💰 So I Tried to Find the Exact Price of Happiness

We've all heard the $75k number. Kahneman and Deaton dropped it in 2010 and it spread like a cold in a daycare. Talk shows, self-help books, LinkedIn posts about "finding your purpose", all citing that one number like it was handed down from the mountain.

Then in 2021, Matthew Killingsworth came along with a mood-tracking app, 33,000 employed adults, and a paper that basically said: nope, happiness just keeps climbing, no plateau, rich people are simply happier, deal with it.

They then wrote a joint paper together in 2023 and still couldn't fully agree. Two of the most cited behavioral economists alive, co-authoring, going "well, it depends." You genuinely love to see it.

So I decided to do it properly. Causal inference. Real controls. 52 years of survey data. Here's what I found.

🔍 What They Got Wrong

Let me be clear: Kahneman is a Nobel laureate and I'm a person with a laptop and a free NORC dataset. But methodology is methodology, and there are three concrete things missing from all their work:

1. No cost-of-living adjustment

$75k in rural Mississippi and $75k in San Francisco are not the same number. One gets you a house and a savings account. The other gets you a studio with questionable plumbing. Using nominal income without regional price adjustment is like comparing temperatures in Celsius and Fahrenheit without converting. You'll get something, just not the right answer.

2. No causal identification

Correlation between income and happiness could mean money makes you happy, happiness makes you productive and earn more, some third variable causes both, or all three at once. You can't disentangle that from a regression. You need a quasi-experimental design.

3. Short time windows

Killingsworth's data covered roughly two years. The GSS goes back to 1972. If the threshold shifted over decades, and spoiler: it did, you can't see it without longitudinal depth.

📊 The Dataset: 52 Years of "Are You Happy?"

The General Social Survey has been asking Americans one question since 1972: "Taken all together, how would you say things are these days, would you say that you're very happy, pretty happy, or not too happy?" 75,699 respondents. 63,667 clean rows after dropping missing data.

I pulled happy, realinc (already CPI-adjusted by NORC, thanks), plus age, education, region, marital status, and employment. Then I added one thing none of the prior papers did: regional price parity data from the Bureau of Economic Analysis to adjust for cost of living.

🛠️ Three Causal Methods, One Answer

Instead of running one analysis and calling it a day, I ran three, then checked if they agreed. All pointing the same direction? That's a finding. All pointing different directions? That's a problem.

The IV estimate being higher than OLS is the interesting detail. The usual assumption is reverse causality inflates correlations (happy people earn more, so OLS overstates the effect). My IV says the opposite: when income goes up due to exogenous wage shocks, happiness goes up more than the raw correlation suggested. Kahneman's estimate may have understated the causal effect.

The DiD estimate is smaller but still highly significant across 50 years of policy variation. All three pointing the same direction: money does cause happiness, at least up to a point.

💡 The Actual Number: $117k, Not $75k

After applying cost-of-living adjustment and running piecewise spline regressions to find the actual inflection point, instead of just assuming one like Kahneman did, the happiness curve peaks at approximately:

That's 56% higher than the $75k quoted in every article for 15 years. And before you say "well, inflation", the GSS realinc variable is already inflation-adjusted. The gap isn't time. It's geography. It's the fact that nobody accounted for where people actually live.

📉 The Part Nobody Published: It Actually Drops

This is the finding I didn't expect, and the one I'm most confident is real. At very high incomes, above roughly $120k in COL-adjusted terms, happiness doesn't just plateau. It declines.

Killingsworth's entire 2021 argument was "it never plateaus, it keeps rising forever." My data says: it rises, peaks around $117k, then turns back down. Hedonic adaptation is real. So is the stress, identity pressure, and social comparison with even wealthier peers that tends to come with very high incomes.

To be fair, this could be selection effects. High-stress careers may attract people who are already less happy. My IV design helps but doesn't perfectly solve this. I flag it because that's what honest analysis looks like.

📅 The Threshold Has Been Moving

The $75k number wasn't just wrong because of geography, it was also a snapshot. The inflection point has shifted substantially over five decades:

The relationship between income and happiness has strengthened over time, likely as inequality increased and relative income comparisons became more psychologically loaded. The $75k number was already aging badly when it was published in 2010. By 2024, it's genuinely misleading.

🚀 What This Actually Means

Yes, money causally buys happiness , up to a point. This isn't just "rich people happen to be happy." Three different methods all confirm it's directional.

The real threshold is higher than you've been told , and it depends on where you live. If you've been using $75k as a mental benchmark, it's time to update your model.

But chasing income past the threshold appears to backfire. The marginal happiness return above $117k (COL-adjusted) looks negative in this data. That doesn't mean money is evil past a certain point. It means optimizing purely for income at the expense of everything else probably stops paying off.

The Lambo Memo is half right. Money up to a real threshold does buy happiness, causally. But the crying part at very high incomes? The data supports it more than whoever wrote that memo probably meant.

🔬 The Methodology (For Those Who Want It)

Data: GSS 1972–2024, n=63,667 after listwise deletion. realinc is CPI-adjusted by NORC. Added BEA regional price parity adjustment across 9 census regions. Happiness coded 1–3, treated as continuous for regression with ordered probit as robustness check.

Causal identification: IV used state minimum wage as instrument under the assumption that wage law changes affect income independently of individual happiness factors. DiD used policy variation across states 1972–2024 as natural experiment. Threshold identified via piecewise linear regression with data-driven breakpoint, not assumed in advance.

Limitations worth naming: GSS is cross-sectional, different people each year, not a panel. The drop at high incomes could reflect selection. Regional price parity is at census region level, not city level, so NYC vs. Buffalo variation is still underfit. A panel dataset like PSID with MSA-level price data would sharpen this considerably.