When does deterrence work?

An interesting piece by father and son team, David and Robert Levine, Deterrence in the Cold War and the ‘War on Terror’ deals with the issue of when 'big stick' military deterrence policies will work for countries like the US and when they will not. I had problems with the exclusive emphasis by the authors on deterrence theory since approaches offering positive incentives for refraining from terrorism seem appealing - for reasons discussed by Bruno Frey in Dealing With Terrorism – Stick or Carrot? But the Levine/Levine piece is interesting. This is my first draft at reading a difficult paper and comments are welcome.

The success of US cold war policies in preventing Soviet aggression was based on deterrence - on the threat of bilateral massive retaliation with nuclear weapons. Thomas Schelling was a major thinker behind this approach with mathematicians like Herman Kahn and Albert Wohlstetter also being important.

Mutual deterrence worked effectively because of the possibility of ‘Mutual Assured Destruction’ or MAD. If the other side offered a certain level of aggressive activity and the other country could make perfect commitments, with no constraint on ability to retaliate, then the home country''s best deterrence plan was to commit to a sufficiently high level of response so enemy activity was always deterred. Because enemy activity never occurred as a result of this stance, the cost of going through with the commitment – although not the sunk cost of making it credible - never had to be borne.

Deterrence theory works if the enemy is a rational calculator with consistent values. This is not so for modern, non-state sponsored terrorism where the enemy are religious fanatics and thugs. Moreover, the stakes are lower with such terrorist thugs - potential deaths and property damage from terrorism are likely to be markedly lower than those from even a limited nuclear war. Levine/Levine modify the deterrence story to capture the idea of modern terrorism by assuming the enemy has ‘bounded rationality’ so, with some probability, it will take actions less advantageous than the ideal course of action given an implied response. A probability density parameterizes the ‘rationality’ of the enemy by measuring the enemy’s incentive to take non-optimal aggressions. Then the ‘friendly country’ has to choose a degree of retaliation to maximise its net benefits assuming the enemy plays from this distribution.

Again a simple deterrence policy proves appropriate. For enemy actions of less than some critical level intensity there should be no response while, for actions above this, retaliation should be massively fearsome.

The sense and limitations of this rule are apparent. Suppose there are ‘modest’ and ‘extreme’ levels of enemy attack. Theories of marginal deterrence suggests modest attacks should receive lower intensity responses than extreme attacks to limit scale of attack. But this ignores the fact that some, who previously made modest attacks, will now be deterred from making any attack at all. If low scale responses discourage modest aggressors from aggressing at all and encourage extreme aggressors to become more modest then a low scale response work best and simple deterrence theory fails. But, if a low response does not deter either modest or extreme aggression then a ‘let ‘em have it’ policy works better.

The ambiguity here stems from the fact that increasing response penalties may encourage a change in the mix of types who carry out various types of terrorism and this might be advantageous or otherwise. Generally, neither all-or-nothing policies nor policies of moderation need make inevitable sense.

All-or-nothing punishments also don't make sense if an enemy only ever learns too late that they have gone too far and have crossed a threshold. And the possibility of launching a maximum attack may not be credible – enemies may not believe modest attacks will lead to maximal responses so the pain of a substantial response may produce no additional gain. Finally, a third-party enemy of the enemy (Israel) might engineer modest attacks (such as car bombings), attributed to its enemy (e.g. Lebanon), to initiate a maximal response to that enemy by a third-party such as the US. The enemies’ enemies then have incentives to encourage the appearance of aggressive activity which limits incentives for a maximal response.

When applying deterrence theory to terrorism rather than the Cold War, the claim is that terrorists are less rational and a more heterogenous enemy. The effects on (say) US welfare of dealing wiirth terrorists rather than the old Soviet Union are therefore ambiguous. The US is worse-off dealing with terrorists because of the randomness of the enemy’s response and of the enemy's consequent need to exert more aggression to trigger a maximal response from the US. However terrorists cause the US less harm than even a limited nuclear war so this increases US welfare implying less tolerance of aggression.

Terrorist states such as Syria, Iran and North Korea can be dealt with using the basic deterrence theory. Threats of an overwhelming response will keep such countries in check given the evident desire of their political elites to remain in power. They can cause great harm and will threaten to do so but won’t because they will be totally squashed if they do and realise this.

However dealing with non-state terrorist groups, such as Al-Qaeda, is more difficult. Terrorists are a diffuse group so enemy behaviour has a high random (irrational) component. The enemy is mobile and elusive so threats to retaliate will be less precise. Finally the enemy have imprecise motivating values ranging from thuggish misogyny to ignorant Islamic fanaticism based on zero understanding of any belief systems other than that of their fathers. Finally it will often just not be credible to retaliate massively against limited actions by a bunch of terrorist lunatics.

If the lunatic terrorists have terrorist state support then this increases their danger but then the deterrence theory should be applied to the state supporters if it cannot be applied to the terrorists. It is only with state support that such groups can do relatively significant damage so that fearsome retaliation should be threatened against countries providing such support.

One is finally left with the difficult case of a terrorist group operating without state support. This is difficult to tackle using the deterrence theory but fortunately these terrorists can do much less damage. Levine and Levine argue that a 15 kiloton nuclear bomb in Manhattan could kill upward of 200,000 people which, while terrible, is of a much smaller magnitude than a limited nuclear war between superpowers of the type that could haver occurred during the Cold War. Leviner/Levine claim the main cost of such a threat to the US is the threat of overreaction through lost respect for civil liberties. It requires a response but certainly a community response proportionately smaller than that appropriate to a bilateral nuclear attack. Capture them and/or kill them but don't ruin the democratic values the terrorists are seeking to obliterate in open societies - the cost is too high.

It is here however that deterrence theory does lose attraction and the approach of Bruno Frey, mentioned above, makes sense. Carrots work better than sticks. I’ll post on the Frey viewpoint separately as my review of the Frey book which I'll finish over the next week or so.

Thanks Damien Eldridge for the Levine/Levine reference.