Arnold Kling

More Privacy Luddites

Arnold Kling, Great Questions of Economics
Previous Entry Next Entry

Linda Gorman and Dave Kopel make the Luddite case against a national database that could be used for security screening.

I am not going to argue that a national identity card is some magical cure for security. However, I do believe that security screening needs to take advantage of the data that is available to sort people into risk classes, and that this process will be more effective if identity can be established with greater confidence.

I want to quote the article in detail and give my rebuttal.

Consider the problems with Colorado's Central Registry of Child Protection...

When the auditor's office compared a sample of 31 incident reports with their registry entries, it found 44 data-entry errors. Fifty thousand of the 107,848 records were incomplete.

Errors by themselves are not an indication of a flaw in a database. If in one entry my middle initial is incorrect, that counts as an error. It is possible that a statistical screening program could be very effective, even if the underlying data has numerous errors.

On the other hand, the authors also point out the following:

Forty percent of a sample of 48 registered sex offenders known to have committed crimes against children were not listed in the registry.

What this tells me is that there is a lot of mis-classification in this particular database. However, this database relies on one factor (past conviction records) to carry out classification. That is like trying to evaluate a person's credit risk based on the performance of their Sears account. One-factor classification systems are unreliable.

Another example cited by the authors:

One product of the matching program that the Social Security Administration (SSA) runs with the IRS is a pool of invalid Social Security Numbers created either by mistake or by people using false identities... There are currently more than 227 million records in the file...

What this shows is that we should not use social security numbers for identification purposes. The social security system was not designed to serve that function.

the unforgeable document has yet to be invented.

This is true. Relying on a single card or identification number would be hazardous. Again, single-factor systems are unreliable. However, screening systems that rely on multiple sources would be much more difficult to evade. I may be able to forge a U.S. passport with a name "John Smith," but if "John Smith" living at the address listed on the passport has no credit history, no record of tax payments, no entry with the motor vehicle administration, etc., a red flag will show up.

Crooked government employees present another practical problem. In 1997, 29 Social Security Administration employees were convicted of crimes ranging from creating fictitious identities, to fraudulently selling Social Security Administration cards to abusing confidential information.

Once again, the authors are acting as if social security were designed to be part of the national security infrastructure, when this is not the case. A fraudulent social security card may be easy to get. Obtaining a fraudulent FBI identity card or NSA identity card is rather more difficult. There has to be a focus on secure processes.

Since evildoers are a small fraction of the general population, this boils down to building a security system that concentrates its limited resources on monitoring the innocent. Freed from intensive surveillance, real criminals should have no problems hiding in the system's inefficient cracks.

Our current security system is the one that infringes on the innocent, subjecting airline travelers to long lines and intrusive searches. A data-driven security system would sort people into low risk and high risk. Low-risk people (the innocent) would move more freely than they do today through airports, sporting events, and so forth. High-risk people would be subject to more scrutiny than they receive today.

Gorman and Kopel do not offer a single constructive suggestion for making national security more effective. As I wrote in Privacy Luddites, the net result is that they marginalize themselves from the debate. There is a valid reason to worry about ineffective security and about the potential for abuse of power. But unconstructive opposition to data-driven screening makes no contribution to addressing the problem.

Return to top