Trusting Government Data
|

Evaluating Data: a TED talk

Trusting Government Data

Lately, there's been a lot of talk lately about fake news and conspiracy theories. People are believing some weird stuff such as the child porn ring at a Washington pizzeria that supposedly included Clinton folks. Then there are folks who don't believe in the government data such as the unemployment rate. I get that they may not believe in that number because a) somebody keeps repeating that those numbers are fake, b) where they live, they may know a lot of people who are unemployed or c) their living prospects are declining and that gets translated into "job prospects are bad" but they don't catch that declining living standard is not the same as unemployment.

I've been wondering, how does one spot fake news? How does one evaluate information being given to you?

Right now I'm reading 2 books with a third waiting in the wings. One of them is Thank You for Being Late by Thomas Friedman and the other Mindware by Richard Nisbett. I'm not done with either but I think I will start talking about them just to get the conversation going. The second book is about tools for better thinking.

But before I get into that book, I wanted to post the TED talk on how to look at data or presentation of data as a prelude. The speaker specifically spoke about government data and how there is a plan to do away with government data. That's scary. We need unbiased data so we can analyze them for ourselves. I can't believe these "business" types want to do away with them.

The speaker also spoke about polls, you know, the ones that said Clinton would win. There's a huge distrust in polls because first, they mislead the Republicans who thought Mitt Romney would win and then they mislead the Democrats when they said that there was no way that Trump could win. Polls are tricky in that you have to get the right mix of people that are representative of those who will be going to the polls and you have to ask the right kind of questions. If you phrase the question differently, you can get a totally different response.

Now, contrast polls to anecdotal stories. There's been a lot of anecdotal stories about Trump supporters changing their minds. But, these are anecdotal stories, not a gathering of data of how many people may have changed their minds. While the Democrats may be salivating that the Trump voters could be having buyer's remorse, there's just not enough there to say there is a mass movement away from Trump. I can't rely on these anecdotal stories, no matter how interesting or emotionally persuasive. For all I know, those stories could represent just onesies and twosies.

On the other hand, last week Washington Post had an article that said only 3% of the Trump voters regret their vote. To me, that seems plausible because Trump hasn't done enough yet to do things that would hurt his supporters. The healthcare repeal and replace might have changed the dynamics if the new act had passed because that act would have likely hurt them. Only until his actions start hurting his supporters will they then change their minds. It's like George Bush: people voted him in again despite being tired of the war and no evidence of weapons of mass destruction having been found, but Katrina changed the dynamics because that incident caused real pain.

The Washington Post disclosed their methodology and it seems pretty solid. I'm no expert in the weighing procedure and the posing of the questions so I would have to study this further and more deeply, but on the surface, it looks like the approach is as solid as it can get. The weighing would be a possible area of weakness because I think you have to apply some judgement on the kind of weights. The questions seem pretty good though, but again, I'm not an expert.

All in all, the Washington Post's poll seem to me more plausible than the anecdotal stories. It is stronger because it is based on data and not wishful thinking.

Look at the TED talk: the speaker has some useful tidbits on evaluating data, charts or statistics.

Similar Posts