Andreas Haupt


Market Design in the Age of Machine Learning

Dear class,

two weeks and a revolutionary change of how our world feels and acts later, I continue what Lukas began with a second essay on "Why haven't we solved climate change yet?" Given the resources that many countries spend in the fight against COVID-19, it might seem that global action is possible now, also action against climate change. In the following, I show you a model, a simplified rule of how the world works, that explains what about fighting climate change is so hard. In the second part, I show that these challenges might hold to a lesser extent for action against COVID-19.

Ok, here is our model:

If desirable actions are costly and provide a positive externality, then they will be underprovided

Wow, that's a complicated sentence. Therefore, I reserved the next five paragraphs to parse it for you.

Let us get into the first word that might be unclear in this sentence. When is an action desirable, and when is it costly? Here is how you can test whether an action is costly to you: Imagine that a good fairy approached you and told you that it could make a thing real without you doing anything. If (a) would you want that (b) you sometimes prefer not to do it (the second fails mostly for breathing: You have to breathe), then the action is desirable and costly. Take a group project as an example. Having all work done is a wish you might want to tell to a fairy (that checks a), but still, in the real world, it turns out, you might not work as much on a group project (that checks b). Hence, working on a group project is a desirable, costly action. So are many things in this world: Healthy eating (giving up the donuts :( ), action against climate change, quarantine against COVID-19.

Let us come to the second word in the sentence that could be unclear. What are externalities? An externality is an effect of an action on others you might not account for in your decision-making. There might be several reasons for not-accounting, depending on the scenario. You might be under such a pressure yourself that you might not care (COVID-19). Or you might by legal construction not care (as companies that need to "maximize shareholder value" on stock exchanges). Yet another example is if you forget to think about others (group projects), or because of "party mode" ([]{.underline}).

The third part that might be unclear is underprovision. For this, we first need to find what "provision" means. The provision of action is how much everyone together did this action. Let us come back to the example of your group project. Several members of the group could work on it. The provision of this action is then the work of all members of the group together.

Now we can say when something is underprovided. A weak and credible criterion for something being underprovided or not enough provided, if, when everyone did more of one action, everyone could be happier.

So why is it that desirable costly actions are underprovided in the presence of externalities? Let's again go back to your group project. If there is one person that really really wants a good grade (and you know this; probably you had such a situation already), then you could think at each day: I could work one more hour, or I could wait a bit more and let them do the work. You only work if one more hour of your work pays off for you, not pays off in total. If everyone worked a bit more on the project, we could all be happier, but everyone on their own would prefer to work less. If you work less, we (Economists and other people thinking about groups) call this you taking a free ride on someone else. In the extreme, this results in only one person working, and it is stressful for all. If all of you worked, you would all be happier. Therefore, the action is underprovided. This "free rider phenomenon" we see again quite frequently: Nations that think they will not be struck by climate change might not invest in sustainable technologies. Also, countries could not invest in fighting COVID-19 and let other countries do the work.

But, wait: Countries are spending like crazy on fighting COVID-19! Is our model wrong? Let's see what we might be missing and compare the fight against climate change and COVID-19.

But wait: Countries are spending like crazy on fighting COVID-19. Is the model wrong?

Let us first be sure that in both actions against climate change and COVID-19, we can apply our model. In both cases, action is costly — if there were a fairy that could effectively contain COVID-19 or make energy plants clean, we would take it, but, still, action is not unlimited in the real world. Hence, action in both cases is costly and desirable. Acting against COVID-19 and climate change also has an externality. Research on renewables would help other countries fight climate change and, say, quarantine against COVID-19 would bring less infected persons into other countries, both of which do not have a direct effect on the country that conducts the research. Hence, the if-part of our sentence is satisfied, and the then part follows.

The provision of action against climate change and COVID are very different, though. This disparity does not mean that our model is wrong. The model might only identify crucial features of only one of the stories. We claim it does -- for climate change.

Both the costs and the externality are small compared to the self-interest of countries in the case of COVID-19. The size differences make the effect that our model predicts negligibly small. Costs of not fighting COVID-19, such as closing down universities and schools and laying off people because of a more substantial outbreak, are humongous. Despite the effect of action against COVID-19 coming with some delay, the expected local impact is significant and likely to be experienced within weeks. Hence, the cost of action is small compared to how desirable it is (the situation is terrible very soon if one does not act), and also the externality on other countries is small compared to the local effect.

The situation is different for action against climate change. Within weeks or months, action against climate change is barely noticeable. Hence, if politicians and the electorate are interested in periods of months, and not of decades, the action against climate change might be pretty costly compared to the perceivable benefit. The small perceptible benefit also makes the externality relatively strong. Hence, our model might not predict reality well if the action is too desirable for oneself -- it might happen despite externality and costs.

What can we learn from the above discussion? We formulated a concise rule/a model in if-then shape, which seems to explain why action against climate change happens less than would be optimal. Action against COVID-19, however, differs in the importance of parts of our model. Acting is much more desirable compared to the cost and the externality/effect on others. And, indeed, in contrast to our model's prediction, there is a lot of action against COVID-19. As an Economist (Harrod) puts it: "Economics is the Science of selecting models relevant to the contemporary world." We saw a model that fits one situation well and not another, and with that already have seen an essential part of what researching in (quantitative) social sciences is like: Selecting the right models.