The purpose of this essay is to examine what constitutes a useful explanation. I am using the recent presidential election as an illustration because it is so fresh in our memory, and so many commentaries have been written. The polls and the pundits predicted a Clinton victory, with a probability of 80% or higher at 8:00 pm EST when the voting ceased on the east coast. So there was a lot of explaining to do. 

I will be using the Causal Landscape as a means of sorting out the various reasons offered for the surprise outcome. The Causal Landscape has two phases. First, to provide perspective, it broadens the set of potential causes for an event and in the second phase it narrows to highlight the high-impact causes.  

Note: This essay is not a political statement. It is not intended to be pro-Clinton or pro-Trump and I have tried to be careful not to make any partisan statements. I simply want to use the election as an example of what we require in a causal explanation for an event — when we ask, “Why did that happen?”

Further, I am not attempting to catalog a comprehensive set of causes. I have collected a small number of causes for Clinton’s defeat. I expect that most readers will be able to identify causes that I don’t mention.

All I want to do is show that there were a number of reasons, rather than just one. This seems so obvious, but most of the accounts I have read zero in on a single cause for Clinton’s defeat, which is a tendency Robert Hoffman and I encountered in our research on what constitutes an acceptable causal explanation. People like to have everything boiled down to a single cause, which almost always is a great oversimplification.

Now, some simplification is necessary. We would find it too overwhelming to deal with every possible cause for an event. The tendency to simplify is natural and is necessary. But at some point, useful simplification tips over into misleading oversimplification, and that’s the problem I think we should try to avoid. In the case of the 2016 presidential election, boiling the explanation down into a single cause seems like just such an oversimplification. Presumably the pundits know better, and choose to oversimplify in order to convey a message more crisply, but they are doing a disservice to their readers, encouraging slogans rather than thoughtful analysis.

The Causal Landscape is a means of resisting the tendency to oversimplify. The first phase portrays a set of possible causes as a starting point, in order to get people out of the single-cause mindset. The diagram shows a Causal Landscape for explaining why Clinton lost the election.

Gary Klein
Source: Gary Klein

We don’t need to go into all the details shown in the diagram. Here are the primary causes I have culled from an unsystematic review of news commentaries. They are causes identified by experienced political analysts.

Systemic factors such as the 13 keys identified by Allan Lichtman for predicting presidential elections. These 13 keys include items such as;

—Having a party mandate. Did the incumbent party gain seats in the last mid-term election?

—Is there a serious contest for the incumbent party nomination?

—Is the incumbent party candidate the sitting president?

—Has there been any major policy change (did the incumbent administration make major changes in national policy)?

—Social unrest.

—Foreign/military failure.

—Is the incumbent party candidate charismatic?

—Is the challenging party candidate charismatic?

Based on the full set of 13 keys, Lichtman bravely predicted a Trump victory in September 2016, relying on his model rather than on the prevailing wisdom.  

Personal issues, especially the perception of untrustworthiness based on the classified emails in Clinton’s private account and the concerns about influence-peddling using the Clinton Foundation, plus the secrecy surrounding Clinton’s televised collapse, later attributed to  pneumonia.

Gender issues/sexism that made some voters unwilling to choose a woman.

One-off events such as the FBI Director Comey’s letter just days before the election, or Clinton’s public comment that half of the Trump supporters belonged in a basket of deplorables.

The Trump effect — the appeal he exerted, including the media fascination with him that allowed him to dominate the news coverage despite the Clinton campaign’s much more effective fundraising.  

A lower-than-expected turnout from groups such that had been expected to be strong Clinton supporters: African Americans and Hispanics, and a less-than-expected backing from women.

Strategic decisions the Clinton campaign made, such as, in retrospect, an over-reliance on polls that led the campaign to take Michigan, Pennsylvania and Wisconsin for granted; energy spent by Clinton in fund-raising instead of campaigning; a failure to present a compelling vision, relying instead on reminding voters of Trump’s negatives; a more leftist agenda adopted during the nomination process in order to fend off Bernie Sanders in the Democratic primaries; a heavy reliance on identity politics that excluded lower-class white voters and painted many of them as a basket of deplorables.

Some of these factors overlap. For example, I have listed the “basket of deplorables” issue as a one-off event, the comment itself, and as a strategic decision to distance the campaign from certain groups of voters.

As stated above, this set of causes is not comprehensive, but it does show why any single explanation is inadequate. There were lots of reasons why the election swung in Trump’s favor.

And, re-iterating the point I made at the beginning, I do not intend this list of causal factors as a criticism of Hillary Clinton, the candidate for president. The list is intended to explore the reasons why Clinton lost an election that seemed within her grasp.

One of the problems with this type of analysis is that it can make Clinton’s defeat seem inevitable, and it wasn’t. This election was not a blowout on the order of Nixon vs. McGovern or Reagan vs. Mondale. At 7:59 pm Tuesday night, November 8, 2016 just before the election results started to be tabulated, we could have even more easily have prepared a Causal Landscape of why Trump lost, which was so widely expected.

Another problem with Causal Landscape diagrams is that they include so many causes. If the Clinton campaign team wanted to imagine what it could have changed, the diagram can be overwhelming. Now we get to the second phase of the Causal Landscape: focusing on the highest payoff actions. In this phase, we ask two questions: which of the causes shown in the diagram would have been the easiest to reverse, and which of the causes, if reversed, would have had the greatest impact?  We make both of these estimates for each node in the diagram. Then we highlight the causes that were easiest to reverse and whose reversal might have tipped the outcome in favor of Clinton.

The next diagram shows my own judgments. (And I am not claiming any political sophistication — I am only illustrating how the process works.) I am suggesting four entries: Clinton’s comment about the basket of deplorables and some of the Clinton campaign’s strategic decisions. 

Gary Klein
Source: Gary Klein

The other influences, such as the systemic factors identified by Lichtman, perceptions of Clinton as untrustworthy, Clinton’s gender, Trump’s approach, all seem difficult if not impossible to alter.

Hillary Clinton has publicly blamed the Comey letter for stalling her momentum and costing her the election. But Diane Hessan, who served the Clinton campaign by tracking 250 undecideds in swing states in the final months of the campaign,  found that Clinton’s remark about the “basket of deplorables” had a much stronger impact than either of the Comey statements. 

In conclusion, all explanations are simplifications, but single-cause explanations seem to be oversimplifications that are unnecessarily misleading. The Causal Landscape seeks to portray a wide variety of causal factors, running the risk of presenting too much complexity. The next phase is to focus on the actionable explanations, running the causes through the filters of ease of reversal and likely impact. The goal is to establish a reasonable level of simplification. By highlighting these high-payoff entries within a field of influences, we can keep things in perspective.

If we can’t be smarter in hindsight, it reflects badly on us. When political commentators resort to single-cause explanations, they are not doing us a service. Hopefully we can use the election post-mortems to learn how to simplify without oversimplifying.

You are reading

Seeing What Others Don't

Retiring the Dreyfus Five-Stage Model of Expertise

It has served its purpose—time to move on.

The Age of Centaurs

Instead of building smarter machines, let's build machines that make us smarter.

The War on Experts

Five professional communities are trying to discredit expertise.