Speaker: Identification strategy: Now it’s personal
184 Responses
First ←Older Page 1 2 3 4 5 … 8 Newer→ Last
-
BenWilson, in reply to
Yeah, I know. Even bothering with null hypotheses is a bit of an add-on for Bayesian analysis, which doesn't really need them. Statistical significance is not even an idea that makes sense under the Bayesian approach. They don't talk about confidence in that way.
-
Oh yeah, and here's the last chunk of my submitted post that Russell probably cut for brevity, so that everyone could start doing the Frequentist vs Bayesian debate (personally, I LOVE that I've kicked off a thread that has descended into a Frequentist vs Bayesian debate):
Man, for Chinese immigrants in the West, buying houses is almost on a par culturally with food. It’s like you’re giving us shit for eating.
With Phil Twyford lying in a bloody puddle somewhere, my stats rage is somewhat cooler. But my old friend race rage is creeping back. It doesn’t matter if 90% of those ‘Chinese sounding names’ really are PRC hot-money investors; for the rest of us under suspicion of being foreign, we know Labour was ready to throw us under the bus.
The Labour study is one of those bits of analysis that comes together because no other data is available to answer the question you really want answered; and is convenient simply because – and this is very important – Chinese people are identifiable.
If your identification strategy for locating the cause of the housing bubble boils down to ‘What minority group can we effectively single out because of their weird names?’, this is NOT SOMETHING YOU SHOULD EVER ADMIT. Not only does it make the Labour Party sound inherently racist and not to be trusted, it makes social science sound inherently racist and not to be trusted. In a perfect world, I’d be able to love them both unconditionally, and stats rage and race rage would not be a problem.
I can’t actually remember the last time that this much effort has been put by any political party into singling out Chinese people (as opposed to ‘Asians’) for their probability of being ‘foreign’.
And if Labour put this much effort into programming an algorithm to identify us, I wonder if it also estimated how many New Zealand Chinese votes this study would cost them.
The real question is not ‘oh, ha ha this doesn’t look like a very reliable method for guessing whether someone is Chinese.’ (It probably is). It is not ‘could local Chinese possibly be buying all of these houses?’ (It is possible, but not very likely).
The real question is what did the Labour Party think it was doing taking this public. If they just fucked up, so far so familiar. If they did this on purpose for well-calculated reasons – and it works – we Chinese-sounding named people are in way more trouble in New Zealand than we ever thought we would be again.
-
Russell Brown, in reply to
Oh yeah, and here’s the last chunk of my submitted post that Russell probably cut for brevity,
Argh! I thought I’d selected all the text except for the footnote, which wouldn’t select, so I just pasted in what selected + footnote.
I’ll fix that now. Sorry!
-
BenWilson, in reply to
If they did this on purpose for well-calculated reasons – and it works – we Chinese-sounding named people are in way more trouble in New Zealand than we ever thought we would be again.
The most chilling part is that it might work. But I do have to ask how else the question of massive foreign investment in NZ can be approached statistically, when the data isn’t collected at all. I’d prefer it if we could get a real handle on it, but it doesn’t look like something we could really get from any data sources available. So the result “there’s something quite anomalous going on with ethnic Chinese investment in NZ” could literally be all we know beyond frustrated buyer anecdotes and the opinions of real estate agents. This against the endless denials issued by National that there is anything real to see here at all.
But yeah, not being ethnic Chinese, it’s easy for me to take this in a whole abstract mathematical way, rather than being personally worried that Big Labour is looking at me, and not at all worried about the ten thousand foreign Wilsons that probably own property in NZ.
personally, I LOVE that I’ve kicked off a thread that has descended into a Frequentist vs Bayesian debate)
Sorry about that. It really was quite irrelevant. Decent statisticians can use whatever tools are good for the job. I guess Bayesian just has a bit of an uphill battle, being the newer kid on the block, at least in the mainstream.
ETA: BTW, it may be true that Chinese love property ownership. But more than, say, Pakeha in NZ, for whom it's practically their only life goal beyond having children? I find that hard to believe.
-
Russell Brown, in reply to
If they did this on purpose for well-calculated reasons – and it works – we Chinese-sounding named people are in way more trouble in New Zealand than we ever thought we would be again.
The most chilling part is that it might work
The distant panicky noises I hear suggest to me they're not as well prepared as they would be if they'd planned it this way.
-
Tze Ming Mok, in reply to
After some discussion with my house-loving dad in Roskill, I think that as long as the National Party stonewalls on bringing in data collection on foreign ownership (to avoid political problems with the PRC); the Labour Party will go to even more ridiculous lengths to try to put a number on foreign PRCs owning houses – and the local Chinese population is going to always get stuck in the middle of their flawed calculations.
I totally admit that there *might* be no difference in cultural behaviour and propensity towards house-buying among different ethnic groups; but no-one has bothered to find out if this is the case. If we want to figure out the impact of foreign PRC buyers - or more to the point, foreign anyone buyers - on the housing market without the straight up government data, you should probably just do a large-scale sample survey on house buying with a lot of ethnic minority oversampling. Then you could actually have a baseline from which to say that ‘local Chinese buy X%, local Indians buy Y%, local white people buy etc etc etc’ (amazing, ASKING PEOPLE whether they bought a house recently, revolutionary approach I know), and THEN I suppose you could compare the percentages from your slightly dubious ‘name provenance’ analysis, (BUT this still can’t accurately show the extent to which we are talking about PRC buyers, but in this case it would be ‘likely’ Chinese buyers from countries other than China, such as Taiwan, Singapore and, horror of horrors, white countries too, who would be tarred with the ‘crazy PRC money’ brush). Costs money to do a survey though. But then, how much this just cost the Labour party in Chinese donors? Like I said, I suppose they did the numbers.
-
Tze Ming Mok, in reply to
The distant panicky noises I hear suggest to me they’re not as well prepared as they would be if they’d planned it this way.
Which is worse, conspiracy or incompetence? I mean, in my post, I concluded the former. The latter still is pretty painful though.
-
Tze Ming Mok, in reply to
Though of course if you did a reliable sample survey, then still had to compare it against a real estate register based only on last names, there would still be serious problem with how racist it is. Because it’s impossible in such an approach to differentiate white, Maori, or any number of other ethnic groups formerly colonised by white people, from white people in data based on last names, meaning that the only defined groups that can be reliably identified by last name are Chinese people and Muslims (not an ethnic group), which means that it’s unavoidably racist to do it at all.
-
Kumara Republic, in reply to
The most chilling part is that it might work
The distant panicky noises I hear suggest to me they’re not as well prepared as they would be if they’d planned it this way.
So in other words, it might not be going horribly wrong, but it might actually go horribly right.
That it’s likely strategic implies that Labour has given up on getting the ‘missing million’ to the polls, and instead is doing what could be its Orewa Rotary moment in order to win back voters who might have decamped to NACT or NZF.
It’d be unfortunate if Labour has to resort to pandering to base prejudices, in order to stay competitive with the NACT bloc. Just like Aussie Labor?
-
BenWilson, in reply to
Yes, an expensive survey lies between racial data mining and official data collection in reliability. The issue of how accurate you could really expect self-reporting to a telephone survey to be of something as sensitive as where you sourced your mortgage from arises. And surely there are other ways besides - if the purpose is really just to understand the level of foreign mortgages in total, something I'm much more interested in than the amount of those that are Chinese, then what David Hood said on Rob Salmond's thread said makes sense. It's kind of interesting how that breaks down in terms of country of origin, but I don't see much point in making policy out of it.
My only actual point of objection on a piecemeal basis to foreign mortgages is to nations that don't reciprocate. There is something quite shitty about the fact that people from various nations (and China is not alone in this) can buy property here, but I can't buy property there. I can remember feeling quite gutted when I looked through a huge bloody mansion in Thailand that would easily be worth 2 million here that was going for about 300K, when they pointed out that I personally couldn't actually buy it. On reflection I thought that was probably not a bad thing for the Thais to do, because their property would rapidly become unattainable to locals if rich bastards with Ozzie paychecks like I was at that point could just come in and laugh at paying twice the price.
So with the boot on the other foot, I don't see it as entirely objectionable to think controlling foreign ownership of NZ more tightly is probably a good thing for the poorer NZers. But it shouldn't be piecemeal, certainly not racially profiled. Reciprocity or not, foreign money is far too big a pool to just allow unrestrained access. There does need to be a debate about tightening it, as one of the probable drivers of massive property inflation.
-
I won’t be describing that process further, as that is sensitive IP for Labour.
Perhaps I am old fashioned in this respect but I don't hold researchers who do this in high regard. When it comes to politics especially on contetious issues with racial implications, this is writing a credibility cheque (trust us we know what we are doing coz we have teh statz). My point being, if you don't want to show it all then keep it to yourself, there can be no middle ground here.
-
giovanni tiso, in reply to
Just like Aussie Labor?
Aussie Labor is voting in favour of the ban on reporting cases of abuse of imprisoned asylum seekers, including children. New Zealand Labour is not quite that monstrous at present.
-
BenWilson, in reply to
My point being, if you don’t want to show it all then keep it to yourself, there can be no middle ground here.
Private researchers often have no choice, though. You can't freely give away IP that belongs to other people. It's not on Rob to let the method out of the bag, it's on Labour.
-
simon g, in reply to
So in other words, it might not be going horribly wrong, but it might actually go horribly right.
It may be. But it can only go horribly right if it is maintained. "Orewa" is shorthand, not just for a single speech, but for an election strategy that National kept pursuing until polling day 2005 - almost 2 years.
That's why the counter-argument can't work. If it's simply legitimate research by Labour, misrepresented by media and opponents and gullible Public Address commenters, then it will soon fade away. Remembered by too many who felt like targets and were repelled, but not by those swing voters who were targets for attracting to Labour. To achieve and sustain any short-term poll boost, Labour will have to keep this front and centre.
If they just dip toe into water and then retreat, they've doubly failed (antagonized some, but not converted others). Sadly, as of now, that's the best we can hope for. The alternative - that this is only the start - is really nasty.
-
81stcolumn, in reply to
Ben - The problem remains that the method isn't public/transparent, which may satisfy Rob and it's fine for him to draw inferences as he chooses. But when you go public with stuff like this its not fair for anyone to have to take his word.
-
BenWilson, in reply to
But when you go public with stuff like this its not fair for anyone to have to take his word.
It might not be "fair", but that might be his deal with his employer. Presumably you've occasionally done research that you weren't allowed to just release to the world without someone else's say so before?
-
Okay while we are here:
I like the idea of including priors it makes a lot of sense. But I think Bayesian analysis has to do its time on the wheel of epistemological rigour.
i) The scope and impact of assumptions made in Bayesian analysis still need some testing, for example history effects on sampled ratios.
ii) The tractability of the frequentist approach is a grand asset which is not trivial.
It may take some time to develop a way of reporting and auditing assumptions made in Bayesian analysis that supersedes the evolution of the frequentist approach.
@Tze Ming Mok - Cool poeple are still doing quants in our wonderfully post positivist world. Sure do miss you.
-
Tze Ming Mok, in reply to
I repeat: Hahahaha I instigated a Bayesians vs Frequentists comments battle!
-
TracyMac, in reply to
estimating Asian ethnicities [methodology] is sensitive IP for Labour.
I'm embarrassed to be a Pakeha right now. But relieved I haven't voted for this lot (nor any party that would have Brash, either).
-
@Tze Ming Mok I haz a curious.
If I posit that Chinese buyers will cluster round particular real estate agencies based on previous personal experience, referral by a friend or a willingness to deal in Mandarin (or Cantonese), would I be accurate? And that some agencies will have lots of Chinese customers and others will have fewer?
Because it seems to me this was a leak of a subset of one agency's customer list. For Labour to claim this has relevance against the total Chinese resident population of Auckland, wouldn't they need a much wider data set?
-
nzlemming, in reply to
I’m embarrassed to be a Pakeha right now. But relieved I haven’t voted for this lot (nor any party that would have Brash, either).
Damn straight, and I've let Twyford know that I'm personally disappointed in him as a person as well as Labour as a party.
-
Presumably you've occasionally done research that you weren't allowed to just release to the world without someone else's say so before?
Yes indeed and I would see anonymity and redaction an important part of the process, especially with respect to qualitative data. The goal here is to act ethically and where possible, avoid harm to those who provide data. At the end of this I still have to decide on my own terms whether the remaining data is a fair reflection of the processes and question(s) involved. But that is not the same as obscuring methodology.
The problem with IP on analytical procedures is that you cannot gain any sense of the robustness of that approach, consequently the worth of what you are seeing. Moreover it removes an important point of agreement regarding truthfulness of the conclusions drawn.
nb. Stats programmes use IP to process data but the method is open and testable. I can test the tool using standard data sets and know what should come back.
I guess because this is "sensitive IP for Labour" the credibility issue really sits with the Labour party (ahem - a group of politicians). Consequently my ire may have been mis-directed - but Rob did publish this in a post defending the method used.
Consequently, for a political party that seeks an electoral mandate to obscure statistical method in pursuit of political goals......well my comments would still very much apply. Indeed I'd go further, the findings as they stand are not really credible, which leaves the inflammatory approach. Labour, Phil and others should not have released this analysis.
-
BenWilson, in reply to
must....not....comment....
Funny thing is, JC* was probably right all along - is it really Bayesian analysis, when you're just working out proportions? We didn't get any juicy details like what distributions he set for his priors just before MCMCing it all.
*Jim Cathcart that is, not the more famous JC.
-
Dear Russell, when the hell are we going to get an NZ English spell checker on PAS??? I’m tired of the squiggly red lines every time I time I type “Labour”
[edit] UK English would be fine #justsaying
-
81st promises to put the ghost of Ronald Fisher to sleep...
Maybe over one of those missed cups of coffee eh?
Post your response…
This topic is closed.