Government Accounting

[Solid]

Hi everyone,
I am currently working on a study attempting to determine whether there is a correlation between Presidential popularity and the amount the government spends on contractors over a 20 year period (preferably security and FID contracts but I'll take what I can get). The numbers are publicly disclosed somewhere, but my problem is that I don't know where and my googlefu is weak. I was hoping that through a miracle some of you more savvy types might have an idea about where I can look.

I've been through all of the industry groups, had a look around the GAO (they investigated PMCs in Iraq but don't publish their data), the bibliographies in Corporate Warriors, FP magazine reports, general academic texts, and war college publishments, and have also tried to look directly at civilian manpower contracted (SIAD has this) but the numbers only stretch to 1997 and thus there are not enough data points.

I don't like asking for handouts, but I hate the idea of not exhausting all of my resources in trying to piece this puzzle together.

Thanks for even reading the thread, and have a good day.

Solid

[x-factor]

Dude, I don't want to rain on your parade, but even if you do find a statistical correlation (and I doubt very much that you will) you're not going to be able to make any kind of judgement about causation. Presidential popularity is simply too broad a variable.

Its like trying to find a statistical relationship between an NFL team's merchandise sales (a measure of overall popularity) and the amount they spend on their punt return specialist. There are way, way, WAY too many other things that effect it.

Before you put in anymore work you need to reexamine a) what issue you are trying to investigate and b) how to specifically address that issue.

If you're trying to get at public sentiments regarding private military corporations, you're going to have a very hard time because before about 2004 (the Blackwater incident in Fallujah) pretty much no one outside of the national security community knew what a PMC was.

[Solid]

X-Factor,
I'm trying to determine (or actually rule out) the notion that Presidential popularity effects his force application choices-- thus I'm comparing Presidential polling to number of PMCs deployed at the time. I don't actually expect to find any kind of correlation, but that won't effect the study.

My problem is that I can't for the life of me figure out where the accounting channels through. I was trying to look for the contractors the state department is paying, and couldn't find any documents beyond their budget, which of course doesn't separate the Dyncorp contract in Sudan from the other humanitarian missions the State dept might be pursuing. In all honesty, I'm surprised by the degree of obscurity the government has allowed in this matter. With very specific exceptions, it seems to me that knowing how the govt is spending money and who that money is going to is in the interest of the American people.

Either way, any help would be appreciated.

Cheers,

Solid

[Razor]

x-factor, the term you're looking for is "lurking variables" (yes, I'm currently taking Probability & Statistics, yipee :boohoo).

[x-factor]

Solid...

If the correlation doesn't effect the study then why are you doing it?

To my knowledge the PMC vs military choice is made by operational commanders in State and Defense, relatively far below the presidential level. Plus, in addition to the lurking variable (thank you Razor) issue, in alot of cases you're going to be arguing counterfactuals which isn't going to do you any good.

I think you're addressing an interesting issue, but in terms of evidence you're chasing you're tail. You're not going to find much and anything you do find is going to be so far distanced from the point of causation as to be irrelevant. I recommend you refocus your thesis to something that you can actually do effective research against.

[jatx]

Solid,

1. Why would you expect a linear relationship?

2. Supposing you do the regression (statistical correlation is not the tool you want) and fit a line, your r-squared will be minuscule.

3. Look at it another way. Research regression analyses of presidential popularity and find out which multilinear regressions have the most explanatory power. Grab the data and set up the same test yourself, but drop a "contractor" variable into the mix. Does the R-squared go up or down? What happens to the error term? Do you observe multicollinearity? Does the multicollinearity make any sense?

I expect that this exercise will teach you a lot more about statistics than it will about presidential popularity. Junk in, junk out.

[GratefulCitizen]

I've got to echo x-factor.

Causality is a problem.

The amount spent on contractors is an effect.
Presidential popularity is an effect.

You are trying to determine if Presidential popularity is also a cause, and what that cause may affect.

Some problems:

How do you allow for of the other potential causes involved in the amount spent on contractors?
How do you even determine what those causes are?
How do you allow for all of the causes which affect Presidential popularity?

If Presidential popularity affects the amount spent on contractors, are there ancillary issues associated with said spending which, in turn, affect popularity?

Even if there is some correlation, it doesn't necessarily imply a direct relationship.

e.g.,
There is likely a correlation between lung cancer and yellow-stained fingers.
This doesn't mean yellow-stained fingers cause lung cancer.
(nor does correlation mean C02 causes global warming, for that matter...-rant off )


It is an interesting hypothesis which may have merit, but it looks to be quite untestable by the suggested means.