9

A report from The National has been widely cited on the Internet recently:

A dramatic fall in traffic accidents this week has been directly linked to the three-day disruption in BlackBerry services.

In Dubai, traffic accidents fell 20 per cent from average rates on the days BlackBerry users were unable to use its messaging service. In Abu Dhabi, the number of accidents this week fell 40 per cent and there were no fatal accidents.

On average there is a traffic accident every three minutes in Dubai, while in Abu Dhabi there is a fatal accident every two days.

Lt Gen Dahi Khalfan Tamim, the chief of Dubai Police, and Brig Gen Hussein Al Harethi, the director of the Abu Dhabi Police traffic department, linked the drop in accidents to the disruption of BlackBerry services between Tuesday and Thursday.

[...]

The precise statistics for traffic accidents in the two emirates this week were not revealed to The National.

A quick examination will reveal that no fatal accidents in three days, when there is an average of "only" one fatal accident every two days is not statistically significant. (I used a back-of-the-envelope calculation to get a 22% chance of this happening if the Blackberry outage had no effect, using a Poisson distribution and without checking my work.)

My question is: Have any evidence been published to suggest that this Blackberry effect is real? Alternatively, has any evidence been published to suggest that this Blackberry effect was NOT real (i.e. we have the figures and the differences in a 2-3 day sample are not statistically significant)?

My gut feel is the whole claim was made up based on anecdote - not that I am claiming for a second that texting while driving is safe.

Community
  • 1
Oddthinking
  • 140,378
  • 46
  • 548
  • 638
  • Including only fatal accidents will get you a significantly smaller dataset. If the numbers are correct, that there is an accident every three minutes, and that fell 20% during the outage, my gut feel is that that would be statistically significant. – David Hedlund Oct 24 '11 at 07:32
  • More ammunition: The outage was on the 10th-13th of October (US time). The newspaper article was on the 15th. Basically, that gave one day for the police reports of all of the accidents to be filed, collated, compared to a typical three day period, analysed for significance, considered against confounding sources, written up and reported to the police chief and then to the reporter. Feasible? Yes. Likely? No. – Oddthinking Oct 24 '11 at 15:51
  • 1
    Seeing as the outage expanded beyond Dubai I would think that if the effect was real we should see a decline elsewhere that correlates with the prevalence of Blackberries. I would guess that Dubai has a higher than average distribution of Blackberries(due to average income). But I would expect a reduction everywhere during the outage if this is a real effect. – Chad Oct 24 '11 at 20:33
  • The Executive Director of the Illinois Tollway has used this statistic in a letter to the editor in the Chicago Tribune to self-congratulate on a "Drive now. Text Later" tollway promotion. No doubt reducing, better yet, eliminating texting while driving is a good thing. However, a public official relying on internet gossip to support policy decisions shows a lack of thorough fact-checking. This could explain why this same official recently promoted and passed an 87% increase in tolls. How accurate were the "facts" to make this decision? –  Nov 11 '11 at 14:00

0 Answers0