Tag Archives: measurement

If you follow basic news you have, no doubt, heard prognosticators speak of the “long”, “medium”, and “short” run possibilities for our economy. Theoretically, the “long run” is an abstract time period in which there are no fixed factors of production. Or, in other words, the time period over which a theoretically competitive firm will not earn supernormal profit. The short run, then, is the such a time period on which at least one factor of production is fixed. Macroeconomists might say the long run is that over which expectations have adjusted to the fundamental state of the economy. I suggest a better heuristic: sensitivity to time-dependent forecasting. But first, background.

In practice, no one can actually observe such conceptual definitions. So when an “economist” enlightens you about his opinion of the future, “short run” is now, and “long run” is later. Some economists might further specify their belief by constraining the short run to factors governed by demand, and the long run as those governed by supply. While a conceptually more useful approximation, economists will dispute ad infinitum the existence of a shortfall. Indeed, the question and debate becomes largely tautological as measures of demand shortfall – like output gap – are measured on a historical “trend” growth rate.

As an intellectual framework, statistical estimates are unsatisfying, as they fail to capture the tectonics that moderate a dynamic economy. For example, an econometrician may note a significant number of peak-trough flows lasted fewer than n years. But do we believe there is something “real” about n? Not really.

On the one hand, there is the purely theoretical and hence rather useless. On the other, we have either ad hoc definitions designed more prominently to suit a particular economists own pet beliefs rather than capturing the business cycle or impure statistical estimates. 

However, what if we – for all intents and purposes – define the long-run as “the point at which future forecasts will not effect current consumption and investment definitions”. Let’s conduct a thought experiment. Let’s say I’m an economic forecaster and, by some stroke of dumb luck, the market actually trusts what I have to say. If I produce a report that promises on high confidence booming growth next year the result is obvious. Purchasing managers, entrepreneurs, and investors will suddenly update their confidence about demand tomorrow and hence increase their investment todayAu contraire, if I suggest a high likelihood of recession next year, the market will update its confidence negatively, decreasing investment today.

While I’m using this example as a thought experiment for another point altogether, it’s important to note this shows precisely why “recession predicting” is an idiot’s game (to the extent you want people to believe what you have to say). A forecast on the future is self-fulfilling today. It is epistemologically impossible to have a good forecast that is also credible insofar as recessions are concerned. 

Back to the experiment. What if I, magically, at the same confidence level produced a forecast for the economy fifteen years from today. The reaction to my report – whichever way – would be dampened. I’m not sure to what extent, but few of us would expect this sort of forecast to have much effect. This is not trivial, especially if you manage to convince yourself (it’s hard) that the market places the same Bayesian likelihood (“trust”) in this report as my short-term prediction. You find it hard to convince yourself of this, of course, because uncertainty is ipso facto correlated with the extent of the forecast.

A longer forecast fails to elicit the same energy from the market for the following theoretically-sound reasons:

  • If my forecast tells you little about the interim between the periods (that, after all, is the point of the thought experiment) the longer the window, the greater the chance of an intercepting recession.
  • Without knowing about the path to the future, capital depreciation make immediate investments unprofitable.
  • If my forecast is ten years out, there is no reason to wait nine years on the expectation that my forecast will improve with improved information.

Of course, there will be some activity which derives from something economic theory has a harder time explaining. Investments take time to build. If I expect a huge increase in energy demand in ten years, I might invest in a nuclear energy plant as this would take about as long to build. This voids the above set of uncertainties by evaporating the relevance of the interim.

Now consider something called a “forecast yield” curve. This measures the markets response (y-axis), given a certain indicator (stock returns, consumer confidence, job creation, etc.) and the time period (x-axis). The response is of course a qualitative feature which may be rather easily quantified using a variety of indicators such as immediate change in stock prices or the purchasing managers index.

The response will be very high if the forecast is on one year, and diminish – in some form, I do not know which – over time. The long run is then defined as the point at which the response becomes insignificant. This is not easy to measure, especially in a methodological way. However it is, theoretically, possible to measure, unlike the rigidity of various “factors of production” which is an entirely epicyclical phenomenon (that is, tautological). It makes great sense in explaining and capturing the idea of market dynamics, but is less useful when married with real numbers.

A slight modification of this definition is very in tune with some currently used interpretations of “short” and “long”, but removes the ad hoc nature thereof. Let’s say the forecast purports to guess the demand side constraints only – say, by proxy, nominal consumer expenditure. (We assume that all long run price adjustments are structural in nature). The time at which the current forecast becomes irrelevant, by nature, then is the point when the market believes the supply side will dominate the demand side. Therefore, the demand forecast sensitivity curve would provide a good idea of when the market expects supply to dominate demand – which many economists would ascribe to long run superneutrality. 

The intellectual benefit from this sort of definition is its ability to be measured, to say nothing of the many statistical and logistical challenges that will follow! I see several rough approximations:

  • Ask purchasing managers what they would purchase given a future indicator, and vary that by time. This is a direct approximation, and perhaps the most logistically sound. It is vulnerable, however, to investors’ inability to know what they would do. (Which violates rationality, but that’s another story).
  • Note market reactions to various government reports and observe sensitivity to time (like the need to invest in more green jobs in ten years, etc.) This suffers quite a bit from: a) the inability to control the “indicator” (specific government policy) and b) extremely small sample size.

The flaws associated with the first method – i.e. the chance that investors do not know their true own belief – are more fixable, and are not at all a theoretical challenge. Further, there is reason to believe such errors are systematic and hence would effect only the value, and not slope, of the sensitivity curve. Ultimately, in calculating the end of the short run, it is the slope that matters. Moreover, many such biases will cancel out over a large sample size, and hence the aggregated curve – weighted by investment value – should be an important indicator.

If nothing else, it is more specific than what we have today. Conducting this type of survey would also provide extremely useful insight into dynamics of economic structure. We’ve heard some say that the “short run is getting longer”. If the magnitude of the downward slope is falling, it would lend evidence to this argument. On the other hand, as of now, we have little reason to believe one story or the other. It’s about time we get a more precise and observable idea of these definitions crucial to economics. The long and short of it all is that practitioners today love manipulating these definitions to serve their pet theory.