A lot of people write in with questions about the top one percent in net worth and income. Soon after, I often see a common question: how well does income predict net worth? In this post, I'll help clear it up and take a look at the correlation of net worth and income.
In short, it turns out income does not predict wealth all that well. While there is a relationship, it's relatively weak. In 2020 in America, the correlation between income and net worth was .5036, an R^2 of .2536.
Read on for the correlation of income and net worth by age – and arguments for why this isn't a great measure.
The Correlation of Income and Net Worth is Weak in America
An R^2 of .2536 is a weak coefficient of determination.
Suppose you accept this correlation matters at face value (and it isn't just "one weird math trick"). In that case, 74.64% of the variability between income and net worth isn't accounted for in the relationship.
Now, it's quite evident that income matters. You can even see the relationship in the graph – it's quite apparent income is an essential input (despite the significant spread in results).
But there is more at play in net worth than a household's income at the point they take a survey.
Visualizing the Relationship Between Net Worth and Income
Above, we've graphed individual responses in the 2019 Federal Reserve SCF with income on the X-axis and net worth on the Y-axis with a logarithmic scale. And definitely click to zoom in – there are a lot of data points.
(Read our net worth research for methodology details.)
A survey can't sample all of America – here, the size of the point reflects the 'weight' of the response. Weight maps to the number of households in America that response should be considered "represented" by each data point.
Further, the point's color represents the age of the householder (grey is set to the weighted average respondent age: 51.725 years old).
Logarithmic Income and Net Worth
I'll continue to play ball with the correlation question. As you see from our axes in the bubble plot, if we took the Pearson correlation coefficient of the log10 income and log10 net worth, it will probably be better than the naive linear correlation.
Now, there's a significant issue here – you lose almost 13% of the data since income or net worth of $0 (or negative) needs to be discarded. Let's accept that's an okay state and push onward.
Using positive numbers and log base 10 numbers:
What Variables Matter for Net Worth?
I don't believe correlations – adjusted for log base 10 or not – make much sense for this data, as we'd ideally like an input like income history.
(Also, I don't want to be the target of this joke:)
However, this exercise is useful to show that a single year snapshot of income isn't enough to explain net worth. That's despite how the media conflates income and net worth.
Here are other things that matter:
- Age – More time in the market almost always beats timing the market
- Consistency – Earn a large income over time
- Frugality – Save a high percentage of your income, annually
- Luck – Sometimes it's better to be lucky than good*
* Yes, luck is controversial. But especially for high variance situation – think a startup seeking funds or similar – luck matters. Luck's definition may be the combination of preparation and opportunity, but you can't discount the effect of the coin flip or the dice roll completely.
Age and Income vs. Net Worth Correlation
You can see in the above visualization that age has a pretty significant effect on the survey results. Even without a full income history variable, it's evident that older folks at the same rough current income level tend to have more net worth.
Although 'events' such as home sales, stock sales, business windfalls, and the like can cause huge year-to-year swings, earned – or wage – income is relatively inelastic.
Let's say we split this up by age. What would we expect to see?
A lovely story might be:
- lower correlation for the youngest groups; high net worth here is probably windfall related (e.g., an inheritance or other inter-generational transfer)
- higher correlation in middle age
- falling correlation in older groups as consistency and propensity to save start to dominate income (also: retirees have less income)
Do we see that? Kind of if you squint – but it falls apart at age 40-44 where people are generally near their highest earnings. (Otherwise it seems to make sense!)
So, what's the deal with mid-career folks?
Just a fluke? Something with the Great Recession? Stock market or real estate returns throwing it off? Is the convenient narrative wrong?
I don't know here, and I really don't even have a guess. Maybe you have a theory?
Net Worth vs. Income: How do they differ?
I'll now mention an excellent way to look at income and net worth that I've discussed in the past:
- Look at income as acceleration - how quickly a household adds to its total wealth... assuming the household can save and invest.
- Net worth, on the other hand, is speed or how fast the car is currently moving. Net worth has inertia.
I know – it's not perfect. It is, however, a decent way to frame this issue. Acceleration matters in gaining speed, but it doesn't necessarily mean you'll reach a high speed.
Lump Sums and Income Streams
Another way to look at income vs. net worth?
It's relatively easy to convert net worth to a stream of income. However, converting an income stream to net worth is not so straightforward. (More discussion in the average net worth post).
Instead of just one year's income, variables like age matter a lot – play with the net worth by age calculator to get a feel.
You should expect a 25-year-old and a 65-year-old, both making $100,000 a year, to have wildly different household wealth. And – although we don't know the future – the 25-year-old can have a substantial net worth at age 65 if they sustain those earnings and are good at saving.
Furthermore, there is a large group of people with little to no income and substantial wealth. That's right - retirees.
(Did you say, "personal finance bloggers"? I laughed!)
Methodology and Source on the Correlation Of Income and Net Worth
We start with the 2019 SCF Data from the Federal Reserve. For this series, we export all five imputed sets and average across them. So, the total dataset sums to 5,777 real data points, and 28,885 to run the math.