Why the 90-9-1 rule works: inequality and community engagement
It is generally accepted that when it comes to community engagement our citizens show very different propensities to contribute. We see this in the frequent observation that in any consultation there will generally be a few people with much to say about the topic, some people with something to say and a large proportion with little or nothing to say. This widespread phenomenon is sometimes characterised as the 90-9-1 rule, which has also been adapted for online citizen engagement. On any given topic, 90% of us have little or nothing to say, 9% have something (but not a lot) and 1% have a great deal to say. In practice, you rarely get this exact split, but 90-9-1 is a useful rule of thumb: you usually get something that has the same basic pattern of inequality of contributions. In this article I put forward an hypothesis as to why this should be.
Patterns of inequality: the ‘lognormal distribution’
It turns out that the 90-9-1 rule approximates quite closely to a more general pattern of inequality which is found over and over again in human affairs. The technical name for this underlying pattern of inequality is the ‘lognormal distribution’, and I don’t believe that its appearance in the context of community engagement is a coincidence. On the contrary, I believe that it tells us something fundamental about community engagement itself.
It is known that the lognormal distribution is what results when you have a large number of influences all operating on the same entity at the same time, and in a mutually reinforcing way. Can we describe the propensity to contribute to community engagement in terms of a lognormal distribution? Because if we can then we have discovered something important about how community engagement works at a fundamental level. For the conditions for a lognormal distribution to be met, there needs to be inequality in people’s propensity to contribute, for the pattern of inequality to take a particular shape, and for there to be a credible mechanism for generating this pattern.
Is there inequality? Yes, as we frequently observe. Does the observed pattern of inequality follow the shape of a lognormal distribution? Well, the 90-9-1 rule or something like it seems to work, in my experience and that of others, so yes the lognormal distribution does seem to fit most of the time. Two conditions down, one to go.
As to the mechanism, let’s assume that an individual’s propensity to contribute is influenced by a range of factors. Not an unreasonable assumption. The most obvious factor is perhaps ‘having some personal stake in the subject of the engagement’. Building a new road in the next county is not as engaging to me as building it at the bottom of my garden.
But we could add some other possible factors, such as the amount of time the citizen has available, the stage they are at in their life cycle, the amount of prior knowledge they have of the topic, the views of their social peers, their familiarity with the processes we use to engage our citizens, whether or not they are about to go on holiday, the views of their immediate family, whether or not they have just come back from holiday, what else is going on in their lives at the time, whether the engagement process uses their first language, their sense of democratic duty, their level of boredom at the moment they are asked to participate, whether the weather outside is good or bad, the lines taken by the news media they are most exposed to, their access to and familiarity with computers, the state of their digestion at the moment, the time of day at which the engagement takes place, the consultation venue … there are many possible factors which could affect someone’s propensity to contribute. Feel free to add your own!
Fortunately, it doesn’t matter, for our present purpose, that we can’t actually measure these independent factors, or even list them all. It is enough to be able to assert that there are a lot of them, that they act independently of each other, that they act randomly (i.e. you can’t predict the extent of their influence in advance) and that they reinforce one another. If we can make these assertions, and I believe they are all reasonable, then we have a mechanism and thus the third condition for a lognormal distribution.
The 90-9-1 rule: not the exception
With all three conditions met it is reasonable to conclude that the propensity to contribute does follow a lognormal distribution, and that therefore something like 90-9-1 should be the rule and not the exception. One of the advantages of the lognormal distribution is that it can give a much better fit to your data than just the three numbers of the 90-9-1 rule. But we now have a plausible explanation for why some people have so much more propensity to contribute than others, and why this inequality approximates to a 90-9-1 pattern. It is simply because this is what happens when a large number of independent, random influences operate simultaneously.
What can we conclude from this? Principally that the propensity to contribute is unequally (and specifically lognormally) distributed. There is no point, therefore, in lamenting people’s unwillingness to contribute, and little point in spending large amounts of time and money trying to change human nature.
However, all is not completely lost. It is possible to buck the trend to some extent by counteracting some of the factors which reduce propensity to contribute. One possibility is to be selective about whom you attempt to engage, and to seek out and cherish the ‘usual suspects’ who will readily contribute. If representativeness is a major concern which rules out selectivity in choosing whom to engage with then another possibility is to supply the citizens with plenty of information about the topic. There is some evidence from deliberative engagement methods that this can be effective in raising the propensity to contribute.
More generally, it can do no harm (and some would argue it is best practice) to pay attention to barriers such as language, time, place and personal inconvenience. By tackling these barriers we might be able to increase individuals’ propensity to contribute. But, if my hypothesis about a large number of factors acting simultaneously (the lognormal mechanism) is correct, then, although we might shift some people up the 90-9-1 ladder, we won’t alter the underlying pattern of inequality. In fact, far from being surprised at finding a lognormal pattern like 90-9-1, we should be surprised if we didn’t find one.