Journalism & Fact-Checking: Verifying Numerical Claims

The Numerical Smell Test

A politician says a program costs "$400 million." A CEO claims their platform has "2 billion users." A press release states that a factory produces "50 million units per year." Are these numbers plausible? Do they pass basic sanity checks? Most journalists, trained in words rather than numbers, lack the intuition to evaluate these claims quickly.

Common Red Flags

Here are patterns that should trigger skepticism:

• Numbers that are suspiciously round. Real data rarely comes out to exactly "$5 billion" or "10 million users." Round numbers often mean estimates, projections, or fabrications.
• Per-person numbers that don't make sense. If a state program claims to have helped "3 million people" in a state with 4 million residents, that means 75% of the population was served. Possible, but worth verifying.
• Growth rates that imply impossible totals. "Growing 50% year-over-year" sounds moderate. But 50% compounded annually means 7.6x in 5 years and 57.7x in 10 years. A company with $100M revenue growing at 50% annually would reach $5.7 billion in a decade. Is that realistic for their market?
• Mixing up millions and billions. This happens in press releases and political speeches more than you'd think. A "$2 billion" program described in detail actually costs $2 million, or vice versa. The three-order-of-magnitude difference often goes unchallenged.

The Division Test

The most powerful fact-checking tool for large numbers is division. Take any large claim and divide it by something meaningful:

• Government spending ÷ population. A "$15 billion infrastructure bill" for a city of 500,000 people means $30,000 per resident. Does that seem right for road repairs? That's suspiciously high.
• Revenue ÷ users. A company claiming "$5 billion revenue" with "100 million users" is claiming $50 per user. For a free app? That's high. For an enterprise SaaS? That's low.
• Production ÷ time. "50 million units per year" is about 137,000 units per day, or about 5,700 per hour (assuming 24/7 operation). For a car factory? Impossible. For a chip foundry? Plausible.

Using Visualization for Verification

When you encounter a numerical claim, enter it into the How Big? tool. The time and physical comparisons instantly contextualize the number. If a number "feels" wrong when converted to time or distance, trust that feeling and investigate further.

Case Study: "The US Wastes $165 Billion in Food Per Year"

This is a real, frequently cited statistic. Let's check it:

• US population: ~335 million
• $165 billion ÷ 335 million = ~$493 per person per year
• That's about $1.35 per person per day in wasted food
• The average American spends about $10-15 per day on food
• So wasting $1.35/day (about 10-13% of food spending) seems plausible

The number passes the smell test. But without doing this division, "$165 billion" is just a scary-sounding number that could be off by a factor of 10 and most readers wouldn't notice.

Building Numerical Intuition

The best fact-checkers develop reference points they can quickly access:

• US population: ~335 million
• World population: ~8.1 billion
• US GDP: ~$28 trillion
• US federal budget: ~$6.2 trillion
• Median US household income: ~$75,000

With these anchors, any numerical claim can be quickly converted to a per-person, per-GDP, or per-budget figure for a basic plausibility check.