If you’re following the American political election campaign closely, you’ve probably run across Nate Silver’s weblog (now subsumed under the NY Times rubric) called Five Thirty Eight (named for the 438 Congressional and 100 Senatorial seats that are contested). He is one of 7 or 8 forecasters that were smart enough to invest time and energy into looking at races at a more local level (as well as national polls) than most news organizations, who were content to take the output of national polls and pass them on to news consumers. Nate Silver and his competitors seem to be doing a better job of predicting past results and frankly I trust them much more than the national polls alone.
The reason Mr. Silver and those like him are beating the big pollsters is that looking at polls together and including polls of smaller geographic regions like states allows for more granularity in the results achieved. Where a Gallup poll or Pew survey can tell you what 1,000 people respond across the U.S., some state polls get 1,000 responses for the individual unit. That’s better. So Mr. Silver can not only get more accurate information about the country’s opinion, he can tell us how the Electoral College results are shaping up.
There’s a parallel between what Mr. Silver has done with U.S. politics and what I am trying to do with global energy consumption. The forecasts most people are familiar with are large, model-driven efforts from organizations like the U.S. Department of Energy’s Energy Information Administration or the International Energy Agency. They incorporate a lot of inputs about fuel supplies, changing portfolios and projected prices and create supply-driven forecasts.
However, many of their results are reported at the regional level, which tends to hide the very different market dynamics between, say, South Korea and Indonesia, or South Africa and Zimbabwe.
This figure shows the difference in quadrillion BTUs (Quads) projected for consumption in 4 countries. The DOE’s EIA figures are compared with the projections from national or independent organizations, and my estimates are in the right hand column. The estimates are for 2030.
I think Nate Silver would say that a more granular look would provide more accurate results. As it happens, many countries have government departments or independent countries that provide estimates of future energy needs.
Those estimates are very different to what the models from large organizations project. My suspicion is because the local estimates understand that fuel consumption is driven by more urgent needs and a willingness to pay than the global estimates that consider fuel availability at what they think is an affordable price.
Because the local estimates are in fact very close to the totals I came up with in my report Energy Consumption in the Developing World in 2030, I am naturally predisposed to favor their figures. But I think that logic as well as my personal bias argues in favor of trusting the boots on the ground. The territory they cover is smaller. They spend just as much time getting their country’s figures as the mega-model makers do for the world. They understand their people and their markets better.
And now I can say, well, Nate Silver would do it my way if he were to do it at all!