In the paper “Climate Policy Must Favour Mitigation Over Adaptation”, I develop the argument that the world must prefer mitigation to adaptation at the global level. The argument rests on the observation that mitigation is a public good while adaptation is a private one. I have the feeling that the academic literature has completely missed this point, and especially the representative agent literature, but also integrated assessment models that introduce adaptation and mitigation.

Read on for more details, why I have trouble to get this published in a journal, the referees’ comments and my rebuffals.

In the paper I develop a theoretical model that shows how important it is to distinguish between the public and private good nature of mitigation and adaptation. I also take the theoretical model to the data and find that a representative policy maker who does not distinguish between private and public goods would invest in both adaptation and mitigation, just as the previous literature recommends. However, a representative policy maker who properly models mitigation as a public good and adaptation as a private one would optimally only invest in mitigation. I then discuss what this implies for the current state of the art literature and what should be the lesson for future research.

Please take a look at my paper for more details: MitigationAdaptation2017.

This paper is what I view as my nicest policy contribution for now, but at the same time it is also my problem child. I wrote it in 2015, but could not get it published since. I tried Journal of Environmental Economics and Management, JAERE, Resource and Energy Economics, Ecological Economics, Environment and Development Economics, even Climate Policy. It took three years to submit and get reports from these journal.

Now, dear reader, I would be happy if you can help me. What do you think about this paper? What are the problems? Dear editors – would you want to publish this paper? Why, why not? I would of course publish it immediately, but I am also a tiny bit biased.

Here is a list of the (slightly abriged) referee comments up to now, with my remarks, so if you are a referee of my paper in the future, and you happen to believe one of the points below is worthwhile to write in your report, please read my remarks/rebuffals first:

Referee: The suggested problem is not (as argued in the paper) that the public good character of mitigation is not observed, but that “representative agents models up to now” consider adaptation to be a public good. The formalism considered implies that “representative agents models up to now” assume that an agent does not only benefit from reinforcing one’s home against storm damages, but also from the reinforcement of other peoples’ homes. I really believe that this is not a correct description of the “representative agents models up to now”. I doubt very strongly that there is single representative agents model that has made the mistake of assuming that everybody benefits from reinforcing a single home against storm damages.

Me: I refer this reader to section 4 in my paper, where I show that all representative agent models have used this assumption. If one is a referee and forwards such a strong claim, at least one should be aware of the literature and not just believe.

Referee: The results are also intuitive, and I’m sure correct. The paper does indeed make a contribution, and it ought to be published somewhere. The model is simple, and maybe too simple. One point, for example, is that mitigation expenditures come right away, to prevent future climate change, while adaptation expenditures can be delayed until facing threats of rising sea levels, flooding, droughts, etc. Thus the discount rate must affect this choice in a major way that cannot be captured in your static model.

Me: Yes, the discount rate will have an effect. But only a quantitative one, not a qualitative one. Thus, the main argument forwarded in the paper still holds. See my points on p18 and 19 of my paper.

Referee: Your introduction states that your question is as follows: “If our climate policy were to be motivated by the first best, should adaptation be part of the policy mix or not?” That is a reasonable question, but of course the answer depends a lot on the assumptions, as you show using all the corner solutions in four results of proposition 1 on pages 9-10. Given the need for assumptions and calibration, I think the conclusions in your final paragraph on page 33 are probably too strong (that climate policy “must primarily consist of mitigation actions”).

Me: Of course, results always depend on assumptions. But this model is so general, uses so few assumptions, that the results should be general enough. Plus, I extend the complete representative agent literature on adaptation versus mitigation, and the integrated assessment models, to this public versus private good setting, and show that my results continue to hold. Furthermore, the calibration I only use to give a quantitative idea about the result. The theoretical result of course still goes fully through.

Referee: The way you implement that point on pages 12-13 is by looking at optimal provision with increases in N, the number of people. But it’s not clear that more N inherently changes the “public good character of mitigation”. It is still a pure public good, which means that the social marginal cost of another user is zero. Adaptation has an additional cost per person, and mitigation does not, so clearly having more people would favor mitigation over adaptation. But the question being answered is “what is the effect on optimal policy when the world has more people?”, and it’s not clear how your answer to that question relates to the question laid out in your introduction.

Me: This is a point I admittedly have trouble to bring across. I am not changing the number of people in the world. I am teasing out how different assumptions on public and private goods affect the optimal policy mix. Think of it this way. At the one extreme, a representative agent model has only one representative agent, and thus this whole literature models adaptation and mitigation as being directed to one single agent. Thus, there is no difference between public and private goods. At the other extreme, if we write a social welfare function that takes into account all individuals on this planet, then adaptation would be a purely private good while mitigation would be a purely public good (in the extreme). My modeling approach allows us to look at any combination between these two extreme cases (as I want to e.g. be able to compare to the RICE model, or a country-level model). Thus, I am not looking at the optimal policy when the world has more people, but I am looking at the optimal policy if we believe adaptation has a fully public good nature, or it works at the regional level, at the country level, or the individual level.

Referee: It is assumed that world income is exogenously given. This is in contrast to previous theoretical models which incorporate the fact that income generation is determined endogenously with mitigation since emissions are typically modeled as a by-product of either production or consumption, which is not the case for adaptation. It seems important to acknowledge that mitigation and adaptation affect world income differentially. Taking this into account may affect the main result, and the numerical example.

Me: This may lead to a quantitative change, but not a qualitative one. See also my extensions in section 4, p.22ff, where I should that the results still hold

Referee: The model does not account for heterogeneity across regions in terms of damage from climate change. Even at the globally optimal level of mitigation, it may be beneficial for some regions to spend on adaptation, given that they are more vulnerable to damage. Thus, it may not be appropriate to argue that “adaptation is… only a testimony of mankind’s inability to cooperate”, as stated in the conclusion. In fact, international organizations’ efforts to fund adaptation in the most vulnerable developing countries may reflect the opposite. Moreover, taking heterogeneity into account would yield more meaningful numerical results.

Me: I am only looking at the social planner perspective. And yes, heterogeneity in regions could be taken into account, but there is not going to be a qualitative change to my results then. I think it is necessary that my point, that adaptation is predominantly a private good while mitigation is a public one, needs to be accepted in the literature, and that this has vital implications especially for the representative agent models.

Referee: The full effects of mitigation are uncertain and would only take effect in the distant future, whereas those of adaptation are often more immediate and certain. Taking this into account, a global social planner may be more open to adaptation than presented in the paper.

Me: Yes, I discuss this in the paper. Still, there is no qualitative change to the results.

Referee: The model assumes that mitigation and adaptation are essentially substitutes across which an exogenously given budget must be allocated. This may not always be the case. As the paper mentions, while mitigation is typically funded publicly, adaptation is often funded privately. Therefore, the budget sets from which the two are funded are often different.

Me: Looking at the representative agent literature, there does not seem to be this distinction. I agree that, in practise, this is relevant on many cases. But of course this is not important if the policy maker fully invests in mitigation, as then the need for adaptation will be virtually zero.

Referee:  The analysis is in my view flawed, and it does not support the conclusions drawn from it.  The argument that socially optimal adaptation must be zero in this model rests on the implicit assumption that one can vary the number of agents in the model freely without making changes to other assumptions. That is wrong. In the representative agent version of the model, a dollar spent on adaptation or mitigation benefits everyone. The same dollar spent on adaptation in the “N=8 billion”-specification benefits 1 individual only, but 8 billion individuals benefit from a mitigation-dollar. So necessarily, as N increases but the other functional specifications don’t change, adaptation becomes relatively less attractive. But that’s because everything else is held constant, which makes no sense. Varying the number of agents in the model changes what adaptation represents (likewise, what the welfare/utility and damage functions can represent varies with the scale of aggregation). Arguing that the comparative statics carried out by the author suggest that optimal adaptation is zero is simply mistaken.

Me: Again, let me emphasize: changing N does two things: it changes the level of aggregation, and thereby changes the relative publiceness of privateness of adaptation and mitigation. Nothing more, and it does not need to do anything else. I am not changing the number if individuals! Also, why should other functions change?

Referee:  Relatedly, a representative-agent model cannot be used to directly investigate socially optimal adaptation that is region-specific; one needs to “translate” the benefits that accrue to a subset of the world population to the world-population level.

Me: Clearly, this is what this paper is about. I think the referee fully gets my argument but does not see that this is precisely what I show. I argue that the way the representative agent models have approached adaptation is flawed. Of course, the implication of my paper is that if one properly translates the benefits that accrue to a subset of the population, then adaptation at the global optimum is zero.

Referee:  Likewise, this same model structure with 8 billion agents cannot deal with adaptation projects that benefit an entire nation or region (such as a dam or geoengineering). The latter specification implies that only individual-specific adaptation exists. Yet even that counterfactual assumption doesn’t mean that adaptation cannot be socially beneficial per se. That result is a product of the erroneous model specification.

Me: If we model adaptation at the national level, e.g. in the case of a dam, then this corresponds to my case where N=196 countries. Numerically I show that this already implies adaptation should be equal to zero. Of course, a much more careful calibration could arrive at a result where adaptation is marginally positive, but the main point is that the optimal policy mix will still be significantly tilted towards mitigation.

Referee: It seems to me that the paper builds on a confusion between the level of aggregation of the welfare function, and the adaptation cost function. When the authors say that they are changing the former, they are in fact changing the latter. When the authors say that they are assuming that the policy maker splits the world into regions, countries, or individual people, what they are actually doing is changing the adaptation cost function, increasing adaptation costs—for the same total adaptation expenditure, damages go up! This leads to the nonsensical conclusion that the level of analysis affects policy in first best. The above applies both to parts 2 and 3, but is most obvious in part 3, where it is extremely clear that when the authors make ‘different assumptions about how the policy maker aggregates the social welfare function’ they are actually just changing the parameter γ: what they describe as splitting up the world into smaller units is actually just raising γ for the same total abatement expenditure.

Me: This referee simply confuses causality. I change the number of regions that a policy maker uses to evaluate the optimal policies, the implication being that this changes the publiceness or privateness of mitigation and adaptation, and the result is that this obviously changes the relative benefits/costs of mitigation and adaptation.

Referee: I think that the author has to make a stronger case for why we should see adaptation as a private good/investment when we also know that adaptation is necessary in certain places…

Me: I note in the paper that, in case current mitigation actions are too late and cannot reduce climate impacts quickly enough, then of course this requires adaptation. My model results only concern the case where there is actually a trade-off between adaptation and mitigation possible! What this referee wants is that I extend the model to cases where there is no trade-off possible. In that case, however, there is obviously no need to do any modelling and the result is as follows: Do as much adaptation as is optimal until there exists a viable trade-off again, and then invest in mitigation.

Referee: … and that such protection in one place can actually be seen also as a public good (=environmental justice) in another place by people who are less vulnerable to CC.

Me: I have not seen representative agent models, or integrated assessment models, forward the point that adaptation should be viewed as a means of achieving environmental justice, and thus have not discussed. But it is an interesting point that may be worthwhile looking at.

Referee: The debate on response to climate change has swung between mitigation and adaptation as being the most urgent and/or effective strategy. Mitigation is obviously needed to halt climate change – and is therefore not only a public good but a global public good that everyone would benefit from. Adaptation, on the other hand, is needed in places and for people who are already impacted by climate change or climate variability. To say that adaption ‘imposes a negative externality on all others’ may be true according to strict economic theory but at the same time it is an odd stance if we take environmental justice into consideration.

Me: See my poin and argument above.