Smoker's Risk by DNA

The NYT has a smaller article in today's Saturday edition that has a great example of the problems with the basics of correlation versus causation. The Synergenz study discussed in the article state that the scientists took a sample of 239 smokers with lung cancer and 200 without, and used that sample set to find the 20 genetic variations most correlated with the presence of the disease.

To 'validate' these results, the team then followed individuals with chronic obstructive pulmonary disease around for 4 years, finding that those who developed lung cancer were more likely to have been given high risk scores by the test. COPD's leading cause is smoking, but my problem is nowhere in this article does it say that the individuals followed with COPD were also smokers.

While it makes a very nice argument for likely factors indicating higher risk of lung cancer for smokers, the article does not make a very clear distinction between correlation and causation. Further, 4 years is a short time frame, relative to the average age of humans, to draw accurate conclusions. It would be interesting to see, over the span of 10 or 15 years, how the results of the study held up.

Leverage Cycle

The Wall Street Journal has a nice discussion of the "leverage cycle" that underpinned, to a large extent, the credit bubble and crunch. Summarizing, the article points out that over the years 2000-2006, collateral requirements were drastically reduced for home purchases, and compares the economy wide margin-call on home loans to the collapse of LTCM in 1998. Ultimately, the theory developed by Professor Geanakoplos boils down to:

• "When banks set margins very low, lending more against a given amount of collateral, they have a powerful effect on...buyers, whether hedge funds or aspiring homeowners, who for various reasons place a higher value on a given type of collateral....Using large amounts of borrowed money, or leverage, these buyers push up prices to extreme levels."

What is not discussed are reasons that banks would set margins lower in the first place. Yes, the model of a bank is to take in money (deposits) for which they offer a minimal interest rate, and loan that money back out at a higher rate (spread). So, the more money they can loan out, the more money they can make. There are a few ways to do this, with some of the most obvious being take in more deposits, or increase the amount that people can borrow. Banking efforts to draw in more deposits are very apparent, with what feels like every other commercial detailing how your money would be better served at a particular institution.

However, what isn't as obvious are the lending standards, which the article above touches on as the source of the current leverage cycle. Key to lending standards are the regulations set by central banks and government authorities (capital requirements being a prime example). As those standards were relaxed over the past decade, banks took advantage of the new standards to maximize their business. Note, maximize, not optimize. The WSJ article completely ignores this relationship, and in fact states that the economic models should be adjusted to allow central banks to better manage the cycles. Would this not just start the inflation of another bubble, of a different asset class, as the regulations try to balance the growth of the economy?

Bay-Es-Ian

One of the blogs I read has an excellent discussion of Bayesian probabilities and one of their more prevalent applications, testing results. The example runs as such:

You have just tested positive for a disease that effects 1 in 10,000 people. The test is 99% accurate. At first pass, this sounds pretty bad and most are 99% confident that they would have the disease. However, you have hope! In a population of 1000 people, there would be 1% of people diagnosed with the disease, for which only 1 would actually have the disease and 99 would be false positives. So, for people diagnosed with the disease, there is actually only a 1% chance of having the disease, and 99% chance of being a false positive.

Now you're saying, this doesn't make any sense, there was only a 1% chance of the test being wrong, so how could I not have it if I was diagnosed? Yes, this is true, but that 1% chance of error is much larger than the 0.01% chance of having the disease.

Initial Post

Welcome, whomever may stumble across this. Just created a blog, and postings will be sporadic at first until I get the swing of this.