Notes

Chapter 1. Introduction: Toward Experimentalist Governance

1. Benedick 1998; Parson 2003.

2. Hoekman 2016; Hoekman and Sabel 2019.

3. Dewey 1927, 207.

4. Cullenward and Victor 2020.

5. Experimentalism shares with the work on governing the commons, associated with Elinor Ostrom, the assumption that local knowledge is indispensable to the solution of a broad range of complex collective action problems. Ostrom 1990.

6. In this sense, our arguments overlap with the literature on adaptive governance and active learning. Lee 1993; Social Learning Group 2001.

7. Alter and Meunier 2009; Drezner 2009; Raustiala and Victor 2004. Related are literatures that see the process of governance being forced open to accommodate many different groups. See, for example, Hoffmann 2011. That study also points to the value of experimentation, although it has a causal logic around experimentation quite different from ours.

8. Hale, Held, and Young 2017; Biermann et al. 2009; Abbott et al. 2012.

9. Barrett 1991. For a time, these debates played out as well through proposals for a World Environment Organization. See the review in Biermann and Bauer 2005.

10. Such surfeits exist in many areas of international governance and help explain the observation, made long ago, that international governance often takes the form of a “regime complex” of partially overlapping regulatory regimes. Alter and Meunier 2009; Drezner 2009; Raustiala and Victor 2004. Extending these ideas, including with application to climate, are many studies such as Hale, Held, and Young 2017; Biermann et al. 2009; Abbott et al. 2012; Keohane and Victor 2011.

11. In this sense, we resonate with Mazzucato 2013 in showing that experimentalism (in this study, a competent state steering the economy) is alive and well in what is widely thought to be a least likely case.

12. For a complementary view of dynamic politics, see Mallet and Khalaf 2020; Lauber and Mez 2004. For this logic applied, for example, to the cluster of “direct air capture” technologies, see Meckling and Biber 2021.

13. Oreskes and Conway 2011; Stokes 2020; Mildenberger 2020.

Chapter 2. Lessons from the Path Not Taken: Montreal and Kyoto

1. The agreement built on a “framework” agreement from two years earlier: the 1985 Vienna Convention for the Protection of the Ozone Layer. It was purely a framework agreement and contained no substantive commitments, and thus most experts see the Montreal Protocol as the onset of real cooperation around the ozone layer. UNEP 1985. For fuller diplomatic histories, see Benedick 1998; Andersen and Sarma 2004; Parson 2003; Hoffmann 2005.

2. For the treaty, see UNFCCC 1992. For a fuller history of climate diplomacy, see especially Bodansky 1993.

3. For example, one study published at the completion of the UNFCCC specifically outlined the similarities with Montreal regarding the formal goals and procedures for adjusting treaty commitments in light of new scientific and technological information. Weiss 1993, 688.

4. See, for example, the comments of Winfried Lang. ESDO 2018. The definitive US-focused diplomatic history of the ozone talks, by Richard Benedick (the chief US negotiator on ozone during the formative years), identifies at least eleven major international environmental institutions, including the UNFCCC, that were heavily influenced by the ozone model. Benedick 1998, 331.

5. Victor 2011.

6. For an essay in counterfactual history on a large scale, see Sabel and Zeitlin 1985.

7. Perhaps the best articulation of the merits of a global market response was offered by someone who wasn’t in Kyoto doing the diplomacy but nonetheless had the big picture of how governance was changing: Janet Yellen, at the time (early 1998) head of the Council of Economic Advisers. She testified in support of the recently drafted (late 1997) Kyoto Protocol. Yellen 1998.

8. Victor 2011.

9. Only four developing countries (Egypt, Kenya, Mexico, and Nigeria) actually joined the original Montreal Protocol. Three were special cases: Egypt was the home country of Mostafa Tolba (the United Nations Environment Programme’s [UNEP] boss at the time and one of the key diplomatic entrepreneurs in the Montreal negotiations), Kenya hosted UNEP’s headquarters and thus needed to show solidarity, and Mexico was the home country of Mario Molina, a favorite son given that he was one of the two scientists who discovered the impact of chlorofluorocarbons (CFCs) on the ozone layer. In crafting the original Montreal Protocol, the G77, the club of nearly all developing countries, was not active, nor were China and India. The treaty did recognize the need for a plan regarding these countries, defined the terms under which a country would be considered “developing” (Article 5), and deferred until later provisions on technical assistance to these countries (Article 10). Developing countries, so defined, were allowed a ten-year delay on any control measures. In effect, at Montreal the developing counties were directed “primarily to maintain maximum usage of CFCs for the longest possible grace period” and not much beyond that. Quoted in Benedick 1998, 148. But when the developed countries started to tighten controls, the developing countries needed a more elaborate plan.

10. Moreover, the UNFCCC was negotiated in the context of the 1992 Rio Earth Summit that focused heavily on the problems of sustainable development in emerging countries. Rio was organized, in no small part, around the increasing political prowess and diplomatic organization of developing countries, their demands that environmental protection benefit development, and their corresponding wariness of measures that would impose costs and constraints on their development pathways.

11. On the then-current understanding of ozone depletion, thinning—but not a hole—should have occurred, if at all, not in the cold of the Antarctic night but rather in the sunny middle latitudes, where solar energy could initiate ozone destroying in the upper atmosphere. Even when the science came to understand how reactions on the icy surfaces of clouds could destroy ozone at alarming rates at the dark wintery pole, calculations of the magnitude of human damage to the stratosphere diverged enormously. Models of ozone depletion only began tracking observed patterns of depletion in the mid-1990s, nearly a decade after the first treaties were agreed on. Parson 2003, 55.

12. Parson 2003, 55–56.

13. Benedick 1991, 83–88.

14. On “environmental conscience,” see Ivanova 2021. On the central role of UNEP as the orchestrator of environmental cooperation at the time, see Tolba and Rummel-Bulska 1998. The huge diversity in experience at UNEP (and other international bodies working on environmental issues) yielded many alternative models for cooperation and was the subject of a major review of “lessons learned” in international environmental cooperation by Peter H. Sand, the principal legal officer for the 1992 Rio Conference on Environment and Development. Sand grapples with how to keep international cooperation from getting stuck in consensus, and encourage firms and governments to push frontiers. Sand 1991. The Sand analysis was cited by academics, but it was not heeded by diplomats as governments raced ahead to craft agreements on topics like climate change that superficially emulated the Montreal Protocol yet lacked understanding of how Montreal worked.

15. Bodansky 1993, 23; Haas, Levy, and Parson 1992.

16. Barrett 2005.

17. Najam, Huq, and Sokona 2003.

18. Andersen and Sarma 2004, 44.

19. US Congress 1977, section 126.

20. Andersen et al. 2018.

21. Andersen and Sarma 2004, 44.

22. Parson 2003, 53–55, 193. For the case for looking to the potential benefits of regulation, see Oye and Maxwell 1994.

23. Benedick 1998.

24. Under the 1977 Clean Air Act Amendments, Congress funded periodic scientific assessments, and the National Academy of Sciences played a central role. Conveniently for historians, the periodic academy studies—which became a partial blueprint for a systematic assessment of the ozone in the 1980s (which in turn became a partial model for the international assessment of climate change)—documented changes in the alarm level of the “consensus” over time. Among the many changes in ozone science were updates in the reaction rates for key steps in ozone-destroying catalysis, which in turn led to updates in ozone-thinning models. By the early 1980s, the trend was toward projections for lesser ozone depletion than originally feared. National Research Council 1983.

25. The talks were slow and meandering. Formally, the UNEP Governing Council authorized talks on a convention about the ozone layer in 1980 (Decision 8/7B); those talks began in 1981.

26. UNEP 1985; Lang 1991.

27. Sand 1991.

28. Parson 2003.

29. Sunlight still played a role, and that is why the hole appeared in the spring (October for Antarctica) just after sunrise. Cold was a key ingredient, for it allowed the formation of polar stratospheric clouds. On the role of polar clouds as the sites for the special heterogeneous chemistry that causes rapid ozone loss in the Antarctic, see Hofmann et al. 1989. For a key early paper on the depletion of the Antarctic ozone, see Solomon et al. 1986.

30. A huge thanks to Ted Parson, who has been generous with his insight as we learned how the Montreal method really works. What we call an “experimentalist” approach overlaps with what Parson calls an “adaptive regime.” For a discussion relating adaptive thinking in environmentalism to experimentalism, see Cosens et al. 2020.

31. UNEP 1987, Article 6.

32. In addition, a distinct panel looked at the effects of ozone thinning on humans and the environment. But as the Montreal Protocol evolved, most of the stocktaking functions were done principally by the Scientific Assessment Panel, which was focused not only on the evolving state of the science but also the broad options for cutting ODS. The ozone hole revealed that even at low concentrations of chlorine and bromine in the upper atmosphere, there would be a lot of harm to the ozone layer, and thus de facto the goal for the ozone regime became approximately zero human-caused chlorine and bromine in the upper atmosphere.

33. Formally, any changes that would alter the substantive commitments of the protocol—such as changing phaseout schedules or allowing time-limited exemptions—required a vote of the members of the protocol. The TEAP/TOC process provided the information needed to frame those votes and gained its influence, formally, by the fact that its recommendations nearly always had the largest impact in shaping the votes.

34. To underscore the fluidity of the committees, at first TOCs were established in refrigerants, foams, solvents, aerosols, and halons with a large number of members, yet little clarity around who would do what. Initially, in fact, they weren’t TOCs at all but rather teams of authors charged with writing chapters in an assessment report. See UNEP 1989a. For more detail on balance across the five areas that became the TOCs, see UNEP 1989b, Annex V, 5. At the outset, TEAP wasn’t TEAP but instead two panels with overlapping functions that later would be fused.

35. Parson 2003, 190.

36. Parson 2003, 189.

37. The US Air Force was a member of the Industry Cooperative for Ozone Layer Protection too since fully half of CFC solvents usage was mandated by US military specifications. The US military’s participation meant changes for allies, and the US Department of Defense was soon coordinating technologies with militaries from Australia to Canada to the United Kingdom, and getting the North Atlantic Treaty Organization to endorse the shift. With the help of the EPA, the Department of Defense also cooperated with the Soviet military to elicit similar changes in their alliance. Andersen and Sarma 2004; Parson 2003, 190.

38. In 1988, just a few months after Montreal, the keynote speaker at the first conference was Montana senator Max Baucus, a central figure in US national regulation of ozone pollutants. While the topics under debate, such as refrigeration compressors and CFC-free soldering, were obscure, the deliberation occurred under the shadow of regulatory policy. Baucus 1988.

39. Parson 2003, 188–90.

40. Parson 2003, 190.

41. Christensen 2016.

42. Formally, the exemptions are granted by the parties to the treaty because they are time-limited alterations of the treaty obligations. But in every case, the party-decided exemption process begins with (and is framed by) the expert assessment described here. Generally the TOCs and system of panels have this framing function with regard to “adjustments” and “amendments” to the treaty. Adjustments are needed when the existing rules are tightened or loosened; amendments are necessary to change the scope of the treaty, such as by subjecting new chemicals to regulation. In both instances, the parties defer almost completely to the judgments of the panels in deciding whether to formally consent to the proposals.

43. Victor and Coben 2005.

44. Parson 2003, 194–95.

45. Parson 2003, 193.

46. As the science evolved, it became clear that healing the ozone layer would require essentially zero emissions of any chlorinated or brominated compound. The ozone hole was triggered at even low concentrations of chlorine and bromine in the upper atmosphere.

47. Benedick 1998, 163.

48. Zhao and Ortolano 2003, 710.

49. Biermann 1996.

50. Benedick 1998, 152–57.

51. On the funding mechanism, see especially Biermann 1996; DeSombre and Kauffman 1996. In China, one initial estimate put the cost of ODS controls at $1 billion while a preliminary estimate for India imagined funding at double that level. Such numbers, anchored in no serious analysis, were far higher than donors were willing to pay. Benedick 1998, 187. The India estimate is reported in Parson 2003, 203, along with an estimate of $4 billion over a decade for all developing countries.

52. Greene 1998, 101–5.

53. Benedick 1998, 247.

54. Zhao and Ortolano 2003, 714, 717.

55. Zhao and Ortolano 2003, 717.

56. Benedick 1998, 247.

57. Andersen et al. 2018, 13–15.

58. Benedick 1998, 207.

59. Zhao and Ortolano 2003, 713.

60. Benedick 1998, 207–9, 220–30.

61. Parson 2003, 211, 218, 227–28. Among other things, the first full assessment of methyl bromide control options finished in 1994 revealed why the first TOCs had been so effective: they were generally smaller, more focused, and in the case of CFCs, didn’t include producers. The Methyl Bromide TOC, with a membership of sixty-five, included manufacturers and some users who steadfastly “fought to have the report conclude that there were no alternatives to [methyl bromide].” Quoted in Parson 2003, 228.

62. MBAO, n.d.

63. UNIDO 2015.

64. As climate diplomacy was gathering steam from 1989 to 1991, the Austrian diplomat Winfried Lang—known as the father of the Vienna Convention and the man who helped organize early European support for cooperation on the ozone layer—singled out the ozone depletion regime as the model to be emulated for the climate, though he allowed that progress on climate change would be slower. His account of the workings of the ozone regime did not mention the TEAP or TOCs; industry, where it figures at all, is noticed for its opposition to the regime rather than its engagement in the joint search for solutions. Lang accurately reflected what most architects of environmental cooperation thought at the time. Lang 1991.

65. A historian of climate diplomacy, Dan Bodansky, dates the end of Western nations’ dominance on the climate agenda to about 1990. Prior to that debate, much of global diplomacy was focused on control strategies and debates among Western countries. After that, developing countries were much more organized and influential. See Bodansky 2001, 28, 30.

66. Hecht and Tirpak 1995. For the first Intergovernmental Panel on Climate Change report, see IPCC 1990. Still other countries saw the World Meteorological Organization, the host of the 1990 Second World Climate Conference, as a logical host for climate diplomacy. Bodansky 2001.

67. In addition to the traditional voice of developing country interests—the G77, which typically caucused and spoke in tandem with China—a new subset of these nations formed in 1990: the Alliance of Small Island States. The group, representing the countries most vulnerable to a changing climate, was so effective that it soon became recognized as an official grouping of nations within the UN system. Ronneberg 2016.

68. UNGA 1988, 1990.

69. Vogler 2007. On the relation of the newly independent developing countries to the former colonial powers and their place in the global economy, see Getachew 2019.

70. Bodansky 2001.

71. Bodansky 1993.

72. UNFCCC 1995.

73. The main decision from Berlin was called the Berlin Mandate, and it set out the process and expectations for those negotiations. UNFCCC 1995, 1/CP.1. Formally, targets and timetables were called, in the language of the Berlin Mandate, “quantified emission limitation and reduction objectives” (QELROs)—a reminder that even the simplest idea can be conveyed by mangled, complex language.

74. UNFCCC 1995, 1/CP.1. This exemption was celebrated by many because the Berlin Mandate thus gave precise, institutional meaning to the principle of “common but differentiated responsibilities and respective capabilities” introduced in the UNFCCC four years earlier. Rajamani 2013. The sharp distinction between country groupings was taken to the limit in Kyoto, which rejected proposals that would have allowed developing countries to make even voluntary commitments to emissions reduction.

75. For the Montreal language, see the operative provisions of Article 6 in UNEP 1987. For the UNFCCC, see the authorizing language for the Subsidiary Body for Scientific and Technological Advice in Article 9 in UNFCCC 1992. In chapter 6, we will show that even where technically oriented meetings have been held, the focus has been diplomacy, not technical analysis.

76. Indeed, these same groups had evaluated and rejected concepts like “pledge and review” that might have evolved into a more experimentalist system of governance for climate change. Targets and timetables—without opportunities for revisions and the coadjustments of pledges in light of implementation experiences, were the centerpieces of their demands at Berlin. Victor 2001.

77. For the origins of the enthusiasm for emissions trading starting in this period, see Hahn and Hester 1989; Boyd 2021. For a discussion of the limits of the market-based approach to problems of this type, see Sabel and Simon 2011, 53. For a discussion of market mechanisms when applied to the rigidity of US environmental law, see Ackerman and Stewart 1984.

78. On the real-world experience with bubbles and other early market-like concepts, see Hahn and Hester 1989. See also Stavins 1988. On the sulphur program, see chapter 4.

79. Boyd 2021, 460. Advocates of emissions trading were well aware of the hot spot problem and suggested that it could eventually be addressed through elaborate refinement of the basic scheme. Sabel and Simon 2011.

80. Stewart and Wiener 2004.

81. Yellen 1998.

82. Victor 2001.

83. Victor 2001; Cullenward and Victor 2020, 87–102.

84. Streck and Lin 2010.

85. While the problem of CDM quality attracted press attention in the United States as an example of a UN program gone awry, in Europe concern was even greater because CDM credits were flooding into the ETS and were one factor in eroding prices in the ETS—in effect, Gresham’s law. For work emblematic of the European Union’s investigation into CDM quality, see Cames et al. 2016. That study, while it appeared after the key decisions were made, is a good overview of the methods and issues that EU policy makers were grappling with, and that led to the conclusion that the CDM can’t be considered a single mechanism; rather, each project type and institutional setting required distinct analysis. Staff work summarized around 2010 outlined a series of reforms to the CDM itself that might have improved quality (we are skeptical because the problem of setting baselines is intractable, no matter how clever the reforms), but such reforms would have required changing the CDM itself (which would require diplomatic consensus) or erecting a parallel EU system to oversee the CDM (which would have been administratively impractical). European Commission 2010. For the EU policy that has curtailed usage of the CDM in Europe, see EU 2013.

86. Victor 2001.

87. Canada and Australia were in similar situations. Canada joined Kyoto, but then withdrew when its largest trading partner (the United States) did not join. Australian politics on climate change were similar to, if not more toxic than, those in the United States, which made joining Kyoto impractical.

88. Victor and Salt 1995.

89. Revkin 2009.

90. In China that decade, the carbon intensity of the energy system actually went up—coal grew disproportionately faster than all other energy sources—and emissions soared. Victor et al. 2014.

Chapter 3. Theory of Experimentalist Governance

1. Chandler 1993.

2. Christensen 2016.

3. On the distinction between mass production by specialized tools and flexible production using reconfigurable capital goods under the control of skilled workers, see Piore and Sabel 1984.

4. From this perspective, decarbonization is a key piece of the larger program of overcoming the dualist separation of the economy into an advanced sector—the knowledge economy—that thrives on innovation—and a stagnant, rearguard sector that creates dead-end jobs and blighted communities. The Green New Deal aims to connect the two projects, but details are scarce. For a proposal to create more good jobs, extending the scope of the advanced sector, in line with the thinking here, see Rodrik and Sabel, forthcoming.

5. But this illustration suggests that the incentive-design problem is easy to solve. In fact, it turns out to be formally unsolvable. The principal’s best strategy is to in effect bribe agents into disclosing one another’s true costs of alternative actions by offering them shares in the project’s returns. The only scheme that generates incentives large enough to induce disclosure, however, also leaves the principal with the discretion to cheat agents by withholding the promised payoffs, undercutting cooperation before it begins. See Holmstrom and Milgrom 1991; Holmstrom 1982. Actors in the real world solving principal-agent problems wing it.

6. Fama and Jensen 1983.

7. On stakeholder views, see Henderson 2020. On the failures of current corporate governance and the possible advantages of the private equity model, see Gilson and Gordon 2020.

8. The distinction between risk and uncertainty was developed by the economist Frank Knight in the 1920s. See Knight 1921.

9. Since the 1990s, a wave of “new” governance writing has called attention to the spread of loosely defined, posthierarchical structures, often in the form of networks, and frequently without explicit reference to the breakdown of principal-agent relations. See, for example, Rhodes 1997; de Búrca and Scott 2006; Scharpf 1997. Much of this literature focused on developments in the European Union, where the push to create the single market led to the creation of regulatory agencies that had to pool the experience of the member states and continue to learn from their differences while maintaining common frames for action. Experimentalist governance developed as a response to changes in the United States and was then applied to an analysis of innovations in the European Union. See Dorf and Sabel 1998; Sabel and Zeitlin 2008.

10. On the logic of mass production and its vulnerability to shocks, see Piore and Sabel 1984.

11. Smith 1976, 21–25.

12. Ben-Shahar and White 2006.

13. Trebilcock 2017.

14. Ward et al. 1995.

15. Womack, Jones, and Roos 2007.

16. Ausubel, Wernick, and Waggoner 2013.

17. For example, hazards to supply chains. See Kinghorn 2016.

18. For a prominent case study of the reaction of the US nuclear power–generating industry to the Three Mile Island disaster, see Rees 1996. On incident reporting in civil aviation, see Mills and Reiss 2014.

19. Holling 1978; Lee 1993.

20. Gilson, Sabel, and Scott 2009, 2014.

21. This section draws on Kessler and Sabel 2021.

22. The legal terminology is confusing. The Administrative Procedure Act of 1946 calls the traditional method of rule making by hearing, with presentation of evidence as in a courtroom, formal, and notice-and-comment rule making informal. In practice most rule making is of the notice-and-comment type and it has become the de facto standard of formality. In what follows we mean notice-and-comment rule making when we speak of formal procedures in contrast to the informality of guidance. See 5 U.S.C. §553.

23. Elliott 1992.

24. Vermeule 2016.

25. Parrillo 2019, 168–169.

26. See, for example, American Bar Association 1993. Nicholas Parrillo speaks in this connection of “principled flexibility.” See Parrillo 2019.

27. Parrillo 2017, 34.

28. Parillo 2017, 34.

29. Parrillo 2019, 167–68.

30. Lindsay 2018.

31. Kessler and Sabel 2021 suggests some doctrinal and institutional reforms in this direction.

32. Callon, Lascoumes, and Barthe 2009.

33. For a concise portrayal of this skeptical literature, see Fishkin 2018.

34. Sunstein and Hastie 2015.

35. For a concise description of how deliberative polling is organized and the role of experts in providing sound information to frame discussion, see Luskin, Fishkin, and Jowell 2002, especially 458–60.

36. On deliberative polling in the radically polarized circumstances of Northern Ireland, just after it emerged from outright violence between Protestants and Catholics, see Luskin et al. 2014. There is, as can only be expected, disagreement among those studying deliberation as to whether deliberative polling reliably proxies what public opinion would be under plausibly favorable conditions. See Neblo, Esterling, and Lazer 2019.

37. See, in general, Fishkin et al. 2021.

38. Ho 2017.

39. Ho 2017, 43.

40. To eliminate the possibility that background features of the situation—the participants’ length of service perhaps, or the experiences of one or another group—might in operationalizing the experiment come accidentally to shape the results as well as to enable the precise measurement of the effects of reform, the entire staff was randomly assigned to a treatment group that would conduct the weekly peer reviews and a control group that would continue inspection as usual. To further reduce the likelihood of accidental, extraneous influence, the peer pairings were randomly reassigned each week, with restrictions to ensure that pairs did not repeat and that assigned establishments were outside the pairs’ home inspection territories. Within each pair, the peers alternated in the role of lead inspector, who took primary responsibility for guiding the visit, and whose report, corrected in light of discussions after the inspection, became the official record of violations, while the “nonlead” inspector filed an unsigned report coded for the purposes of the experiment as peer review. A great advantage of this procedure was to allow Ho to trace the evolution of judgment inspector by inspector and week by week. Ho 2017, 31.

41. Ho 2017, 45, 54.

42. Personal communication, Daniel E. Ho, December 8, 2021.

43. Ho 2017, 93.

44. Ho 2017, 57.

45. Goodin and Niemeyer 2003 emphasize the role of exposure to new information, credibly presented, in promoting reconsideration of views.

46. Ho 2017, 92, 86, 67.

47. Ayres and Gertner 1989.

48. Hiscox and Smyth 2006.

49. Gunningham, Kagan, and Thornton 2004.

50. Copeland 2012.

51. Clean Air Act 1963.

52. On the California effect, see Vogel 1997. On the Brussels effect, see Bradford 2020.

Chapter 4. Innovation at the Technological Frontier: Three Policy Icons and a Common Approach to Uncertainty

1. Gardner 2014.

2. Under Section 177 of the Clean Air Act, states can opt to follow California’s clean air vehicle standards under a waiver granted by the EPA. For more, see Smith 2019.

3. California Health and Safety Code 1975, §39500, §39510. See Reed 1997.

4. CARB 1994. California has consistently failed to meet either standard. Bedsworth and Taylor 2007.

5. Hughes 2015.

6. Vogel 2018.

7. Reed 1997, 785.

8. International Harvester Co. v. Ruckelshaus 1973; Natural Resources Defense Council, Inc. v. US Environmental Protection Agency 1981.

9. Natural Resources Defense Council, Inc. v. US Environmental Protection Agency 1981.

10. Dixon, Garber, and Vaiana 1996, 7.

11. As one senior regulator put it in an interview, “[A manufacturer might say,] ‘Look, I have this product, and I think it can be really cost-effective and I think it can really lower emissions’ or ‘I have the data to show you I can do it, and I can do it better than my competitors and I can do it cheaper.’ ” Wong 2015.

12. Cherry 2016.

13. Because of this expertise, CARB can “meet with manufacturers and we can roll up our sleeves and talk technical talk with them. It got us a lot of respect with them.” McCarthy 2015.

14. CARB 1996, 4–5; Hughes 2015. See also Magnusson and Berggren 2011, 7–8: “Hence, systems integration still constitutes a critical activity in automotive development and design. Even though automotive manufacturers do not have to manufacture the complete vehicle in-house, they still need a broad technological knowledge base to be able to absorb new technologies (Cohen and Levinthal 1990), as well as to understand interdependencies in the product system and cope with imbalances due to different rates of development in different technologies (Brusoni and Prencipe 2001).”

15. One regulator described the push-pull dynamic created by CARB’s feasibility requirement this way: “We have to know when we adopt the more stringent standards, there’s a way to get there. And of course, they have to know there’s going to be a market for the product they come up with.” McCarthy 2015.

16. CARB 1996, 4–5.

17. Reed 1997, 722–23.

18. Reed 1997, 723.

19. Reed 1997, 733–34, 737–40.

20. BTAP 2000.

21. Reed 1997, 57.

22. Bedsworth and Taylor 2007, 7.

23. Reed 1997, 771.

24. Reed 1997, 774.

25. That year, all battery and plug-in hybrids accounted for 8.9 percent of new car sales; pandemic turmoil make data from 2020 less indicative.

26. Reed 1997, 774–75.

27. Bedsworth and Taylor 2007, 7–8.

28. Reed 1997.

29. Reed 1997.

30. Rocky Mountain Farmers Union et al. v. Corey 2013.

31. Dixon, Garber, and Vaiana 1996, 11.

32. CARB 2012.

33. For California sales, where about 8 percent of new vehicles are electric (including plug-in hybrid vehicles), see Kane 2020. And for global patterns, which remain robust despite big reductions in subsidies in China, thanks to expanded policy support in other parts of the world, see Holman 2020.

34. Chestnut and Mills 2005.

35. Environmental Defense Fund 2018.

36. “The Invisible Green Hand” 2002.

37. Ruckelshaus at the EPA’s National Compliance and Enforcement Conference, January 1984.

38. Turner and Isenberg 2018.

39. Depending on the plant design, there will be other pollution devices as well—baghouses, filters, and in the most modern plants, chemical systems for removing nitrogen pollution.

40. Gypsum in theory has value as the key ingredient in wallboard, but only if there aren’t heavy metals and radioactive particles in the gypsum. Most of the sulfurous slurry from scrubbers is simply a waste product—that is, another cost.

41. Barker 2012.

42. EPA 1990.

43. Environmentalists favored some data and ideally wanted to cut emissions nearly in half; industry did not mind a study program that delayed action; neither was powerful enough to force its preferred solution on the other. Formally, the US approach to acid rain looked like a rational study-then-regulate program. In practice, the arrows of causation ran in all directions, and the science played almost no role in shaping the politics. The science that could have mattered—estimates of the cost for cutting sulfur—turned out to be wildly wrong because analysts did not update their models to reflect a trove of rapidly shifting factors such as improvements in scrubbers, expansion of coal leasing in the West (mainly Wyoming), and the deregulation of the railroads. All told, the United States spent $500 million on the National Acid Precipitation Assessment Program, which issued its formal report after the 1990 Clean Air Act was completed. The program sponsored a lot of good research on air pollution sources, transportation, and ecological effects. None of it affected policy because political pressures ran faster than the scientists.

44. Association of American Railroads 2011; Hecker 2007, 6.

45. Stavins 1988.

46. Newell and Rogers 2003.

47. Hahn and Hester 1989.

48. Schmalensee and Stavins 2017; Ellerman 2010.

49. Bedsworth and Taylor 2007.

50. Stavins 1988.

51. Turner and Isenberg 2018.

52. By the time the program was designed, the emissions from the covered plants were about sixteen million tons. Some plants, due to other pollution controls, were not covered. And the total US emissions of SO₂ was higher than twenty million tons due to other industrial sources, including sulfur in petroleum fuels used for road transport.

53. A draft report was circulating while the Clean Air Act amendments were finalized, but that draft offered no crisp advice on where to set pollution limits. An earlier “interim report” had been released in 1987, but it was pilloried as hewing to the Reagan administration’s probusiness position and thus ignored politically.

54. Schmalensee et al. 1998.

55. Turner and Isenberg 2018.

56. Schmalensee and Stavins 2017.

57. The availability of those options varied with exogenous conditions as well—leading, as market theorists rightly expected, to variations in prices. In 2005, sulfur prices peaked at more than $1,200 per ton in response to a series of exogenous shocks, such as derailments of coal cars that made it harder to deliver low-sulfur coal.

58. Studies of patenting data show that during the era of command-and-control regulation, there was a much higher level of innovation in sulfur control than during the market period. Taylor, Rubin, and Hounshell 2003.

59. “Smog” is the word frequently used for photochemical ozone pollution in cities. But the same chemical processes that make ozone (which is harmful to humans and plants when contacted in the lower atmosphere, but good for ecosystems up in the stratosphere) also run regionally; source pollutants such as hydrocarbons and carbon monoxide mix with NO_x and sunlight in the atmosphere, and make smog downwind and regionally. If a locality must meet smog rules (per the NAAQS) locally, it often can’t have a full impact without getting cooperation from states upwind.

60. Schmalensee and Stavins 2013.

61. Chan et al. 2018.

62. Schelling 1983; Sabel, Fung, and Karkkainen 1999.

63. EPRI 2005; American Electric Power 2017.

64. Cohen and Noll 1991; Lester and Hart 2012.

65. For an exception that proves the rule, see Albert Hirschman’s (skeptically received) suggestion that large development projects are governed by the principles of the “hiding hand”: they are undertaken because their advocates underestimate the challenges to be faced, and succeed because difficulties, providentially, lead to the discovery of unexpected problem-solving capabilities. Hirschman 2014.

66. International Energy Agency 2012, 2020a, 2020b.

67. Chan et al. 2018.

68. National Academies of Science, Engineering, and Medicine 2017.

69. Bonvillian, Van Atta, and Windham 2019.

70. There is attention, certainly, to what potential customers might want; indeed, the early stages in creating a new DARPA program involve extensive canvasing of possible use cases. Bonvillian, Van Atta, and Windham 2019.

71. Rodrik and Sabel, forthcoming.

72. Azoulay et al. 2019.

73. Rodrik and Sabel, forthcoming.

74. Rodrik and Sabel, forthcoming.

75. Azoulay et al. 2019.

76. Howell 2017.

77. Doblinger, Surana, and Anadon 2019.

78. National Academies of Sciences, Engineering, and Medicine 2021b.

Chapter 5. Experimentalism in Context: Ground-Level Innovation in Agriculture, Forestry, and Electric Power

1. The idea that the final, decisive interpretation of price signals and other unambiguous guides to action is rooted in mute, local circumstance is itself deeply rooted in economics. The kind of statistics expressed in prices, economist Friedrich Hayek writes, “have to be arrived at precisely by abstracting from minor differences between the things, by lumping together, as resources of one kind, items which differ as regards location, quality, and other particulars, in a way which may be very significant for the specific decision. If we can agree that the economic problem of society is mainly one of rapid adaptation to changes in the particular circumstances of time and place, it would seem to follow that the ultimate decisions must be left to the people who are familiar with these circumstances, who know directly of the relevant changes and of the resources immediately available to meet them.” Hayek 1949, 83–84.

2. Hirschman 2014. For an exception in the study of technological borrowing, see Yi 2020. For a view of implementation as reinvention in use akin to the one here, see Eric von Hippel’s work on the innovative modification of software or capital goods, such as machine tools, through lead users in collaboration with one another or producers. von Hippel 2006.

3. Stokes 2020.

4. Hutchins 2000; Latour 1999.

5. The consolidation of Irish milk processing is less pronounced than in Ireland’s major dairy export competitors, such as Denmark, the Netherlands, and New Zealand, where one company processes as much as 70 to 80 percent of the milk pool. Dairy Industry Prospectus Report 2003.

6. Eurostat 2017; Food and Agriculture Organization 2018, 5; Bord Bia 2016.

7. Fitzgerald 2019.

8. Thorne et al. 2017.

9. Alothman et al. 2019.

10. It is estimated that in 2008, the wider agri-food (which includes beverages, infant formula, and food ingredients) accounted for 40 percent of the total net foreign earnings of all primary and manufacturing industries. Riordan 2012.

11. The Food Harvest 2020 strategy of 2010 set a target by 2020 of a 50 percent increase in the volume of milk production over the average of 2007–9 milk supply (4.93 billion liters). The volume of milk production in 2017 had reached 7.27 billion liters—an increase of 47 percent. Central Statistics Office 2018.

12. Poikane et al. 2014.

13. For the guidance documents, see “WFD Guidance Documents,” n.d. For a detailed discussion of the Common Implementation Strategy as an experimentalist institution at the heart of the Water Framework Directive, see Scott and Holder 2006.

14. Hering et al. 2010; Voulvoulis, Arpon, and Giakoumis 2017; Giakoumis and Voulvoulis 2018.

15. Shortle and Jordan 2017.

16. Burgess 2018.

17. Daly, Archbold, and Deakin 2016, 161.

18. For a good overview of the role of LAWPRO’s Catchment Assessment Team and its place in the new governance structure, see EPA Catchments Unit 2018.

19. On the Agricultural Sustainability Support and Advice Programme, see “ASSAP—Farming for Water Quality,” n.d.

20. For the most current review of LAWPRO in operation, commissioned by the Irish EPA, see Boyle et al. 2021.

21. Clarke et al. 2014. Exactly how far electrification must go is unknown. If other energy carriers, like hydrogen, prove cost-effective and reliable, then they may play a big role in deep decarbonization. If not, then more of the work of deep decarbonization will need to be done by electrifying end uses and then decarbonizing the power sector. Azevedo et al. 2020a.

22. Among the studies looking at the early days of renewables, including in California, Gregory Nemet’s is notable. Nemet 2019.

23. “SB-1078 Renewable Energy” 2002; Jacobson 2020.

24. Pyper 2019.

25. Jacobsson and Lauber 2006; Hockenos 2015.

26. Sivaram 2018.

27. Nemet 2019.

28. Jacobson 2020.

29. The governor’s own Office of Planning and Research orchestrated the effort because no state agency had the authority to convene all the right actors, which included not just solar developers but also city officials, organized labor, and fire departments. (Solar panels, badly installed, were a fire hazard.) Governor’s Office of Planning and Research 2014.

30. Governor’s Office of Planning and Research 2012.

31. For example, different kinds of rooftop projects—different sizes and local grid configurations—required different kinds of interconnection and top-level rules to govern that process. It was feared, with some reason, that the local utilities that controlled the local grids were using risk-averse grid interconnection rules to prevent the implementation of rooftop solar. Changing and streamlining those rules was thought to advance implementation in the traditional sense. Governor’s Office of Planning and Research 2014.

32. Annual Levelized Cost of Energy Analysis (LCOE 14.0) 2020.

33. We describe the macro patterns here, but the real world of course is more complex. In particular, there are scenarios in which large numbers of small rooftop generators have synergistic impacts that could propagate across the grid—for example, if a frequency disturbance propagates into rooftop systems and the “droop” set points on the inverters cause them all to trip simultaneously. (There are interesting scenarios that start with cyberattacks that lead to similar outcomes.) Such concerns, in part, have led some jurisdictions to make costly upgrades to the existing inverters and inspired a lot of attention by engineers to inverter stability at all scales. All that said, scale is more important—and what grid-connected solar offers is scale.

34. CPUC 2015.

35. Simulation occurs over multiple timescales. Some models are used for long-term planning; others run real time alongside the grid to help operators manage problems as they arise.

36. Roberts 2018.

37. CPUC 2015.

38. The full story is more complex, of course. There is a third agency, the California Energy Commission, which oversees policy strategy and works jointly with the other agencies to redirect investment for the long term; the commission also administers various funding mechanisms (as do CPUC and CAISO) to support advanced technologies.

39. Indeed, curtailment was assumed to be such a nonproblem that CAISO didn’t track or report it systematically until 2017. Nonetheless for grid operators and planners accustomed to traditional thinking—that grid operators would simply find ways to integrate renewables, and any excess would be small and easily dumped—the curtailment levels of 2017 would seem acceptably small. Only as real-world deployment advanced in the following few years did the numbers climb steeply. Moreover, once modelers caught up with the speed of renewables deployment, they showed that a massive reliance on renewables alone would lead to extensive curtailment along with other side effects, such as huge land areas needed for the overbuilding of renewables and huge needless costs. On the curtailment data, see “Managing Oversupply” 2021. And for a recent look back at the problem of curtailment and overbuilding, see Cohen et al. 2021.

40. CPUC 2008.

41. Formally, the planning process is done by opening a proceeding through which utilities make proposals. For regulated utilities (known formally as “electric corporations”), CPUC can mandate participation in a proceeding and thus de facto procurement. For public power agencies, CPUC’s authority is more limited, and therefore the law requires that those organizations go through their own local planning process and feed that information to a different state body. Given this, as a practical matter, the experimentalist approach to battery deployment has focused principally on the utilities because they interact most extensively with CPUC. “AB-2514 Energy Storage Systems” 2010.

42. CPUC 2013, 6.

43. CPUC 2013, 15, table 2. The question of battery deployments on the high-voltage system was outside CPUC’s oversight—a matter for CAISO, which more slowly developed its own program.

44. CPUC saw its mission as “transformational” in creating a new market for batteries, not just deploying a quota of storage devices. Thus the original plan was to evaluate projects against the potential for transformative impacts, not simply near-term costs and benefits. See CPUC 2013, 6.

45. Letting such projects compete, CPUC said, “would inhibit the fulfillment of market transformation” that CPUC saw as a central purpose of its storage policy. See CPUC 2013, 34.

46. Personal communication, former executive at a major California utility.

47. On the use of experience to drive changes in procurement, see CPUC 2013, 25.

48. See CPUC 2013, 26. Moreover, as a recognition that local details mattered, CPUC adjusted target compliance utility by utility to reflect the pipeline of storage projects that utilities already had funded with support from the Self-Generation Incentive Program and other incentive programs. Thus top-level goals, even at the outset, were highly contextualized to local circumstances. Once CPUC began this process, it learned that lots of special projects claimed exemptions; hence, like the Montreal Protocol TOCs, it identifies top-level rules that would align with reason and then sets those rules, provisionally, as the metrics for compliance. CPUC 2013, 32. See also CPUC 2013, 27, section 4.5.

49. It was known as the “power charge indifference adjustment,” which CPUC adjusted in a time-limited fashion (explicitly rejecting arguments that the changes should extend beyond a decade since CPUC had no idea how the battery market would evolve). See CPUC 2014. For a decision that looks back at the procurement experience and also addresses issues with value stacking, discussed below, see CPUC 2018, 3.

50. CPUC 2016.

51. The subsidy scheme is known as the Self-Generation Incentive Program and it paid some of the costs for advanced battery projects; CPUC adjusted, utility by utility, the procurement targets to reflect its assessment of which Self-Generation Incentive Program projects should “count” toward the goals—a detailed regulatory oversight that is reminiscent of what the Montreal TEAP does when it recommends time-limited exemptions for critical uses of ODS, discussed in chapter 2. CPUC also adjusted the rules to make it easier for Community Choice Aggregators to participate in battery deployments. On the former, see CPUC 2013. On the latter, see CPUC 2016. It is worth noting that CPUC clarified that Community Choice Aggregators must assure “resource adequacy” for their customers—that is, to have contracted enough power to supply expected loads (and then some). Resource adequacy is one of the major mechanisms through which regulators help assure network reliability, and the scramble to meet resource adequacy rules has led some Community Choice Aggregators and other power providers to invest in more battery projects (and projects that integrate batteries with solar generators) for resource adequacy purposes.

52. Fitzgerald et al. 2015, 7.

53. California regulators, based on the deployment experience in the state, convened a workshop in 2018 to evaluate what was being observed and learned. See CPUC 2018, appendix B.

54. In practice, there are several different types of interconnect agreements, depending on the voltage, configuration, and application of the device in question. One of the many concerns of regulators that are trying to create opportunities for new technologies is that incumbent utilities will use the interconnect process as a barrier to entry—thus the process is overseen strictly (as is the modeling that feeds into the interconnect agreement) with an eye to evenhandedness. Moreover, when a utility installs a battery on its own network—thus raising such concerns about self-dealing—there is a requirement for an interconnect between the arm of the utility that deploys the device and the arm that manages the grid—a requirement designed in part to reduce the self-dealing risks. Nonetheless, griping is rampant, especially when an interconnect process leads the utility to find that a costly grid upgrade will be required. One advantage of storage projects is that they have, in most cases, been designed to use the services of the battery system to avoid the need for grid upgrades.

55. Formally in California, these rate and cost recovery decisions are taken after the expenditure is made, and thus the procurer is especially attuned to the likely interpretation of rules and experience regarding the prudence of its spending.

56. The process we outline here is most extensive for the largest projects—those that typically offer the greatest value (and potential risks) to the grid. For the smallest projects, such as most behind-the-meter deployments, the utility and regulators have much less oversight, which helps explain why the behind-the-meter market has seen the most expansive growth.

57. It is also emblematic of the attention to how models (to a limit) can help advance the “adaptive management” of renewable resources. See Holling 1978.

58. The answer required navigating federal regulatory standards (which defined grid-connected resources) along with challenges in the power flow modeling of resources that were unlike anything that had been connected to the grid. CAISO first grappled with these issues in 2013 when it looked at the initial cluster of battery projects (thanks to CPUC’s procurement mandates); at the time, there was essentially zero experience anywhere in the world to offer an empirical guide for the modeling, and hence the modeling techniques were developed from logic and comparing batteries with other, more familiar resources. See CAISO 2014.

59. For more on how these “non-generator resource” deployments might affect the energy markets, see CAISO 2011. For a discussion of how the interested parties affected the setting of interconnect rules—and how uncertainties in battery operation would affect the reliability of the grid, see CAISO 2014. And for more on the factors affecting deployment of batteries over time, see Hart and Sarkissian 2016.

60. That project was a large Tesla battery system located at a wind farm and put into service in 2018. For a study in 2017 that made predictions and focused on the battery adding value by arbitraging power, see MacKinnon 2017. For a retrospective on how the battery actually delivered value, which was with voltage support and other “ancillary services,” see Parkinson 2020.

61. Indeed, worries emerged within the regulator that its battery program was not pushing enough experimentation. As the idea of value stacking was put into practice, it became easier to identify projects using the known values that were the most cost-effective. This was leading to an overinvestment in one class of technologies (notably lithium chemistry) and known applications around the medium-voltage grid. According to interviews, in the late 2010s, CPUC became concerned that this approach was narrowing the scope of search and the rules might need adjustment. Part of the problem was that the review process at CPUC was mainly backward-looking. The investor-owned utilities would advance their battery projects, and get approvals for rate base and other costs only after committing the resources, and CPUC reviews of the whole program looked back after that. How to rectify this situation, which is distinct from the experience with most of the Montreal Protocol TOCs that also did forward-looking evaluations of projects, is an ongoing challenge.

62. CPUC 2018, 15–18.

63. The CPUC order that created rule 6 (and other rules) explicitly called these provisional, and established a process to investigate performance and make future adjustments, if needed.

64. See CAISO, CPUC, and California Energy Commission 2021. While the main troubles in August were not related to batteries, a large part of the analysis focused on how batteries providing resource adequacy, the key measure of whether a power supply is reliably available, actually functioned on the grid—a novel topic because the battery market was changing so quickly. (The resource adequacy provisions are one way that regulators make sure that at every hour of the day, there are sufficient generating resources on hand. In practice, the shift to renewables and the need for more power that can be dispatched quickly are forcing a rethinking of the whole concept of resource adequacy, and also forcing the regulator to develop new methods for measuring resource adequacy—methods that overlap, in concept, with the concept of value stacking; one of the many values to be stacked is power flow at specific moments in time, which is what resource adequacy measures.) Prior to 2018, CAISO and CPUC rules in effect did not allow batteries to provide resource adequacy. Until the operational experience under the first wave of procurement, there was neither the installed base nor the experience to craft effective resource adequacy rules for batteries. In August 2020, CAISO could observe, for the first time under conditions of severe grid stress, how the initial two hundred megawatts of batteries procured for resource adequacy actually performed.

65. See CAISO, CPUC, and California Energy Commission 2021. This study finds that batteries might have been able to provide greater reliability services than they actually were called on to supply in August. The conservative approach embodied in rule 6 may help explain that outcome. For more, see Penn 2020. Nonetheless, CAISO remains cautious—as any regulator with the responsibility for grid reliability must—because of the ways that the operation behavior of battery systems remains murky yet increasingly important as they get larger and more material to grid management.

66. One big open question is whether markets will be established for all or most stackable services, or whether investment will instead proceed principally under a rate-of-return regime. But regardless of how that question is answered, contextualization under a regulatory regime of the kind CPUC has devised is likely to continue into the indefinite future. The decentralization of power generation and management will continue to cascade within many of the (larger) local nodes of the grid, such as big-box stores, hotels, hospitals, or universities.

67. For an example of two projects underway, see Center for Sustainable Energy 2020. For a more systematic look, see the “ARD, S8” projects, summarized in California Energy Commission and Doughman 2020.

68. On the projects underway, see, for example, Center for Sustainable Energy 2020. Meanwhile, CPUC procurement has focused on a new frontier: encouraging the mass deployment of local networks of batteries and generators known as microgrids. The logic behind this push, which is destabilizing the status quo again, is that more local control and reliability of service will make it possible to keep electricity flowing to customers even when long-distance lines are turned off in fire-prone areas. The logic for microgrids has been around for a long time. See Hanna et al. 2019. Wildfires have altered the potential value from these deployments, although whether those values are realized is entirely context dependent and highly uncertain at present. Storage is pivotal to the idea, and like CPUC’s battery procurement program, an experimentalist approach to deployment and learning through context is now underway. See St. John 2021.

69. For example, there is enormous variation in the reliability of the existing grid service. In remote communities served by a single power feeder line, grid reliability is often relatively low—especially if that line runs through a mountainous brushy area, and thus is periodically disconnected when it is windy and dry to avoid touching off wildfires. By contrast, communities served by many interconnected feeder lines will have more reliable grid service because the system overall is less vulnerable to the loss of any single component. In principle, these differences in grid configuration will lead to big differences in the value gained to a community from installing microgrids that keep operating even when the larger grid fails. While that variation in value is appreciated in theory, it is not yet possible to measure reliably how different configurations of macro- and microgrids affect the soundness of different microgrid investments.

70. See especially California Energy Storage Alliance, https://storagealliance.org/. See also California Solar and Storage Association, https://calssa.org/. At the federal level, the industry is also increasingly well organized. See Energy Storage Association, https://energystorage.org/.

71. Pyper 2017.

72. “About REDD+,” 2021.

73. The laws, administrative measures, and statistics under discussion here assume different definitions of the Amazon. The most comprehensive is the Legal Amazon, an area of more than five million square kilometers, including the Brazilian states of Acre, Amapá, Amazonas, Maranhão, Mato Grosso, Pará, Rondônia, Roraima, and Tocantins, established by Brazilian federal law in 1953 in order to fix the scope of regional development and environmental policies. Some 25 percent of this area was originally covered by the Cerrado or savanna and other nonforest vegetation. The Amazon biome or habitat is contained within the the Legal Amazon except for a tiny area in the state of Maranhão. We refer to all of these indiscriminately as the Amazon. West and Fearnside 2021, 2.

74. Brandão et al. 2020.

75. Lucas and Warren 2013.

76. For a case study of these developments in a forest district in Indonesian Borneo, see Gecko Project 2017.

77. Campbell 2015.

78. Campbell 2015, 60.

79. Schwartzman 1991, 399.

80. For a thoughtful development of this view, well beyond the limits of the “primitive,” however understood, see Western 2020.

81. Schwartzman et al. 2010, 277. Reference is to Alcorn 2005.

82. Arima et al. 2014.

83. Hochstetler and Keck 2007.

84. Locke 2013; Bartley 2018; Overdevest and Zeitlin 2018; Gereffi, Regini, and Sabel 2014.

85. Hochstetler and Keck 2007.

86. Riofrancos 2020.

87. Schwartzman et al. 2010.

88. Brandão et al. 2020, 4; Nepstad et al. 2014, 1118.

89. Jackson 2015.

90. Thaler, Viana, and Toni 2019.

91. West and Fearnside 2021. But for early signs that the registration program is coming into wider use and reducing deforestation rates among smallholder registrants, see Lipscomb and Prabakaran 2020.

92. Gibbs et al. 2016.

93. Nepstad et al. 2014, 1118.

94. “Based on a panel of Amazon municipalities from 2002 through 2009 … deforestation responded to agricultural output prices. After controlling for price effects … conservation policies implemented beginning in 2004 and 2008 significantly contributed to the curbing of deforestation.” Assunção, Gandour, and Rocha 2015. See also Assunção and Rocha 2019; Arima et al. 2014.

95. Gibbs et al. 2016; Heilmayr 2019; Picoli et al. 2020.

96. Assunção et al. 2020.

97. “About REDD+,” 2021.

98. Since a low in the early 2010s, deforestation has more than doubled by 2020, but is still less than half the peak levels of the early 2000s. See BBC 2020.

99. Gomes, Perz, and Vadjunec 2012.

100. Hoelle 2014, 2015.

101. For details, see Kröger 2020; Hofmeister 2020. “The total for [2019] had already reached 74.5 square kilometers (28.7 square miles), three times more than the average in each of the past five years. (The latter figure was itself twice as high as registered before 2013, when the annual deforestation rate didn’t exceed 10 km² [3.8 mi²].)”

102. “Half of protected areas in the Brazilian Amazon have no approved management plan and 45 percent have no management council.” Butler 2011. For a case study of disorganization, see Freitas, Rodrigues, and Silva 2016.

103. “Isso porque a maioria das iniciativas aplicadas era de natureza regulatória ou meramente repressiva, incapaz de alterar a dinâmica das atividades produtivas vinculadas ao desmatamento e, ao mesmo tempo, fomentar uma nova base eco- nômica sustentável na região. Além disso, a sociedade local, fortemente impactada com essas medidas, não percebia como vantajosa a questão ambiental e, por isso, não se engajava no processo de ordenamento. Foi nesse contexto que surgiram iniciativas locais ou regionais de combate ao desmatamento com potencial enorme de sustentar e aprofundar os ganhos até então obtidos, como é o caso do Programa Municípios Verdes no Estado do Pará.” Whately, Campanili, and Veríssimo 2013.

104. Schmink et al. 2019.

105. For a comparison of Paragominas and São Felix, see Brandão et al. 2020. For a full account of São Felix, see Schmink et al. 2019.

106. Varns et al. 2018.

107. Schmink et al. 2019.

108. Thaler, Viana, and Toni 2019.

109. Brandão et al. 2020.

110. In the increasing weight of smallholder deforestation after the command-and-control reforms, see F. Seymour and Harris 2019; Schneider et al. 2015. See also Schielein and Börner 2018.

111. Vargas 2020.

112. Wallace, Gomes, and Cooper 2018.

113. Kröger 2020.

114. “Povos Indígenas do Rio Negro Planejam o Futuro que Desejam” 2017; Navarro 2020.

115. Which is not to say, of course, that even with regard to the regulation of larger property, the measures have been a panacea. For ways to view them, see Branford and Torres 2017.

116. Schmink et al. 2019.

117. For an account of the widespread embrace of no-till agriculture among US farmers in the Midwest and Great Plains concerned about soil erosion but reluctant to discuss climate change, see Horn 2017.

118. Santiago, Caviglia-Harris, and Pereira de Rezende 2018.

119. Skidmore 2020.

120. “Smallholder Crucial to Preserving the Amazon Rainforest” 2019; de Paiva Serôa da Motta et al. 2018.

121. For the potential uses of latex and its by-products, see Jaramillo-Giraldo et al. 2017.

122. Coslovsky 2021.

123. Coslovsky 2014.

124. Buainain, Lanna, and Navarro 2020; Correa and Alkmin Junqueira Schmidt 2014.

125. Guanziroli et al. 2019.

126. For details, see Guanziroli et al. 2019; Rada, Helfand, and Magalhães 2019. The figures are from 2006. A census of agricultural holdings was scheduled for 2016, but it was postponed because of domestic turmoil. Between 1985 and 2006, the productivity of land increased with the size of farms, reflecting economies of scale in the use of that factor. But the size distribution of the total factor productivity, taking into consideration the productivity of labor and capital as well, was U shaped, indicating that both small and large farms can combine endowments with increasing efficiency. Some of the efficient small producers supply oranges, chickens, or pigs to large food processors, which provide them with key inputs and prescribe production methods that must be rigidly followed. Such industrial supply chains are as a rule environmentally dirty. Other efficient smallholders grow fresh produce by cleaner, more skill-intensive methods. Sustainable, economically viable smallholder ranching resembles the latter.

127. Pacheco et al. 2018; Schoneveld et al. 2019; Dharmawan et al. 2021.

128. Nature Conservancy 2018; Seymour, Aurora, and Arif 2020; Stickler et al. 2018; Buchanan et al. 2019; Fishman, Oliveira, and Gamble 2017; Brandão et al. 2020.

129. For an incisive discussion of the choices that rural society in Brazil faces, see Navarro 2020.

Chapter 6. International Cooperation beyond Paris

1. Victor and Jones 2018; Victor 2019.

2. The Swedish strategy involves a partnership between Swedish steel firms and an energy supplier, Vattenfall, that has a specialty in hydrogen. Hoikkala and Starn 2020.

3. The value of scale is greater for small countries, of course, but even for large countries cooperation can expand the range of the technological ideas, policy strategies, and business models tested. Other methods include capturing carbon dioxide during hydrogen production and the use of electricity to transform, chemically, iron into steel without carbon dioxide as a by-product. “Accelerating the Energy Transition” 2016; Victor, Geels, and Sharpe 2019.

4. Dolphin, Pollitt, and Newbery 2020; Cullenward and Victor 2020.

5. Dolphin, Pollitt, and Newbery 2020.

6. Abnett 2021.

7. In one of these areas, forestry, there is a formal link back into the Paris Agreement through the Reducing Emissions from Deforestation and Forest Degradation system discussed in chapter 5.

8. Raustiala and Victor 2004; Alter and Meunier 2009; Oberthür and Stokke 2011; Biermann et al. 2009. Related ideas are advanced around the governance of problems that don’t implicate international cooperation. See Ostrom 2010; Hawkins et al. 2006.

9. Sabel and Victor 2015; Sabel and Victor 2016.

10. Sabel and Victor 2015, 2016; Victor 2015.

11. See, for example, Morales 2015; Victor 2015; Falkner 2016; Bodansky 2016.

12. See paragraph 24 and subsequent decisions in UNFCCC 2018.

13. Victor 2015; Hale 2017.

14. Hale 2016. Formally, some assessment of these subnational efforts does exist, organized partly in response to the Marrakech Partnership for Global Climate Action by the High-Level Champions in 2016. See, for example, UNFCCC 2020. For an appraisal of the efforts at assessment, see Hale et al. 2021.

15. Of course, in the case of Montreal, national commitments were codified in a different way. All members subscribed to common commitments in the form of targets and timetables; national and sectoral differentiation arose through the effort to cut emissions, which often revealed places of difficulty such as essential uses that required exemption (see chapter 2). This kind of differentiation of common commitments through local, reviewable experience is used in many different international agreements and institutions—for example, through Article IV of the International Monetary Fund. See Chayes and Chayes 1991. In Paris, by contrast, the differentiation is built into the pledges—the NDCs—along with the differentiation revealed through a national effort.

16. Not only was Paris itself negotiated under consensus rules, but the negotiations elaborating most of the key procedures (e.g., the detailed reporting via the NDCs and facilitative multilateral consultative process review) were subject to the same procedural rules as well. Those negotiations were launched right after the Paris Agreement was established. Hale 2016; Rajamani and Bodansky 2019.

17. At this writing, the rulebook negotiations on Article 6, which implicate international emissions trading and other topics, are still incomplete.

18. The role of consensus remains on display, at this writing, around Article 6, which many countries see as the mechanism through which emission trading might occur. There the controversies continue to simmer—partly because international emission trading remains an abstract possibility with few tangible benefits—and thus the rulebook has not been finalized. Stavins 2019; Cullenward and Victor 2020, chapter 6.

19. What can be observed so far is the evolution of the NDCs, which is not encouraging. We note, however, that some language of Article 13 of the Paris Agreement (as opposed to the narrow rules of Article 4) read together with the “country reports” produced under the UNFCCC, might be construed as enabling NDCs detailed enough to serve as the basis for experimentalist review.

20. “Update of Norway’s Nationally Determined Contribution” 2020.

21. OGJ editors 2020. For a discussion of Norwegian contributions to the Amazon Fund, see chapter 5.

22. Formally, the Global Climate Action predates Paris and was originally known as the Nonstate Actor Zone for Climate Action portal. It emerged, in part, from the massive attention to subnational and corporate action on climate change displayed at the UN General Assembly in September 2014. At the next COP following that event, in Lima, governments agreed on a process that became known as the Lima-Paris Action Agenda and from there the NAZCA portal.

23. In Glasgow the UN Secretary General announced plans for a new program that would review subnational policies—a good idea in principle, although one that can’t be made effective if it follows UN consensus rules.

24. For, more generally, the differentiation of commitments and expectations in the Paris process, see Voigt and Ferreira 2016.

25. Substantial resources are devoted to these, although the agendas are so broad that the procedures are quite far from the detailed postexperiment review that characterizes how the Montreal TOCs work. For a summary of the meetings in 2019, see UNFCCC 2019.

26. The climate change secretariat, which organized the Technical Expert Meetings (TEMs), was established first and foremost to facilitate negotiations, not to delve into actual substantive issues. The secretariat would basically line up a long list of speakers—parties and nonparties that, over a TEM, would talk about what they were doing and what worked (and didn’t work), but this involved long panel discussions in empty rooms. What could have evolved into something more experimental—a framing of possible new directions and advice to the parties about ways to adjust top-level commitments—was further encumbered by the fact that many governments sent diplomats to the TEMs, as opposed to actual experts or regulatory decision-makers grappling with the substance.

27. Adjacent to the Paris Agreement, however, are funding mechanisms that are, to varying degrees, gathering this information and, in their own diverse ways, operating more along the lines of experimentalism. For example, the long-standing Global Environment Facility includes some of this experimental learning, as do some of the multilateral development banks that operate in the same space. The Green Climate Fund, if not perennially beset with political difficulties and contestation, could evolve in this way. (That “if” is important, for an intergovernmental body designed like the Green Climate Fund might be incapable of operating in a nimble fashion.) Future research on the funding regimes that are adjacent to Paris—most of which have their own decision-making systems—should look at and explain the degree of experimentalism in these kinds of institutions. Experimentalism and learning in international funding mechanisms may prove to be one of the most important topics in international cooperation on climate change.

28. In advancing this argument, we resonate with the assertions made in Hale 2020; Young 2017. Where we differ is in our view of how the machinery that uses these goals is experimentalist in its functioning.

29. Of course there are many hybrid or intermediate cases, such as soybeans—a commodity—grown from seeds optimized to highly local conditions, or a standard piece of mining equipment customized to local geological conditions and regulatory requirements.

30. In the real world, technology, business, and social science will involve blends of these different functions; a revolution in electric vehicles, for example, involves a mixture of the first functions (e.g., battery and vehicle manufacture) and local contextualization (e.g., local charging networks). We articulate here the bookends to allow a focus on the functional attributes.

31. Emissions from electricity are harder to categorize. On the one hand, many key electric technologies are manufactured to global standards and sold in global markets (e.g., refrigerators and solar panels). On the other, the core challenge of decarbonizing electricity and expanding electric networks is a matter of contextualization, as we show in chapter 5. On the breakdowns of emissions by sectors, see Victor et al. 2014.

32. International Energy Agency 2020a, 2020b.

33. A few sector initiatives are emerging—for instance, the Oil and Gas Climate Initiative, a group of firms that aim to advance carbon capture and storage as well tighten controls on leaks of methane, a potent warming gas. The initiative invests in new technologies (and works with members that invest in those technologies as well). As is typical of such collaborations, there is frequent stocktaking (e.g., annually). In the world of Paris, by contrast, the stocktaking process, as we noted, runs every five years; controlled by governments, it is slow and broad, and likely to be encumbered by rules established through intergovernmental consensus.

34. Buchanan and Tullock 1962; Nordhaus 2015; Victor 2011.

35. This is a familiar problem in industry groups. See, for example, Rees 1996. It is also a problem internationally, often addressed in the past through hegemonic cooperation. For a look at this problem applied to climate change, see Victor, Geels, and Sharpe 2019.

36. On aviation, see, for example, Carroll 2021.

37. Maersk has pledged net-zero emissions by 2050, which means nearly immediate changes in its fleet since a typical ship lasts twenty to twenty-five years. See “IMO Agrees to CO2 Emissions Target” 2018.

38. As a practical example, see Martin 2020. For the theory, see Kahler 1992; Buchanan and Tullock 1962; Victor 2017. For the theory applied to the observable record, see Downs, Rocke, and Barsoom 1996.

39. The North Sea and Montreal Protocol cases—problem-solving through fine-grained sectoral approaches—are discussed in chapter 2.

40. Parson 2003.

41. Energy Transitions Commission 2020b; “Accelerating the Energy Transition” 2016; Victor, Geels, and Sharpe 2019.

42. A weakness for ARPA-E is that the DOE does not have reliable programs to help pick up these promising combinations of technological innovation and help them cross the “valley of death.” ARPA-E feeds them, gets them to the valley, and many then die for lack of demonstration support. For more on this, see Buchanan and Tullock 1962. And for proposals aimed at boosting energy innovation that are cognizant of this problem and thus aim for a huge ramping up of demonstration support, see, Sivaram, Friedmann, and Cunliff 2020; National Academies of Sciences, Engineering, and Medicine 2021a, 2021b.

43. National Academies of Sciences, Engineering, and Medicine 2021a; Larson et al. 2020; Sustainable Development Solutions Network 2020; Clarke et al. 2014.

44. In the lingo of deep decarbonization, these sectors are often called “hard to abate” precisely because the solution that is a likely winner in other sectors, namely electrification, is not plausibly available for these sectors. Energy Transitions Commission 2018.

45. At this writing, the European effort is the most credible, but complementary efforts in other countries are also emerging, notably in Japan and the United States. A degree of competition in creating what is widely seen as an industry of the future has spurred investment even as global supply chains for key hydrogen technologies (e.g., electrolyzers) will bring down the costs with scale just as occurred with solar power. On the solar experience, see Nemet 2019.

46. This problem of keeping a consensus club together exists not just for investing in technologies of the future but also for initiatives that help speed the exit of the old guard. For example, the European Bank for Reconstruction and Development has operated a pioneering program aimed at showing how to help countries move beyond coal by shutting down old facilities while taking care of affected communities—a program that where successful, would help open up frontiers for noncoal sources of energy. When the Trump administration came to power, it worked to block such programs, as the United States was a founding member of the European Bank for Reconstruction and Development, and only because other pioneering members of the bank and secretariat found ways to navigate around US intransigence was it possible for the beyond-coal campaign to continue.

47. Energy Transitions Commission 2020a.

48. The one notable exception to this statement is in China, where there are substantial frontier-pushing investments in climate-related areas (e.g., advanced nuclear power, electric vehicles, solar, and wind) that have historically had little to do with climate change, and are more firmly anchored in other overlapping concerns such as energy security, industrial development, and local air pollution. Over time, partly due to export opportunities, the Chinese investment has become more climate focused.

49. For example, in the oil and gas industry, the consortium invests in decarbonization technologies, including carbon capture and storage. Oil and Gas Climate Initiative 2019. And there is a parallel initiative focused just on methane—which began with detailed reporting about methane control methods, and then adjusted the central rules to require more information and action about the absolute volumes of methane emissions. UNEP 2020.

50. National Academies of Sciences, Engineering, and Medicine 2021b, especially chapter 3.

51. For more on the Regulatory Assistance Project, see https://raponline.org/.

52. Skjærseth 1998.

53. Ogden and Marano 2016; G20 Working Group on Energy and Commodities 2012.

54. For the best overview, see Ogden and Marano 2016. For the self-reviews and peer reviews, see China et al. 2016; Germany et al. 2016. For the Chinese and US self-reports, see United States 2015; China 2015. The US presidential election in 2016, just a month after the reviews were published, helped put the bilateral relationship into a deeper freeze, from which it has not recovered—a reminder that volunteer-based activities can help to demonstrate models and catalyze broader efforts, but are fickle when the volunteers focus on other tasks. Volunteerism must be followed by institutionalization. At this writing, it is not clear that the United States and China would be the ideal starting partners for this kind of mutual review because both countries are increasingly wary of the other; a bigger tent, such as a quadrilateral approach in the Pacific or transatlantic US-EU-UK trilateral system, might be more effective and less dependent on the already-stressed bilateral US-China relationship.

55. Azevedo et al. 2020b; Rabe 2004; America’s Pledge Initiative on Climate Change 2020.

56. Azevedo et al. 2020b.

57. C40 (with ninety-seven affiliated cities at this writing) has created a series of forums where city officials can compare notes on topics like land use planning and building efficiency; it has also created some standards for measuring emissions, with minimal impact. See C40 Cities, https://www.c40.org/research. In a few cities (eight at this writing), C40 sends advisers to consult on best practices. Local Governments for Sustainability (ICLEI), bigger and broader (twenty-five hundred cities, towns, and regions at this writing), is a little more active in assessing city-level performance, mainly for the purposes of offering thin case studies on success stories on topics such as efficient street lighting, fundraising, and urban planning—but there is no systematic performance review of the organization’s members and no requirement for review as a condition of membership.

58. Chayes and Chayes 1998.

59. Stone 2018.

60. Mallet and Khalaf 2020.

61. In particular, the Biden plan would “condition future trade agreements on partners’ commitments to meet their enhanced Paris climate targets.” “The Biden Plan” 2020. It is important not to overstate the credibility of this position, however, since at this writing (mid-2021), the Biden administration, now in power, also realizes that the growing momentum for trade measures in Europe is partly intended to punish the United States if it drags its feet. See Geman 2021.

62. Because Paris is a visible rallying point for action, it is mostly helpful in focusing political energy, but in some settings it has the opposite effect—such as with the Far Right in the United States, where the visibility of Paris made it easier to mobilize opposition, ultimately leading to the US announcement of withdrawal under Trump. That opposition and the conspicuous withdrawal, however, created a counterresponse (e.g., the “We Are Still In” movement) that induced further effort within the United States in states and localities as well as among financial regulators and firms. All told, that subnational action—itself inspired by Paris as the conscience of the world—is probably proving more durable than many federal actions that have been easy to reverse. Victor, Frank, and Gesick 2020.

63. Indeed, at this writing a “carbon border adjustment mechanism” is advancing rapidly in Europe with possible countermeasures also contemplated in the United States.

64. Bacchus 2018; WTO 2001a, 2001b.

65. The procedures for information provision and review are designed to make discrimination and tailoring to country circumstances difficult, except where the country itself might invite that. Some of the funding mechanisms active on climate-related topics are investing in local capacity building similar to the MLF-type funding mechanisms, among them the World Bank, European Bank for Reconstruction and Development, Inter-American Development Bank, and Asian Development Bank. A full analysis of the constellation of funding mechanisms relevant to climate is outside the scope of this book, but we think it’s a high-priority topic for future research.

66. Chayes and Chayes 1998.

67. Notably, see UNFCCC 2015, Article 15. This article is reminiscent of the “multilateral consultative process” of the UNFCCC (see UNFCCC 2015, Article 13), and offers a “facilitative” approach to implementation that in principle, is “expert-based” and attentive to “respective national capabilities and circumstances of the Parties.” A pioneer country or countries could demonstrate facilitated implementation along the lines we suggest and point to consistency with Article 15. We remain skeptical that a formally established Article 15 process—which would need to follow rules and procedures established through a mechanism like the rulebook—could actually function the way we suggest here.

68. Bagehot 1867.

Chapter 7. Piecing Together a More Accountable Globalization

1. This is a paraphrase of the aim of the European Union’s Carbon Border Adjustment Mechanism, which aims to induce all countries to adopt market-based mechanisms for controlling pollution and then adjust prices at the borders to equalize the impact of these policies on industry. While simple in theory, the market mechanisms proposed for comparing abatement levels across borders are similar to those introduced in the CDM and suffer, despite much subsequent refinement, from the same defects. The mix of national and local policies that affect the behavior of firms is impossible to observe from the outside, as are the counterfactuals, and thus identifying the border tariff equivalents will be unworkable. For more on the Carbon Border Adjustment Mechanism, see Abnett and Twidale 2021. For a more protectionist variant—a US proposal—see Coons and Peters 2021.

2. For an insightful review of the technocrats’ overconfidence in their predictions of the benefits of trade, obstinate lack of interest in evidence to the contrary, doubts about the ability of government to identify those harmed by its policies, but the conviction that compensation, if any, should be through the tax-and-transfer system, not the regulation of the economy, see Raskolnikov 2021.

3. Autor, Dorn, and Hanson 2013.

4. Case and Deaton 2015, 2020.

5. Rodrik 2011.

6. Rodrik 2011.

7. This continues to be the standard view among economists. For a recent restatement, see Nordhaus 2021.

8. Jennison 2013. For background developments in the European Union, see Soekkha 1997. For an overview of the agreement as an example of regulatory equivalence, see Sabel 2019. Like many such collaborations between national regulators outside preferential trade agreements, the EU-US Bilateral Aviation Safety Agreement was formalized as an executive agreement. Such agreements have their own legitimacy deficits. See Hathaway, Bradley, and Goldsmith 2020, 629. The implementation of joint regulatory programs across national borders can also be difficult because of the same kinds of organizational problems and conflicts of interest that make domestic reform difficult. For a sobering assessment of US-Canadian cooperation on food safety, see Kerr and Hobbs 2021.

9. On Australia and New Zealand as leaders for the adoption of Hazard Analysis Critical Control Point food safety standards starting in the 1990s, see Ropkins and Beck 2000.

10. This draws on Hoekman and Sabel 2019. For the OPA concept as it is emerging in trade talks, see, for example, Buchanan 2020.

11. Just how closed PTAs can be is demonstrated by Great Britain’s continuing difficulties in finding any commercially feasible alternatives to the continuing, deep engagement with EU regulators—at odds with the reassertion of national sovereignty that motivated Brexit. The European Union’s accession and neighborhood policies are the exceptions that prove the rule as it focuses on building the capacity of candidate members to meet regulatory requirements. For a nuanced discussion, see Lavenex 2015.

12. Hoekman 2016; Hoekman and Sabel 2019.

13. For an account of how the imposition of such a choice on the peoples of eastern Europe after the fall of the Berlin Wall allowed populist parties to present themselves as champions of national values against predatory cosmopolitan elites, see Krastev and Holmes 2019.