Government policymakers have been hoping for twenty years that companies will be driven to good cybersecurity by the threat of tort liability. That hope is understandable. Tort liability would allow government to get the benefit of regulating cybersecurity without taking heat for imposing restrictions directly on the digital economy.

Those who see tort law as a cybersecurity savior are now getting their day in court. Literally. Mandatory data breach notices have led, inevitably, to data breach class actions. And the class actions have led to settlements. And those freely negotiated deals set what might be called a market price for data breach liability, a price that can be used to decide how much money a company ought to spend on security.

So, how much incentive for better security comes from the threat of data breach liability? Some, but not much. As I’ve been saying for a while, the actual damages from data breaches are pretty modest in dollar terms, and the pattern of losses makes it very hard to sustain a single class, something that forces up the cost of litigation for the plaintiffs.

You can see this pattern in recent data breach settlements. I put this chart together for a talk on the subject at the Center for Strategic and International Studies. While the settlements below all have complications (Sony’s settlement was mostly in free game play, for example), they all cap the defendants’ total liability. And what’s striking about the caps is how low a price these agreements set, especially on an individual basis, where $2.50 per victim looks to set the high end and 50 cents the low. Of course, to determine how much you spend annually to avoid that liability, a company would have to discount the settlement price by the probability of a breach in any given year. Even Sony doesn’t have a breach every year, so a probability adjustment cuts the value of avoiding liability to something between a half and a tenth. At those prices, I wouldn’t expect much change in corporate cybersecurity budgets.

(I know that these charts don’t account for the biggest claims in cases like Target and Home Depot — banks suing for the cost of reissuing credit cards. That’s a very different theory of liability mainly applicable to a limited number of big retailers. In the end I doubt that liabilities to issuing banks will drive much cybersecurity either, not because the claims are low — they’re more likely to be in the $50 per card range — but because establishing liability will not be all that easy and because things like tokenization will likely prove much cheaper than improving security.)