Since 1871 the real return on wealth beat economic growth in 177 of 202 country-decades

What Jordà, Knoll, Kuvshinov, Schularick and Taylor found

Their Rate of Return on Everything assembled real returns on equities, housing, bonds and bills for 16 advanced economies back to 1870, and combined them, at each country's portfolio weights, into a single return on wealth. Comparing that return to real GDP growth, they reported that since 1870 the weighted-average return on wealth r has been about 6.0%, against a weighted-average real GDP growth rate g of about 3.1%, an r minus g gap of 2.9 percentage points over the full sample and around 3.6 points in peacetime. Over the past 150 years, they wrote, the return on wealth exceeded growth in 13 of 15 decades and was below it only in the two decades of the world wars: in peacetime, r has always exceeded g. Those magnitudes are the sanity band this note checks itself against; nothing below is copied from them.

Source: Jordà-Schularick-Taylor Macrohistory Database Release 6 (16 economies with a wealth-portfolio return): real return on wealth (capital_tr deflated by cpi) and real GDP growth (rgdpmad times pop). JST is released for non-commercial use with citation; displayed here, not redistributed. The claim and the quoted magnitudes are Jordà, Ò., K. Knoll, D. Kuvshinov, M. Schularick, and A. M. Taylor (2019), "The Rate of Return on Everything, 1870-2015," Quarterly Journal of Economics 134(3), 1225-1298.

What we run it on

Both sides of the comparison come from one file, the JST panel view jst (data/parquet/jst.parquet, Macrohistory Release 6), so r and g are dated by the same source and no second compiler's levels enter the ratio.

• r is JST’s own return on wealth. The column capital_tr is the total return on all investable wealth, a portfolio of equities, housing, bonds and bills weighted by each country's wealth shares, exactly the aggregate the paper built. It is nominal, so it is deflated by JST's own cpi; the German 1923 hyperinflation makes deflation non-negotiable, and the same real-return arithmetic runs behind the site's rate of return pages.
• g is real GDP growth. Real GDP is rgdpmad (Maddison real GDP per capita) times pop, the aggregate output series, and g is its year-over-year growth. Using aggregate rather than per-capita growth follows the paper and is the harder test for r above g: per-capita g would be lower and the gap wider.
• Every annual change is calendar anchored. The wealth-return series has genuine holes, war and hyperinflation years are simply absent, so a year-over-year figure is built by joining each year to the row for the year before it, never by stepping back a fixed number of rows. A row-offset step over a series with gaps silently bridges them and fabricates multi-year changes labelled annual, a defect this site has hit before and now guards against everywhere.
• Decade averages are geometric. Within each country-decade, r and g are the compound (geometric) annual means of the real return and the real growth over the years present, because returns and growth compound; a country-decade enters the count only when it has at least one year of each. The JST database panel is 18 economies; two of them, Canada and Ireland, carry no wealth-return series and drop out, leaving the 16 the Rate of Return on Everything itself covers, exactly as on the /returns pages.

Source: Jordà-Schularick-Taylor Macrohistory Database Release 6 (16 economies with a wealth-portfolio return): real return on wealth (capital_tr deflated by cpi) and real GDP growth (rgdpmad times pop). JST is released for non-commercial use with citation; displayed here, not redistributed. The calendar-anchored self-join and the geometric mean are the same conventions the /returns module uses on the identical panel.

The headline: r above g, decade after decade

Pooling all 16 economies by decade, the real return on wealth sits above real GDP growth in 13 of the 15 completed decades since 1871. The two exceptions are the 1910s and 1940s: in the 1910s the pooled real return was 0.04% against 0.95% growth, and the 1940s are the other, the two windows in which the world wars destroyed capital faster than they cut output. In every other decade the return line runs clear above the growth line.

The finer count is by country-decade rather than by pooled decade, and it tells the same story with room for national exceptions: across the 202 country-decades that carry both a return and a growth figure, r exceeds g in 177 (87.6%), with a median gap of +3.65 percentage points. Dropping the world-war years from every decade average barely moves it: 174 of 201 peacetime country-decades (86.6%) still show r above g, at a median gap of +3.62 points. Pooled over the whole panel the real return on wealth is 5.79% against 2.61% growth, a gap of 3.18 points; in peacetime it is 6.33% against 2.92%, a gap of 3.41.

Every economy, every decade

Each cell below is one country-decade, teal where the real return on wealth beat real GDP growth over that decade and amber where it did not, the shade deepening with the size of the gap. The amber cells cluster in two columns, the 1910s and 1940s, and in a scatter of fast postwar catch-up decades where growth briefly ran with the return. Read across any row and the teal dominates: in all 16 economies the full-sample return on wealth exceeds full-sample growth.

Source: Jordà-Schularick-Taylor Macrohistory Database Release 6 (16 economies with a wealth-portfolio return): real return on wealth (capital_tr deflated by cpi) and real GDP growth (rgdpmad times pop). JST is released for non-commercial use with citation; displayed here, not redistributed. Full-sample compound real return on wealth and compound real GDP growth per economy; r minus g is the difference of the two rounded figures.

How this lands against their numbers

The magnitudes recomputed here sit inside the paper's sanity band without being fitted to it. Our pooling is unweighted over country-years and the paper's is GDP-weighted with decadal moving averages, so the levels need not match to the decimal; that they land this close is the check.

Source: Jordà-Schularick-Taylor Macrohistory Database Release 6 (16 economies with a wealth-portfolio return): real return on wealth (capital_tr deflated by cpi) and real GDP growth (rgdpmad times pop). JST is released for non-commercial use with citation; displayed here, not redistributed. Right column: Jordà, Ò., K. Knoll, D. Kuvshinov, M. Schularick, and A. M. Taylor (2019), "The Rate of Return on Everything, 1870-2015," Quarterly Journal of Economics 134(3), 1225-1298. Left column: computed in src/lib/rVsG.ts from the JST columns.

Where ours differs from theirs

• Unweighted, not GDP-weighted. The paper weights each country by its real GDP and smooths with decadal moving averages; this note pools country-years unweighted so every figure is one a reader can reproduce straight from the file. A large economy therefore counts the same as a small one here.
• Counting, not their decomposition. They decompose r into its asset legs, study its co-movement with growth, and weigh the inequality mechanism at length; this note counts where r exceeds g and reports the gap. The +3.65-point median is an association in the panel, not a claim that the return caused anything.
• Through 2020, not 2015. This JST release runs to 2020, five years past the paper's window, so the recent decades carry data the original did not. The single 2020s year is shown on the chart but excluded from the completed-decade count.

What this cannot tell you

• r is an index return, not anyone’s. The return on wealth here is an aggregate market return on a national portfolio, before taxes, management costs, idiosyncratic risk and leverage. No household earned it, and it says nothing about the returns of the rich versus the rest, which is a separate and contested measurement.
• r above g is not, by itself, rising inequality. Piketty's mechanism is that a persistent gap between the return on capital and the growth rate lets accumulated wealth grow faster than income and so tends to concentrate it; this note measures only that the gap has existed, not whether wealth in fact concentrated, which turns on saving rates, taxation, inheritance and demography that this data does not carry. Nothing here forecasts inequality or endorses any prediction about it.
• The exceptions are informative. The country-decades where g beat r are not noise: they concentrate in the world-war decades, when capital was destroyed, and in fast postwar reconstruction booms, when output rebounded quickly. The gap is a peacetime, steady-state regularity, and the note counts the exceptions rather than hiding them.
• Sixteen advanced economies, annual. The panel is annual and confined to the long-run advanced economies of the JST database; it says nothing about emerging markets, and Canada and Ireland drop out for want of a wealth-return series. Decade cells vary in how many years they rest on, and the year counts are printed so a thin decade is visible.