Policy Research Working Paper 8781 Firms’ and States’ Responses to Laxer Environmental Standards Tito Cordella Shantayanan Devarajan Development Economics Vice Presidency March 2019 Policy Research Working Paper 8781 Abstract On June 1, 2017, President Trump announced the United profit-maximizing firms. It finds that a relaxation of States’ withdrawal from the Paris agreement on climate emission standards (i) may increase firms’ incentives to change. Despite this decision, American firms continued adopt clean technologies, but not to pollute less; (ii) may investing in low-carbon technologies and some states com- negatively affect industry profitability if it is perceived as mitted to tougher environmental standards. To understand temporary; and, when this is the case, (iii) the unilateral this apparent paradox, this paper studies how a weaken- adoption of stricter standards by large states may increase ing of environmental standards affects the behavior of the expected profitability of every firm. This paper is a product of the Development Economics Vice Presidency. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at tcordella@worldbank.org and sdevarajan@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Firms’and States’Responses to Laxer Environmental Standards Tito Cordellayand Shantayanan Devarajanz JEL Classification Numbers: F53, L51, Q54, Q55, Q58, H75 Keywords: Environmental Policies, Technology Adoption, Paris Agreement, Subnational Regulation, Policy Reversal We would like to thank Quy-Toan Do for constructive comments and suggestions. Woori Lee provided outstanding research assistance. The usual disclaimers apply. y The World Bank, tcordella@worldbank.org z The World Bank, sdevarajan@worldbank.org 1 Introduction President Donald Trump’ s announcement on June 1, 2017 that the United States would withdraw from the Paris agreement was a setback to the global e¤ort to mitigate climate change. Curtailing greenhouse gas (GHG) emissions is a global public good that, in the absence of public action, will be underprovided by the private sector. Furthermore, the U.S. is a large economy, so increased GHG emissions by American …rms in the wake of relaxed environmental regulations could accelerate climate change further. Against this background, the International Energy Agency’ s (2018) report on Global Energy and CO2 showed that the share of renewables in electricity generation reached a record level of 17 percent in the U.S. The major oil companies, including ExxonMobil, have continued to invest in low-carbon technologies. Moreover, in September 2018, the then-governor of California (the largest U.S. state), Jerry Brown, signed an executive order committing the state to being carbon-neutral by 2045. As Hultman and Bodnar (2018) put it, “Trump tried to kill the Paris agreement, but the e¤ect has been the opposite.” These actions may be due to the public-spirited behavior of …rms and state governments. We show in this paper that they could also be the result of pro…t-maximizing behavior and successful lobbying e¤orts of …rms in the face of weaker environmental standards. First, we …nd that a decision to relax emission standards could increase …rms’incentives to adopt clean technologies. However, the overall level of emissions may increase, since the cleaner technologies enable …rms to produce more. Secondly, industry pro…tability may be reduced by the weakening of environmental standards if there is a chance that the decision will be reversed later. In this case, …rms in large states have an incentive to lobby state governments to implement environmental standards that are stricter than the federal ones. If the chances of policy reversal go down, the state’ s regulations will approach those of the federal government. In short, while the seemingly “green” actions following the U.S.’ s withdrawal from the Paris agreement can be interpreted in terms of pro…t-maximizing behavior of …rms, there is no guarantee that they will translate into lower emissions. Of course, we are not the …rst to study how emission standards a¤ect …rms’technology choices.1 Van der Zwaan et al. (2002) introduce endogenous technological change in a macroeconomic model of climate change, and argue that the development of carbon-free technologies plays a critical role in carbon reduction and allows for lower optimal carbon taxes. Acemoglu et al. (2012) show how temporary environmental regulations– a combination of research subsidies and carbon taxes– can redirect technical change toward clean technologies and avoid environmental disasters. Acemoglu and Rafey (2018) study the e¤ects of geoengineering in a model in which the government cannot commit to future policies. They show that, when this is the case, the anticipated reduction in future taxes may discourage …rms from adopting cleaner technology.2 Particularly relevant to our analysis is Krass et al. (2013) who look at …rms’reaction to environmental regulations (tax, subsidy, and rebate levels) and show that increasing carbon taxes does not monotonically increase …rms’adoption of green technology. The intuition is that when taxes are very high, …rms’production quantity can be so small that switching to cleaner technology generates insu¢ cient additional pro…t to o¤set the …xed cost of adopting the technology. Perino and Requate (2012) also show that the relationship between policy stringency and the rate of technology adoption is an inverted U-shape for a broad class of technologies. Our model shares such features. On the empirical side, Aghion et al. (2016) show that, in the auto industry, …rms are more likely to innovate in clean technologies if they are confronted with higher tax-inclusive fuel prices. Yang et al. (2012) examine whether stringent environmental regulations induce more R&D and promote productivity in Taiwan, China’ s manufacturing industry and …nd that stricter regulations are positively related to R&D expenditures and industrial productivity. 1 See Ja¤e, Newell, and Stavins (2002), Löschel (2002), and Requate (2005) for an overview. 2 Theidea that the possibility of a policy reversal may turn detrimental an otherwise sensible reform was …rst discussed by Rodrik (1991). 1 2 The Model We consider an economy where a continuum of …rms (of mass one) produce a …nal good, which is sold in the world market at a price p, which we normalize to one. Denoting output by x, and assuming a convex cost function c(x) = 1 2 2 x , the pro…ts of …rm i at time t are given by x2 it it (xit ) = xit : (1) 2 Production causes carbon emissions, eit , which are proportional to output levels and depend on the technol- ogy adopted by the …rm. More precisely, we assume that eit = kj xit , j 2 fd; cg, where d stands for dirty, and c for clean technology,3 and kd > kc . The cost of adopting the clean technology, i , varies across …rms and we let i be uniformly distributed over the interval [ i , i ]. We further assume that there are no private costs associated with emissions, and that there is no cap and trade system that allows …rms to sell “pollution rights.” Hence, …rms adopt the clean technology only if it yields higher pro…ts. The timing of the model is as follows: At time t = 1 – The government announces a carbon emission limit ! BTt , for t = f0,1g, that all …rms should abide by. The limit is consistent with a carbon emission target that is exogenously decided (e.g., as part of an international agreement); – Firms announce whether they are considering adopting a (costly) clean technology in period 0 or not (cheap talk). At time t = 0 – Unexpectedly, a new government gains power and tweets a new carbon emission cap ! T t , which is not consistent with , that is, ! T t > ! BT t . – Firms revise their plans of whether to adopt the clean technology or not; – Firms make production plans. At time t = 1 – With probability 1 , the current government is replaced by a new government that sets a carbon emission limit ! AT that is consistent with ; – If this happens, …rms adjust their production plans, but not the technology that is already set, according to the new limit. 2.1 Before Transition (t = 1) Recall that the government is committed to cap the sum of total carbon emissions in periods 0 and 1 to . Absent carbon emission limits, …rms have no incentives to adopt the clean technology and they produce bt = arg max x t (xt ) = 1: (2) x Denoting total “unregulated” carbon emissions by e , we have that e = kd (x b1 ) = 2kd ; no …rm will b0 + x upgrade to the clean technology because it is costly (see below). If a limit ! t is put in place, the output of a …rm of type j is given by !t xjt (! t ) = M inf1; g: (3) kj 3 For real world examples, one could look at industries where emissions are mostly associated with the shipment of the …nal good and …rms can either use fuel e¢ cient or fuel ine¢ cient means of transportation. 2 To simplify the analysis, we consider situations in which the government is committed to a target that is stringent enough. More precisely, we assume that < 2kc , (A.1) so that, even if all …rms switched to the clean technology, a carbon limit will still be needed to meet the target.4 Notice that this implies that kc > ! t . We also assume that the government cannot verify the technology adopted by the …rms and cannot impose a limit that is contingent on the technology adopted. Under such assumptions, and assuming no discounting, total pro…ts of a …rm of type j are given by !0 !1 j = j0( )+ j1( ): (4) kj kj We further assume that the government is benevolent, and that it chooses ! 0 and ! 1 to meet the target at the smallest cost for the …rms. In our set-up, where all …rms share the same concave pro…t function, the least costly carbon limits that are consistent with are necessarily BT ! = ! 0BT = ! 1BT = =2: (5) Substituting now (5) into (4), we can express the variable pro…ts of clean and dirty …rms as BT (4kc ) c = 2 ; (6) 4kc BT (4kd ) d = 2 ; (7) 4kd respectively. Hence, a …rm adopts the clean technology i¤ the cost i of adopting the clean technology for …rm i is low enough, that is, BT BT (kd kc ) (4kc kd (kc + kd ) ) BT i < c d = 2 k2 ; (8) 4kc d BT and the fraction of …rms that adopt the clean technology is given by 8 < BT0; > if BT < i ; BT = i ; if BT 2 ( i ; i ); (9) > : i i 1; if BT > i : 2.2 Transition We now consider a situation where a new government, unexpectedly voted into o¢ ce in period 0, decides that the carbon target is no more binding and that carbon emission limits can be increased. How would such a policy a¤ect …rms’decisions to invest in clean technologies and, ultimately, their pro…tability? T To answer this, assume that the government decides to depart from and sets a new emission cap such that T = + z: (10) For the sake of simplicity, and recognizing that it is often di¢ cult to completely rewrite a regulatory frame- work, we start by considering the e¤ect of small changes in regulation. Hence, we assume that z < 2kc ; (A.2) 4 Suchan assumption greatly simpli…es the analysis ensuring that the emissions of …rms adopting clean and dirty technologies are the same. When this is the case, the limit that is consistent with the target does not depend on the fraction (see below) of …rms that adopt the clean technology. We will relax this assumption in section 5. 3 so that, even if all …rms switched to the clean technology, a carbon limit will still be needed to meet the revised target. Again, if the government continues choosing ! 0 and ! 1 to meet the target at the smallest cost for the …rms,5 we have that +z ! T = ! 0T = ! 1T = : (11) 2 As we already mentioned, the change in the regulatory framework can be temporary. Indeed, in period 1, with probability (1 ), the government will be voted out of o¢ ce. When this happens, the new government sets an emission limit ! AT that restores the original commitment of capping total emissions at . Of course, such a limit is stricter than ! BT because it has to compensate for the increase in period 0’ s emissions. We thus have that z ! AT = !T = ; (12) 2 and we work under the assumption that z< ; (A.3) so that ! AT > 0. The expected variable pro…ts of a …rm of type j are then given by !T ! AT E[ j] = (1 + ) j( ) + (1 ) j( ), (13) kj kj where E is the expectation operator and !T (4kj z )( + z ) j( ) = 2 ; (14) kj 8kj AT ! (4kj + z )( z) j( ) = 2 : (15) kj 8kj Using these expressions, (13) can be written as 2 4kj ( + z ) 2 z z2 E[ j] = 2 : (16) 2kj Hence, …rm i will adopt the clean technology i¤ (kd kc )( (4kc kd (kd + kc ) ) 2 z ((kd + kc ) 2kc kd ) (kd + kc )z 2 ) T i < E[ c] E[ d] = 2 k2 ; 4kc d (17) T and the fraction of …rms that adopts the clean technology is given by 8 < T 0; > if T < i ; T = i ; if T 2 ( i ; i ); (18) > : i i 1; if T > i : 3 Firms’Responses In the remaining of the paper, we discuss how a relaxation in the regulatory standards is likely to a¤ect …rms’incentives to switch from the dirty to the clean technology and, ultimately, their pro…tability. We also discuss whether, and under which circumstances, a subset of …rms may …nd it to be in their self interest to abide by standards that are stricter than the prevailing ones. 5 In setting its policies, the government assumes that it will remain in power in the next period with probability one. 4 3.1 Cleaner or dirtier? In order to assess how a relaxation in environmental standards may a¤ect …rms’ decisions regarding tech- nology adoption, we focus our attention on the interesting case in which BT 2 ( i ; i ) and thus a fraction BT > 0 of …rms would be interested in adopting the clean technology before the change in regulation occurs. We now ask the question of whether the relaxation in the emission cap will increase or decrease the number of …rms that decide to switch to the clean technology. Our main …nding is that: Proposition 1 I¤ the original carbon emission target is stringent enough, < b 2 kc k d kc +kd , a small e increase in the emission cap z , z < z (b ), induces more …rms to adopt the clean technology; a large increase has the opposite e¤ ect. The higher is the probability that standards are not reversed in the next e. period, the larger is z Proof: In Appendix. Notice that Proposition 1 does not mean that a relaxation in the emission cap, by inducing …rms to adopt a cleaner technology, is good for the environment. In our set-up, the total amount of emissions is independent of the choice of the technology and higher caps necessarily mean more emissions. The reason why …rms may be induced to switch to the clean technology is that, when < b , an increase in the emission limits increases the pro…ts associated with the use of the clean technology more than those associated with the dirty one.6 This follows directly from the concavity of the pro…t function and the linear cost of adopting the clean technology.7 In addition, the shift in technology adoption induced by the relaxation of the emission standards is magni…ed, in either direction, if the probability that the government remains in power increases, so that policy reversals become less likely.8 Figure 1 below, illustrates the di¤erent forces at play. If emission standards are pretty tight to start with, = a , their relaxation by an amount z increases the pro…ts of clean …rms (in green) more than those of the dirty ones (in brown). This induces more …rms to adopt the clean technology and the more so if current policies are likely to stay. However, if emission standards are initially laxer, = b , their relaxation by the same amount z , increases the pro…ts of the dirty …rms more than those of the clean …rms, and this induces more …rms to keep the current dirty technology. Indeed, what drives the …rms’ technological decision is the comparison between the …xed cost of adoption and the di¤erence in the pro…ts associated with the two technologies. Let us now consider the e¤ect of a change in emission standards on industry pro…tability. Total expected pro…ts E [ ] are given by Z T T T T 1 E[ ]= E[ c] + (1 )E [ d] id i, (19) i i i where the last term denotes the cost of adopting the clean technology (for the …rms that adopt it). We can now prove that Proposition 2 If the probability that the current government remains in power in period 1 is small enough, < (2k z ) , then a relaxation of the emission limit decreases the pro…tability of the industry; the opposite d is true if > (2k z ) . c 6 Indeed, @E [ T c ] (2kc ) z @E [ T d] (2kd T ) z @E [ c ] @E [ T d] we have that @z = 2 2kc and @c0 = 2 2kd , @z > @z e. () z < z 7 To get a better understanding of the di¤erent forces at play, we considered a more general model, à la Perino and Requate (2012) where pro…t functions are concave, and …rms choose among a continuum of technologies whose cost of adoption is increasing and convex in how green they are. In such a model, the e¤ect of an increase in an emission quota depends upon the third derivative of the pro…t function. In the case of a quadratic pro…t function, a small increase in the emission quota leads to an increase in the adoption of green technologies, while a larger increase has the opposite e¤ect (as in our model). Instead, with constant returns to scale, …rms always adopt greener technologies when quotas are relaxed. We thank Quy-Toan Do for pointing this out. 8 It is immediate to verify that @ BT > 0 () < b. @z@ 5 Figure 1: Change in environmental standards and pro…tability Proof: In Appendix. Proposition 2 suggests that changes in emission standards that are designed to increase …rms’pro…tability may end up having the opposite e¤ect if the probability that they are reversed in the future is high, and technology cannot quickly adapt to the new regulations. The reason is that, in order to undo the damages that current policies are going to in‡ict on the environment, …rms may end up facing draconian emission cuts in the future and this may negatively a¤ect their (expected) pro…tability; this despite the fact that (expected) emissions also increase. Finally, it may be worth remarking that the conditions set in Proposition 2 are su¢ cient conditions for an increase and a decrease in pro…tability, as they require all …rms to gain or lose from the relaxation in the emission standards. This greatly simpli…es the analysis as it allows us to ignore the e¤ects of technological changes (induced by the change in policies) on the industry’ s pro…tability. We will discuss this in the robustness section below. 4 The California E¤ect The question we discuss in this section is whether a subset of …rms may …nd it in their self-interest to abide by emission standards that are stricter than those imposed by the government. We also discuss whether such a strategy is more likely to pro…t those …rms that adopt (or are prone to adopt) the clean or the dirty technology. The fact that …rms may want stricter standards (if they anticipate that the current ones may be reversed) follows directly from Proposition 2. Indeed, when all …rms are made worse o¤ by the changes in the emission standards, they are necessarily better o¤ if they can reverse them. Hence, the grand coalition of …rms would necessarily …nd it in its self-interest to abide by stricter emission standards. The problem is that such a coalition is not stable: any (in…nitesimal) …rm will be better o¤ by deviating from the stricter standards 6 and free riding on the grand “green” coalition. This means that self-imposed standards may not work, and stricter standards may need to be enforced by subnational authorities, such as states. And if a state is large enough, it may …nd it in the interest of its …rms to issue tougher emission standards even if other states do not follow. But let us proceed by steps. First, we want to …nd out what is the smallest subset of …rms , 2 [0; 1], that would …nd it optimal to abide by stricter emission standards, if such standards were externally enforced. If a subset of …rms decides to abide by a tougher standard ! T " and join the “green coalition,”we would have that e AT = ! (! T ") (1 )! T : (20) Now, the total expected variable pro…ts of a …rm of type j belonging to the “green coalition” are given by T AT t " ! " " !e E[ j] = (1 + ) j( )+ j( ), (21) kj kj with " ! " (4kj + 2" z )( 2" + z ) T j( ) = 2 ; (22) kj 8kj AT " !e (4kj 2 " + z )( + 2 " z) j( ) = 2 : (23) kj 8kj We can then prove Proposition 3 If the probability that the current government remains in power in period 1 is small enough, < 2k z , there exists a < 1, such that any coalition of …rms of type j and of size > would gain j from a stricter emission cap. is smaller for the …rms adopting a clean technology. Proof: In Appendix. According to Proposition 3, if current lax standards are likely to be reversed in the future, then a subset of …rms may prefer to abide by stricter emission standards today and tomorrow to be able to better “smooth” emissions (and production) over time. The problem, as we already mentioned, is that such a coalition is not stable. However, states may impose tougher standards (even if this is currently disputed in the U.S.). Do they have an interest in doing so? If states are small, they will face a free-rider problem similar to the one faced by individual …rms. However, if a state is big enough, then it can unilaterally decide to impose stricter standards, and this may be in its …rms’interest even if other states do not follow. Notice that, in Proposition 3, we only considered coalitions of homogeneous …rms, that is, of …rms that have adopted either a clean or a dirty technology. The analysis can easily be extended to the case of mixed coalitions. In this case, the larger is the share of …rms adopting the clean technology, the smaller is the coalition of …rms that …nd it in their self-interest to abide by stricter emission standards. This, in turn, implies that the larger a state is, and the larger is its share of clean …rms, the more likely it is that such a state is willing to impose stricter environmental standards. This is what we christen as the California e¤ect. 5 Robustness 5.1 Large Changes in Regulation In the previous analysis, we focused our attention on small changes in regulation assuming that z < 2kc . In this section, we relax such an assumption and allow for any regulatory change that is compatible with assumption (A.3). First, it is important to notice that if the new emission cap is such that + z > 2kd , 7 then it is non-binding and we can ignore it. Hence, in what follows, without loss of generality, we restrict our attention to the case in which 2kd > z > 2kc . When z belongs to such an interval, the emission limit is binding in period 0 only for the …rms that T operate with the dirty technology. This implies that x bc T = 1, xbdT = 2+ z kd and hence that b c = 1 2, (4kd z )( +z ) bd = 2 8kd . b = bkc + (1 In addition, we have that e b) + z 2 , so that, in period 1, if a new b AT = government is voted into o¢ ce, it will set a limit ! b. Thus, b T is the i that satis…es both e i = E [bc T (! b AT )] E [bdT (! b AT )]; (24) AT b ! = b(b); e (25) bT where b = i . i i While we are able to solve explicitly for the share of …rms that adopt the clean technology and thus for the industry pro…ts associated with di¤erent changes in emission standards, the expressions are quite convoluted. We thus present our results with the help of numerical simulation. They are summarized in the …gures below, where we plot b and E [ ] as a function of z , in the interval [0, ], for di¤erent values of the probability that new emission standards would last, . In Figure 2, we set < b , and in Figure9 3, > b . In the case of stringent emission caps, Figure 2, the relation between the number of …rms that adopt the clean technology and the emission limits is hump shaped. The higher is the probability that policies are not reversed, the higher is the fraction of …rms that adopt the clean technology. The e¤ect of the relaxation of the emission standards on industry pro…tability also depends upon the probability that the new standards are going to be repealed. The higher the latter is (low value of ), the more likely it is that laxer emission standards negatively a¤ect industry’ s pro…tability. Let us now move to the situation, depicted in Figure 3, where the original emission targets were less stringent. Also in this case, higher emission caps increase expected pro…ts if they are likely to be permanent, and they reduce them if the probability that they are reversed is high. The number of …rms switching to clean technology decreases the laxer the new standards are. Finally, we …nd that, for low values of z; the higher the likelihood of a policy reversal, the higher is the number of …rms adopting the clean technology. The opposite happens for high values of z . 6 Conclusions In this paper, we have attempted to explain two, somewhat surprising, developments following President Trump’ s decision to withdraw the United States from the Paris accord on climate change. First, private …rms have continued to invest in clean technologies. Second, some state governments have decided to impose stricter environmental standards in the wake of weaker standards at the federal level. Using a simple model, we showed that pro…t-maximizing …rms could increase the use of clean technologies because the increased output that these technologies permit raises pro…ts by more than what the dirty technologies allow. This e¤ect is stronger if …rms believe that the change in the standards is permanent. However, we also showed that, if the relaxed emissions standards are likely to be reversed, industries will face a decline in expected pro…ts— unless …rms are able to form a coalition to adhere to stricter environmental standards. Since such a coalition is unstable, one possibility is for state governments, especially those of large states, to impose the standards. While these explanations stem from a simple model, they provide important insights not just about the mechanisms through which environmental policy a¤ects …rms’and industries’decisions, but also in how we should interpret these developments following the U.S.’ s withdrawal from the Paris accord. As the model shows, the increased use of clean technologies could be accompanied by an increase, rather than a decrease, in overall emissions. Hence, we should not be complacent about the fact that private …rms are investing 9 As per the parameter values, we set kd = 2, kc = 1, = 2, = 0. In Figure 2 we chose = :75, while in Figure 3, we i i have that = 1:5. 8 Figure 2: Stringent Initial Emission Targets Figure 3: Less Stringent Initial Emission Targets 9 more in clean technologies. Similarly, if the stricter environmental standards at the state level are the result of lobbying by …rms whose expected pro…ts would otherwise be lower, then we should be leery about the underlying reason for the lower expected pro…ts— the possibility that the relaxed environmental standards may be reversed. If this possibility is reduced, then the mechanism could work in the opposite direction and states may adopt laxer environmental standards.10 In sum, notwithstanding some encouraging developments about which we now have a better understand- ing, President Trump’ s decision to leave the 192-nation coalition on climate change may still undermine progress towards mitigating global warming. 1 0 Of course, this is would not be the case if …rms acted in a socially responsible way, see, for instance, Campbell, (2007). 10 References [1] Acemoglu, D., Aghion, P., Bursztyn, L. and D. Hemous (2012), “The Environment and Directed Tech- nical Change,” American Economic Review 102: 131– 66. [2] Acemoglu, D., and W. Rafey (2018), “Mirage on the Horizon: Geengineering and Carbon Taxation Without Commitment,” NBER Working Paper Series 24411. [3] Aghion, P., Dechezleprêtre A., Hémous D., Martin,R., and J. Van Reenen (2016), “Carbon Taxes, Path Dependency, and Directed Technical Change: Evidence from the Auto Industry,” Journal of Political Economy 124: 1– 51. [4] Ja¤e, A., Newell,R., and R. Stavins (2002), “Environmental Policy and Technological Change,” Envi- ronmental and Resource Economics 22: 41– 70. https://doi.org/10.1023/A:1015519401088. t,” The [5] Bloomberg M. and J. Brown (2017), “The U.S. Is Tackling Global Warming, Even if Trump Isn’ New York Times, Nov. 14, 2017. [6] Campbell, J. L. (2007), “Why Would Corporations Behave in Socially Responsible Ways? An Institu- tional Theory of Corporate Social Responsibility,” Academy of Management Review, 32: 946-967. [7] Hultman, N. and Paul Bodnar (2018), “Trump Tried to Kill the Paris Agreement, but the E¤ect Has Been the Opposite,” Friday, June 1, 2018 [8] IEA (2018), “Global Energy and CO2 Status Report,” https://www.iea.org/geco/. [9] Krass, D., Nedorezov, T. and A. Ovchinnikov (2013), “Environmental Taxes and the Choice of Green Technology,” Production and Operations Management, 13:1035-55. [10] Löschel, A. (2002), “Technological Change in Economic Models of Environmental Policy: A Survey,” Ecological Economics 43: 105– 26, https://doi.org/10.1016/S0921-8009(02)00209-4. [11] Meyer, R. (2017), “Most Americans Support Staying in the Paris Agreement,” The Atlantic, May 31, 2017. [12] Perino, G. and T. Requate (2012), “Does More Stringent Environmental Regulation Induce or Reduce Technology Adoption? When the rate of Technology Adoption Is Inverted U-Shaped,” Journal of Environmental Economics and Management,” 64: 456-67. [13] Plumer, B. (2017) “What to Expect as U.S. Leaves Paris Climate Accord,” The New York Times, June 1, 2017. [14] Requate, T. (2005), “Dynamic Incentives by Environmental Policy Instruments— a Survey,” Ecological Economics, Technological Change and the Environment, 54: 175– 95. [15] Rodrik, D. (1991), “Policy uncertainty and private investment in developing countries,” Journal of Development Economics, 36: 229-42. [16] Van der Zwaan, B. C., Gerlagh, R. and L. Schrattenholzer, (2002),“Endogenous Technological Change in Climate Change Modelling,” Energy Economics 24: 1– 19. 11 7 Appendix Proof of Proposition 1 @ T 1 @ T From (18) we have that @z = @z . Hence, di¤erentiating (17), we have that i i BT @ (kd kc )(2kc kd (kd + kc ) (kd + kc )z ) = 2 k2 ( ; (26) @z 2kc d i i) from which it follows that BT @ 2kc kd z > 0 () < : (27) @z kc + kd Hence, if < 2 kc k d b ; there exists a non-empty interval [0; z e e), with z (b e), ) such that, if z 2 [0; z kc +kd BT @ @ze @z > 0. The fact that @ > 0 completes the proof. Proof of Proposition 2 T T @E [ ] Notice that a su¢ cient condition for @E@z[ ] Q 0 is that @z j Q 0 for j = c; d. Indeed, if the relaxation of the emission limits increases the pro…tability of all …rms for given technologies, then, a fortiori, it increases the industry pro…tability when …rms can optimally choose technology. Moreover, if the relaxation of the emission limits decreases the pro…tability of both the …rms that adopt the clean and those that adopt dirty technology, it is trivial to show that allowing …rms to switch technology cannot completely reverse the e¤ect. Di¤erentiating (14) with respect to z , we have that @E [ T j ] (2kj ) z = 2 : @z 2kj @E [ T d] z Hence, a su¢ cient condition for @z < 0 is that (2kd ) z < 0 () < (2kd ) e1 : Instead, a T @E [ j ] z su¢ cient condition for @z > 0, is that (2kc ) z > 0 () > (2kc ) e2 , with 0 < e1 < e2 < 1, because of (A.2). Proof of Proposition 3 A su¢ cient condition for the existence of a coalition of …rms of type j that is willing to abide by stricter regulatory standard is that @ "j Lim > 0. "!0 @" Substituting (22) and (23) into (21), we have that " 1 2 2 E[ j] = 2 (2(1 + (1 ) ) " 2(1 + m + (1 ) )"2 (28) 4kj 2 z + 2(1 + + (1 ))"z z 2 + 4kj ( (1 + (1 ) )" + z ): Di¤erentiating (28) with respect to ", we have that @ "j 2kj (1 + (1 ) )+T 2" + z + ( 2 " + z ) + ((1 + )T 2"(1 2 ) + z (1 ) = 2 ; @" 2kj (29) so that @ "j (2kj )((1 ) (1 + )) + ( (1 ) + (1 + ))z Limit = 2 : (30) "!0 @" 2kj 12 Furthermore, we have that 2 (1 + )(2kj z) > 0 () > 2 ; (31) (1 )(2kj + z) and z < 1 () < : 2kj @ This, together with the fact that @kj < 0, completes the proof. 13