How to predict a policy failure

As soon as the Coalition Government started cutting the flood defence budget in 2010, it was as predictable as the water cycle what would happen next: within no more than a few years and probably within the lifetime of the government which made the cuts, there would be heavy rainfall, resulting in massive floods, and a backtrack on the cuts – emergency spending if not a change to the planned budget – either way, a tacit admission of failure.

This sort of thing seems crashingly inevitable to me. There’s an obvious trajectory, of reduced budgets, reduced regulation or reduced oversight, followed by conspicuous calamity, followed by attempts to mop up the mess which generally involve reimplementing whatever system had originally been in place to prevent the calamity.

I’m not here to congratulate myself on uselessly predicting the flooding crisis (also, because I never went on record predicting it, so there’s no proof I ever did). I want to teach you how to predict similar balls-ups in the future, because the depressing thing is, it’s not that difficult.

Future policy failures can be predicted by applying two rules: maximal exploitation and multiplied risk.

Any new powers or freedoms introduced by a policy will be fully, and often excessively, exploited by the organisations which lobbied for the policy.

You can guarantee that if a particular industry or organisation lobbied for the policy, when it comes in to effect, they will exploit it to the maximum extent they can get away with, and probably even more. Otherwise why would they have spent money pushing for it?

This applies even if the powers introduced were meant to be for emergencies, special cases or last resorts only. For example, when the Regulation of Investigatory Powers Act (RIPA) was introduced in 2000, it was justified as a tool for protecting national security, but in the decade following, it was systematically abused by local authorities and other public bodies, in up to 3 million cases and for often very trivial matters.

That’s powers. When it comes to freedoms, the ones which are lobbied hardest for are the freedoms from control and oversight given to industry by deregulation. While the industry will almost certainly have argued that they will self-regulate to maintain the same standards without regulation as they did with it, it’s equally inevitable that as soon as they’re no longer under scrutiny, they’ll make maximum use of their new freedom and exploit every profitable strategy which was disallowed under the old regime. Sooner or later, disaster will strike: the precise disaster which the regulations existed to prevent.

Across a large enough number of cases, ‘risk’ is not about whether or not failure will happen, but about an expected failure rate.

As an individual, we usually think about risks as single cases: “what’s the chance of falling off this ladder and injuring myself?” But a policy is a set of rules to be applied across a large number of cases. A policy maker mustn’t think, “what’s the chance that some people will fall off ladders and injure themselves,” but, “how many people will?”

To do that, you have to apply a very simple rule, which is:

Probability of failure x Number of cases = Predicted number of failures

If you think I’m being patronising by explaining such a simple bit of statistics, then all I can do is encourage you to run for Parliament, because if you already understand it, you’re better qualified to administer a country of 60 million people than our current MPs, 60% of whom couldn’t tell you the chance of tossing a coin heads twice in a row.

An imaginary example: swimming pool regulations

How does it apply to public policy? Well, consider how a policy might have come about for a particular issue. Let’s take safety regulations for public pools, as a purely illustrative example: by which I mean, I’ll be inventing entirely fictitious numbers.

Once upon a time, there would have been public pools with no regulated safety supervision. Suppose there are 10,000 pools in the UK, and back in the bad old days before health and safety, there was a 0.01 (or 1%) chance each year that any given pool would have a fatal accident. At your local pool, there were hundreds of people using it every day, and only a one-in-a-hundred chance of there being an accident all year, so most likely, no accident ever occurred to your knowledge. And that’s true for most people. So there’ll be plenty of old codgers around who have happy, accident-free memories of relatively dangerous pools, willing to write in to the Daily Telegraph decrying all the modern red tape and risk aversion.

A 0.01 chance of a fatal accident within a year is fine for one pool, but multiply it by the 10,000 pools across the country, and you have an annual fatality rate of 100 drownings, which doesn’t look so good. So at some point in the past, as a society, we’ve decided that 100 deaths a year in public pools isn’t acceptable, and we’ve introduced safety regulations to bring that number down.

When I say that we’ve decided as a society, obviously we haven’t all gathered together, debated it and voted on the matter. Instead, here’s a more realistic mechanism of how that regulation might have come about:

Person A drowns in a pool. The family of A is angry about their preventable death, and demands a change in the law to improve safety standards at public pools. Person B drowns in a pool. The family of B does the same. Person C drowns in a pool. And so on. Every year, there are up to 100 more families clamouring for something to be done. Some of them come together to form pressure groups; others just add to the general volume of clamour. Eventually the campaigning reaches a critical mass, the issue goes up in priority on political agendas, and legislation is passed. Now all public pools must be attended by trained lifeguards while in use. This reduces the probability of a fatal accident to 0.0001, and the fatality rate drops to just 1 a year.

Note that no-one, at this point, explicitly says, “we’re satisfied with one drowning death per year in pools. No further regulation is required.” Further regulation could be written. Stricter safety requirements could be made mandatory. The probability of a fatality would never be reduced to 0, but with extra rules and spending, you could certainly reduce it by another factor of ten, meaning only one fatality every ten years across the whole country, and saving nine lives within a decade. But that doesn’t happen. Somebody may have weighed up the cost and decided those nine lives aren’t worth it, but there’s never a need to state that justification out loud. Regulation ends there, simply because one angry family a year isn’t loud enough to get noticed above all the other, more pressing issues.

So now we’re in the present day, with lifeguarded pools and a very low fatality rate. But we’ve also got all those old codgers moaning about how pool safety was fine as it was, and children are too mollycoddled, and we’ve also got pool owners lobbying the government to reduce the costly burdens of lifeguards and safety requirements. The government agrees there’s too much red tape, and decides deregulating pools would be a good idea.

Completely abolishing the need for lifeguards is a pretty implausible policy; let’s suppose instead that they make two lesser changes. First, they remove the requirement for any particular lifeguard training standard (because they trust industry to be ethical and self-regulate, so they can rely on responsible pool owners to provide, and audit, their own training). Second, they stipulate that having on-site security staff monitor CCTV coverage of the pool is sufficient to meet the lifeguarding requirement, even if those staff aren’t physically at the poolside and are monitoring other cameras simultaneously.

It sounds worryingly believable doesn’t it? So much so, I’m tempted to delete this whole post right now, in case a Coalition policy wonk stumbles across it and misinterprets it as a genuine suggestion.

What our hypothetical government glaringly fails to make clear, during its bonfire of the health and safety standards, is that, although the fatality probability probably won’t go back up to its pre-regulation level, it will go up. If any minister or civil servant is ever pressed on this point, they give some waffle about “accepting the risk”, which is a blatant weasel phrase, because it implies he either hasn’t calculated the probability x cases equation, or he’s pretending not to have.

Within a year, all the pool managers across the country have taken full advantage of the cost-saving slackening of safety standards that they lobbied so hard for, are paying the merest lip service to the idea of in-house lifeguard training, and have overworked, undermanned security half-heartedly monitoring their pools, often from entirely separate buildings, along with dozens of other tasks competing for their attention. In these circumstances, the probability of a fatal accident goes up by a factor of ten, to 0.001. It’s not as bad as it was when there were no lifeguards at all, some managers do resource their security staffing and training better than others, and most security staff are reasonably conscientious, try to do their job well, and many of them save lives in the pools.

Nevertheless, the fatality rate is now 10 deaths per year. There are a few other factors worth noting too. First, as a society we’ve become accustomed to 1 death a year in pools, so 10, though less than the 100 our grandparents were used to (in a time we rightly regard as less enlightened) is a shocking rise. Second, we live in a more connected, media-saturated age, so journalists quickly and enthusiastically pick up on those deaths and connect them together into a narrative of lethally declining pool standards. Third, those journalists aren’t shy about digging down into those stories to uncover a few particularly shocking details (which may or may not have had any causal link to the accidents): the pool manager who employed his brother-in-law as a security guard despite a history of convictions, the two security staff who were having sex in the monitor room while a child drowned, etc, etc. Fourth, it’s absolutely impossible for the minister in charge to avoid the implication that since these deaths can be directly attributed to his deregulation, he is personally responsible for nine deaths this year, will be for another nine next year, and for nine every year after that until public pools are re-regulated.

And so, the inexorable course of events plays out, and the minister, if he survives, will have to yield to pressure and bring in new regulation to tackle the problem, although he’ll try to avoid ever admitting the original deregulation was a mistake, and so the new solution will probably be some complicated hotch-potch (compulsory registration of in-house training, a semi-independent auditing body, etc) which is more unwieldy and expensive than ever, and still doesn’t quite get the fatality rate back down to 1 per year.

Real life examples

It’s a fictional example, but I’m guessing that everyone reading can recognise the same pattern in events which have already happened, and events which haven’t happened yet, but inevitably will.

  • Policy: Cut the budget for flood defences.
  • Result: Flooding.
  • U-turn: Emergency spending. Budget increase for flood defences.
  • Policy: Introduce new job advertising website with limited manual checking/filtering for fraudulent adverts (to cut costs).
  • Result: Increased abuse of website by fraudsters to con vulnerable job seekers.
  • U-turn: More manual checking/filtering introduced (increasing costs again).
  • Policy: Abolish requirement for risk assessments in office working environments.
  • Result: Increase in serious accidents in office working environments.
  • U-turn: Reintroduction of requirement for risk assessments?

The second has already happened: the Universal Jobmatch website launched by the DWP in 2012. The third seems to be just an idea that’s floating around at the moment, but it’s definitely something the current government have talked about doing.

As soon as you’ve intuitively grasped the idea that failure is an inevitable result of risk multiplied by large numbers – and become cynical enough to believe that any policy weakness will be ruthlessly exploited by those it gives power or profit to – the predictability of these failures and U-turns becomes obvious.

  • Policy: Increase age requirement for driver re-testing from 70 to 80.
  • Result: Increase in road accidents caused by elderly drivers between 70 and 80.
  • U-turn: ?

This example was a proposal made by the Minister for Roads, Stephen Hammond MP, in his review of the DVLA last month. I haven’t seen the statistics on which the DVLA base their claim that the change would have “little or no impact on road safety”. Presumably they’ve looked at the accident statistics for older drivers and done some cautious predictive modelling. But I’d be very surprised if, following an increase in the age threshold, there wouldn’t be a number of accidents involving dangerously incapable 70-80 year olds, who would have been stopped from driving by re-testing. I’d also be surprised if, when those cases are reported, there wouldn’t be some kind of U-turn on the re-testing policy.

Another policy where people are currently discussing risks versus regulations, without admitting that they’re talking about greater or lesser rates of failure, is euthanasia. One of the objections is that, with euthanasia legalised, some unscrupulous family members may be tempted to put pressure on older relatives to commit assisted suicide, in order to remove them as a burden and get their hands on their inheritance.

The pro-euthanasia camp says that suitable legislation would provide enough checks and balances to stop that happening. But no system of checks and balances will completely eradicate the risk. Stricter regulation and more checks would reduce it, but there would be diminishing returns, and the more bureaucracy there is, the fewer genuine cases would get through, voiding the purpose of the policy.

I’m pro-euthanasia myself; I believe that a person’s right to consensually end their life trumps the risk of abuse. But it doesn’t help the pro-euthanasia argument to pretend that the risk of abuse means ‘doing our best to stop it, then crossing our fingers that it won’t happen’. It means there will be cases of abuse, and we have to legislate to minimise them within reasonable bounds.

That’s what angers me most about these policy failures: the dishonesty inherent in the introduction of the policy. When an industry says, “self-regulation will prevent abuses”, I know they’re lying and I know that abuses are on their way. If they were to say, instead, “we admit that members of our industry will try to abuse this policy; here are our methods of preventing them, and here is our prediction of the number who will abuse it anyway”, I’d be more sympathetic. At least we’d have a reasonable criterion against which to judge the success of the policy, and be better prepared to deal with the inevitable failures.

Leave a Reply

Your email address will not be published.