The subject of performance targets continues to evoke strong negative reactions. In my last blog I suggested that targets can be useful if they are understood and treated in the right way. This was far from an unqualified advocacy of targets. I know their downsides very well, having worked with police forces in the 1990s, when it seemed that every performance indicator came with a target attached. Back then my own position on targets, espoused in courses and conference papers, was highly critical.
But then in those days everyone loved them. Now everyone seems to hate them, I’m tending to argue in their favour. I’m a natural contrarian I suppose, and for that reason I was pleased at the Twitter squall (not big enough to be a storm) I provoked.
Not only pleased, but interested. Some reactions were well argued and articulate, others less so (to put it mildly). Either way it seemed to be that I had caused offence by mentioning targets and not denouncing them as the devil’s work.
I evoke the devil here for a good reason. There are too many things that people do to each other that are wholly bad. Genocide, slavery, racism – the list of moral outrages is depressingly long. If anyone were to describe any of these as the devil’s work I wouldn’t argue with them, even though I don’t actually believe in the devil.
OK, so I haven’t actually heard or read someone say that targets are the devil’s work. But all I said in my blog was that in some circumstances targets can be useful, and the reaction was in effect, no, targets are always bad, irredeemably, under any circumstances. Which seems to me akin to a theological, or at least an ideological denunciation.
So I’d like to set out the reasons for my qualified position. And if anyone wants to continue the debate, that is to be welcomed. But I’m inviting you to engage with the reasons, rather than simply say ‘no, you’re wrong, targets are always bad.’
People’s dislike of targets is founded on one (or usually both) of two grounds. The first is that targets are arbitrary. The second is that they cause dysfunctional behaviour (which is a euphemism for corruption). I will deal with these two ideas in turn.
In defence of the arbitrary
In the heady days of rampant target setting, there was a standard joke among police officers involved in what in the 1990s passed for performance management. ‘How do we set targets?’ went the rhetorical question. In response, the joker licks an index finger and holds it in the air.
The implication is that targets are guesswork. They are unscientific. For sure, there is no standard methodology for setting targets. If you want to know how to work out a statistic, such as the standard deviation (an important but overused statistic in police performance analysis) or the chi squared test, you can find the formula online or in any statistics textbook. But you would look in vain for a standard algorithm for setting a target.
This is perhaps why targets are often suspiciously round numbers – a ten per cent reduction in this, a five per cent increase in that.
So targets are arbitrary.
But is that fair? And does it matter?
The word ‘arbitrary’ conjures up the idea of something randomly plucked from the air, or simply made up. In the very early days of police target setting this was true to some extent, and I witnessed quite a few examples. In an annual planning meeting in one nameless force (OK, the force had a name, but I’m not going to tell you what it was) I witnessed the Chief Constable saying he wanted a 20 per cent reduction in burglary. Because I was involved in the meeting as an independent adviser, I asked him, ‘why twenty per cent?’ In reply he referred me to one of many graphs in a document provided to inform the meeting, and said, ‘this shows burglary has increased by 20 per cent over the last five years, so if it’s risen by 20 per cent, then it can come down again by 20 per cent’.
I don’t have the space here to list everything that is wrong with this answer. But let’s not be too hard on the Chief Constable. This was nearly 30 years ago, at a time when police forces were learning to get to grips with the idea of using statistics and analysis to manage performance. They were expected to set targets in the absence of a coherent conceptual framework for understanding what they were doing. Misunderstandings and mistakes were inevitable.
In the following years, senior managers and analysts began to learn from the mistakes – in some places more than others, of course – and target setting slowly became more rational. Targets tended to be set after consideration of the limits of what might be achievable, albeit often erring on the side of the challenging. And they were often informed by analysis of what had been achieved in the past, and what the determining circumstances were. Furthermore they were increasingly negotiated with staff responsible for achieving them.
So gradually we came to have a little more confidence in the way targets were set. If a target was set as a 10 per cent reduction, then it was likely that a 20 per cent reduction would have been unrealistic.
And yes, targets are often round numbers, and that is indeed arbitrary.
But what is wrong with arbitrary? Speed limits are arbitrary. There is no scientific reason why the national speed limit shouldn’t be 71.3 mph or 68.7 mph. But 70 mph is widely accepted as reasonable. And it’s a nice round number.
Similarly – and closer to home in this discussion – the application of the standard deviation to give ‘control limits’ is equally arbitrary. We know that in certain circumstances in a random process, 95% of variation will occur between the limits set by two (strictly speaking 1.96, but two is a nice round number) standard deviations either side of the mean. This is widely used as an objective definition of exceptional. But there is no rational basis for using 95% as the criterion for exceptionality. Why not 96.2%? Or 93.9%? No reason. But as with the speed limit, 95% is widely accepted as reasonable. And it’s a nice round number.
So there is nothing wrong with arbitrariness in itself. If chosen sensibly, an arbitrary point provides us with an objective basis for making a judgement. In which case arbitrariness can be seen as the price we pay for the privilege of claiming that our judgement is objective.
What matters more is what we do with that judgement against a target. Which brings me to the second objection to targets.
In defence of police integrity
There is no denying that numerical targets can have undesirable consequences. That is not at issue. What we need to think about is whether the perverse and undesired consequences are inevitable.
Corruption (an ugly word, but an apt one) is not unique to performance targets. It is difficult to think of any human system that is immune to it. If rewards are high and (some) people are dishonest, cheating is likely. And if a system is perceived as unjust or irrational, and the chances or consequences of detection are low, people are likely to manipulate the system for a desired outcome. This behaviour is often described as ‘gaming’. The more unreasonable the system, the more likely gaming becomes. There are of course circumstances in which a target is self-evidently unreasonable. A target may be set on an outcome over which staff have no control. Or staff may not have the necessary skills or knowledge to achieve it. Or the target may be so demanding as to be beyond the reach of reasonable effort. Or staff may be set so many targets that their efforts in pursuit of each one are hopelessly diluted.
In any of these circumstances, if the reward for hitting the target is high, or if the sanction for failing to hit the target is severe, then gaming will be more likely. Indeed, in such circumstances gaming can even be seen as a rational response to an irrational system.
But to conclude from this that targets inevitably drive undesired consequences is to mistake symptom and cause. There is nothing intrinsic in targets that produces corruption, unless we assume that people are natural cheats. If we assume that most people are honest, and that they want to do a decent job, then gaming in response to targets must reflect the way targets are used, which in turn is likely to reflect weaknesses in the underlying performance ‘culture’.
The solution is not to abandon targets, but to ensure that they are rationally set, used in an enlightened way, with appropriate safeguards against gaming.
Starting in the early 1990s, I spent more than a decade warning police forces of the potential problems of targets. I don’t claim to have been a lone voice, but mine was very much a minority opinion.
Now I’m arguing in favour. How perverse is that? But my opinion hasn’t changed. All along I have argued that unless we have a clear conceptual framework for understanding performance – based on the fundamental distinction between apparent performance and actual performance – then our efforts at managing performance will founder, regardless of whether we set targets.
Targets are not malign. The problem is that it is too easy to treat them as an end in themselves, rather than as a starting point for a rational, informed discussion about why things have happened, what we have learned, and what we should do next. The underlying problem is with the performance culture. Getting rid of targets won’t make that problem go away.