The independent resource on global security

What happens when the traffic light never turns green?

Olaf de Groot and Neil Ferguson

A world where traffic lights never turn green would be a strange and dystopian place, and yet this is the world created by the United States’ recently-retired ‘traffic light’ security indicator—a five-stage indicator of threat that never went green. Rather, it spent its entire life fluctuating between amber (‘elevated’ threat) and red (‘severe’ threat). In essence, this suggested that the USA was permanently under a higher-than-average threat—not just a mathematical impossibility but also, ultimately, an entirely useless source of information.

The disconnect between threats and real-life events

Thankfully, the USA has replaced this indicator with one less prone to ‘threat inflation’, although it still suffers from political manipulation. This two-stage indicator now has a base level and can be temporarily elevated by specific (yet undefined) threats. However, other countries including the United Kingdom and Sweden still follow the ‘traffic-light’ approach.

While a ‘low’ threat level in Sweden may be credible, it is more difficult to believe that a small band of dissident Northern Irish terrorists poses the ‘substantial’ threat to Great Britain suggested by a secondary UK indicator. This is not to suggest that these groups do not threaten the lives of individuals in the UK. There is, however, a major difference between the likelihood of an attack and the scale of the attack such groups currently appear capable of mounting.

The primary problems, therefore, appear to be that the developers of indicators do not have a strong idea of what they are supposed to be measuring, and that they have created a disconnect between the ‘likelihood’ of an event occurring and the ‘effect’ of such an event, should it be realized. Threat relates both to events that occur and to events that are deterred or prevented from occurring, while the expected scale is also important. In other words, the relevant threat level depends on the effect of incidents that would occur in the absence of deterrence.

The impact of threats and acts of protection

Although not exactly the Newtonian ‘equal and opposite reaction’, if there is threat, it is likely that governments will aim to protect themselves, and their electorate, from it. In this respect, any meaningful indicator must also account for the role of ‘protection’. Like threat, however, protection is not simply a catch-all term. It relates both to the scale of protection and to its effectiveness. For example, it is, in principle, possible to avoid skyjackings by grounding every airplane in the world (‘the scale of protection’). However, the direct and indirect costs of such a policy would be many times greater than the benefits of the skyjackings avoided (‘the effectiveness of protection’).

When it comes to protection, therefore, one must strive for ‘balance’. That is, the balance between the outlay spent on protection and the value of that protection—or, as we economists would describe it, the balance between marginal costs and marginal benefits.

Threat and protection are relatively objective ideas, while ‘security’ is more normative. Accordingly, a third aspect—individual perceptions—is also important. An individual safe from all dangers but still scared to leave their house, for example, could hardly be described as ‘secure’. In Germany, for example, recent research has shown that the perceived threat of transnational terrorism is higher among population groups that are statistically less likely to be victims.

How can economics better measure security?

In our forthcoming book chapter, we suggest a methodology for making the measurement of security more transparent and accurate. Of course, the measurement of realized threats is unproblematic. Measuring events avoided, however, is an entirely different ball game. This is where econometrics comes into its own. The ability to isolate causal determinants means that, in principle, it is possible to accurately define the effects of protection. From this, by repeating our analysis with all such deterrents set to zero, we can create a counter-factual showing the level of threat that would have occurred had no protection been provided.

Protection, too, may seem complex to measure. That said, our concern with the effectiveness and scale of protection provides our rationale. Simply put, the effectiveness of protection should be regarded as the level of threat effectively avoided. The value of protection, therefore, is already defined within the threat analysis and is the total threat minus the realized threat. Such positive effects of protection can then be compared with government spending to determine whether it provides value for money.

Finally, while individual perceptions may seem like the most difficult of our notions to estimate, economics can still provide our answers. When individuals' behavioural patterns cannot be rationally explained by the observable incidence of threat, ‘willingness to pay’ becomes important. The consumption foregone to purchase items deemed ‘protective’ (e.g. burglar alarms or locks) reflects something of a cash value of that individual's perceived security. A person may end up better off feeling ‘protected’ than spending the same money on other, ‘non-protective’ goods. Of course, such consumption patterns may not be wholly observable. Other behaviours, such as voting for political parties with tough security policies, can be observed and become important measuring devices.

Towards a more meaningful measure of security

By bringing these notions together through causal regression analysis, we provide a new rationale for objectively and meaningfully measuring security. Although a series of statistical outputs, or even some clever algebra to normalize them into a 1–100 scale, may be less palatable for general consumption than a traffic light system, the process yields more useful information. To begin designing meaningful security policies, and to provide evidence-based evaluations of their effectiveness, we must know what is to be measured. This ‘what?’ is entirely missing from the indicators developed to date.

By beginning with robust definitions and a transparent and replicable methodology, we not only measure variations in security (and security policies) across time and space, we also provide a definite answer to the hitherto intangible ‘what?’. From this, the presentation of outputs can follow. Perhaps we can assign each of our numbers to a colour on the pantone scale. Or, maybe, our outcomes will be numbers that already resonate with the population.


This blog post is published as part of a collaborative partnership between SIPRI and Economists for Peace and Security (EPS).