. Rather, they were states capable of learning rapidly, adjusting policies pragmatically, and building institutional memory over time.

An adaptive model of policy implementation has become more urgent now as the world becomes more uncertain, the global economy more volatile and the local economy rapidly changing. Economic management in such a world cannot rely solely on fixed blueprints and static assumptions. It requires continuous learning, experimentation, and revision.

It is not that policy makers in Bangladesh have not been adaptive at all. We do have some examples of adaptive behavior within government. For example, rural electrification, driven by the Rural Electrification Board, did not follow a rigid centralized model alone.

The cooperative-based Palli Bidyut system evolved over time through experimentation. And when the Infrastructure Development Company Limited (IDCOL) rolled out the solar home systems, it learned from its implementation experience to gradually modify financing mechanisms, technical standards, partner organization structures and subsidy design. Field-level feedback was critical.

We see examples of such approaches in other sectors as well. The garment industry has benefited from policy measures such as back-to-back LCs, bonded warehouses, export incentives, and customs simplifications. But these did not come in one go. They were introduced incrementally as policy-makers learned better what exporters needed.

But, often, such adaptive practices have happened in a somewhat ad hoc, informal manner without structured monitoring, evaluation and feedback mechanisms in place. And, in any case, these have been exceptions and have not been mainstreamed.

It is time for Bangladesh to move in that direction and mainstream the practice of adaptive policy-making.

Syed Akhtar Mahmood is an economist, previously with an international development agency.