I was recently reminded of a joke from the days when I was an ad executive (before I became a lawyer).

Client: What time is it?

Ad Exec (looking at watch): What time would you like it to be?

It made me realize the similarity between advertising advice and mathematical modelling used to justify decisions about both climate change and COVID-19.

Politician: How scary is the model forecast?

Modeler: How scary would you like it to be?

What they have in common is that the client gets the answer they want. This brings me to the misuse of models by much of the media and many politicians, to make really uninformed political decisions about important issues like climate change and Covid-19.

Start With the Answer You Want

Scott Adams, author of the popular Dilbert cartoons, used to be a model developer in a company selling scientific devices.  In his recent podcast #903 (brought to my attention by reader Isabel Gibson) he explained that if someone could build a model to predict the future more accurately than by chance they would soon be very rich.  In his forecasting job if he had given management a model showing declining demand for the company’s product he would have been told to go back and create a new model showing increasing demand.  Computers can print out whatever model answers you want.  You start with the answer you want and reject any modeling results that won’t get you that answer.  As Adams says, models aren’t created for information, but for persuasion.

There Are No Facts in the Future

Politicians and the media often treat the model forecasts as facts, when there are no facts in the future, only opinions.

DILBERT © Scott Adams. Used By permission of ANDREWS MCMEEL SYNDICATION. All rights reserved.

Be First, Be Scary

If the model writer wants to become famous and influential with the media and politicians, they must be the first to produce a model that will be truly scary.  Don’t wait for better data to become available months or years later; get the model out there now, before someone else beats you to it.  If the data to be used in the model is questionable, then just estimate the data and make whatever assumptions you need to achieve your preferred result.

Even if the scary model is later criticized, the initial panic will stick in the public’s mind. There’s little risk in this ‘be first, be scary’ strategy because predictions of the future cannot be proven to have been wrong at the time they were made. If they are ultimately shown to have been wildly pessimistic, the defence is that circumstances changed after the model was created but the model would have been accurate if nothing had changed. There is no way to prove or disprove that.  The other defence is that “I exercised an abundance of caution”, as if such abundance had no upper limit.

Forecasting COVID-19 Deaths

A current debate is over the Neil Ferguson (Imperial College of London) model estimating 2.2 million US and 510,000 UK deaths from COVID-19 if nothing is done.  These numbers frightened US President Trump, UK Prime Minister Johnson and our Justin Trudeau into imposing lockdowns.  Shortly after Ferguson’s forecast, Oxford Professor Sunetra Gupta created another model with lower estimates, but her timing was too late as Ferguson had already grabbed the spotlight.  Last week, Nobel laureate Professor Michael Levitt estimated that Ferguson’s prediction was 10 times as high as it should have been, even at the time the model was made, based on his analysis of the data available to Ferguson at the time Ferguson’s model was released. He suggested that the lockdown, in response to Ferguson’s model, was a huge mistake. Instead, we should have implemented less extreme and much less costly mitigation measures (i.e. as the Swedes did).

I don’t know whether Levitt’s estimates are more reasonable than Ferguson’s, but the debate shows the wide range of modelers’ opinions.

Choose the Evidence You Want to Make the Decisions You Want

This wide range casts doubt on politicians’ assertions that they have listened to “the science” and are making evidence-based policy using the “best science available”.  Politicians say that their decisions are based on “the” evidence, when actually, they (or their advisors) have selected their evidence to justify their decisions. Listening to “the science” implies that there is only one scientific view, when obviously there are several widely-ranging views.

Science is used to explain things in the physical world, science is not about forecasts of the future.  Therefore, modelling the future is not science.

The Guardian recently published a critical editorial opinion that the ‘science advice’ on COVID-19 is essentially political rather than scientific. In a previous article it debunked the political idea of the best science available.

Panic Doesn’t Make Good Decisions

Climate change and COVID-19 are both serious public policy issues.  Decisions about them need to be made in an atmosphere of calm consideration of the costs and benefits of various alternative policies.  Panicked responses to scary models are not likely to be helpful.

Politicians Can’t Hide Behind “the Science”

I don’t blame a modeler for having an effective marketing plan for their work.  We would all like to be famous and internationally influential.  However, I do blame much of the media for treating model forecasts as scientific facts about both climate change and COVID-19.  I also blame the politicians for hiding behind meaningless clichés like “best science available” or “we have listened to the science” to avoid political accountability for decisions about climate change and COVID-19.  The politicians’ decisions may be wise or foolish, but they should not escape critical judgement without justifying why they chose to act on one model forecast — probably the scariest one — rather than another.

>>>>>>>>>>>>>>>>>>>>>>>>

If you would like to be notified of my future blog posts please enter your email address above the FOLLOW button on the left side of the page and then click on the button.

If you like my blog please click LIKE and share it via Twitter, Facebook or email.