I asked ChatGPT a question about white people. Here's what it answered.

Pictured: a representative of a tiny minority who thinks he is the majority.

For some weird reason, a question about demographics popped into my head this morning while I was vacuuming my living room. 

What percentage of the world population are white, non-Hispanic, male, heterosexual, native-born American citizens?

Doing some quick math in my head, I came up with about one percent. If the US population is about four percent of the world population and half are male, and half of them match these other criteria, that is about one percent. I got on the internets and started clicking around for about two minutes to test some assumptions and try to figure out a more accurate number, before I decided to just ask ChatGPT.

Sure, ChatGPT gets stuff wrong and makes stuff up all the time. Yes, every query contributes to the destruction of our world’s ecosystem. Yes, it’s investors have become insanely wealthy by stealing the intellectual property of the world’s creative minds. But hey, I need an answer. (This is where I mention that I am a vegetarian, I drive rarely, my car is a hybrid, all my household power is solar, and I buy carbon credits to offset my occasional but rare air travel.)

So I asked:

What percentage of the world population are white, non-Hispanic, male, heterosexual, native-born American citizens?

Here is what ChatGPT answered:

Calculating the percentage of people worldwide who fit the description of being white, non-Hispanic, male, heterosexual, native-born American citizens requires layering demographic data. Here's a breakdown of the key points:

1. World Population: The global population is approximately 8 billion as of 2024.

2. White Population: Roughly 16% of the global population is white, with the majority living in Europe, the Americas, and other Westernized nations.

3. Non-Hispanic Whites: Within the U.S., non-Hispanic whites account for about 59% of the population. Globally, this percentage would decrease significantly due to the diversity of the white population outside the U.S.

4. Males: Males constitute about 50% of the global population.

5. Heterosexual Individuals: Estimates suggest that around 90-95% of people globally identify as heterosexual.

6. American Citizens: The United States has about 4.25% of the global population, equating to roughly 340 million people.

7. Native-Born Americans: About 86% of the U.S. population are native-born.

Combining these factors:

• Start with the global population: 8 billion.
• Apply the percentage of whites (16%): 8 billion × 0.16 = 1.28 billion.
• Adjust for non-Hispanic whites (approx. 75% of whites globally): 1.28 billion × 0.75 = 960 million.
• Account for males (50%): 960 million × 0.5 = 480 million.
• Adjust for heterosexuals (95%): 480 million × 0.95 = 456 million.
• Focus on American citizens (4.25% of the global population): 456 million × 0.0425 = 19.38 million.
• Account for native-born Americans (86%): 19.38 million × 0.86 = 16.67 million.

Finally, divide by the global population to find the percentage:

(16.67 million divided by 8 billion) ×100 ≈ 0.21%

Answer: About 0.21% of the global population fits this description.

That answer was much lower than my ball park calculation. But notice how ChatGPT calculated it differently than I did? I decided to ask another question:

Do you get a similar answer if you start by applying the percentage of the world population that live in the United States and then apply the other percentages after that?

Here was the answer:

Let’s calculate by first applying the percentage of the world population living in the United States and then layering the other demographics.

Breakdown

1. U.S. Population: The U.S. has about 4.25% of the global population.

   8 billion×0.0425=340 million8 \, \text{billion} \times 0.0425 = 340 \, \text{million}

2. Native-Born Americans: Approximately 86% of the U.S. population are native-born.

   340 million×0.86=292.4 million340 \, \text{million} \times 0.86 = 292.4 \, \text{million}

3. Non-Hispanic Whites: Non-Hispanic whites make up about 59% of the U.S. population.

   292.4 million×0.59=172.5 million292.4 \, \text{million} \times 0.59 = 172.5 \, \text{million}

4. Males: Males account for about 50% of the population.

   172.5 million×0.5=86.25 million172.5 \, \text{million} \times 0.5 = 86.25 \, \text{million}

5. Heterosexual Individuals: Approximately 95% of people identify as heterosexual.

   86.25 million×0.95=81.94 million86.25 \, \text{million} \times 0.95 = 81.94 \, \text{million}

6. Percentage of World Population: To find the final percentage, divide by the global population.

   81.94 million8 billion×100=1.02%\frac{81.94 \, \text{million}}{8 \, \text{billion}} \times 100 = 1.02\%

Comparison

The answer is approximately 1.02% when starting with the U.S. population and applying other factors, higher than the previous estimate (0.21%) where we began globally.

Why the Difference?

The discrepancy arises because starting with the U.S. population assumes all characteristics (white, non-Hispanic, male, heterosexual) occur within the subset of people living in the U.S., which skews the calculation. Starting globally gives a more diluted result, as it accounts for the rarity of each trait worldwide.

I was officially confused. So I asked one more question:

Which answer is more accurate?

ChatGPT answered:

The first calculation ( 0.21%~0.21\%) is more accurate, as it begins with the global population and progressively narrows down to the specific group by applying probabilities for each characteristic globally. Here's why:

1. Global Perspective: The first method considers the rarity of each trait across the entire world population, which is more representative of the global demographics.

2. Avoiding Overconcentration: The second method starts with the U.S. population, which inherently assumes all other traits (e.g., being male, white, heterosexual) occur in their average proportions within the U.S. This skews the result because these traits are not equally distributed across the world and do not scale directly from the U.S. demographic proportions.

Correct Interpretation

To estimate the percentage of the global population fitting this specific combination of traits, the first method avoids overrepresenting Americans and is therefore closer to reality. Thus, the global-based approach (~0.21%) is the better estimate.

Honestly, I don’t know which of these answers are closer to reality. But regardless of the exact number, this statistical fact makes for some interesting realizations.

Thoughts? Please share in the comments.

Thank you for your time, and have as good a day as possible in these insanely trying times.

Happy New Year.

Update:

​​​​I just want to say thanks to everyone who commented on this post. 

One thing that is now clear, thanks to some of the comments, is that ChatGPT behaved like a Republican. It did a terrible job at being accurate and an excellent job at sounding absolutely certain it was correct. 

The second estimate method was closer to the truth, and the explanation it provided of why it thought the first method was more accurate was complete bullshit.

I am happy this made it to the top ten and also happy it did not make it to number one. There were several other posts that were more important than this one. But this was fun. Again, thank you, and I wish you all a better year in 2025!