01 January 2012

The Art Of Prophecy And Its Broader Implications

There are several basic ways to game the art of predicting the future.

* One is to be sufficiently vague to make your prediction a sure thing in a way that sounds meaningful anyway. Predicting that the singer with the most weeks at the top of the Billboard Top 40 charts in 2012 will be between 15 and 39 years old in a rock genre from an Anglophone country who is a tenor or a soprano sounds much more impressive than a prediction that the singer will be a human being who isn't mute, with only a slight degradation in certainty. Some of this is mere rhetoric and gamesmanship, but the more you know that more you can do to be meaningful while still being sufficiently vague.

* A second trick is to make lots of predictions. If you make enough of them, some of them will end up being right and you can truthfully say that your prediction was on the money while omitting all the times that you were wrong. In particle physics, they call this the look elsewhere effect but they fall victim to it just as much as anybody else anyway. Indeed, this potential makes it worthwhile to go out on a limb and be specific now and then, particularly when there are only a few options which are plausible if you've looked into the situation at any length. A corollary of this trick is to regularly update your predictions as new information becomes available with stated reasons that absolve your past predictions for being almost surely wrong and keep you in the running as a pundit.

* One of the other fine points is timing. Lots of predictions, for example, that there will be more than 100% inflation, or that the world population will increase by at least a billion, will almost certainly be true in some time frame. There are sweet spots when it comes to timing that are domain specific.

For example, when predicting the weather, it is possible to be quite accurate for two or three days out and to make quite reliable statements about seasonal variation. Throw in El Nino/La Nina effects and a few key indicators like rainfall in Africa and local rainfall in the last few years, and you can characterize the weather over a period on the order of a year relative to seasonal averages with meaningful accuracy, and climate models can also say meaningful things about probabilities of certain amounts of deviation from the norm in specific future time frames of five years to centuries.

On the other hand, there are also blind spots in weather prediction. For periods of more than a week and less than a month, it is virtually impossible to beat seasonal averages by much, which is unfortunate, because there are a host of situations when this would be a really useful trick.

The same holds true in geopolitics. It was necessarily true that Kim Jong Il would cease to be the Supreme Leader of North Korea at some point. A good actuary, informed by some individual specific details and a little data set of past head of state life expectancies and causes of death in history could have probably confined the date to plus or minus three or four years with a pretty high degree of accuracy. But, nobody outside of his closest inner circle could have predicted that this would have happened in December 2011 rather than, for example, January 2012. The person likely to take the helm when Kim Jong Il ceased to serve has likewise been clear for several years and been possible to narrow to no more than half a dozen or so candidates at any give point in the past decade, and the total group of plausible successors at all given times in the entire period of the last decade has probably never had more than ten or fifteen people in it. The outcomes, in other words, have been quite a bit more predictable than the timing of those outcomes.

Similarly, betting that the Dow and S&P stock indexes will be higher twelve years from now is a pretty sure bet, but predicting how those indexes will fare in the next six months is far more of a gamble.

On the other hand, sometimes, timing is easier to predict than outcomes. I can tell you to plus or minus an hour or two when we know who the President of the United States will be in 2013. I can even narrow the list of who that person will be to a list of people small enough to count on my fingers with something like 98% plus certainty that is broad enough to even throw in a candidate or two on the Democratic side in the highly improbable but not unprecedented event of a mishap that causes Barack Obama to be unable to serve and the possibility of a last minute white knight GOP candidate. But, I certainly wouldn't bet the house on whether or not President Obama will be re-elected. The probabilities this year are far too close to call. The best political science models and pundits have margins of error to large to call it with even crude accuracy at this point with the facts we have because the best models and pundits are predicting a close race due to a combination of a bad economy, a trend towards peace and recovery, a GOP opposition in Congress that has shot itself in the foot politically and ideologically, and a parade of nutcase GOP Presidential nomination contenders. Throw in predictions about the outcome of all of the pending U.S. House and U.S. Senate races and control of statehouses nationwide, and one has to be pretty vague indeed to make a prediction with much accuracy ten months out.

It is much easier to predict the Congress will pass an omnibus appropriations bill late a night near the end of the current fiscal year for the following fiscal year than it is to predict who the winners and losers in the budget process will be in the coming year.

* One easy way to chalk up a lot of accuracy predictions is to identify a small number of very likely outcomes upon which a large number of other outcomes are dependent. Economists, in particular, do their own discipline a disservice by failing to make clear enough just how good they are at this piece of the prediction business.

Timing business cycles and predicting their severity is very hard. But, there is actually a considerable consensus among economists who are true social scientists as opposed to politicians making political-economy suggestions with an air of respectability, regarding an immense number of outcomes that are associated with particular phases of the business cycle.

Foreclosures and bank failures are highly concentrated in busts. Employment is a lagging indicator - it recovers during booms, but more slowly than GDP. Busts shift consumption from luxury consumption to less expensive substitutes - they help Budweiser and PBR relative to premium beers.

Economic steps backward in the business cycle take at least as long as it does to recover from them, and as a first order approximation, the time it takes to get back where you started is a little longer than the time it took to hit bottom. This certainly isn't a firm rule, but it is at least as accurate empirically as far more sophisticated models. If you want to refine your model a bit more, you could opine that the shorter the time you spend going down, the closer the recovery time will be to the decline period. In contrast, the hypothesis of symmetry tends to break down in more severe recessions. The longer the fall, the more likely it is that the recovery will take significantly longer than the fall did. The financial crisis, which has continued to morph over time with the European sovereign debt struggles currently taking center stage, has also underlined the fact that nobody with power over economic policy anywhere knows how to prevent recessions from happening entirely over the medium term.

The practical implication of this fact is that if one has for whatever reason accurately called the business cycle, which leading indicators can help to do unless a collapse comes on very suddenly (or at any rate, not long after a downturn has begun) one can make a whole raft of other accurate predictions. One bet can generate a great many payoffs if you know what to bet upon.

Another practical implication of this fact, which is familiar to MBAs, hedge fund analysts, and the better financial planners and investment analysts, but not to most policy makers or educated lay people, is that we know a great deal about what kind of events will or will not happen at the same time with much more accuracy than we can predict when any one of them will happen.

For example, you simply aren't going to see high unemployment and homelessness for in tact families at the tail end of a long economic boom. Similarly, increased demand for social safety net programs from disability insurance to housing assistance to welfare to unemployment insurance to means based health care programs to the most heavily subsidized parts of the higher education system are generally going to coincide with reduced government revenue from cyclic sources like income taxes and even more so, sales taxes. Securities fraud suits are always going to be more common when stock prices have crashed than they are when the stock market has surged.

As an example, the Peak Oil hypothesis, that the world is going to hit a point where production starts falling despite increasing global demand due to economic development and improving technologies for oil extraction, is every bit as certain now (indeed, much more so) as it was when it first started to be widely discussed in the wake of the OPEC oil embargo in the 1970s. And, from the peak oil hypothesis, it doesn't take terribly sophisticated economic and technological analysis to figure out a great many consequences that will flow from that event as economies change their behavior in response to rising oil prices. But, timing the question of when exactly we will hit peak oil and how abrupt the need to transition away from oil as a result will be is a much more subtle venture, and overconfidence on the issue of timing greatly undermined the credibility of its advocates, despite the fact that they were basically right on everything else that they had to say.

Economists are not the sort to look for what they've lost under the streetlight. They insist on looking in the dark spot in the alley where the glory is most likely to lie, even if it is pitch black, they have earplugs in their ears, mittens on their hands, swampy mud under their feet, and dark sunglasses over their eyes. They are not alcoholics, and as a result, are not familiar with the Serenity prayer. They do not routinely engage in the introspection necessary to conclude that there are some things that they can predict, some things that they cannot predict and do not aspire strongly enough to have the wisdom to know the difference.

The rare individuals who do engage in that kind of introspection, Schumpeter comes to mind among economists, and the founders of the index fund movement in the financial markets, often develop very powerful and profound insights about practical conclusions that can be reached in dealing with the economy.

* I'm inclined to think that Keynesian macroeconomists would be considerably more effective in the world of public policy if they abandoned the pretense that their conclusions are necessary and rationally necessary implications of a solid theory, and retreated instead to the safer ground of merely claiming historical empirical proof for their conclusions, and casting themselves as phenomenologists while downgrading the theoretical basis for their conclusions, in public debates anyway, to mere heuristic arguments for the results that we have seen in real life. An example heavy, empirical argument that anti-government, anti-tax, anti-regulation policies don't work is more powerful than an effort to convince the public and policy makers that an intricate theory full of subtle assumptions is more likely to be true than a simple theory that can be reduced to emotionally satisfying sound bites.

This tactical suggestion is intimately related to a final well validated bit of evidence of what makes for good predictions. Hedgehogs, i.e. people who know one big thing, have far worse track records than foxes, i.e. people who know many things which don't necessarily have a lot of ideological coherence.

Troublingly, the media has learned that hedgehog pundits have more entertainment value and politicians guided by media informed voters are only a step behind talk show producers. The best predictor of a pundits' likelihood of being right is his prominence.

No comments: