Most honest cities | Hacker News Reader

Most honest cities

rd.com

20 pointssnambi26 comments

Any statisticians want to chime in here? How many wallets should be dropped in each city before you can reasonably rank them? Or, is it statistically significant that Lisbon was 1/12, while Helsinki was 11/12? In other words, even with the small sample sizes, are those two values far enough apart to state the difference with 95% confidence?

On another note, that site sprang 20 separate trackers/beacons/etc. in Ghostery. I know at least one was Disqus, but 7 were different beacons, and the slideshow itself didn't work correctly without disabling it.

I guess the rate would also vary widely depending on which part of the city or time of the day you drop the wallet.

Also, quite possibly: day of month (contrast a week before payday vs just after), temporary local events and conditions (high unemployment one year vs the previous), and if the wallet contained a tourist's details: the level of 2nd language knowledge predominant in the country (e.g. the old lady in one of the first slides probably didn't speak English, meaning a barrier to contact or an opportunity for a second individual to pocket the wallet).

Overall though, if you randomised the drops nicely, increased sample size and repeated them a few times, it might be an interesting study.

I believe the main interest is not the statistical study itself, but rather the spontaneous explanations people give for returning a lost wallet.

A lot of people said they returned it because of the family photo in it. That seems to confirm something I heard years ago: that wallets with photos of children were more likely to be returned.

So if you don't have kids, put a photo of a niece or nephew, or even a complete stranger's kid in there.

assuming everything else were controlled for (and i have no idea how you do that) then you'd expect the "noise" in num-wallets-returned to be about sqrt(num-wallets-returned). in other words, in a city where the "real" rate is 4 returned, you expect to see that number varying by about 2 each time, so maybe 4, 2, 7, 3, 1...

so the results here really don't mean much.

Well, I'm not a statistician, but since nobody seems to be answering, I'll give it a shot.

Choosing the statistical method is slightly tricky; we have binary data (which is a bit different to the methods most people are used to). Guidance from measuringusability.com [1] tells us that our scenario is very similar to their "example 2", essentially where we have two groups with simple totals (1/12 and 11/12). Their advice is to use a N-1 two proportion test [2] for this.

Plugging in the values 1/12 and 11/12 here, we get a 99.994% probability that the distributions are different. This is because although the sample sizes are relatively small, 1/12 and 11/12 are very far apart. Slightly surprising, nevertheless.

A more reasonable comparison would be 5/12 vs 8/12, and in this case the same test yields a 77.108% probability that they are different (i.e. not particularly significant).

[1] http://www.measuringusability.com/blog/what-test.php [2] http://www.measuringusability.com/ab-calc.php

Sorry but this is almost a random test.

What neighborhoods? What days? What hours?

All that can make a difference and quite a big one.

Wallet is not a good testing device. It's a common sense in a lot of people that if you see a wallet lying around on the ground, you should never pick that up. If you do, there will be a bunch of thugs showing up from nowhere, claiming it's their wallet (it is) and there used to be 1000 dollars in that wallet (there's not). It's one of the oldest tricks.

Except in honest cities with few thugs...

I have never heard of this trick. It must not be popular in Washington, DC.

It would be great to see this methodology taken to its logical end. Decent sample sizes and some transparency in how drop sites were chosen, what the wallet contents were, etc.

The measurement of this kind of empathy seems intuitively valuable. I wonder what it would correlate with. Probably a higher average standard of living, but maybe more homogenous societal values? A smaller population?

Meh, article created to maximize clicks, not an objective study by any means. When you have to click for the next city, that should've been a giveaway.

The site has according to ghostery 66 trackers on it

that's insane

Except for London, the honest to dishonest ranking seems to equate to a prosperous to impoverished ranking.

Mumbai #2, Zurich #14... That seems out of place as well.

I disagree, there doesn't seem to be a clear relationship with affluence. Here's the list along with GDP PPP per capita at a country level [1]:

  Helsinki, Finland   11/12   35771
  Mumbai, India        9/12    3843
  Budapest, Hungary    8/12   19497
  New York, US         8/12   51704
  Moscow, Russia       7/12   17518
  Ambsterdam, NL       7/12   41527
  Berlin, DE           6/12   38666
  Ljubljana, Slovenia  6/12   27837
  London, UK           5/12   36569
  Warsaw, PL           5/12   20562
  Bucharest, Romania   4/12   12722
  Rio, Brazil          4/12   11747
  Zurich, Switzerland  4/12   44864
  Prague, CZ           3/12   27000 
  Madrid, Spain        2/12   30058
  Lisbon, Portugal     1/12   23047

n.b. How about 'least honest source of pageviews' for Readers Digest on this article?

1. International Monetary Fund, 2012 data.

I think you need to look at the City level, not country level - I know from experience that one city in that list is a poor dilapidated hell hole in a wealthy country - but you're right, London isn't the only outlier.

As an Indian both humbled and surprised to see Mumbai at top of the list.

Where's the complete list of cities they tried it in?

... or did they only try the 18 cities included in the slide show, almost all in Europe?

Yeah, just those 18 cities, 12 wallets in each, for a total of 192 wallets.

16 cities, 16*12=192. The first, 18th and 19th slides did not refer to any city.

I got to the bottom of the list thinking "damn it, rome is the most dishonest?", but it wasn't even in the list :)

As anecdotic and not rigorous this is, I fully expected Madrid (or any Spanish city) to be one of the last :(.

I'm surprised Zurich did so much worse than Moscow. I need to adjust my prejudices.

As someone from Lisbon, ouch. But it doesn't surprise me, there have always been plenty of pickpockets in the city, particularly in the more touristy areas. The cops have identified more than a 1000 in the last five years.

That said, as far as I know we fare much better on violent crimes, and in general I've rarely felt unsafe on the city, even walking alone at 3am and later.

Reply on news.ycombinator.com