Prepare for the False and Misleading Denials

It is as predictable as clockwork.  Just as every horrific mass shooting is immediately followed by a rapid-fire chorus of declarations from the gun lobby that the unchecked proliferation of guns is not part of the problem, after every major hurricane, Americans are bombarded by hurricane-force denials that the event was caused by climate change.

Particularly after – or during – each weather disaster, professional climate change deniers are joined by others, including representatives of respectable news organizations, who look and sound like reasonable individuals as they patiently explain the complex atmospheric and water conditions that precipitated the latest calamity.  The apparent goal of these reasonable individuals is not to serve the cause of the deniers but to provide an in-depth look at exactly what happened in order to provide Americans with fact based information rather than politically-charged rhetoric.  Unfortunately, many of these analyses are based upon a misunderstanding that leads to a string of reasonable sounding but false, or at least deeply misleading, statements.

The misunderstanding that leads to eventual falsehoods begins with an innocuous question: “did climate change cause this hurricane (or drought, or fire season or other disaster)?”  Given that question, a conscientious scientist will inevitably respond with a statement along the lines of “we cannot directly link this weather event to climate change.”  While some reporters will report that answer verbatim, others will interpret and rephrase the answer so that it winds up as a headline declaring that “climate scientist declares that climate change is not behind hurricane Florence” or some other weather disaster.  Those headlines will be wrong, perhaps not maliciously wrong, but wrong nonetheless.  What lies at the heart of the rapid progression from reasonable scientific statement to misleading or false headlines is a recurrent miscommunication between the version of English used by scientists and mathematicians and every day English that is rooted in the use of statistics.

The misunderstanding arises because statistical methods can be used to very accurately describe the cumulative actions of a large group (of people, atoms, storms, or whatever) even when they may not understand the causal linkage and cannot predict the activity of any single individual inside that group.  The inability to accurately predict a single instance is often perceived by people who do not understand statistics as ‘proof’ that the entire statistical relationship is wrong or unproven.  That is a fundamental error.

The classic coin flip model helps to explain the difference between predicting long-term trends and predicting specific, individual results.  Anyone with basic familiarity with statistics can tell you with absolute certainty that if you flip a coin 1,000 times, there is a better than 99% chance that the coin will land with ‘heads’ up between 450 and 550 times.  Despite being able to predict the long term results of a marathon coin flipping session with great certainty, that same person would have absolutely no idea whether the next flip in the sequence will come up heads or tails.  Or, to be more precise, he or she will have the exact same 50-50 odds as any other random person in guessing the outcome of the next coin toss.  It does not matter if the toss came up heads ten times in a row, the odds of the next toss coming up heads is still exactly 50-50 and thus completely unknown.  (And, by the way, there is a 1 in 1024 chance of flipping heads ten times in a row, but knowing that does not change that fact that the 11th toss still has a perfect 50-50 chance of coming up heads.)

To expand the simple coin toss example to real life, think of the example provided by the tobacco industry.  Americans of a certain age will likely recall the heated debate in the 1950’s through 1970’s over the health impacts of smoking cigarettes.  For decades, professional deniers of the adverse health consequences of smoking would state that no one has ever proven that a single case of lung cancer was caused by smoking.  They would then shake their heads as they assured Americans that the medical science just wasn’t there on the risks of smoking.  To drive their point home, they would point to numerous individual cases of non-smokers who suffered lung cancer as well as other cases involving heavy smokers who lived to a ripe old age.

While their statements were so misleading as to have the impact of lies (and were eventually found to be such in courts of law), the deniers weren’t necessarily technically incorrect.  Even though by the 1960’s doctors knew beyond any doubt that people who smoked had a much higher rate of lung cancer than those who did not, there was never a perfect one-to-one relationship nor was the exact chemical and genetic mechanisms that lead to lung cancer precisely understood.  As a result, even though doctors could state with virtually 100% certainty that if Americans stopped smoking then thousands of lives would be saved each year, they could not look at an individual case of lung cancer and state with certainty the patient would not have contracted lung cancer if he/she did not smoke.  Thinking back to the coin toss example, doctors based their certainty on the health consequences of tobacco use on the results of tens of millions of “coin flips” in the form of known health results for Americans who either smoked or didn’t.

While the tide of both science and public opinion eventually turned against the tobacco companies, for decades many perfectly reasonable people, including many dedicated journalists, were seduced by the verbal sleight-of-hand employed by tobacco industry spokespersons whose rhetoric was designed to portray the lack of perfect medical certainty in a specific case as a lack of any medical certainty about the overall, society-wide impacts of smoking on health.  That’s a bit like saying that because a statistician cannot tell me whether the next flip of the coin will come up heads or tails then it stands to reason that the statistician cannot really tell me anything about how many ‘heads’ will come up as a result of 1,000 — or 100,000,000 — coin flips.  That was the lie that eventually led to the legal downfall of the deniers of the health risks of smoking.

Today, a very similar scenario is playing out with climate change.  Nearly all geophysical scientists in the USA and around the world agree that burning fossil fuels and other human activities that increase the concentration of greenhouse gasses in the atmosphere are leading to a rise in the average surface temperature of the earth.   They also (nearly) all agree that even though the net aggregate rise in temperature may sound small – just a few degrees – the rise will precipitate enormous changes in weather patterns.

Rather than the whole world warming up just a bit, climate change is playing out in very different ways in different places.  Rising temperatures are already inducing significant changes in established weather patterns, but even as some areas of the globe are already seeing fairly large increases in average annual average temperature, others are experience cooling.  As major ocean currents and jet streams change in response to rising temperatures, one very common result will be much more extreme weather, meaning that a small net increase in annual average temperature will be experienced as a combination of much colder winters combined with a much hotter summers.  Changes in cloud patterns also mean that established rainfall patterns are changing and will continue to do so.  Some areas are wetter than ever while others are caught in historic droughts.  Climate models can even predict which regions are likely to become drier and which wetter — but they still can’t predict the weather for next Tuesday or next month.  The models can, however, explain that hurricanes are packing more of a wallop because rising sea temperatures are causing more water to evaporate into the atmosphere.  As all of these changes unfold, there will be many, many more severe weather events, including both droughts, floods, hurricanes, hard freezes and heat waves.

Ultimately, however, all of this scientific certainty about what is happening due to climate change and what lies in store is based on predictions involving statistical averages.  Just as knowing the exact statistical probability of tossing a certain number of ‘heads’ in 1,000 coin tosses does not help in knowing whether a specific toss will come up heads or tails – or just as a doctor who is certain that smoking greatly increases the risk of lung cancer cannot say with certainty if a particular case of lung cancer was directly caused by lung cancer or if it might have happened anyway — knowing that human induced climate change is leading to many more, and much more severe, weather events still does not mean that climate scientists will be able to say whether any given storm is “caused” by climate change or would have happened anyway.

So, when you begin reading or hearing the inevitable reports that a renowned climate scientist states that hurricane Florence was not caused by climate change, take it with a huge grain of salt.  If you can find the actual original words and place them in the context of a conversation, you will almost certainly find that what the scientist actually said was something like “there is no way to say if hurricane Florence is a direct result of climate change” or even “climate change models did not predict this hurricane.”  Those are both reasonable statements but neither means that there is no relation between hurricane Florence and climate change.  A more in-depth conversation with virtually any geophysical scientist on earth would lead to the broader, and more policy-relevant observation that conditions created by a changing climate, including warming ocean surface temperatures, increased evaporation, and shifts in air and water currents all play a role in making destructive weather events such as hurricane Florence much more common.

 

 

 

Advertisements

A Classic Example of How to Lie With Statistics

The March 3 Republican debate showcased a classic example of how to lie with statistics.  (Note: No partisanship here; it was all Republicans attacking Republicans.  No matter who says it, however, lying with statistics is still lying.)

Senator Rubio kicked off a marathon lesson in how to lie with statistics fairly early in the debate when he attacked Donald Trump by stating: “Two-thirds of the people who have cast a vote in a Republican primary or caucus have voted against you. They do not want you to be our nominee.”

What a classic!  Rubio’s statement starts with an accurate statistical fact and then spins it into deceitful story line that is not supported by the facts he cites.  Let’s look deeper.

The factual statistic at the heart of Marco Rubio’s incredibly dishonest statement is that while Donald Trump has been “winning” (to use the dominant media term) most primaries, he has not won a majority of the votes cast.  Instead, he won a plurality – the largest slice of the vote pie but still less than a majority of all votes cast.  In a multi-candidate field, a 30% to 40% share of the pie is frequently enough to be declared the winner.  Senator Rubio would have been entirely truthful if he had pointed out that Donald Trump only won a plurality of the vote, not a majority.  Senator Rubio even would have been 100% correct had he stated, for example, that “you have received well less than half the votes” or that “taken as a group, the candidates opposing you have won more votes than you in every primary held so far.”  That second version is a bit tortured, but still entirely accurate.

Rather than making statements entirely supported by the statistics he cited, however, Mr. Rubio chose to draw – and attempt to convince viewers to draw – a conclusion that is completely unsupported by the statistics he cited.  To reach his conclusion that “[t]wo thirds of the people who have cast a vote in a Republican primary or caucus have voted against you,” Rubio made the assumption that every person who voted for another candidate specifically voted against Donald Trump.  While it may seem somewhat logical that if you don’t vote “for” someone then you vote “against” him or her, that is not necessarily true in a multi-candidate race.  In a multi-candidate contest, a random voter may have had a very hard time deciding between Mr. Trump and one of the other candidates, but, in the end, wound up voting for the other candidate.  This hypothetical voter did not vote against Mr. Trump but for someone else.  If his or her favorite candidate later drops out, our hypothetical voter may well transfer his/her loyalty to Mr. Trump.

I may be stepping into a mine field here by moving from statistics and logic to religion, but perhaps Mr. Rubio hoped that his comments would remind religious Christian voters of Matthew 12:30, which states “Whoever is not with me is against me…”  Matthew was not referring to politics when we wrote that quote, and he certainly was not talking about a multi-candidate primary race!  Matthew’s observation might have some applicability to a U.S. general election in which only two candidates have a realistic chance of winning.  In early primary voting among a multitude of candidates, however, it is entirely likely that many voters have a 1st choice, 2nd choice, 3rd choice and so on and that those choices are only narrowly separated in their minds.  In such cases, a vote for one candidate is not a vote “against” the others but simply for the person at the top of a list.

Based on my own ‘gut feeling’ (which has no statistical significance) and a look at numerous polls attempting to measure not only which candidates Republican voters support but which candidates they most strongly oppose, there is no doubt that a number – probably even a big number – of Republican voters genuinely “oppose” Mr. Trump and would support any other candidate besides him.  Not one of those polls, however, supports Marco Rubio’s contention that every person, 100% of them, who favored another candidate in a primary or caucus chose that candidate because they oppose Mr. Trump.

Another way to highlight the faulty logic behind Senator Rubio’s claim would be to apply the same logic to Mr. Rubio’s own showing in the primaries and caucuses that have been held to date.  That same false logic that leads to the conclusion that two-thirds of Republicans “voted against” Donald Trump would also lead to the conclusion that over 85% “voted against” Senator Marco Rubio.  Ouch!

Unfortunately for him, the normally quick-witted Donald Trump seemed to fail to zero in on the source of the lie – the disconnect between statistical fact and utterly false assumption that every vote he failed to win as a vote against him – at the heart of Senator Rubio’s charges.  As a result, Trump’s initial rebuttal was so confusing as to be difficult to analyze.  Rather than focusing on the deception at the heart of Rubio’s charge, the exchange devolved into a meaningless exchange of random poll numbers about how each candidate might fare in a hypothetical match-up with Hillary Clinton.  Given how incredibly variable and inaccurate such polls are this far out from the general election, it is entirely unsurprising that each candidate participating in this pointless exchange (Rubio, Cruz, Kasich) was able to cite “polling data” purporting to show that he would do better against former Secretary Clinton in a general election.  I have not bothered to fact check the polls each cited since that type of poll means nothing at this point in the contest.

Much later in the debate, Senator Cruz essentially launched the same attack – based on the same unstated faulty and misleading assumptions that allowed Rubio to go from statistical facts to an outright lie.  By that time, however, Mr. Trump seemed to have zeroed in on the core fallacy at the heart of the argument and used it to turn the tables by noting that, by Mr. Cruz’s own logic, the fact that Cruz polled just 15% in some poll means that 85% of voters are absolutely opposed to him.  Typical for the entire debate, the ensuing exchange was a bit garbled as candidates yelled over each other, but in the melee it seemed clear to me, at least, that Mr. Trump had grasped the fundamental fallacy in the argument presented by both Senator Cruz and Senator Rubio.

Postscript:  Just to be clear:  this is not a political argument.  I am not trying to tell anyone who to support.  Instead, I just found it interesting how such a classic example of “how to lie with statistics” popped up in the March 3 Republican debate.  There were many other lies told that night, but I won’t go into those.  As someone who is a student of statistical reasoning and its abuses, however, I found that particular exchange in the debate highly amusing.