Econometric modeling

pab, It's impossible to develop any economic mathematical model, simply because it's not possible to include all the variables. Future events do not necessarily follow from the past. Nobody that develops a model can control the price of energy, nor the fluctuations in the world's economies. Forget it. Do what I do; use your "gut feeling" on how the economy will react in the future, and how that will impact your investment portfolio. Always think long-term. Short-term fluctuations only confuse.

I must say I disagree. While imperfect, as are all works of humans, econometrics has a most respectable track record. I would say it has so far been demobstrated a credible forward projection tool - one among many, but one with considerable credentials. I call to your attention in particular, c. i. , the long running debates in which the two of us have been involved relative to the US Economy and to the US Elections - and point out that one of us based conclusions in part on econometric indicators, and that one of us has seen conclusions borne out by events.

timber, Rather than debating me, you should offer pab what he's looking for. It seems you have all the answers; I don't.

Dunno as I'd say I have all the answers - more like I question or disagree with some answers provided by others, and offer my basis for question or disagreement. Is there something about that you find disagreeable?

Always - especially with you! pab asked a question, and I answered him honestly with my opinion. Your post is nothing but an attack on me. Rather, you should be answering his question. Just maybe, both of us might learn something from you. Your SOP is not appreciated.

Again I must disagree. While I did address you, and your response to PAB's question, my response was a specific response to PAB's question - honestly reflecting and representing my own opinion. The reference to our own previous interactions relevant to the matter at discussion were simply by way of example assumed to be familiar to the other participant in this discussion, who expressed an assessment with which I disagree, to whit; yourself. Any "attack" would be inferred, and incorrectly, as none such was inteneded, expressed, or implied. There was nothing other than disgreement with the two positions thus far taken, with expansion on the basis for that disagreement, in which I indicate econometrics strikes me as valid, legitimate, and proven for its purpose, as exemplified by the congruence of events with positions and projections I had offered any number of times in the past on these boards, including discussions in which you participated and in fact interacted specifically with me from time to time, generally in disagreemnet with my own conclusions, conjectures, and positions. Would there be something of factual basis regarding the cited example you wish to dispute?

pab should not have to go back to our past disagreements/opinions. That's between you and me. OUr disagreements adds nothing to pab's inquiry.

No suggestion that PAB do anything was expressed, implied, or intended - the reference specifically addressed your contention. In fairness, I note I'm in complete agreement with your assessment of short-term fluctuations. I disagree with the negative assessment of econometrics. As for "gut feeling" versus analytical evaluation and projection of economic trendings, I believe both my position on such and record of performance relevant to pertinent conjecture, assessment, and projection have been documented extensively on these boards. Where many here have been at odds with my bullish assessment of the US Economy, those who were unwilling - or at least reluctant - to recognize and acknowledge the economy was robust and steadily improving over the past few years are left with no support whatsoever for their arguments in such regard.

"yada yada yada ... in which I indicate econometrics strikes me as valid, legitimate, and proven for its purpose, as exemplified by the congruence of events with positions and projections I had offered any number of times in the past on these boards,....yada yada yada..." And you expect pabs to have read all those past opinions that you illucidated. Get real, man. Your statement accomplishes only one thing: it deminishes my opinion. If you want to do that, just say so right out. No fancy wording is necessary.

If you say so - I can see what we got here is a failure to communicate, to steal a line from a movie.

Aaaaaaah Timber,

It seems that you communicated a lot. The problem is, what did you communicate? It sounds remarkably like Cicerone's opinion is not wanted here, that you think everything he posts is dumb, etc. Don't you think that should have been left up to the author of the thread?

I can't help but notice that the author hasn't returned. Could it be because of your attacking everything posted by CI?

To the original poster - I am a mathematical modeler and can assure you that the expression included in your essay

[D = F (P, Y, G, S, O)]

isn't an econometric model at all. It's the implicit form of a function and can't be tested as written. In order to make an econometric test of a function you have to write out the function explicitly. E.g.

D = aP + bY + cG + dS + eO + r

where D is the dependent variable, a, b, c, d, e are coefficients to be calculated by the model, r is an error term, and your independent variables P, Y, G, S, O are inputs to the model.

In this particular function you will find that testing first and second derivatives will result in better econometric fits for D, the demand as calculated by the model, with D', demand as actually measured by statistical observation.

The explanatory text you provide should be reworked: it's wrong mathematically, syntactically, and grammatically.

Let me clarify: In all of the economics classes I have studied, and my personal observation of the ability of businesses or governments to use economic models to forecast future events is at best an art and not science. It has been my personal (humble) opinion that any economic model cannot possibly include all the variables that impacts any economy - micro or macro. That's the reason why we see professional economists disagree on many issues. If I am wrong, then I apologize to timber. c.i.

On that basis, no aplology warranted, IMO; opinions are opinions, yours, mine, or those of professional economists who disagree. I just happen to agree with those who understand, endorse, and appropriately employ econometric modeling as one of many analytic and projective tools. It works for me, and to those who dispute the validity of the method, I point by way of challenge and rebuttal to numerous instances I feel validate the methodology - including some of my own observations and projections as recorded on this board.

Absent disagreement, and the attendent exploration thereof, what else is the point of debative discussion?

Trying to find common ground between Timber and Cicero is not an easy thing to do. Their backgrounds, world view, politics and means of expressing themselves works against much agreement ... on anything. In the present case, I think Cicero's belief that economics can not be successful modeled is wrong. Both Conservatives and Liberals do it all the time, and we rely upon the projections everyday. Are the projections often wrong? Certainly, it was supposed to be rainy today but the sun is shining. Our plans to work at our desk today may now include a picnic out on the back deck.

Poorly designed models can be very deceptive. Defining as narrowly as possible what we want to know, and what the key variables are that affect outcomes is not an easy task. Often the data needed either doesn't exist, or it is contaminated in some fashion. Analysts often deceive themselves and lose their objectivity in the face of expectations. Those who depend entirely on the hard numbers overlook and ignore the common sense questions that more experienced analysts bring to bear as "art". Straight line projections, as in least squares, do not persist indefinitely, for instance. As time and weight of trends continue they affect the environment within which the operate, and the equation shifts. Students and laymen often make the mistake of thinking that one equation might suffice for projections going out to a thousand years. Bah. In my experience, even the best projections done by the most sophisticated and gifted analysts rarely survive even a decade without hopeless drift.

Projections are at their best usually when dealing with large trends and numbers where advanced inferential statistics can be brought to bear. No one is likely to "predict" what any individual will do in the future. Projections that fail to provide a range within which the expected score/value will fall are the most risky of all. The wider the range (all else being equal) the greater the probability that the forecast will be correct. The further a projection is pushed into the future, the greater the likelihood that it will be on spot.

The "model" at the top of the thread in my opinion is pretty badly flawed, but I'll leave the details of improving the equation to others. For what its worth, I and my team once projected a number of complex social trends within the criminal justice field for five years. At the end of that period, it was found that we were over 97% correct in the total forecast. On some elements we were dead on within a percentage point, though on some others we were spectacularly wrong. The political powers received our report without much confidence, and mostly failed to consider the alternatives we offered to mitigate. God, I love to be right.

Cicero ... My wife and I will be arriving in San Francisco on Thursday, April 28th (3:14 p.m). Wedding rehersal is Saturday, the wedding Sunday, and we expect to fly back to New Mexico on May 2nd. We would very much like to share a dinner with you, your wife, GeorgeOb1 and his wife. Be thinking of it, and let me know if we can arrange dinner, perhaps Friday evening.

Asherman, A well written post - even when it disagrees with my personal opinion. The problem with statistical modeling is the fact that nothing remains equal. That you were able to create a model that projected with 97 percent accuracy is quite a feat even if it was based on social trends. For example, I know that locally (in Northern California) we seem to have an increase in Autism. The social scientists and the folks in the health field are investigating this phenomenon to try to figure out why. As for your arrival date to the bay area, I'm afraid I shall already be in Malta from April 15. That will be followed by a visit to London for three days, then a chunnel ride to Amsterdam with dys and Diane. In total, I wil gone for 30 days beginning with my Malta trip. One of these days, we're going to connect for that meal. ITMT, take care and blessings, c.i.

Paul, another take that I might add is the use of "similitude systems" or setting up a linear equation of one system in terms of another physical system entirely. (Many times we have to use erf notation but , for example drumhead theory was used to model the curvature of a surface at instantaneous times as energy is added or released-used many times in physics) SO , you have a multifunction linear equation that may, or may not fit a set of assumptions. Try to look at it from an entirely different perspective wherein you have something like , for example.
ALL THE WATER THAT RAINS/unit time= f(runnoff+infiltration -evaporation)
If you can set up an equation that sets the dependent variable in terms of addition or subtraction of a number of independent variables, you can , by this similitude, locate where your needs for more robust data lie.
Basically, is the equation follows a specific form, you can substitute just about anything else and see where it takes you.
Im certainly ignorant of econometrics but we can speak the same equation languages

I remember in one of our economics class, we used differential calculus to determine the most efficient level of inventory. When one can control most of the variables, economic models works fine, but when we have no control over most of what occurs in economics, and nothing remains static, the challenges are overwhelming. Even in budgeting for a small (and probably large) company, most five year budgets are reviewed every year to make adjustments as current variables change. I was able to develop budgets accurate to +/- 5 percent on an annual basis, but that's because the turnover rates for employees could be determiend quite accurately, and staffing costs made up over 80% of our budgets. The revenue side was a little trickier, because over 40 percent came fron donations from individuals, foundations, and corporations.

Cicero,

We'll miss you, but you might ease the pain by picking up the dinner tab by leaving us your credit card.