Dec

26

The long-heralded MIT Billion Price Project , which monitors the daily price-changes on 5 million items is now available. This is a valuable data for TIPS and other traders, and a step forward in the dissemination/analysis of market-based data….independent of the Labor Department.

Worth a visit.

Jim Sogi writes:

The data is surely helpful, but the the graphic on the site using colors and a separate legend is nearly impossible to decipher, especially for the colorblind. They would learn much from Tufte.

Rocky Humbert adds:

I'm sure if you make a sufficiently generous donation to MIT, they will send you a copy in whatever color you'd like.

Your comment is reminiscent of when I was a first-year associate at GS (in Fischer Black's group), and delivered one of the first option-pricing models to the new oil trading desk. The app ran on an IBM 3270 terminal (hooked to a System /370). The trader said, "This is really nice. But can you do it in yellow?" [IBM 3270 terminals were monochrome and informally known as "green screen terminals."

Epilogue: That trader didn't last very long at GS.Personally, I find the following text from the MIT website more interesting than the color: " As part of the BPP we are not only constructing average price indexes – such as those you can see in the web page – but also construct tracking indexes with the purpose of predicting macroeconomic variables. We are working on predicting the CPI seasonally-adjusted announcements, the PCE, and even the GDP. These tracking indexes use the information from daily price changes to improve standard models that predict the macroeconomic performance of the US economy.

We started in the most obvious place – trying to forecast what the BLS will announce in terms of monthly inflation. Although the BLS publishes the inflation of November in mid December, today is the last day of the month so we already have a measure of the inflation rate of the retailers in our data. So, here we go… Our prediction is that the BLS will announce November's inflation rate (headline, seasonally-adjusted) of 25 basis points. Our models produce estimates between 21 and 27 depending on how the different prices are aggregated. Well, we now have to wait 15 days to see how close we are… that long? yes… that long.

An important point: this estimate is different from the inflation you can compute in the average of online prices (the indexes we are showing on this page). They are related but not identical. We are not going to release the tracking indexes yet. We are at the stage in which we are mostly testing and researching the best way to do the prediction.Epilogue: The CPI actually came in at 0.12% … so there's a divergence that will either resolve or cause MIT to tweak their methodology….


Comments

Name

Email

Website

Speak your mind

1 Comment so far

  1. Sam on December 29, 2010 11:42 am

    Hal Varian’s team is doing similar work at Google. http://www.theatlantic.com/technology/archive/2010/10/google-price-index-highlights-slowness-of-economic-data-collection/64393/

Archives

Resources & Links

Search