domingo, 1 de mayo de 2011

Are simulations useful?

I am taking a class called Human Ecology, the class is taught by Jeffrey D. Sachs. I love the class although I feel there is much disorganization in the materials and we don't go into the level of depth into each model as we should. The class is very theoretical and studies the interactions between the economy and natural ecosystems.

The class is hard. Most students don't like the class due to its theoretical content. I understand why they don't like the class. But I love the class. I remember why I came to Columbia because of this class.

Our last problem set made us study three trends:

1- The population growth projections:
2- The GDP per Capita projections:
3- CO2 Emissions per Capita projections:

Question k of our final problem sets says:
Are simulations of this kind useful? What are some key elements that are missing from this simulation? How would you change it to improve it?

As a joke some of my classmates mentioned that this simulations are not useful and that we should be investing our time in action more than models.

I think that in some aspect they are right, these simulations are so over simplified that they can lead to wrong decisions. This is my poorly written answer of a sunday night:


The exercise of simulations is very important to make smart policies that avoid perverse effects. Intuition and opinion-based decisions are often well intended but lead to huge mistakes. For example, in the case of subsidies in water and energy for agriculture in Latin America in the 70’s that was intended to help smallholder farmers modernize their agricultural practices. The long-term results of these policies have led to market distortions that have helped the richer landholders and dis-incentivized modernization of small-scale agriculture. Close monitoring of key indicators, and models can help to identify the perverse effects of poorly designed public policies before it is too late to correct them. Plotting business as usual scenarios can help to understand and forecast the outcomes of inaction. Decision-makers need to rely on this type of simulations to avert disastrous outcomes and correct trends to effectively reach desired outcomes.

This simulation has strong limitations; it does not take into account the important amounts of uncertainty that exist about earth dynamics. CO2 sinks are not considered in this simulation and that is a very important aspect of the climate change policy.  Also the simulation fails to consider that there are actually different types of energy intensities to GDP per capita and also different levels of carbon intensities to energy use. Although some European countries have the same GDP per capita of the United States, the differences in consumption behaviours leads to large differences in actual per capita emissions that this simulation is not considering. Americans tend to have more energy intensive consumption behaviour than some very rich European economies that have a less intensive use of resources. On the same line of thought, even if two populations use the same amount of energy per capita per unit of GDP per capita, the amount of emissions of CO2 required to produce the energy have a great amount of variability. Indeed the emissions released per kilowatt of energy produced with coal, oil, nuclear or biodiesel have different carbon intensities. It is not the same to compare for example a country like the US, China or Australia that rely of highly inefficient use of carbon based power plants to for example the natural gas or nuclear power plants that Europe and other developing countries are currently using.

A third problem of these simulations is that they are based on the past behaviour of societies and are just an extrapolation of past trends. The danger of such simulations is that it fails to consider events that have not yet happened. The colloquial “black swan” concept that you cannot predict what has not happened is a great limitation of linear simulations. In this sense today’s computer power and artificial intelligence algorithms can help us to make models that take into account uncertainty and non-linear relationships. Simulations that take the complex-systems approach are much more likely to forecast scenarios closer to future reality. The only problem is that by doing this the results from the model will probably be a range of possibilities and that can lead politicians to mistake the uncertain nature of the climatic crisis with a lack of scientific knowledge of what is really occurring. In the end the quality of simulations has to go hand-in-hand with the quality of our politicians and decision-makers.


Some additional trends that scare me:





So! What should we do? Do we know where we are going? Do we want to go there? I am unsure but it seems it is terra ignota...