Tuesday, 29 October 2013

More than 15%

GST underexplains the price difference between New Zealand online retailers and foreign competitors. A notable current example.

Eleanor Catton is the NZ author who just won the Booker Prize. The Luminaries was published by Victoria University Press. They're selling the paperback online for $35. It's out of stock and they promise to have more in two weeks. They don't say anything about shipping charges; I'm not sure if you can buy directly from them. They link to meBooks, a new Zealand e-books supplier. They're selling the electronic version for NZD $19.95.

NZ online retailer Fishpond has it for $33.72, including delivery, but shipped from a UK supplier rated as low reliability (by Fishpond) and subject to delays. MightyApe has it for $29.99, but also on backorder, and with shipping that brings it up to $33.89.

Without GST, Fishpond would be selling the book, including delivery, for $29.32. meBooks would have the electronic edition for $17.35 if GST didn't exist.

Book Depository has it, delivered to NZ, for NZD$22.65. Their website doesn't suck, it's easy to find what you want, and they're reliable. Amazon has it for USD$14 (NZD$16.87), but I never shop there as figuring out the international shipping charges is always a hassle. Amazon has a Kindle edition for USD$10.65, or NZD$12.83.

So a New Zealand book from a New Zealand publisher, in a world where GST didn't exist, would still be $6.67 cheaper sent to me from Book Depository: the Book Depository price is 77% of the ex-GST Fishpond price. GST makes up 40% of the total price difference. The rest is the standard "bought it in NZ penalty". And the e-book is $4.52 cheaper from Amazon than the ex-GST meBook price: GST is 37% of the total price difference. The rest is not GST.

For the marginal customer, the extra bit of price discount drives the decision to shop online from abroad. But I'd be awfully surprised if any substantial portion of online buyers are marginal. And especially not when it's typically easier to source things from international retailers than to find a Kiwi one online.*

Lance Wiggs suggested in comments at NBR levying the tax but placing the onus on taxpayers to comply. This is less bad than other options, but it still imposes a large hassle cost on online customers who want to be honest.

Owen Matthews noted that imposing the burden on foreign retailers risks their refusing to ship to New Zealand at all. This shouldn't be discounted. When I lived in Canada, a lot of American retailers simply refused to ship to me because hassles at the border for customers turned into headaches for them. If you wanted to bring stuff across and found a retailer who'd sell to Canada, couriers would impose something like a $30 brokerage fee at the border to get it across. This is likely what NZ retailers want to have happen here as well.

NZ retailers are pushing hard because they know that a lot of them just shouldn't exist. If you can have stuff shipped here, from abroad, including shipping, for much less than the NZ retail cost, then that part of the retail industry should shut down and have its inputs flow to higher valued uses. A small retail section would remain for those products for those who really need things in a hurry. Otherwise, warehousing and distribution costs in small markets mean that some local retail should simply cease to be. Expect much lobbying for protection in the interim.

* Exception: beer and wine. The Beer Cellar is great. So is Regional Wines.


  1. Wow based on a sample of 154! Only 14 from one event.

  2. Yeah. I wouldn't draw strong conclusions of any kind from this one.

  3. Do you have enough data yet to do a post on the evaluation of WASP? I am curious to see how it performs.

  4. There is probably not enough data to use from games where Sky has used the WASP. We could simply apply it to all games that have been played since WASP first appeared about a year ago, but it is an open question what we would be looking for. This is because:
    1. WASP requires a judgement call input into what the par score would be; so any evaluation is a joint assessment of the model and the par score. I think the Sky team used past history at the particular ground to determine a par score, but they didn't talk about that. The only way to tell what they chose is to wait for the WASP worm to appear in the first innings and see what score it started from.
    2. WASP is not a predictor in the TAB sense. It is a measure of how the average team would have done against the average team on that pitch, to get a sense of "who is winning", in the same way that Ireland leading the All Blacks 10-9 at half-time means that Ireland is winning but is not the team you would predict to win. This raises the question, what is a good performance. Models that took into account who was playing to predict the first innings score or the result would likely (should) out-predict WASP, but they would be asking different questions.
    3. If not comparing to an alternative predictor, what would be the measure of performance. In the first innings, WASP will be very accurate (of course) with only a few balls to go, and as reliable as the par score before the match starts. So I might be inclined to go with WASP's prediction at the 25 over mark, and see what the distribution of the difference between the prediction and the actual final score and see if the error averages to something statistically indistinguishable from zero. If it does, how small a variance would we need to say it is a good model? In the second innings, I would like to again look at the 25 over mark, and do a probit regression of winning (1 or 0) on the WASP, and see how close the fit is to a 45-degree line. But this will require a large number of games.

  5. I love that they baked "very" into the question for the moral-panic one. How do you even navigate that? Can you "strongly disagree" with the "very" to indicate that you just slightly agree with the rest of the statement?

  6. Seamus, I think we could keep a whole class of your students occupied with questions such as Peter's.

    One way of addressing the point you make about valuing ODI performances could be to use the ICC player ratings, which I think do take some account of run rates. I wouldn't know, however, how to base a player's first class average against their ICC ODI rating.

    I suspect Peter's query is the 'George Bailey conundrum'. If so, sample selection would need to be precise - i.e. Players who have enjoyed a decent run in ODIs before playing test cricket.

    I'll let you know if I devise something to cast some light on this.

    Chris (Declaration Game)