Thursday, April 29, 2010


I was listening to This American Life and their episode on Urban Legends. There were a lot of things in this episode that really irritated me. Mainly because the reporting presented conclusions, but did so by missing pretty important alternative explanations. Explanations that left their conclusions on less solid footing than they presented.  I’m only going to talk about one of them.

If you’re familiar with TAL, you know that they break their show into “Acts”. An act is a single story that revolves around the central theme that they’ve chosen for that week. The theme for this particular week was urban legends that turn out to be true. Act 3 was called “Sleeper Cell”.  And it was about the urban legend that cell phones are dangerous.

At one point, the person being interviewed, Christopher Ketcham, notes that there are lots of studies on each side of this debate. But he notices that if you draw a line around the source of funding of the studies, you discover that of those studies that are funded by industry, 75% show no harm. But those studies that are not funded by industry have 75% of them showing harm.  The implication is that industry is buying the results that they want.

And let me say right now: maybe that’s true. But the big question that came to my mind was this: who’s funding the studies that aren’t funded by industry? There are really only two possibilities: these studies were funded by individuals (extremely rare) or they were funded by the government. And here’s where Ketcham makes an assumption that I think is false: that government is independent, working for our better interests, with no ulterior motives involved, and that their funding of studies is for pure scientific results only, and is not in any way influenced by an agenda.

I don’t believe this assumption. Maybe my friends on the left do. To which I would say to remember that republicans are part of the government, too. It’s pretty easy to see how they’ve got ulterior motives, isn’t it? Sure, you might say, they’re funded by industry. Ok. But some of them lose to democrats. Who funds the democrats campaigns? By people ardently opposed to industry? If you believe that republicans are beholden to the sources of their funding, why do you then not believe the same thing about democrats?

My point is this: if you assume that the source of funding for a study invalidates the results, then don’t you have to call into question the studies that are funded by the government? Aren’t politicians at least as politically motivated to lie and get the results they want as is industry? Why do we automatically assume ill gotten results when industry funds a study, but automatically assume validity when government funds a study?

And none of this even goes to the heart of the problem: it is patently false to say that the source of funding is sufficient as the *only* means to invalidate a study. If you want to correctly invalidate a study, you have to find fault with one of the following:
  • The methodology used to gather the data in the study
  • The data
  • The conclusions drawn from the data
A critique of the funding source is only valid *AFTER* you’ve found some flaw inside the study. At that point, and only at that point, can you claim that the funders bought the results that they wanted. If your only critique of a study is the funding source, what you’re demonstrating is that you were either too lazy to read the study or too ignorant to be able to critique it on its merits.

Watch me now feign surprise that Ketcham, the guy who found the connection between the funding and the results, was a journalist and not a scientist.

And one more thing. Ketcham’s article on this topic says the following: “Interphone researchers reported in 2008 that after a decade of cell-phone use, the chance of getting a brain tumor—specifically on the side of the head where you use the phone—goes up as much as 40 percent for adults.” Two things: first notice the use of weasel words “as much as”, second it provides no context for what 40% means. That looks like a really big number so it must be a really big risk, right?

Let me show you a completely hypothetical example of how numbers can be used to mislead.  Suppose that in any given year, you have a 1 in 1,000,000 chance of dying prematurely. That’s a 0.0001% chance. If you introduce something else and the risk of dying now is 2 in 1,000,000 the percentage is now 0.0002% chance. However since the number of people who died has doubled, you can say that the risk of the thing you introduced increases your chances of death by 100%. And this is not false. But the risk of death is still only 0.0002% after introducing the change. In absolute terms, the new risk is still incredibly small, only slightly larger than the old risk.

Journalists are prone to this type of reporting.  Numbers like this are really big and create a reason for people to read their story. Note that by this logic, a risk that goes from 0.0001% to 0.001% is a 1000% increased risk. But a 0.001% risk is still an incredibly small risk.  So when you read that the risk of some bad thing increases by some percentage, the thing to ask is this: what was the risk before, and what is the risk after? If they’re both really small numbers then you probably shouldn’t worry about it.

Wednesday, April 07, 2010


I've been re-watching one of my favorite TV series of all time, Firefly. It's unfortunate that this series was canceled before even one season could complete. It was absolutely brilliant and Fox did it a great disservice in the way they aired it.

Anyway, I like it mostly because the cast of characters in this 14 episode series are incredibly complex and rich and well thought out. My favorite character is Malcolm Reynolds, the captain. He would be - if he lived here - a libertarian. This quote from him is really quite insightful:
A government is just a group of people, usually notably ungoverned.
And it reminds me of those who put a great deal of trust in government. The government isn't magic. It's just people. People who are given a lot of power. And with almost no accountability. Sure, there are elections... every 4 years!  But in between those elections we have very little power to contain the hubris of those we've elected.

And I'm reminded of this as I contemplate a piece of news that came out today:
Today, both The New York Times and The Washington Post confirm that the Obama White House has now expressly authorized the CIA to kill al-Alwaki no matter where he is found, no matter his distance from a battlefield.
The person in question, Anwar al-Awlaki, is an American-born Islamic cleric. In other words, he's an American citizen. Now one could argue that this doesn't change anything. And I have before. I believe that the rights described by our constitution are rights that all people have - not just US citizens. So it makes no difference that he's an American or not. I believe that his right to due process is a constraint on the behavior of our government regardless of his citizenship. Of course, not everyone agrees with this.

But there is little disagreement that the constitution requires due process as a protection for American citizens, and he is one. And the president has just authorized his execution without due process! This is crossing a line beyond which is a short step to totalitarianism. I was angry when the Bush Administration held what it called "enemy combatants" for years in Guantanamo. That was, in my opinion, an egregious violation of those people's civil rights. But so far as I'm aware, no one was summarily executed. If the people in power can kill this man without due process demanded by the constitution, what else are they willing to do? Is there no line they're unwilling to cross?

My hope for this is that somewhere, something was grossly mis-reported. That there's some fact here not yet known that changes this whole story. The alternative - that it's accurate - is horrifying. It means that we don't just have a government that's fiscally out of control, we have people in power who do not respect the rule of law. And they have the keys to the weapons.

This Bears Repeating

The United States Code -- containing federal statutory law -- is more than 50,000 pages long and comprises 40 volumes. The Code of Federal Regulations, which indexes administrative rules, is 161,117pages long and composes 226 volumes.

No one on Earth understands them all, and the potential interaction among all the different rules would choke a supercomputer. This means, of course, that when Congress changes the law, it not only can't be aware of all the real-world complications it's producing, it can't even understand the legal and regulatory implications of what it's doing.

Glenn Harlan Reynolds

No Tax Refunds

Megan McArdle has a great post on why it’s important to *NOT* get a tax refund. Long and short: you’re giving the government an interest free loan. Don’t do it. If you really like the forced savings, you can do that on your own AND earn interest. Adjust your withholding on your W4 form so that you don’t give them the interest free loan. Unfortunately, she doesn’t address how to adjust your withholding so that you can control how much you pay in taxes. So I’m stepping in to tell you what I do.

I should mention that I am *NOT* a tax attorney, nor an accountant. I’m simply telling you what I do and what works for me. If you decide to use this, and it doesn’t work out, don’t come looking to me for answers. You’ve been warned.

So here’s the deal. You want to withhold enough to cover what you’ll likely owe in taxes, but not too much so that you don’t give too big of a loan to the government. My personal goal is a $100 refund. My real goal would be a $0 refund. But the problem is that if I estimate wrong, I underpay. And our lovely government, who doesn’t pay interest if you overpay, will charge you a penalty if you underpay. !@#$%*^ hypocrites!

But it turns out that it’s pretty easy to adjust my withholding to be pretty darn close to what I want. Here’s what I do:
  1. On last year’s tax return, I look for the amount I owed in total tax. For 2009 (the one I just filed) it this was line 44 on form 1040. I don't know if it's always line 44, but that's what it was this year.
  2. I will use that number as a guess as to how much I will owe next year. Let’s pretend that the number was $2000 and that I get 24 paychecks per year.
  3. Now, if I want to get a $100 refund, I need to pay $2000 (what I estimate I'll owe) + $100 (my intended refund). And I need to divide that into 24 paychecks. Which means that I need to pay $87.50 per paycheck.
Now I know that I want $87.50 withheld from my paycheck to reach my goal of a $100 refund. To set that up, I get a W4 form from my HR department. It’s got a whole bunch of instructions on it. I ignore them & go straight to the form part:
  1. Line 5 on the W4 asks how many allowances I am claiming. Using the instructions, most people get fewer than 10. With 4 kids, I end up with about 10 allowances. But I declare 30. Yes, really. By declaring that large of a number, it effectively puts my federal withholding to $0 per paycheck.
  2. Now, of course, I don’t want to underpay, so on Line 6 it asks me how much additional I’d like withheld. Since I’m already at $0, and I know from above that I want exactly $87.50 withheld, I put $88.00 on Line 6 (the feds like to round to the nearest dollar, and I don't want to guess if they'll round up or down). And that’s the exact amount that my HR department will withhold for Federal taxes each paycheck.
I then do the same thing for the state withholding form, using last year’s state tax paid. It works well.

My point: I give the government a teeny tiny interest fee loan. Sure, I don't get that big refund every year, but I get a bigger paycheck each pay period.