Conversation

Notices

1. Embed this notice
   Ryan Castellucci :nonbinary_flag: (ryanc@infosec.exchange)'s status on Thursday, 08-Aug-2024 20:43:54 JST Ryan Castellucci :nonbinary_flag:

   DWP provided the data they said was too hard.

   I need to ask for how many national insurance numbers have been issued in total, but my best guess is about 135M since they were introduced in 1948. This figure obtained by summing population that year with live births since then and estimated annual immigration figures.

   Per DWP,

   • 1,673,313 people with duplicate full name and date of birth
   • 33,274 of them could be differentiated by gender

   These figures seem rather high, but the key point is that gender would be the disambiguating factor in fewer than 2% of cases, and is relevant for this purpose in approximately 0.025%.

   https://www.whatdotheyknow.com/request/national_insurance_records_disti/response/2722514/attach/html/2/DWP%20response%20following%20ICO%20Decision%20Notice.pdf.html

   • Embed this notice
     Ryan Castellucci :nonbinary_flag: (ryanc@infosec.exchange)'s status on Friday, 09-Aug-2024 02:37:02 JST Ryan Castellucci :nonbinary_flag:

     in reply to

     • Christian Lawson-Perfect

     @christianp I'm still a little surprised it's that high - I would have thought that middle names would minimize the collisions. Drilling down would be really interesting.

   • Embed this notice
     Christian Lawson-Perfect (christianp@mathstodon.xyz)'s status on Friday, 09-Aug-2024 02:37:03 JST Christian Lawson-Perfect

     in reply to

     @ryanc that is fascinating!

     I have a dataset of first names given to babies in the UK between 1996 and 2015, from the ONS. The probability that two people with the same first name have different recorded genders is about 4%, which isn't far off the number you were given.

   • Embed this notice
     Ryan Castellucci :nonbinary_flag: (ryanc@infosec.exchange)'s status on Friday, 09-Aug-2024 02:37:28 JST Ryan Castellucci :nonbinary_flag:

     in reply to

     • Christian Lawson-Perfect

     @christianp also tempted to do a FOI for how long this actually took them