Saturday, April 27, 2013

Questionable Studies

Academic publication is a dialog, and bad science happens for a lot of different reasons, but it seems like this month, certain papers with problems have become unusually popularized (e.g. given full articles in the WSJ and other newspapers).

Reinhart and Rogoff ask: How does debt relate to growth?

This is an important question, and the way they framed their results (implying a debt/GDP cutoff where growth becomes hampered), got a lot of press when published 3 years ago and was cited as the basis of a lot of austerity measures. Sadly, there are mistakes. In particular, there was an Excel coding error. The authors have also been accused of selectively omitting, or cherry picking, their data, and implying that debt causes slow growth when the causation could easily go the other way, that is, making a correlation implies causation error.

In their response to the criticism, R&R admit the Excel error and strongly deny the cherry picking and causation error. To be fair, it was more of the press and interested governmental parties that were more than happy to claim this paper proved a causation.

Here's a "scorecard" of the saga with a description of how the results change, links to the original paper, the criticism, the response, and other relevant things. Note that it seems like the error may have been pointed out in a way to maximize press coverage and that R&R is just one of the many papers on the debt/growth question.

It's not just macroeconomics, biology has some issues, too.

Sharov and Gordon claim: Life on Earth originated "elsewhere." 

I am just glad this paper is on arxiv and hasn't been picked up yet by a journal (I don't think).  Here's figure 1:



Here's a great description of the ridiculousness of it. The WSJ correctly pointed out they use only 5 data points -- a severe lack of data that bests even macro papers. However, as Myers points out, they also cherry pick those data points, and abuse extrapolation.

Tuesday, April 23, 2013

What's My Password, Again?

Spring cleaning doesn't stop with my apartment. After I clean the physical world, I clean up my digital world, too. And that means, in part, new passwords for my critical accounts. I pick several strong* passwords and make sure (user name, password) pairs are different for all my important accounts.

But, boy, is it ever a hassle.

What's Your Favorite Color?

Even if I had completely random strings for passwords for every single account that I changed every month instead of every year, I still wouldn't feel totally safe. Why? I don't believe that sites secured with passwords are actually secure if they ask basic security questions to "recover" your account in the event you forget your password. Then your account is only as secure as your mother's maiden name and favorite color. And given the amount of publicly available information and stuff people post about themselves on Facebook, that means the account isn't really secure at all. With a little guessing, anyone could answer those questions! To solve this problem I try to do two things:
1) Write my own security question.
2) If I can't write my own security question, I make up a ridiculous or completely unrelated answer to the posed question. Q: "What's your favorite color?" A: "Live long and prosper."

You get the idea. Sure, I may forget my password and also the answers to the security questions, but really, if there isn't a chance that you can lock yourself out of an account permanently, then is the account really secured at all?

Think I'm paranoid? The NSA actually recommends this (see page 6):
To prevent an attacker from leveraging personal information about yourself to answer challenge questions, consider providing a false answer to a fact-based question, assuming the response is unique and memorable.
While I am on board with good password management, and do other things like encrypt my home wireless signal, use anti-virus software and firewalls, and am careful about which websites I visit, I still don't follow all the best practices.

Why? Because, man, is it ever a pain.

What's the Worst that Could Happen?

This is the question computer scientists usually ask, and sometimes it's a question worth pondering.

Here's an account of a Wired editor being hacked and all his data deleted, just because the hackers were attracted to his three letter Twitter handle.

How can you prevent this? You can't, 100%. But following the best practices linked to above gets you a lot of the way there. Here's Lifehacker on undertaking those costly security measures: have good password security, utilize two-device verification, and regularly maintain good backups. 

Actually, you should have a backup system no matter what because digital devices have a high failure rate even without the help of hackers. I think the half life of a hard drive is only on the order of 5 years. And remember the golden rule of backups: If your data isn't in at least two places at the same time, then it's not backed up.

But ultimately, you have to weigh the costs and benefits yourself about what level of security is worth having on line and off and how much and how often to backup and secure digital and physical files. The benefits of security is high, but so are the costs.

What's Likely to Happen? Is all This Security Stuff Worth It?

Microsoft published a good article on this topic a few years ago. If you want to know about common attacks on passwords, details about phishing and the total uselessness of "certificates" read the article.

The main result: "Most security advice simply offers a poor cost-benefit tradeoff to users and is rejected."

They are trying to get across the idea that worst case harm is not the same as expected harm. Really bad things can happen to you online, but they usually don't, so when users ignore security advice, they do it not because they are lazy, but because, boy, is it ever costly. But getting hacked and losing your data is a pain, too, so the question is really, Is it worth it to you?

Sunday, April 14, 2013

Who's the Fattest of Them All?

Answer this: Which region of the country has the highest obesity rate?

You said the South? I did, too. And we are wrong. Southerners self-report the highest obesity rate, but when actually weighed and measured by researchers, the West North Central region is most obese, and the East South Central region ranked fifth (out of nine) most obese! What's going on? In phone surveys, Southerners lied less than other people about their weight.

It's no surprise that lying is a very real concern on surveys, especially about sensitive issues like weight, even when the survey is anonymous. In fact, the article notes previous research has found that women under-report their weight more then men, and men over-report their height more than women, and in other cases, researchers have found people over-report how much they give to charity, state they're registered to vote when they are not, and underestimate how much they drink, among other things. What IS surprising is the fact that there is a regional component to the amount of lying.

Mark Twain said, "There are three kinds of lies: lies, damned lies, and statistics."  Normally, people sympathize with this quotation when facts are presented in misleading ways* using poorly done statistics or when numbers are cherry-picked to support only one side of a position, but the truth is even worse: Even good statistics is done on the lies that make up the survey responses themselves!

(weight research HT: MR)

* An oldie but a goodie on this topic is the book How to Lie with Statistics by Darrell Huff.

Monday, April 8, 2013

Debt to GDP

Greg Mankiw wrote up a really nice article last week in the NYT about the debt to GDP ratio. After a lot of facts, he makes the main point:
Military and economic catastrophes are, by their nature, unpredictable. While we can’t plan on one, prudence requires that we take their possibility into account. In normal times, when we are lucky enough to enjoy peace and prosperity, the debt-to-G.D.P. ratio shouldn't just be stable; it should be falling. 
While the article is great, it really lacks a good picture. Mankiw talks about the debt to gdp ratio over time, so I made a graph using the CBO data to illustrate those facts. I think having this picture makes reading the article a lot easier.
Note that the CBO projections take into account the President's proposal of maintaining the current debt/GDP ratio (rather than allowing it to drop over time, as it has after past spikes).

In addition to pointing out Mankiw's article, I would like to make some additional points.

1) During the Reagan years, defense spending was increased dramatically and tax rates were cut. Laffer and others argued that people would work harder or report more taxable income due to lower rates, and the rate cuts would not only pay for themselves, but would also pay for the increased spending. The increase in the debt to GDP ratio over the Reagan years shows that the increases in spending weren't all paid for.

2) The increases to the debt/GDP due to the War on Terror and W.'s policies in his first 7 years pale in comparison to the financial crisis and policies following.

3) The CBO projections going forward optimistically assume that the economy will still return to the old GDP trend line of steady 2-3% growth in relatively short order (for instance, see the second figure in this document). This requires some years of 4-5% growth very soon. This is opposed to the idea put forward by Cochrane and others that we are on a new, lower trend line (e.g. here and here) -- the basic idea being that the financial crisis and government policies enacted during and after the crisis prevent the economy from returning to trend. If the economy does not return to the old trend, the spending levels proposed by the administration will continue to balloon not only the level of debt, but the debt to GDP ratio, as well.

4) Mankiw warns of the unpredictable crises that will surely happen in the future which will necessitate a large amount of borrowing, but what about the crisis we already know is coming? Social Security, Medicaid  Medicare, and other welfare programs are drastically underfunded and it's nearly impossible to pay out on promised benefits indefinitely.

Actually, the debt to GDP ratio graph makes clear why those programs have suddenly come under scrutiny as part of a budget deal NOW but they weren't prior to the crisis. Prior to the financial crisis, SS, Medicare, etc. were funded for between 20-40 years before serious problems crept up. Of course, anytime there are under-funded, over-promised programs, it's better to fix them as soon as possible because it gives people more time to adjust, but it's at least understandable why politicians did not want to touch them. Why were they okay for a couple decades, at least, in 2008? While the War on Terror and W's policies were running a deficit, there were medium term budget SURPLUSES predicted. As late as March 2008 (see page 8), the CBO was predicting surpluses from 2012 to 2018 or even later. Not forever, though. Eventually welfare programs would balloon out of control, running up unsustainable deficits. But as long as there was some surplus years, there was credibility that the problems would be dealt with at a reasonable point and the country could take on reasonable amounts of debt to help pay for those programs later. After the financial crisis, all medium term surpluses vanished and were replaced with large deficits as far as the eye can see. There is no rosy medium term, now. Growth disappeared, tax revenue declined, and our ability to borrow heavily in the feature to pay for ballooning welfare programs is gone. We have to be realistic about what promises can be kept and to whom, or the pain of failing to meet those promises in the feature will be even greater. We only have to look to Europe to see that.

Friday, April 5, 2013

Food for Thought

Thought-provoking sentences from worth-while reads around the web: