RWER issue 54: Lewis L Smith
Real-World Economics Review, issue 54
You may read the whole paper here:
I regret to report that Lewis L Smith passed away the day before this paper was published. When it came to knowledge and understanding of the economics of oil and gas he was virtually peerless. His reports on the energy industries have been a feature of this blog. He will be greatly missed.
The epistemology of economic decision making
Lewis L. Smith [Energy Affairs Administration, Puerto Rico]
From knowledge to ignorance in four easy steps
To engage in a business, to undertake a project, to provide long-term financing or do anything “for the long haul”, takes one on a mental journey into a land we call “the future”. Initially, when we are contemplating the short run, the terrain looks familiar. But it becomes less and less so, the farther we look ahead and the more and more uncertainty and ignorance displace certainty and probability. This marked deterioration in our ability to anticipate and understand the future as our planning horizon advances, is the chief problem today of what may be called “the epistemology of economic decision making”.
One approach to this problem is to systematically describe as many as possible of the things that might happen to us “on the road of life”, do so in terms of possible “outcomes” and the possible “impacts” of each outcome and then attempt to estimate for each one, the probabilities that it will be realized. Unfortunately the results of this effort will vary greatly in quality, depending on many factors, especially our time horizon. So it is convenient to group the outcomes in four “domains” — certainty/near certainty, risk, uncertainty and a fearsome domain of high-impact, low-frequency outcomes, one which we may call HILFO’s or “hill fows”.
In oversimplified fashion, the nature of the these four domains may be illustrated as follows —
Certainty/near certainty — There will be no hurricanes before June. [Or, I am 95 % sure that there will be no hurricanes before June.]
Risk — There is a 45% chance of a major hurricane this summer.
Uncertainty — There might be as many as five hurricanes next summer but if we are lucky, there could be none ! [No probabilities ventured].
HILFO’s — I am sure that there will be at least one hurricane next summer, but I have no idea if any of next summer’s hurricanes will be as nearly bad as Hurricane Andrew.
Unfortunately, in forecasting and decision making, most people tend to focus on first two domains as if they were the only ones which count. In these domains, we can usually enumerate each and every one of the possible outcomes and most [if not all] of the several impacts which each outcome might cause. Moreover, we can estimate the great majority of the probabilities, construct a collective probability distribution for all the outcomes and construct individual distributions for many of the sets of the impacts corresponding to particular outcomes. [Feynman 1999, Seife 2003.]
However, such cases as Chernobyl, Deepwater Horizon, Exxon Valdez, Three Mile Island and massive extinctions of dinosaurs remind us that the third and fourth domains are sometimes of great consequence for humanity. So we must give them some thought as well.
Domain of uncertainty
Beyond the domain of risk, lies the domain of uncertainty. And if our long run is long enough, it is probably the most populous. Here we can neither fully enumerate all the outcomes nor all of the impacts, we lack a full understanding of the impacts and we certainly cannot calculate all the probabilities. At best, any estimates of the latter will have wide margins of error and at worst, they will border on the indeterminate. In many cases, the best we can do is to guess at the rank order of outcome probabilities.
Uncertainty has many origins. We mention the ones which seem to be most common and/or important —
 Our imperfect knowledge of dynamic systems and their processes.
This is especially so for those systems which are not only complicated but also complex — such as economies, enterprises, markets, nuclear generating stations, space missions and large telecommunications systems. It is an even bigger problem with those which are predominantly chaotic — such as weather systems.
 Difficulties in interpreting the time series generated by the diverse processes which drive dynamic systems, that is, “the tracks which they make in the sands of time”. For example —
[a] It is frequently possible to characterize a particular line segment of such a time series by a mathematical function or statistical test and so define what one may call the system’s “mode of behavior” during the period. However, we often find that a particular segment looks like it came from somewhere else. That is, that it could have been generated by a process other than the one which actually originated it. This can happen even if the process is deterministic and generates a long run of data.
In fact, each of the major types of dynamic systems — the chaotic, the complex and the random — is capable of imitating several of the modal patterns exhibited by the other two. In particular, any one of the three can exhibit trends, pseudo trends or random behavior on occasion. [Auffhammer, 2005. D’Agostini et al, Apr 2006; Ormerod and Mounfield, Jan. 2000; Phillips, Jul 20O4; Phillips, 2003; Ploberger and Phillips, Mar. 2003; Smith 2002, Sornette, 2003 and Jan 2003; and Zellner, 2005.]
[b] Tests for identifying a process, based on the characteristics of particular line segments, tend to be complicated and require large samples of data.
[Das 2005, Hommes 2006, Srbljinovic 2003;, Voorhees 2006.]
[c] Despite the best efforts of econometricians in academia and the “technicians” on Wall Street, there is no foolproof method for telling when a given time series is going to switch from one mode to another. In securities markets, for example, it is very hard to call more than 70% of the turning points. The NYSE in 1986-87 is a classic case of the latter. [Blackman 2003, Penn 2006, Snead 1999 and Zellner 2005.]
You may read the whole paper here: