Running Hot, or a Moving Ruler? GDP, AI, and Whether 'Normal' Inflation and Interest Rates Change With the Times

The Bureau of Economic Analysis (BEA) publishes second-quarter GDP on September 30. Two stories are circulating ahead of it.

The first is worry: the numbers look strong (the Atlanta Fed's GDPNow model puts third-quarter growth at 5.1%), the economy may be overheating, and the Fed will keep raising rates. That worry has a basis. On September 16 the Fed raised its target range for the federal funds rate by 25 bp to 3.75%–4.00% (effective September 17), its first hike since July 2023. On September 15 the 10-year Treasury yield closed at 5.00%, its first close at or above 5% since July 19, 2007.

The second is optimism: AI will carry the economy into a new boom, as the internet did in the 1990s. If the times have changed, shouldn't our sense of "high" and "low" change too? Rate hikes make people nervous, yet compared with 30 or 40 years ago, today's rates are low. The yardstick for rates has already moved. Should the yardstick for inflation move with it?

This article takes those questions apart and tests each one against data:

1. When people say "GDP is high," which number is high?
2. Has the AI boom started?
3. Does the yardstick for interest rates change over time?
4. Can the yardstick for inflation change the same way?
5. What did the great breaks of the past (the Industrial Revolution, electricity, the internet) change?

The short answer: the questions have different answers, and the reasoning for one cannot be carried over to another. Why that is so is the main thing this article tries to explain.

Data run through September 21, 2026. This is a history-and-methodology article that describes only what has already happened and been published. It makes no forecast of the September 30 figures, and it is not a recommendation about any security.

1. Where the numbers come from

• US government works, public domain: GDP and each component's contribution to growth, and investment by type (BEA); labor productivity, CPI and unemployment (BLS); interest rates and FOMC participants' longer-run projections (Federal Reserve); the natural rate of unemployment (CBO).
• UK government work, Open Government Licence v3.0: the long-run Retail Prices Index from 1800. Source: Office for National Statistics (ONS).
• CC BY 4.0: UK GDP per person from 1270, from the Maddison Project Database 2023. The database's citation policy requires citing the original papers, which are listed in the references.
• This site's own research: US inflation since 1900, from the same data behind The Break-Even Line.

This article uses no stock prices and no index data.

Two caveats up front. First, GDP gets revised. This article uses the vintage published before September 30; after that date the second-quarter figures, and possibly earlier ones, may change. Second, the BLS never published October 2025 CPI (the federal government was shut down). All year-over-year changes are aligned by calendar month, and nothing is filled in.

Apart from figures explicitly marked as quoted (statements by Fed officials, St. Louis Fed research, the GDPNow estimate), every figure in this article was computed by a program from the public data above. None was typed in by hand.

2. First: which GDP is "high"?

The September 30 release is the third estimate of the second quarter, not a new quarter. The advance estimate for the third quarter comes on October 29. The second-quarter number has already been published twice.

Three GDP figures tend to get blurred together: real growth (how much more the economy produced, after removing price changes), nominal growth (the change in dollar value at current prices), and the gap between them, which is prices. The headline "GDP growth rate" is usually the first.

(Every figure in this table is an annualized quarterly rate: growth from the previous quarter, scaled up to a full-year pace and seasonally adjusted. This is the "GDP growth rate" in headlines. The other common measure is the year-over-year rate, which compares a quarter with the same quarter a year earlier. They are different numbers, and the text below says which one it is using.)

Real growth in the past two quarters was 2.1% and 1.5%. Against every quarter from 1947 through 2019 (again at annualized quarterly rates), the median was 3.1%: 65.64% of quarters beat 2.1%, and 71.48% beat 1.5%. Year over year, compared with the second quarter of 2025, real GDP was up 2.10%, almost exactly the 2.0% that FOMC participants estimate as the economy's normal long-run growth rate.

By real output, the last two quarters were not high.

The high number is nominal growth. Nominal GDP grew at an annualized quarterly rate of 8.0% in the second quarter, though single-quarter rates jump around: the table shows 8.3% in 2025 Q3. The steadier measure is year over year. Compared with the second quarter of 2025, nominal GDP was up 6.56%, the highest year-over-year rate since 2023 Q3 (6.60%).

Prices pushed it up. The GDP price index rose 4.36% year over year, the highest since 2023 Q1 (5.32%), and at a 6.4% annualized rate in the second quarter alone.

The distinction matters. If the sense that "GDP is strong" comes from nominal figures, most of it reflects prices rather than more output, and prices are the reason the Fed is hiking. Chair Warsh put it plainly at the September 16 press conference: "inflation is too high and has been for too long." The PCE price index has been rising faster than 2% a year for 65 consecutive months, since March 2021.

How noisy a single quarter is

Here is real growth broken into each component's contribution, in percentage points:

(Percentage points. Each component is rounded separately and chain-weighted, so they do not sum exactly to GDP growth.)

In the first half of 2025, businesses rushed imports in ahead of tariffs, then pulled back. The net-export contribution swung from −4.68 to +4.83 in consecutive quarters, and inventories swung the other way. Government subtracted 0.99 points in 2025 Q4, the quarter of the October–November federal shutdown. None of this says anything about the economy's underlying health, but it is enough to move a single quarter by several points.

So when reading one quarter of GDP, the headline total is the least informative number.

What is the 5.1%?

When people say "GDP is running high," they often mean a different number: on September 17, the Atlanta Fed's GDPNow model estimated third-quarter real GDP growth at 5.1% (annualized).

Three things are worth being clear about.

First, it refers to the third quarter, not the second-quarter figure due September 30. The official first estimate for the third quarter comes on October 29.

Second, it is a model's running estimate, not an official figure and not a forecast. In the Atlanta Fed's own words, GDPNow is not an official forecast of the bank. It is a running estimate built by a model from the data published so far in the quarter, with no subjective adjustments. It moves with every new data release. The next update is September 25, and it will keep updating until October 29.

Third, this time the high number is real growth. Unlike the previous two quarters, if the third quarter ends up near this estimate, that will mean strong real output, not just prices. This article makes no forecast of where it will land.

Is 5.1% high? Again, it depends on the yardstick. From 1947 to 2019, 25.43% of quarters had real growth of 5.1% or more; from 2000 to 2019, only 5.0% did.

The composition matters more than the total, and the GDPNow page shows a component-by-component breakdown (linked in the references). Of all the components, inventories are the noisiest. From 1947 to 2019 the standard deviation of each component's quarterly contribution to GDP growth was 2.47 points for inventories, 2.01 for consumption, 1.59 for fixed investment, 1.35 for government and 1.12 for net exports. Inventories are a small part of the economy, yet they swing the most. They also tend to reverse. In the 53 quarters since 1947 when inventories added 2 points or more, the next quarter's inventory contribution was negative 60.38% of the time, averaging −0.52 points.

So two quarters that both come in near 5% can mean very different things, depending on whether consumption and business investment carried them or inventories did.

Why "a strong economy" makes people nervous

When inflation is the main worry, markets read good news as "the Fed will act." The previous article on Treasury yields measured this. The correlation between monthly Nasdaq returns and monthly changes in the 10-year yield was +0.32 in the low-inflation years of 2000–2020 and −0.48 from 2021 onward. A rising yield stopped meaning "the economy is doing well" and started meaning "rate hikes are coming." Warsh's September 16 opening remarks mentioned both that "economic activity is expanding at a solid pace" and that "inflation remains elevated." A strong economy was itself part of the case for the hike.

3. Has the AI boom started? Look at it in two halves

A new technology shows up in GDP twice.

First, when it is being built: companies buy servers, chips and software and build data centers. This is the demand side. Spending rises, which lifts GDP in the short run and also pushes up prices and interest rates.

Second, when it is being used: the same workers produce more. This is the supply side. Productivity rises, and the economy grows faster without pushing up prices.

The two push inflation in opposite directions, and several years usually separate them. So "has the AI boom started?" needs two answers.

The spending half: already a record

Investment in information processing equipment and software reached 4.96% of GDP in the second quarter, the highest on record since the BEA series begins in 1947 and above the dot-com peak of 2000. At an annual rate, this spending has grown from $857.42 billion in 2019 Q4 to $1,611.52 billion.

Data-center construction is a more direct measure. It was $9.23 billion in 2020 and $40.83 billion in 2025, ×4.42 in five years. In the second quarter of this year it ran at a $53.67 billion annual rate.

It shows up in growth, too. Over the past four quarters (2025 Q3 to 2026 Q2), IT equipment and software contributed an average of 0.70 percentage points a quarter, while real GDP grew at an average annualized rate of 2.12%. That is about 32.8% of growth. From 1996 to 2000 the same category contributed 0.71 points, but real GDP was growing at an average annualized rate of 4.32%, so its share was only 16.4%.

The push is about the size it was in the internet era. It stands out because the rest of the economy is growing far more slowly than it did then.

At Jackson Hole in August, Warsh also said that more than half of this year's growth in business capital spending can likely be attributed to the AI buildout.

One discount applies. Much of this equipment is imported, and imports are subtracted in the GDP accounts. The contributions above are gross, so the net effect is smaller. The BEA does not publish that net figure directly, and this article does not estimate it.

Finding one: the demand-side AI boom has started, and it is large by historical standards. Its direction matters, though: this half pushes interest rates up, not down. When Warsh listed the reasons long-term yields have risen this year, his second was "competition for capital," with the large cloud providers raising money in the markets.

The productivity half: not yet clear

Here is nonfarm business labor productivity (output per hour) by era:

(Annualized. The era boundaries are this article's choice. The first two are business-cycle peaks. 1995 and 2004 are the breakpoints commonly used in the literature for the IT productivity acceleration and slowdown. 2019 Q4 is the last pre-pandemic quarter.)

Recent productivity really is better than in the 2010s: 2.51% annualized over the last 12 quarters, 2.16% over 8 and 2.24% over 4, all above the 1.55% of 2004–2019. Warsh's September 16 opening remarks also stated that "productivity growth is strong."

But it has not reached the internet era's 3.12%. And the pickup began in 2023. How much of it comes from AI, and how much from other forces such as the post-pandemic reallocation of workers, is unsettled, and this article's data cannot separate them.

More importantly, history says that a productivity boom is nearly invisible while it is happening.

• In 1987 Robert Solow, who won the Nobel Prize in economics that year, wrote a line that is still widely quoted: "You can see the computer age everywhere but in the productivity statistics" (Solow, 1987).
• In 1990 the economic historian Paul David showed that electricity took decades to show up clearly in factory productivity. Factories first had to rebuild their production lines around the electric motor before the technology paid off (David, 1990).
• The best example is the 1990s themselves. As a July 2026 St. Louis Fed review recounts, the data published at the time showed productivity slowing, from 3.4% in 1995 Q3 to 0.4% in 1996 Q3. The revised data show the opposite over the same period, an acceleration from 1.6% to 2.5% (Kliesen & Wheelock, 2026).

Finding two: the supply-side AI boom may have started or may not have, and current data cannot tell us which. Even if it has, experience says we usually learn so only after several rounds of revisions. Two quarters of GDP cannot answer the question.

4. The yardsticks really do move: the Fed's own "normal"

That standards change over time does not have to be inferred. Every quarter, FOMC participants publish their estimates of "normal" in the longer run: where the economy settles when there are no shocks and policy is neither tight nor loose. Here they are side by side:

(Real rate = longer-run federal funds rate − 2%.)

The first three rows have all moved, and by a lot:

• The "normal" interest rate fell from 4.3% to 2.4%, then came back to 3.2%. What participants considered normal in 2012 had been revised down to 2.5% by 2019. Within about fifteen years, one institution rewrote "normal" twice, first down and then up.
• "Normal" growth stayed between 1.8% and 1.9% from September 2016 to December 2025, then returned to 2.0% in March of this year. The projections come without reasons, and this article does not guess at them.
• The "normal" unemployment rate was revised steadily down from 5.0% to 4.0%. The CBO now estimates the natural rate in 1996 at 5.29%, while the St. Louis Fed review cited above reports that Fed staff in July 1996 put it at 5.5% to 6%. Even the "normal" of a given year gets rewritten after the fact.

Only the last row has never moved: 2.0% at all 45 meetings since the series began in June 2015.

The yardsticks for interest rates, growth and employment move. The yardstick for inflation does not. That is not an accident but a design choice, and Section 6 comes back to it.

5. The interest-rate yardstick: high or low?

Start with the intuition: compared with 30 or 40 years ago, rates today are low.

(Real fed funds = effective fed funds rate − core PCE inflation, year over year. "Now" is the September 21 rate of 3.88% minus July core PCE inflation of 3.34%. The last column is in percentage points.)

The intuition is right, and not only in nominal terms. The fed funds rate of 3.88% is below the 9.97% of the 1980s and the 5.15% of the 1990s. Net of inflation, today's real rate of 0.54% is also below the 1980s (4.68%) and the 1990s (2.74%). It is even below the Fed's own estimate of the longer-run real neutral rate, 1.2%. That is why Warsh, repeating what he had said at Jackson Hole, told the press conference he would be "hard-pressed to describe broad financial conditions as restrictive."

So why the nerves? Because most people are not comparing with the 1990s. They are comparing with the 2010s. In that decade the fed funds rate averaged 0.61% and the real rate −0.97%. Many of today's mortgages, corporate debt structures and equity valuations were set during that period (as the yields article measured, the 10-year averaged just 2.28% from 2009 to 2021). For those balance sheets, the 2010s are the relevant yardstick.

Is there a yardstick that works across eras?

One common approach compares interest rates with the economy's nominal growth: the cost of borrowing against the pace at which incomes grow. That is the last column above, the 10-year yield minus year-over-year nominal GDP growth.

By that measure, the second quarter of 2026 stands at −2.14 points. Even with the 10-year touching 5%, it is still below the economy's nominal growth rate.

This yardstick has a trap, though. The decades when r − g was most negative, the 1960s and 1970s, were the decades when inflation was taking off. Nominal growth was high partly because prices were rising. Rates that fail to keep up with prices signal that inflation is eroding lenders' money, not that rates are low and safe. In this year's second quarter, nominal GDP was up 6.56% year over year, and prices alone were up 4.36%.

So the question is not only "are today's rates high or low by historical standards?" It is also "which stretch of history are you measuring against, and what was inflation doing then?" A nominal rate is roughly the real neutral rate plus expected inflation. Rates were high in the 1980s because inflation was high. They were low in the 2010s because both the real neutral rate and inflation were low.

6. Can the inflation yardstick move too?

This is the most interesting question. If the standard for interest rates changes over time, should the standard for inflation change as well? Should 3%, say, count as "normal" today?

Both sides have serious arguments.

The case for moving it

One: 2% is a very recent convention. The first central bank to adopt an explicit inflation target was New Zealand's, in 1990. The Fed formally announced its 2% goal in January 2012. Before that, "normal inflation" differed under each monetary regime. From 1900 to 1932, mostly under the gold standard, US inflation averaged 1.93% a year, and prices fell in 11 of 33 years. From 1933 onward it averaged 3.45%. Section 7 shows that Britain's "normal" for the whole nineteenth century was zero.

Two: economists have formally proposed higher targets. In 2010 Olivier Blanchard, then chief economist of the International Monetary Fund, and his co-authors raised the idea of a target of about 4%. Interest rates would then also run higher in normal times, leaving central banks further from zero and with more room to cut in a crisis (Blanchard, Dell'Ariccia & Mauro, 2010). The lower the neutral rate, the stronger this argument becomes.

Three: each generation's sense of "normal" already differs. Here is annualized US inflation during the twenty years after each birth cohort turned 20:

Someone born in 1950 averaged over 6% inflation in their first twenty adult years. Someone born in 1980 averaged 2.16%. The first person may find 3% inflation mild; the second finds it high. Research by the economists Ulrike Malmendier and Stefan Nagel shows that people's inflation expectations lean heavily on the inflation they have lived through (Malmendier & Nagel, 2016). The feeling that "3% is high" is partly a generational effect.

The case against

One: the two yardsticks do different jobs. The neutral interest rate is an estimate of a property of the economy itself: the rate that neither stimulates nor restrains it. When the economy changes, the estimate should change, which is exactly what the table in Section 4 shows. An inflation target is a promise, and all of its value comes from not changing. Raise the target while inflation is above it, and everyone learns that the target is negotiable. That is how inflation expectations lost their anchor in the 1970s: markets learned that inflation would never really be brought down (see Section 5 of the yields article).

The Fed recently paid a price of its own. In 2020 it adopted "average inflation targeting," which allowed inflation to run "moderately" above 2% for a time after a period of undershooting. In August 2025 it dropped that approach from its framework. Then-Chair Powell said the inflation that followed had been neither intentional nor moderate.

Two: compounding does not care what era you live in. Whatever generation you belong to, inflation erodes purchasing power by the same formula:

Move the target from 2% to 3%, and over a 30-year retirement the purchasing power of each dollar falls to 41.2% instead of 55.2%. For people on fixed incomes and anyone holding cash or bonds, that difference is real, and "times have changed" does not make it smaller.

Three: the argument about rates cannot be carried over, because the rate yardstick is built on top of the inflation yardstick. A nominal rate is roughly the real neutral rate plus expected inflation. Rates are "lower than in the 1980s" precisely because expected inflation is anchored near 2%, not at 1980s levels. If the inflation yardstick moved up, the rate yardstick would move up with it, and the comfort that "rates are historically low" would disappear. These are not two parallel theories but two layers. The upper layer (rates) can move because the lower layer (inflation expectations) stays put.

Which yardsticks really should change with the times?

Weighing both sides, the more defensible position is this:

• The neutral rate, potential growth and the natural rate of unemployment describe the economy itself. They shift with demographics, technology and global saving, and our estimates should be updated as they do.
• The inflation target is an agreement whose job is to provide a fixed reference point while everything else is changing. It is not unchangeable (New Zealand's original 0–2% target was later raised), but every change costs credibility, and the cost is highest when the change comes while inflation is above target.

The AI debate belongs to the first category. In standard growth theory, faster trend growth means a higher real neutral rate: when an economy has more good investment opportunities, the price of capital rises. So if AI really does put the economy on a faster growth path, the reasonable expectation is that the "normal" interest rate will be higher, not back at 2010s levels. The Fed's longer-run rate estimate has moved in exactly that direction over the past two years, from 2.5% to 3.2%.

In other words, expecting AI to deliver a boom and expecting rates to return to their old lows are expectations that pull against each other.

7. The Industrial Revolution, electricity, the internet: what the great breaks changed

The last question: did the technologies that changed the world also change how people measure the economy?

Start with the longest record. Britain was the birthplace of the Industrial Revolution and is one of the few countries with estimates of GDP per person going back nearly 800 years:

Lesson one: the Industrial Revolution changed the yardstick for growth

Before the Industrial Revolution, living standards took more than four hundred years to double. A person saw no change in a lifetime, and economic growth was not something anyone expected. Afterwards, doubling took about one lifetime (71 years), and after the Second World War, one generation (29 years). The expectation that the economy should grow every year dates from the Industrial Revolution.

Lesson two: it did not change the yardstick for inflation

Prices tell a different story over the same period. Britain restored gold convertibility in 1821. Over the 92 years to 1913, GDP per person grew ×2.48 while the price level fell 4.85%, an average of −0.05% a year. Prices fell in 40 of those 92 years.

After 1945 the picture reversed. By 2025 British prices had risen ×60.63, an average of 5.26% a year, and they fell in only 1 of 80 years.

What changed the inflation yardstick was not the steam engine, electricity or the internet. It was the monetary regime, the move from the gold standard to fiat money. The US record tells the same story: 1.93% average inflation under the gold standard (1900–1932), 3.45% under fiat money (1933–2025).

Technological revolutions, if anything, have held prices down. In the 1920s, as electrification spread and productivity surged, US inflation averaged −0.12% a year. In the internet era (1995–2004), CPI inflation was 2.47%, down from 5.61% in the era before.

Lesson three: misreading the yardstick costs either inflation or growth

The growth yardstick can move down as well as up. After 1973, productivity growth fell from 2.74% to 1.50%. Policymakers kept treating the 1960s growth rate as the economy's "normal" capacity, concluded the economy was far from full capacity, and kept policy too loose. That is one of the leading explanations of the Great Inflation of the 1970s, advanced by the economist Athanasios Orphanides (Orphanides, 2003).

The 1990s went the other way. At the FOMC meeting of September 24, 1996, some members pressed for a hike because unemployment had fallen close to the estimated "natural" rate of the time. Chair Greenspan argued that information technology had raised productivity and persuaded the committee to stand pat, keeping the rate at 5.25%. What followed:

Unemployment fell well below the "natural" rate as estimated at the time, and inflation fell rather than rose.

What separates the two episodes is not whether the yardstick moved but whether policymakers correctly read which way it moved. When Greenspan made his call in 1996, even the data available at the time pointed the other way (Section 3: productivity as first published was slowing).

Lesson four: booms fade

The internet productivity boom was real, but it did not permanently change the yardstick for growth. After 3.12% in 1995–2004 came 1.55% in 2004–2019 (Fernald, 2015). The internet raised the level of the economy; it did not permanently raise its speed. For AI the same question is open. Nobody knows yet whether it will bring a one-time jump or a steeper new path.

8. Back to September 30: what to look at separately

If you want to use next week's data to test the questions in this article, several things are worth looking at separately. This is not a forecast and involves no investment judgment. It is simply a way of reading the data.

1. Look at real growth, nominal growth and prices separately. If "GDP is strong" comes from the nominal figure, check how much of it is prices.
2. Look at the components, not just the total. Consumption and investment in IT equipment and software say more about the trend. Inventories, net exports and government spending often swing widely from quarter to quarter (Section 2).
3. Remember that this is a third estimate. The second-quarter figures will be revised. The first estimate for the third quarter does not arrive until October 29, and until then the GDPNow estimate will keep moving with new data.
4. The supply side of AI shows up in productivity, not GDP. The BLS next publishes productivity on November 5. Warsh has also said that a Fed task force on productivity and jobs expects to report by the end of the year.
5. Any quarter's number may be rewritten years later. As the 1990s showed, the most important trends are often visible only after revision.

9. What this data can and cannot show

It can show: US real GDP growth in the past two quarters was not high, while nominal growth and prices were. Investment in information processing equipment and software has reached a record share of GDP going back to 1947. Labor productivity over the past three years has run above 2004–2019 and below 1995–2004. The Fed's estimates of longer-run interest rates, growth and unemployment have all moved, while its inflation target never has. And for a century after the Industrial Revolution, British incomes rose substantially while prices did not.

It cannot show: whether AI will bring a lasting rise in productivity, or when. It says nothing about the September 30 figures or any later quarter, including whether the third quarter will come in near GDPNow's 5.1%. It cannot say where an inflation target should be set, which is a question of values and trade-offs rather than data. Nor can it show that the recent improvement in productivity comes from AI.

It also cannot show: how any reader should allocate assets. Everything above is historical measurement published for a general audience and identical for every reader.

10. Methodology and known limitations

Every number above comes from scripts/build_growth_study.py in this site's repository, which produces data/growth-study.json. Known limitations, in order of importance:

1. Every figure is today's vintage. GDP and productivity are revised, sometimes by a lot (the 1990s example in Section 3). This article uses the vintage published before the September 30 third estimate. After that date the second-quarter figures, and earlier ones, may differ.
2. Quarterly contributions do not sum exactly to GDP growth. The BEA rounds each contribution separately, and chain-weighting means they were never strictly additive. They are shown as published, without adjustment.
3. "IT equipment + software" is not the same as "AI investment." It is the closest official category, but it also includes ordinary computer and software spending that has nothing to do with AI. Data-center construction is counted separately. The offset from imports is not deducted.
4. The era boundaries are this article's choice. Different boundaries would change the annualized figures but not the ranking of the eras.
5. The real fed funds rate uses realized core PCE inflation, not the inflation markets expected at the time.
6. The CBO's natural rate of unemployment is today's estimate, not the real-time estimate. The real-time figure is taken from the St. Louis Fed review.
7. UK long-run data: GDP per person before 1870 is a reconstruction by economic historians, not official statistics of the time, and the long-run Retail Prices Index before 1947 is spliced from several historical series. Both cover the United Kingdom only.
8. The cohort table is only illustrative. It uses national average inflation; the prices each person actually faced differ by region and spending pattern.
9. The 2026 events (the rate hike, the press conference, the Jackson Hole speech) are taken from the Federal Reserve's published texts and were not calculated from this article's data.
10. GDPNow's 5.1% is a quoted figure from the Atlanta Fed's website (September 17 update), not something this article calculated. Under the bank's terms of use, this article quotes only that single total and does not reproduce its component breakdown. The historical inventory statistics were calculated separately from BEA data.

References

Data sources

• Bureau of Economic Analysis, NIPA flat files (Tables 1.1.2, 1.5.2, 5.3.5, 5.4.5) and release schedule
• Federal Reserve Bank of Atlanta, GDPNow (the third-quarter estimate of 2026-09-17 is quoted here; GDPNow is not an official forecast of the Atlanta Fed)
• FRED: real GDP (GDPC1), nonfarm business labor productivity (OPHNFB), unemployment rate (UNRATE), CPI (CPIAUCSL), core CPI (CPILFESL), PCE price index (PCEPI), core PCE (PCEPILFE)
• FRED: effective federal funds rate (DFF), monthly (FEDFUNDS), target range (DFEDTARU/DFEDTARL), 10-year Treasury (DGS10)
• FRED, longer-run medians from the FOMC Summary of Economic Projections: federal funds rate (FEDTARMDLR), real GDP growth (GDPC1MDLR), unemployment rate (UNRATEMDLR), PCE inflation (PCECTPIMDLR)
• FRED: CBO noncyclical rate of unemployment (NROU)
• Office for National Statistics, Retail Prices Index: Long run series (CDKO), Open Government Licence v3.0
• Jutta Bolt & Jan Luiten van Zanden, "Maddison style estimates of the evolution of the world economy: A new 2023 update," Journal of Economic Surveys (2024), CC BY 4.0. Original sources for the UK series used here:
  • S. N. Broadberry, B. Campbell, A. Klein, M. Overton & B. van Leeuwen, British Economic Growth 1270–1870, Cambridge University Press, 2015
  • C. H. Feinstein, National Income, Expenditure and Output of the United Kingdom 1855–1965, Cambridge University Press, 1972
  • Angus Maddison's original database (1950–2008, last updated 2010); The Conference Board, Total Economy Database (2008–2022)

Federal Reserve documents

• Transcript of Chair Warsh's press conference, September 16, 2026 (preliminary)
• Kevin Warsh, keynote remarks at the 2026 Jackson Hole Economic Policy Symposium, 2026-08-28
• 2025 Statement on Longer-Run Goals and Monetary Policy Strategy; Jerome Powell, Jackson Hole remarks, August 22, 2025
• Minutes of the Federal Open Market Committee, September 24, 1996

Literature

• Olivier Blanchard, Giovanni Dell'Ariccia & Paolo Mauro, "Rethinking Macroeconomic Policy," IMF Staff Position Note SPN/10/03, 2010
• Paul A. David, "The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox," American Economic Review 80(2), 1990
• John G. Fernald, "Productivity and Potential Output before, during, and after the Great Recession," NBER Macroeconomics Annual 29, 2015
• Kevin L. Kliesen & David C. Wheelock, "Guiding Monetary Policy during a Productivity Boom: Evidence from the 1990s," Federal Reserve Bank of St. Louis, On the Economy, 2026-07-13
• Ulrike Malmendier & Stefan Nagel, "Learning from Inflation Experiences," Quarterly Journal of Economics 131(1), 2016
• Athanasios Orphanides, "The Quest for Prosperity without Inflation," Journal of Monetary Economics 50(3), 2003
• Robert M. Solow, "We'd Better Watch Out," New York Times Book Review, 1987-07-12

Related articles on this site

• Is 5% High? US Treasury Yields Since 1962
• The Break-Even Line: 126 Years of US Inflation

Disclaimer

This article is published by Elnath Finance Academy for general information and educational purposes only. It is written for all readers and is identical for every reader. It is not investment advice, not a recommendation to any individual, and not an offer or solicitation to buy or sell any security. Elnath Finance Academy is not a registered investment adviser (RIA).

All figures are historical measurements calculated from the public data sources listed above, or quoted from texts published by the Federal Reserve and other institutions. They describe the past and published facts. They are not forecasts of growth, inflation, interest rates or any market, and they guarantee no future result. Economic data are revised, and the vintage used here may differ from figures published later.

Investing involves risk, including the possible loss of all principal. Consult a qualified professional where appropriate.