I am a gambling man – aren’t all investors? When I get invited to speak, I like to make a wager with my hosts. I bet them $1 that the first question from the audience will be, “How do you get the data for the Vice Index?” So far, I haven’t lost once.
My stock answer: bring me a lot of cash in non-sequential bills and a ticket to Vegas, and I’ll show you.
(The reality is that economists and prostitution are familiar bedfellows. Ben Bernanke wasn’t the first Federal Reserve governor to whore himself out to Wall Street, and we don’t even have to mention Dominique Strauss-Kahn of the IMF and his shenanigans.)
Now that we’ve had our fun, let’s get serious.
The concept of Hookernomics – tracking vices to gain advance visibility to consumer spending – is not exactly new. Luxury spending as a canary-in-the-coalmine for consumer spending is a basic concept. Rather, the challenge has always been figuring out (1) which discretionary spending activities cut across every socioeconomic demographic and (2) how to track them reliably and consistently.
I can’t completely reveal the data collection methodologies. Not because they are illegal (they aren’t), but because that’s my secret sauce. I can say that it is entirely quantitative and that I use big data tools to harvest the information from the Web. From a statistical sniff test point of view, the Vice Index is more reliable and viable than almost every other macroeconomic data point. Here’s why:
- Large Sample Universe: I use data each month that reflects millions of events. By way of comparison, the Consumer Sentiment benchmark is based on a monthly survey of up to 400 people.
- Diverse Population: The information I gather is not limited by age, geography, sex, or income.
- History: The Vice Index data goes back decades.
- Static Model: The formula behind the Vice Index is always the same. The factors don’t changed and neither does the weighting. That’s an important distinction. The standard Wall Street approach is to use a dynamic statistical model: as new information comes in, the statistical models re-weight each of the variables. For example, if weather was a factor in the formula and there were heavy storms, the model would change. Or if oil prices dramatically fluctuated, that could affect the model as well. The list can go on. For better or worse, I don’t do that. My model is static. I don’t revise it or tinker with it. The data is the data.
The Vice Index is a great idea that comes from great data – but is it useful and practical? To use a technical economic term, hell yes!
Here’s a snapshot of the Vice Index and retail figures going back 20 years.
And here’s a chart showing the correlation between the retail figures published by the Census Bureau and the Vice Index for the last decade. Statisticians would call this very highly correlated (R2 measures the degree of correlation, with 1.00 being perfect).
But the Vice Index has a particular advantage: it leads by several months. This is bar none the best leading indicator of consumer spending. Since I started forecasting retail figures on Bloomberg, I have beaten Consensus most of the time. Better still, I have caught the major inflection points.
Something this cutting-edge should generate questions. What is the data? Where do I get the data? How is the data adjusted? These are all valid. I have decades of data and have been publishing the Vice Index for almost two years, but my reluctance to share my sources and methodology means that I can’t put all of the questions to rest, unfortunately. I know that it’s not fair to say “trust me”, but that’s what I am saying. Once Goldman Sachs starts sharing the details of its models, then I will too.
In the meantime, I expect to be a target for many naysayers with knee-jerk criticisms. For example, the value and validity of the Vice Index was recently questioned by Rafi Farber of CalvinAyers.com. If anyone knows vices, it’s Calvin. Rafi raises great questions about the data – it’s just a shame that he didn’t reach out to do a little fact finding.
Here are some of the key points he raises, and my response to each:
- The Vice Index needs to be at least 10 or 20 years old to be useful.
- The Vice Index claims to track activities that can’t be tracked, because if they could be tracked then the Federal Reserve would be tracking them.
Rafi, your faith in the Fed is refreshing. You’re probably aware that the Bank of England started including prostitution and drug dealing in its GDP calculations. So it can be tracked, and I’m doing it.
- The Vice Index tracks revenue.
Umm, nope. The definition of ‘Index’ means that it is a compendium of activity.
- The latest data isn’t informative and the March slowdown looks like a seasonal drop.
Nope! March happens to be a peculiar month because it is so vulnerable to weather. December is always snowy but March weather could be mild or mean – just look at this year and last year as well. Each year saw unusually late snowstorms, which tend to cool down extracurricular activities. Weather has a way of limiting means (heavier storms = slower seasonal worker hiring = less cash to burn) and opportunity (heavy storms = more family time).
Here’s the point of the Vice Index. As investors we know a simple truth: when fundamentals matter, the markets will move according to acceleration and deceleration in consumer spending. There are many, many different metrics that attempt to provide visibility to that trend. The Vice Index is one more of these, except that it happens to be more statistically robust and offers more forward visibility. Given the choice between a soft survey of 400 people (the Consumer Sentiment benchmark) and a set of hard data that tracks the actual spending patterns of millions of people (the Vice Index), I’ll go hard every time.