One can readily cite examples of gross inaccuracies in government macroeconomic forecasting. Some of these inaccurate forecasts have been critical to policy formation that ultimately produced unintended and undesirable results. (See, e.g., Professor Edward Lazear, “Government Forecasters Might as Well Use a Ouija Board,” Wall Street Journal, Oct. 16, 2014) Likewise, the accuracy of forecasts of long-term global warming is coming under increasing scrutiny, at least among some climate scientists. Second-looks are suggesting that climate science is anything but “settled.” (See, e.g., Dr. Steven Koonin, “Climate Science and Interpreting Very Complex Systems,” Wall Street Journal, Sept. 20, 2014) Indeed, there are legitimate concerns about the ability to forecast directions in the macro-economy or long-term climate change reliably. These concerns, in turn, argue for government officials, political leaders, and others to exercise a degree of humility when calling for urgent government action in either of these areas. Without such humility, there is the risk of jumping into long-term policy commitments that may in the end prove to be substantially more costly than beneficial.
A common factor in macroeconomic and long-term climate forecasting is that both deal with highly complex systems. When modeling such systems, attempts to capture all of the important variables believed to have a significant explanatory effect on the forecast prove to be incredibly difficult, if not entirely a fool’s errand. Not only are there are many known candidates, there are likely many more unknown candidates. In addition, specifying functional forms that accurately represent the relationships between the explanatory variables is similarly elusive. Simple approximations based on theory are probably the best that can be achieved. Failure to solve these problems — omitting important explanatory variables and incorrect functional forms – will seriously confound the statistical reliability of the estimated coefficients and, hence, any forecasts made from those estimates.
Inherent in macroeconomic forecasting is an additional complication. Unlike models of the physical world where the data are insentient and relationships among variables are fixed in nature, computer models of the economy depend on data samples generated by motivated human action and relationships among variables that are anything but fixed over time. Human beings have preferences, consumption patterns, and levels of risk acceptance that regularly change. This constant change makes coefficient estimates derived from historical data prone to being highly unsound bases on which to forecast the future. Moreover, there is little hope for improved reliability over time so long as human beings remain sentient actors.
By contrast, models of the physical world, such as climate science models, rely on unmotivated data and relationships among variables that are fixed in nature. Unlike human beings, carbon dioxide molecules do not have changing tastes or preferences. At least in principle, as climate science advances over time with better data quality, better identification of explanatory variables, and better understanding of the relationships among those variables, the forecasting accuracy of climate change models should improve. Notwithstanding this promise, however, long-term climate forecasts remain problematic at present. (See Koonin article linked above.)
Given the difficulty of modeling highly complex systems, it would seem that recent statements by some of our political, economic, and even religious leaders are overwrought. President Obama and Pope Francis, for example, have claimed that climate change is among mankind’s most pressing problems. (See here and here.) They arrived at their views by dint of forecasts that predict significant climate change owing to human activity. Each has urged that developed nations take dramatic steps to alter their energy mixes. Similarly, the world’s central bankers, such as those at the Federal Reserve, the European Central Bank, the Bank of Japan, and the International Monetary Fund regularly claim that their historically aggressive policies in the aftermath of the 2008 financial crisis are well grounded in what their elaborate computer models generate and, hence, are necessary and proper for the times. Therefore, any attempts to modify the independence of these institutions to pursue those policies should be resisted, notwithstanding that the final outcome of these historic and unprecedented policies is yet unknown.
It is simply not possible, however, to have much confidence in any of these claims. The macroeconomic and climate systems are too complex to be captured well in any computer model, and forecasts derived from such models therefore are highly suspect. At the least, a prudent level of humility and a considerable degree of caution are in order among government planners, certainly before they pursue policies that risk irreversible unintended, and potentially very costly, consequences.
Theodore A. Gebhard is a law & economics consultant. He advises attorneys on the effective use and rebuttal of economic and econometric evidence in advocacy proceedings. He is a former Justice Department economist, Federal Trade Commission attorney, private practitioner, and economics professor. He holds an economics Ph.D. as well as a J.D. Nothing in this article is purported to be legal advice. You can contact the author via email at firstname.lastname@example.org.