[08/17/2007: This site has been retired. It's a hassle to maintain, and after three years I think I've made my point. If you're interested in acquiring the spam corpus for research purposes, please contact me.]
Who buys stuff from spam? Usually, there's no way to tell. Spammers aren't exactly forthcoming with their sales numbers. If only those numbers were required to be made public; then we could correlate the arrival time of the spam to the volume in sales at that point.
A pleasant daydream, you say? Well, actually, there is one type of spam where we can see the results: spam that tries to get you to buy stock. Spammers send out spam promoting a stock to try to drive up the price. People buy the stock, and the spammers sell. Since in this networked age we can get stock quotes in near-real-time, it's easy to match spam arrival time to stock price and trading volume. Two 2006 studies (Böhme and Holz, and Frieder and Zitrain) show how stock spam takes money from those who respond to it, and makes money for the spammers.
This site is my own contribution to science: a daily updated catalogue of stock spam. Below you'll see a graphical representation of the stock spam I received over the most recent day of stock trading. Each stock price graph has been annotated with a vertical red bar every time I received a spam for that stock. On the left side of each graph you'll see any spam that came in after the market closed the previous day.
11 spams sent before market close 2007-8-6:
126 spams sent before market close 2007-8-6:
In addition to my normal spam-filtered email addresses, I've got spam trap addresses that receive an enormous amount of spam. About 15% of the spam I get is promoting some company or other (when I started the Monitor in 2004 it was 3%). It's easy to distinguish this spam from other types of spam because it uses a distinct vocabulary. However it does use many of the general obfuscation techniques as other spam, seemingly with no effect on its credibility among those it suckers.
A script run each evening finds new stock spam in my various inboxes, grabs the latest price charts for those stocks and annotates them. It then uses the accumulated data to write out reports like the one above.
False positives are rare but possible, especially for stock symbols that are also English words. I'm on the lookout for these and get rid of them as I see them.
The aforementioned academic papers: The Effect of Stock Spam on Financial Markets by Böhme and Holz (which is based on SSEM data), and Spam Works: Evidence from Stock Touts and Corresponding Market Activity by Frieder and Zitrain, which is based on independently gathered data. You can explore the Frieder/Zitrain data set at their useful website.
I was interviewed about this site for the public radio program Marketplace.
The very informative Stock Spam FAQ, which mantains a list of Spam-contaminated Stocks.
Spam Stock Tracker: how much money can you lose by buying spammed stocks?
|
This document is part of Crummy, the webspace of Leonard Richardson (contact information). It was last modified on Tuesday, November 11 2014, 12:46:15 Nowhere Standard Time and last built on Sunday, September 24 2023, 16:00:02 Nowhere Standard Time.
| Document tree: Site Search: |
