RAW DATA: Raw riposte to Auckland real estate data defence

Freelance data journalist Keith Ng has called out an opinion piece by political analyst Rob Salmond on Labour’s use of leaked Auckland real estate data as “incorrect and defamatory.”

Last week Mr Ng took issue with the data itself and declared the way in which it was released “cynical, reckless dogwhistling. Like Winston Peters, just without the smirk.”

In the Sunday Star Times column, Dr Salmond – who counts the Labour Party among his clients and did the data work that informed the political party’s interpretation of the three months of Barfoot & Thompson sale information – defended the rigour of his analysis and the intentions of Labour in releasing the information in that form.

He also wrote that “Keith Ng, Tze Ming Mok, and Chuan-Zheng Lee – all skilled analysts, all otherwise critical on this topic – agreed the name-based ethnicity analysis was statistically sound, robust, and accurate.”

Mr Ng begs to differ, however, and has done so in a strongly worded post on the Public Address blog site to which he is a regular contributor. That post is reproduced below with permission as ‘raw data’.

Mr Salmond is also a Public Address contributor and some of his work – including a piece last week about the housing data – has previously been cross-posted on NBR ONLINE.

-----

RAW DATA: Keith Ng's blog post Don't put words in our mouths, Rob

In an opinion piece in the SST, Rob Salmond said:

After I published Labour’s method online, Keith Ng, Tze Ming Mok, and Chuan-Zheng Lee – all skilled analysts, all otherwise critical on this topic -all agreed the name-based ethnicity analysis was statistically sound, robust, and accurate.

Hey Rob, don’t put the words “statistically sound, robust, and accurate” into our mouths to describe your work.

You asked whether I was willing to give a ”more-or-less endorsement”. I said the method was sound but the data wasn’t, and we left it there. I would note that on the original blog I said of the data problems:

these problems only render it unscientific and utterly useless, which is the least of Labour’s problems.

If you need clarification, let me restate it: The method is fine, the data is broken, and those problems render it unscientific and utterly useless. Not sound. Not robust. Not accurate.

Rob also said:

Having said that, one group I think did not overreact – despite their strongly critical stance – was the New Zealand Chinese community, including Keith, Tze Ming, and Chuan-Zheng. Their criticism was less about Labour’s intentions, and more about the impact of these revelations on ethnically Chinese New Zealanders.

Thanks for the flattery, but I was very critical of Labour intentions and I thought I was bloody clear about it.

I said that Phil Twyford was knowingly “straight-up scapegoating” Chinese New Zealanders and offshore Chinese alike and “fueling racial division in this country”. I said it was “cynical, reckless dogwhistling”.

What part of this was ambiguous for you??? Did you think I meant “cynical, reckless, but ultimately well-intentioned dogwhistling”?

But since we’re talking about intentions, Rob said:

Labour’s intention was always to talk about offshore money, and never to conflate ethnicity with nationality, or to make life more fraught for any group of New Zealanders.

Except that the ethnicity analysis (the Chinese-sounding names part) was their “main data analysis”, and the “analysis” of offshore buyers was secondary to this. Here’s what Rob said about the two in our exchange on Twitter (step 1 = Chinese-sounding names):

I presented *step 1* as rigorous, against folk saying I’d sat in the dark asking “who sounds Chinese?”

..then (step 2 = offshore buyers, emphasis added):

I only ever presented step 2 as informed speculation stemming from the rigourous step 1 work.

That is, they claim their intention was to talk about offshoreness, but what they knew about offshoreness only came from “informed speculation” secondary to the main analysis about ethnicity.

And what did Rob concluded from this “informed speculation”?

My conclusion: if my prior for “is there large-scale offshore $?” were X, my posterior post these data is >X

It’s a wanky way of saying: After seeing the Chinese-sounding names evidence, he is more confident that “there is large-scale offshore Chinese buying in Auckland” than he was before. How confident was he before? And how much more confident has he become?

No, I won’t quantify it, because that would be introducing false precision to qualitative reasoning. 

But here’s the problem. He is literally saying his level of certainty is unknown + unknown. Which equals, of course: unknown.

This is the statistical basis on which Twyford is out there using words like “implausible” and “very unlikely”.

In case you still think this is well-intentioned, or that unknown might actually be meaningful in any rigourous sense: 

Rob: But directional evidence has always been par for the course in political debates, as u know from media work.

Me: That’s bullshit Rob. You’re dropping the standard of proof to >0. i.e. Literally infinitesimal. 

Rob: What, you think directional evidence is *not* entirely standard in political debate. Come on.

What Rob means by “directional evidence” is to judge evidence by its direction only – as opposed to its direction and its strength. That is to say, Rob believe it’s okay to use evidence which supports a claim in political debate, explicitly regardless of how weak it is. According to Rob, any shred of evidence is okay in a political debate, because that’s how political debates work.

Please do not mistake me for thinking that this is well-intention. This is a cynical attempt to bamboozle the media and the public by hiding your utter lack of evidence behind fancy jargon. It’s a travesty and a sad excuse for analysis. You ought to be ashamed, Rob.

Also, Sunday-Star Times: These claims Rob made about me are incorrect and defamatory. Please issue an correction in your next issue.

Also also: I’ve set up a permanent tip-jar over on Givealittle. If you feel like giving me a tip, feel free to do so.

• Keith Ng runs a data visualisation company and freelances as a data journalist. This post originally appeared on Public Address.