Home

Predicting iPhone vs. Non-iPhone Users

Can you tell if someone is going to mess up the group chat?

Contents

• Background
• Results
• Summary
• Appendix I: Data
• See Also

Background

In my culture, there are certain types of people that don't use iPhones. Take that how you will, just don't get mad at me, I don't make the rules. And although there is definitely some confirmation bias going on, I don't think that's the entirety of it. I've been making predictions about which phone someone uses before I see it and I feel like I'm correct a majority of the time (again, confirmation bias absolutely plays a role here, but how much is up for debate and what I'm trying to accomplish here).

This is my chance to actual record predictions. For every person that I know or meet but do not know their model of phone, I will predict the phone they use with some degree of confidence, find out which phone they use, then mark it down as a success or failure. (Side note: I only started the confidence tracking halfway through, so the first half of people are confidence-less.)

A few ground rules for fair play:

• If any specifics as pertaining to the model are said about their phone before I make my prediction, that prediction is no longer valid
• I will not ask any probing questions about the phone itself. Personal questions that may help my prediction are allowed
• People who I have texted before are off limits, whether or not I remember what type of phone they use
• People who I know have immediate family members with certain phones are off limits, as I believe it more likely that if a son/daughter has an iPhone, the parents will, too. (I've found that the converse, inverse, and contrapositive are not necessarily true!)

The rules effectively boil down to me focusing on strangers or acquaintances whose personal history and details I know little to none of.

Results

Data was collected from 05 April 2022 to 15 January 2024 during in-person interactions and over a dating app. I started off with just yes-no prediction and transitioned to probabilistic predictions.

See Appendix I: Data for the full dataset.

Statistics

I used the Brier score to calculate my accuracy:

Some basic statistics:

• n: 59 (48 female, 11 male)
• Correct about women's phone: 90%
• Correct about men's phone: 73%
• Correct about owning an iPhone: 90%
• Correct about owning a non-iPhone: 67%
• Average Brier score: 0.112

These numbers mean nothing unless I compare them to the iPhone vs. non-iPhone market shares:

• 67% of 18-29 year-olds use iPhone (source)
• Apple held 55% of the market as of 2Q23 (source)
• I could not find information on age-gender-brand numbers. This would have been the most valuable dataset. I suspect 18-29 year-old women own iPhones at a much higher rate than the aforementioned 67%, with the rate going even higher with income.

Summary

I did decently better than chance when predicting women's phones and only slightly better than chance when predicting men's phones.

Appendix I: Data

More information that could make this dataset better:

• Age
• Income
• Percentage of close friends and family members with iPhones
• Occupation
• Housing location

See Also