«I. Introduction Green growth can be defined as a trajectory of economic development that fully internalizes environmental costs, including most ...»
In all these cases, there are strong a priori, theoretical justifications for policy intervention, but inconclusive empirical evidence on whether policy works “on average.” However, debates typically focus not on whether the governments should have active policies in these areas but on how the requisite policies should be designed – whether government should run schools or simply finance them, or the appropriate mix between monetary and fiscal policies, for example. Green industrial policy needs to be approached in the same manner, as an important government function, that can be carried out better or worse. The useful debate to be had is not whether green industrial policies should exist but how they should be designed.
A serious debate about the design of industrial policy would bring it out of the shadows and allow it to be carried out in an explicit manner. It would save it from being carried out surreptitiously, as an appendage to other governmental functions and hostage to related, but distinct objectives such as employment or competitiveness.
Industrial policies must be built on three key ideas. First, the requisite knowledge on the existence and location of technological spillovers, market failures, and constraints that impede green
communities. Second, private investors and others who are the beneficiary of public support have strong incentives to “game” the government, by bending the rules to their advantage through their informational advantage and political muscle. Third, the intended beneficiary of industrial policies is neither bureaucrats nor business, but society at large.
Each of these ideas has specific implications for the institutional design of green industrial policies. We can summarize these implications as (i) embeddedness; (ii) discipline; and (iii) accountability. I take up each design principle in turn.
(i) Embeddedness When economists think about optimal policy design in the context of industrial policy, they typically hone in on models of regulation based on agency theory. In these models, the principal (the regulator) aims to alter the behavior of an agent (the firm) to pursue a public objective (an investment of a particular kind). The central feature of the setup is that the agent has some private information (e.g., its costs). In light of this asymmetry in information, the principal has to cede “informational rents” to the agent, and cannot obtain its unconstrained (i.e., perfect-information) optimum. Effectively, the principal has to offer the agent a reward to dissuade it from mimicking a less efficient counterpart (Laffont and Tirole 1993).
In this top-down model of interaction between the bureaucracy and business, communication between the two is neither assumed, nor required. From the standpoint of minimizing rent-seeking and lobbying, this can be viewed as a distinct advantage. The policy maker need not consult with business people, and can keep them at arm’s length. Any direct interaction is unnecessary from the standpoint of carrying out the public purpose and hence can be judged as illegitimate. The autonomy that is built into the framework insulates the bureaucracy from pressure from below and protects them for it.
By the same token, the principal-agent perspective severely limits the flow of information
uncertainty is multifaceted and may often take Knightian characteristics. The agency framework assumes the principals already have a very good idea of what needs to be done to achieve public goals, and all that needs to be done is to provide the agents (firms) with the right incentives to carry out the requisite investments. But as Charles Sabel puts it, “what if … there are no principals … with the robust and panoramic knowledge needed for this directive role? Then the problem for reform is at least as much determining ways actors can discover together what they need to do, and how to do it, as determining which actors ought to be the principals in public decision making.” (Sabel 2004) The principal-agent model presumes the existence ex ante of a well-defined social objective function
-- well-defined in the sense of not just what is being maximized, but also the types of instruments and strategies that are available. Businesses cannot communicate information about the constraints they face other than through their actions. Neither can they communicate directly any new opportunities that may arise, or advance proposals as to how these might be pursued with the help of the public sector.
An appropriate industrial policy framework needs to make room for learning by government officials on all these counts. That in turn requires a significant amount of interaction and communication between the public and private sectors. This is what the term “embeddedness” refers to. It was used first in the industrial-policy context by Peter Evans (1995), who described South Korea’s developmental state as one in which the bureaucracy exhibited “embedded autonomy.” The South Korean bureaucracy, he argued, operated along Weberian, meritocratic lines, but it was not insulated from the private sector. Quite to the contrary, it was “embedded in a concrete set of social ties that binds the state to society and provides institutionalized channels for the continual negotiation and renegotiation of goals and policies” (Evans 1995, p. 12). As Evans wrote, “[a] state that was only autonomous would lack both sources of intelligence and the ability to rely on decentralized private implementation” (1995, p. 12).
Clearly, the embeddedness that is required is one that falls far short of bureaucrats being
in bed with, business. The right model lies between arm’s-length and capture. It is one of strategic collaboration and coordination between the private sector and the government with the aim of learning where the most significant bottlenecks are and how best to pursue the opportunities that this interaction reveals. There are multiple institutional settings within which this kind of collaboration can occur: deliberation councils, supplier development forums, search networks, regional collaborative innovation centers, investment advisory councils, sectoral round-tables, private-public venture funds, and so on.
This way of looking at green industrial policy highlights another important implication: the right way of thinking about it is as a process of discovery, by the government no less than the private sector, instead of a list of specific policy instruments. This perspective focuses attention on learning where the constraints and opportunities lie and responding appropriately, rather than on whether the governments should employ tax breaks, R&D subsidies, credit incentives, loan guarantees, and so on. It is important, of course, to evaluate the effectiveness of these specific instruments. But the prior, meta-question on green industrial policy is whether a government has put in place the appropriate processes and institutions of engagement with the private sector.
(ii) Discipline The embedded nature of green industrial policy makes the need for disciplining devices against abuse all the more imperative. Firms and industries that receive help from the government must know that they cannot game the system, and that underperformance will result in the removal of assistance. Carrots must be matched by sticks. This was indeed a key ingredient of East-Asian style industrial policy. In South Korea firms that did not meet their export targets saw their subsidies cut, and in some cases even became targets of government recrimination in the form of aggressive tax audits, for example.
In democracies, discipline has to take a form that is different than the one in which it came
institutionalized. But since each case is different and the nature of green technology is inherently uncertain, a certain element of discretion is unavoidable. The trick is exercise discretion in a manner that can be justified by the facts on the ground.
A principled discipline requires first and foremost clarity in objectives. If the objectives of a program to support green technologies have not been explicitly specified ex ante, it will be difficult to know whether the program is working or needs revision. This seems to be a rather obvious principle, but it is frequently flouted in practice. For example, the DOE loan guarantee program was touted on account of its contribution variously to jobs, global competitiveness, technological benefits, external spillovers, and contribution to curbing climate change. It is certainly possible to meet one or more of these objectives while failing on account of the others. Technological benefits and spillovers can be reaped without attaining competitiveness at the same time, as was the possibly the case with Solyndra. Similarly, jobs may be created without gains either in technology or competitiveness.
Politicians may naturally want to kill multiple birds with one stone. But multiplicity of goals – or confusion about them – does not contribute to discipline. It becomes possible to justify any range of results after the fact, by latching on to the least problematic aspects of performance. The greater the multiplicity of goals and the hazier their definition, the less the ability to recognize failure, remove support, and change course.
What then are appropriate goals for green industrial policy? As discussed in the introduction, public support is justified by the need to foster private investment in green technologies and contribute to reduction of GHG emissions, in view of the likely market failures in both areas. This is a largely technological goal. It has to be distinguished from employment creation, competitiveness, profitability, and other commercial aspects. A promising new technology may be worth supporting even if it does not generate many jobs; employment objectives are better served
bankrupt; if the technological learning and spillovers from the pioneer spawn a new industry, its own commercial failure is of little consequence.
Unlike jobs and commercial profitability, however, a technological objective is very difficult to monitor. Within firms and industries, probably the best single observable indicator would be cost.
The progress of, say, a solar-cell firm in meeting its technological objectives can be measured by its rate of cost reduction in producing energy. Therefore cost-reduction targets make much more sense in general than employment or investment targets. (Interestingly, by this measure, Solyndra’s performance was quite solid.) But other measures of technological development can be used as well, many of which necessarily require judgment and discretion. Patenting activity, cluster development, and indicators of beneficial spillovers to other firms can all be scrutinized. Ultimately, monitoring overall technological progress needs to rely on periodic audits by professional specialists who can render their independent judgment.15 The next step is evaluation, which goes beyond tracking observable indicators and monitoring performance. Evaluating whether a program is meeting its objectives requires an explicit counterfactual: what would have happened in the absence of the program? Technically the most sophisticated (and credible) evaluations of public programs are based on randomized trials, regression discontinuity, or instrumental variables methods (Jaffe 2002; Van Reenen 2013). The first of these approaches compares firms that were randomly selected to receive support with those that did not. The second compares outcomes just above and below the threshold of qualification for public support. The third identifies the program effect through an exogenous component of variation in qualification. These techniques can be quite useful in some settings, when the intervention is relatively specific and the potential number of beneficiaries is large. Criscuolo et al.
As Jaffe et al. (2004, 14) note for the U.S.: “systematic assessment efforts are woefully lacking. Because success is uncertain and difficult to measure, most agencies engaged in support of research and technology adoption have resisted efforts to measure their output against quantitative benchmarks, as is required in the United States by the Government Performance and Results Act” (reference omitted).
-26provide a useful application to regional state aid in Britain. As Jaffe notes (2002), building such evaluation protocols into support programs from the outset are an important safeguard.
But in many instances, such techniques are not easily applicable, either because of small numbers of support beneficiaries or because the program components differ too much across recipients to render comparison meaningful. Inability to undertake evaluations that are rigorous enough to satisfy journal referees should not stop governments from implementing certain monitoring routines which are useful early-warning signals and can flag blatant program failures. A particularly useful practice would be to establish explicit cost, productivity and other targets ex ante.
Such targets cannot eliminate the influence of unforeseen, exogenous changes (such as unexpected technological and market developments) that take place once the project is rolled out. But at least they allow outcomes to be evaluated against a particular benchmark – the baseline established by ex-ante expectations. Significant under-performance relative to that benchmark would then call for either the abolition of support or an explicit countervailing argument as to why unanticipated developments warrant continued support.
As long as there remains fuzziness about objectives, targets, and results – which seems inevitable, in light of the nature of green industrial policies – firms will always try to make a case for continued subsidies – either before the program agency or through political lobbying. As Matsuyama (1990) shows in a related context, the threat to remove support when a firm does not perform as expected is often not credible. Appropriate rules can help alleviate the dynamic inconsistency that bureaucracies face in these circumstances. For example, automatic sunset clauses would reverse the burden of proof by requiring positive action to renew support schemes, and make it harder for failing projects to be propped up. The requirement that agencies must provide an explicit accounting – preferably of a public kind – for continuing support when initial targets are not met would raise the bar similarly.