I’m not sure who said this at the bottom of my last comment in this thread, but whoever it was just made me spit my drink out.

I’m talking about the OECD graph on wikipedia which Steve linked to on this blog, and was claimed to show that the corporate tax rates in New Zealand were low – when the chart showed anything but. Even Rob agreed with me.. but I’ve yet to see Steve admit he was wrong.

[lprent: Dean is correct. Please label yourself when you insert into a comment. Otherwise it gets too confusing.]

If that isn’t a sign of error, I don’t know what is. A rough calculation in a spreadsheet puts the 1991 growth rate about 4 times out of standard deviation from the mean!

With respect to the broad theme, you’re arguing a tautology. Regieme change like we had in the 80′s is ALWAYS going to cause disruption and unemployment. That’s not conjecture on my part, it’s a fairly obvious statement based on simply looking at history. Whether or not we needed the change is a whole separate issue.

The problem for you is that the gap you have described narrows under both parties. It is not suddenly, as if by magic, fixed under Labour

And none of this changes the full story – there was a decrease in employment at the start of the neoliberal era, and then very weak employment growth from 1993. That saw the productive potential of around 2 million person/years wasted.

As ever, if you have a problem with my analysis you’re wlecome to make your own graph.

The size of that break is 120,000 people, give or take a few hundered, which accounts for roughly half of the purple block in that year. I suspect that if the series break is treated in a mathematically correct fashion, your conclusions will be untenable

The Census is every five years, not four.

You are so right. I got confused because I was looking at some data from the 2001, which was labelled as 2002 because that was when it was published, and remembered that the last census was 2006. Ok – actually it is because I was short of coffee at the time.

2006, 2001, 1996, and 1991.. Maybe.

“when something looks interesting or unusual, it’s probably an error’

Yes, but I don’t have access to the graph data so I made a presumption of accuracy on SP’s side.

The next most common problems with any data series are either real data injected into estimated data, or a change in how something is measured. I see these kinds of error all of the time in IT.

vto. no, you’re a towel

Just the same as your stupid thread below re interest rates under the different govts.

“Let me see . 2006 2002 1998 1994 1990. That would fit. Typically the data shifts in the year following the census”

The Census is every five years, not four.

There is no other ‘burp’ in the data shown here – everything else follows a reasonably understandable trend (with the exception maybe of ’83).

I was always under the impression that the first rule of statistics is thus; “when something looks interesting or unusual, it’s probably an error”

This site is more about making comment than being a reference site. It definitely isn’t a academic site. Making assertions should be backed up with data. That is simply to reduce comment that comes from personal opinion igniting flamewars or for perpetuating popular myths.

Look at the bottom of the jpeg Steve put up, the sources are mentioned. Yes – a direct link would be nice, but not critical. The major links usually are in reference sites on the blogroll on the right. I’ll add the rbnz.

IMHO: Argument about what is applicable data is preferable to argument about urban myths. This shows up strongly in this post and the previous one. I also enjoy the effective use of humour, satire and sarcasm in comments. It makes them easier to read when I scan (I tend to laugh and become more tolerant of transgressions).

what I assume is a massive population influx in 1991?

Census in 1990? There are often lumps in statistical data when the census comes through. Sometimes the estimates are off, and sometimes the measurement changes. I’ve seen it a number of times.

Let me see…. 2006 2002 1998 1994 1990. That would fit. Typically the data shifts in the year following the census