street prices for digital goods?

[David Molnar]

John Ioannidis wrote:
> Hmmm... a how about a market-data feed for warez?
> 
That would be useful for research. My colleague Karl Chen pointed out 
that it would probably be more useful for the underground market.

For the case of drug street prices, the U.S. Drug Enforcement Agency 
does keep a database of prices, called STRIDE, obtained from informant 
and undercover agent buys of drugs. These are records from actual buys, 
so they partially address the concern Richard Clayton raises about going 
by advertised list price -- but there are concerns (to which Richard 
alludes) about whether agents systematically overpay or informants 
systematically lie about the  price they paid for drugs in order to 
pocket the difference between money given to them for drug buys and the 
actual price.

STRIDE also includes data on purity of drugs assayed in DEA labs. This 
includes drugs seized by the feds, but not usually drugs seized by local 
agencies. There's actually a trio of papers here in particular that 
might be of interest to people who want to look at possible parallels 
between data gathering on drug street prices and illegal digital goods.

The first is an overview paper that discusses the conceptual and 
practical problems in doing price and purity analyses over time for 
illegal drugs. The paper also points out some interesting features of 
the drug market. For example, the author points out that drugs are 
"experience goods." That is, the purchaser does not know the actual 
quality of the good until after making the purchase. For drugs, quality 
means purity of the drug. What this boils down to is that when looking 
at time series of drug street prices, it turns out you need to model 
what the buyer believes the purity of the drug will be to make sense of 
the data.

"Price and purity analysis for illicit drugs: Data and conceptual issues"
J.P. Caulkins
Drug and Alcohol Dependence , Volume 90 , Pages S61 - S68
http://linkinghub.elsevier.com/retrieve/pii/S0376871606003061
(Unfortunately the article is behind a paywall.)

The second looks at the STRIDE data and argues it is not suitable for 
use in economic analyses of the drug market. The primary criticism is 
that the data are mainly gathered from buys intended to produce evidence 
for busts, except for a smaller program aimed solely at heroin. They are 
therefore not a uniform sample of any kind. More interesting to me, 
however, is the author's contention that the data are not internally 
consistent: he is able to separate out prices reported by the DEA from 
prices reported by the DC metro police, then does a analysis showing 
that the two agencies report a statistically significant difference in 
prices. He concludes that the difference is greater than can be 
accounted for by normal price differences within a single city and that 
therefore something is wrong with the data.

"Should the DEA's STRIDE Data Be Used for Economic Analyses of Markets 
for Illegal Drugs?"
Horowitz, Joel L
http://www.biz.uiowa.edu/econ/papers/uia/STRIDE_rev1a.pdf

The third and final paper is a rebuttal of the second. The authors claim 
that the second paper improperly lumps together retail and wholesale 
purchases of illegal drugs. They also claim that the second paper does 
not properly account for the relationship between price and purity of a 
drug. Once they toss the appropriate magic indicator variables into 
their regressions and stratify by purchase type, the supposed conflict 
between DEA and DC police reported prices disappears.

Why the DEA STRIDE Data are Still Useful for Understanding Drug Markets
Jeremy Arkes, Rosalie Liccardo Pacula, Susan M. Paddock, Jonathan P. 
Caulkins, Peter Reuter
NBER Working Paper No. 14224
Issued in August 2008
http://www.nber.org/papers/w14224
(Also paywalled, unfortunately)

What is the relevance to us? Well, I see a couple of points:

1) Like drugs, compromised PayPal accounts appear to be experience 
goods. In the case of drugs, quality is purity. In the case of 
compromised PayPal accounts, quality is something like the amount of 
money that can be successfully moved out of the account. Therefore, I 
would expect the same kind of modelling the buyer's "expected quality" 
of the good would be useful for us. In particular, failing to take it 
into account when analyzing price series could lead to the same kind of 
internal inconsistencies noted by Horowitz.

Not clear to me where other illegal digital goods stand. Botnets for 
example seem easy enough to test whether they are real. Also as Peter 
Gutmann points out, escrow services are possible and exist with illegal 
digital goods to aid fair exchange -- this is not reported for drugs.

2) Unlike STRIDE, the data sets we have reported so far were gathered 
specifically for research in mind, and not as part of some other 
mission. Unfortunately, they still are almost certainly not uniform 
samples of illegal prices, and unlike STRIDE, as pointed out, they are 
not actual transaction prices.

3) One of the complicating factors in drug data is the lack of 
standardized units. For example, Caulkins notes that 16% of all meth 
data reported in the STRIDE data was sold in units other than 
grams...and a few early analyses of the data didn't notice, yielding 
bogus results. A more serious issue is purity, again; the same $10 bag 
of pot may have wildly different amounts of THC. Similarly, as others 
have pointed out here, it is hard to do an apples to apples comparison 
of "compromised online banking accounts" if the lots of compromised 
accounts come in different sizes, from different banks, etc.

4) Finally, the sheer amount of money spent on drug enforcement and 
market disruption is huge. The NBER paper cites $8.3 billion expended by 
the federal government for the purpose of disrupting illicit drug 
markets, and $13 billion overall. How much do you think is spent, total, 
by everyone everywhere, on disrupting markets for illegal digital goods?

-David Molnar


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 250 bytes
Desc: OpenPGP digital signature
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20080920/07a4faf6/attachment.pgp>