That's right. EVERY Republican administration since World War II experienced lower growth than the previous one. EVERY Democratic administration experienced higher growth than the previous one. The same pattern holds true for jobs, as referenced in a previous post.

Faced with this in-your-face evidence, the right winger can only protest that it's all a coincidence. Expect to hear things like (1) correlation is not causation, or (2) every Democrat enjoyed the fruits of the previous Republican stewardship, and when all of that fails (3) governments can't affect economies that much no how. The alternative, as I'll explain below is to be forced to acknowledge that Republicans really aren't able to or interested in economic growth and they don't have any ethical problems lying to you about it.

To his credit, Jason Williscroft massaged all of the above three points in a statistical argument so airtight, the only problems with it were the basic assumptions, which were totally bogus. So instead of anything meaningful, he's left with a fully illustrated case study of GIGO. That of course, will definitely get you published in a variety of right wing think tanks Jason, so press on dude!

Here's where our intrepid Dead Hander went wrong. In describing the differences between his numbers and mine he says:

*"Then there's the choice of metric. Torgerson is talking about year-to-year delta, whereas I am talking about deviation from the mean. Why would I do that? Well, it's a basic difference in perspective. Torgerson's choice of metric suggests that the government is principally responsible for the performance of the economy. Mine suggests that the economy mostly takes care of itself, and that the government mostly affects it on the margin."*

There's a couple of points here. First off, I am NOT talking about year-to-year changes. I am measuring the growth over an entire administration (four or eight years) and describing that growth in terms of average annual change so that different administrations can be compared. You see, that's a much closer measure of actual reality. For example, in his first year in office, Bill Clinton spent much of it passing his tax and budget program, which by the way was a five year plan. The fact that that first year showed a different pattern of growth than the subsequent two years after the plan was implemented is not realistically significant when measuring Clinton's economic performance, BUT, if you are merely measuring statistical patterns, then you would be measuring some 'noise' in that first year which the statistician would consider significant.

Further, the longer time period you look at, the less "noise" from the ordinary business cycle will obscure the results. Looking at each year as an isolated data point assures that any longer term effects from governmental policy will be drowned out.

Further, he admits his basic difference in perspective. He denies that the government has a principal impact on the economy, and designs a model that reflects that belief. Not surprisingly, he finds little but the statistical noise he sets himself up to find.

Thirdly, Jason goes and does it again when he says:

*"Torgerson has left his implicit assumption unidentified and unchallenged. Take your pick*" . Just because he did not notice my "implicit assumptions" he assumes they aren't there. Let's help out Jason with a link to a recent post which specifically explains why Democratic administrations do better economically. In that post I said:

"Basically, expenditures targeting low to middle income people grow the economy short term far more than expenditures targeted to affluent people. For example, committing more funds to extending unemployment benefits adds $1.74 to GDP for every $1.00 spent. By contrast, reducing taxes on stock dividends only adds 9 cents to GDP per dollar of taxes reduced. So doing a little math here, if you repeal the dividend tax cut, and take the estimated $36 billion in revenues split evenly between increased benefits to the unemployed and to reducing the budget deficit, that action would increase GDP by $28 billion in the first year alone (about a 0.3% increase in growth.) You'd lose $3.24 billion in GDP growth from repealing the dividend tax reduction, (.09 X $36 billion) and gain $31 billion in GDP from extending unemployment benefits, (1.74 X $18 billion).Paying more money in unemployment benefits to reduce unemployment seems to be counter-intuitive. But, if you think it through, it makes sense. An unemployed person receiving unemployment benefits is the person most likely to spend those benefits quickly in ways that keep the money recycling through the domestic economy: food, rent, bus fare, utility bills etc., which are all provided courtesy of employed workers, very few of whom could possibly be outsourced to India."

The specific GDP effects of the policies cited above come from an Executive Summary of a paper produced by Economy.com, the folks behind the Dismal Scientist.

Surprise! Democrats are far more likely to favor policies, like unemployment benefits, tax cuts for the less well off etc. that happen to have a more stimulative effect on the economy than the Republican's favorite goodies for their constituencies, like capital gains tax cuts, corporate breaks etc.. Therefore, it should come as no surprise whatsoever that applying more stimulative policies to the economy, you get...

*Just so my friends at the Dead Hand can keep up, this is what is known as "causation".*

**more stimulus to the economy!!**Now, if this whole line of reasoning holds up, what you would see is a significant difference between the overall economic growth patterns under Democratic government control versus Republican control. And, of course, that is exactly what you see.

You see, this is all basic policywonk-craft in Washington. It is not rocket science, governments have been priming the pump as needed for decades. It works. The implication here is when a George W. Bush pushes economic and tax policies that direct benefits towards the rich, then he darn well knows that such policies won't be very effective in growing the economy, because they have never worked very well. So when he looks into the camera and says with a straight face that his tax cuts should reinvigorate the economy, he is lying to you. And, if he is too dense to understand that he's lying to you, he's got 10,000 policy wonk clerks that work for him scurrying to implement his program who know full well they are committing a fraud.

So right wingers HAVE to hide the evidence of their very own eyes amongst as many layers of bogus assumptions and statistical massaging necessary to protect them from the hard cold reality of the dishonesty of their fearless leaders.

## 4 comments:

Okay, Torgerson, let's take this point-by-point, shall we?

First off, I am NOT talking about year-to-year changes. I am measuring the growth over an entire administration (four or eight years) and describing that growth in terms of average annual change so that different administrations can be compared.Um... yes you are. You are averaging year-to-year changes over the course of an administration. An average—or amean, technically speaking—is a statistical device. In fact, it ishalfof a statistical device; the other half is calledstandard deviation, and it describes the reliability of the mean based on the sample size and distribution. Talking about a mean without talking about its standard deviation is, at best, incomplete, and at worst deliberately misleading.Standard deviation is such a basic concept in the science of counting that, if you are going to disallow references to it, then you may as well throw out averages as well and return to your primary debating tactic: bald, unsupported assertion.

If you are happy to talk about the average year-to-year GDP change across

oneDemocratic administration, then surely it isn't too much of a stretch to talk about the average year-to-year performance underallDemocratic administrations since 1950. After all, as you pointed out yourself:The longer time period you look at, the less "noise" from the ordinary business cycle will obscure the results. Looking at each year as an isolated data point assures that any longer term effects from governmental policy will be drowned out.Amen. The results? As I pointed out here, when you do that, the performance distribution is so broad as to be effectively useless for making predictions.In other words: Democrats to very well in some years, but they do very poorly in others. Hardly a momentous conclusion.

Then there's this:

Further, he admits his basic difference in perspective. He denies that the government has a principal impact on the economy, and designs a model that reflects that belief. Not surprisingly, he finds little but the statistical noise he sets himself up to find.Okay, we'll try an experiment. I just altered the model described in The Historical Effect of Political Alignment on GDP so that the operative metric is the annual change in GDP, rather than deviation from the mean. Now we're using YOUR model, right? So let's compare the resulting correlation coefficients:Linear Model, Annual Change: 0.3990

Linear Model, Deviation from Mean: 0.7215

Third-Order Model, Annual Change: 0.4359

Third-Order Model, Deviation from Mean: 0.7681

Now we see the

numericaljusification for using deviation from the mean: the numbers correlate twice as well. Another way to make that statement: they're less random. And before you decry this as more statistical sleight of hand, Torgerson, please recall that the Linear Model reflects YOUR ideas: that one party is better than the other at managing the economy. The only difference between the Linear Model and your own assertions is that it also allows the Senate and previous years' economies to have an effect.Hardly a big leap of faith. In fact, I kind of expected you to be nodding your head up through that point in my paper. Silly me: I forgot that it isn't sufficient to corroborate Richard Torgerson's ideas. You have to

quotethem.What's the bottom line? The

damned lieselement of the statistical equation happens when people—people likeyou, Torgerson—throw around words like "average" and "trend" with neither approriate context nor proper application. If you're going to use these tools in public, use them correctly. If you're going to criticize somebody else's application of these tools, for G-d's sake have the dignity to know what the hell you're talking about.I had said:

First off, I am NOT talking about year-to-year changes. I am measuring the growth over an entire administration (four or eight years) and describing that growth in terms of average annual change so that different administrations can be compared.

Jason replies: Um... yes you are. You are averaging year-to-year changes over the course of an administration. An average—or a mean, technically speaking—is a statistical device. In fact, it is half of a statistical device; the other half is called standard deviation, and it describes the reliability of the mean based on the sample size and distribution. Talking about a mean without talking about its standard deviation is, at best, incomplete, and at worst deliberately misleading.

RT: Ummm... no I am not. I am measuring the total economic growth over the length of entire administrations. I am only expressing that growth in terms of average growth per year to enable us to compare administrations of different length.

Standard Deviation is indeed a basic concept, and it's always a surprise to me when people misunderstand it's application. Rest assured you don't have to lecture me on standard deviation. I use it as a matter of course in my business continuously. My clients' investment performance is reported using both performance over time as well as monthly standard deviation to measure the volatility of that performance.

However, in this case a yearly fluctuation is noise, non-relevant information for which computing a standard deviation yields nothing whatsoever of any value. For there has never been any economic policy ever devised that has as it's goal economic growth over a 365 day period. The very notion is absurd. Economic programs are intended either for very short term stimulus when needed, or more longer term objectives when desired. So if no one is trying to make a yearly number look good, why in the world would you attempt to measure their 'success' in making a yearly number look good?

Statistics are easily abused if you don't have a clear idea of what you want to measure and how. Let's give an example to let you see where you are going wrong. Take an Olympic swimmer. Let's say in the 300 meter freestyle he has a pattern of saving his energy for a burst of speed at the end. Because of this, the only laps he actually races faster than the competition are the last two, but he wins more than his share of races with this style. Now you could compute a standard deviation of his speed per lap, which would yield a distribution from slow to fast. But, since speed per lap, and the variance thereof, is irrelevant to the point of who wins the race, you are just crunching numbers to no useful effect. (By the way, in another note, adding the Senate to the mix is less than useful. Budget bills originate in the House by Constitutional dictate. Therefore the House is much more relevant to measuring economic effects.)

Now let's take that example to a more direct application. In the economy, things happen. Certainly, 9/11 had a short term effect on growth that skews Bush's 2001 number. Now, let's say he responded with a strong economic program that generated 3% growth in 2002, 4% in 2003 and 5% in 2004. Taking a standard deviation of those four years shows a distribution all over the place. So what? A much better measure would be overall growth over his entire term. If his economic policies worked well, that number would be higher than if they did not work well. Duh. That's what we are measuring here. And by looking at the total record of each administration, we find that no matter what the 20th century threw whichever administration, EVERY single Democratic administration grew the economy better than EVERY single Republican administration.

Then you said: What's the bottom line? The damned lies element of the statistical equation happens when people—people like you, Torgerson—throw around words like "average" and "trend" with neither appropriate context nor proper application. If you're going to use these tools in public, use them correctly. If you're going to criticize somebody else's application of these tools, for G-d's sake have the dignity to know what the hell you're talking about.

Jason, read what you just typed here several times, until you understand what you are really saying. Your "paper" shows me you do not. Good luck getting it published. Check with the guys at Heritage or American Enterprise Institute. They'll pay for anything with lots of charts that show that Democrats suck.

...which leads me to wonder aloud—AGAIN—if you actually bothered to read the damned thing.

I didn't say Democrats suck at managing the economy. The only one of the two of us saying anything like that is YOU, and you're saying that about Republicans.

What I said, and have now repeated times without number, is that (at leat according to my results) economic performance doesn't appear to correlate very well with the party in power... but DOES appear to correlate well with the degree to which the party in power is politically entrenched. WHICHEVER party that may be.

Regarding statistics: if you want to construct your samples of four-year runs rather than one-year runs, of course that's fine. Trouble is, you just cut your sample population by three-quarters... which means that your confidence intervals are going to be even worse than before. Do I really have to work the arithmetic for you, Torgerson?

Here's a better idea: since you understand statistics after all, PROVE it. Give us a complete analysis, one that identifies your sample structure and discloses the confidence limits of your predictions.

Put your money where your mouth is.

Jason G. Williscroft

The Dead Hand

In a message dated 10/31/04 11:50:24 PM Eastern Standard Time, jscroft@thedeadhand.com writes:

Here's a better idea: since you understand statistics after all, PROVE it. Give us a complete analysis, one that identifies your sample structure and discloses the confidence limits of your predictions.

>>S<<

A guy hits himself in the head with a ball pein hammer. He notices his head hurts. He takes some Tylenol. He feels better. He hits himself in the head with his hammer again. It hurts, but not as much. He takes some Tylenol. He feels a little better. He hits himself in the head with a hammer really hard. It hurts a lot. He takes some Tylenol, and feels much better. Along comes an Engineer who took some Statistics 101 some years ago. He asks, "Wazzup?" His buddy reports that whenever he hits his head with a hammer, it really hurts. But when he takes Tylenol, it feels better. The Engineer says, "you don't know that. Your sample isn't large enough to make that conclusion, and based on the variability of pain and relief, you don't know for sure what is happening at all. As a matter of fact, my analysis shows that if you do the same thing repeatedly, whether it be hammer or Tylenol, the longer you do it, the better you'll feel."

His friend backs away slowly, looking to side to side. He says nervously, "....uh...right. I think I'll take some Tylenol now."

Jason, we tried Truman's (D) Fair Deal, and it got us transitioned after WWII. We tried Eisenhower's (R) slow growth policies, and growth slowed down. We tried Kennedy/Johnson's (D) programs and economic growth rose again. We tried Nixon/Ford's (R) programs and growth slowed. We tried Jimmy Carter's (D) policies, and growth sped up. We tried Reaganomics (R) and growth slowed down. We tried Bush I's (R) policies, and growth slowed down some more. Then we tried Clintonomics (D) and growth sped up again. Then we tried Bush II's (R) tax cuts and spending programs and growth slowed down again.

Jason, for crying out loud, how many times does America have to hit itself in the head with a ball pein hammer before it gets that it hurts?

Post a Comment