Economics · Data dashboard

Main Economic Indicators

Key measures of the US economy: the national accounts, the composition of consumption and investment, capital, prices, the labour market, productivity and interest rates (debt and credit are on the Financial, Credit and Business Indicators pages). Each chart shows the full history available. A trend can be added to compare each series with its long-run path.

Trend
  • Shaded bands on every chart mark US recessions, from the NBER-based quarterly recession indicator (USRECQ).
  • Hollow circles mark values estimated where the source published none; see Missing data and estimates.

Loading data…

National accounts: GDP and its expenditure components

Real GDP and its components in chained 2017 dollars, at seasonally adjusted annual rates. Use the switch on each chart for the per-capita measure (divided by the BEA mid-period population).

Real GDP
Measure

Loading chart…

Real personal consumption expenditures
Measure

Loading chart…

Real gross private domestic investment
Measure

Loading chart…

Real government consumption expenditures and gross investment
Measure

Loading chart…

Trade balance: exports, imports and net exports
Measure

Loading chart…

Real net exports are real exports minus real imports, as defined by BEA (published on FRED from 1970; computed here from 1947 and identical where both exist). They are a difference of chained-dollar series, not a chained index of their own.

Composition of real GDP: output, consumption, investment and government spending
Measure

Loading chart…

Exports and imports are omitted for readability, so the chart gives a semi-closed-economy view of the composition of real GDP. Chained-dollar components are not additive, so consumption, investment and government spending do not sum exactly to GDP.

Shares of GDP

Loading…

Growth rates

Loading…

Real consumption

Real personal consumption expenditures and their three components — durable goods (cars, furniture, appliances), nondurable goods (food, clothing, energy) and services (housing, health care, transport, financial and other services). Spending on durable goods is much more cyclical than the rest, because purchases of long-lasting goods can be postponed.

Composition of real consumption: durable goods, nondurable goods and services
Measure

Loading chart…

All series are in chained 2017 dollars. BEA publishes the components in chained dollars only from 2007; earlier values are built with BEA’s own formula from its quantity indexes, which reproduces the published values exactly where both exist (see Definitions and methods). Chained-dollar components do not add up to the total, and their relative sizes are only meaningful close to 2017; for the composition of consumption over time, use the nominal shares below.

Shares of consumption

Loading…

Growth rates

Loading…

Real investment

Real gross private domestic investment and its three components. Nonresidential fixed investment covers business structures, equipment and intellectual property products; residential fixed investment is mainly housing, which often turns down before recessions; the change in private inventories is small on average but swings sharply over the cycle and accounts for a large part of the volatility of total investment.

Composition of real investment: nonresidential, residential and inventory investment
Measure

Loading chart…

All series are in chained 2017 dollars. BEA publishes nonresidential and residential investment in chained dollars only from 2007; earlier values are built with BEA’s own formula from its quantity indexes, which reproduces the published values exactly where both exist (see Definitions and methods). BEA publishes the real change in private inventories directly from 1947; it can be negative. Chained-dollar components do not add up to the total, and their relative sizes are only meaningful close to 2017; for the composition of investment over time, use the nominal shares below.

Shares of investment

Loading…

Growth rates

Loading…

Capital

The capital input of the US business sector: the flow of services from equipment, software, research and development, artistic originals, structures, business-owned housing, inventories and land, as measured by John Fernald for his quarterly TFP series. It is an index (2017 = 100), not a value in dollars.

Capital input (capital services), business sector

Loading chart…

Index built from Fernald’s quarterly growth rates of capital input, 2017 = 100. Capital services grow smoothly, because the capital stock changes slowly relative to its size; they slow after recessions, when investment falls. Like TFP, the series ends one quarter before the national accounts. See “Capital” under Definitions and methods for how it is measured.

Capital–output ratio (annual, BEA)

Loading chart…

A different capital measure from the chart above, shown to give a sense of scale: the current-cost value of BEA’s year-end net stock of fixed assets divided by nominal GDP of the same year. Both are at current prices, the usual way to measure this ratio. “All private fixed assets” adds residential housing (owner-occupied and rented) to business structures, equipment and intellectual property; government capital and consumer durables are excluded. The ratio is per year of GDP: in a quarterly model the capital–output ratio is four times larger.

Prices

Inflation: GDP deflator and consumer prices
Rate basis

Loading chart…

Labour market

Unemployment rate

Loading chart…

When a trend is selected, this chart shows the Congressional Budget Office’s noncyclical rate of unemployment (NROU) instead of a statistical trend. The NROU is CBO’s estimate of the unemployment rate consistent with stable inflation once cyclical forces have faded, so the gap between the two lines measures labour-market slack. It is a model-based estimate, revised by CBO over time; CBO’s projections beyond the latest observed quarter are not shown. HP and Christiano–Fitzgerald trends of the unemployment rate remain in the downloads.

Labour force participation rate

Loading chart…

Hours worked (nonfarm business)

Loading chart…

Real hourly compensation (nonfarm business)

Loading chart…

Civilian population aged 16 and over

Loading chart…

Productivity

Labour productivity (nonfarm business)

Loading chart…

Total factor productivity (business sector)
Series

Loading chart…

Index levels built from John Fernald’s quarterly TFP growth rates (Federal Reserve Bank of San Francisco), 2017 = 100. Measured TFP is output growth not explained by the growth of capital and labour inputs; it also rises and falls with how intensively capital and labour are used. The utilization-adjusted series removes that effect (Basu, Fernald and Kimball, 2006), so it is closer to technology. Fernald’s data are released about a quarter after the national accounts, so the series ends one quarter earlier.

Interest rates

Nominal interest rates: federal funds rate and 3-month Treasury bill
Rate basis

Loading chart…

Ex-post real interest rates: federal funds rate and 3-month Treasury bill
Rate basis

Loading chart…

Definitions and methods

Series, units and frequencies

All series are shown at a quarterly frequency, the standard frequency for business-cycle analysis and for estimating macroeconomic models. Series published monthly (the unemployment rate, participation rate, population aged 16 and over, consumer price index, federal funds rate and Treasury bill rate) are averaged over the three months of each complete quarter. National-accounts series are at seasonally adjusted annual rates.

Charts, the clean CSV and the MATLAB file start in 1950 Q1. Every calculation — estimates of missing values, trends, growth rates and shares — uses each series’ full published history, and the raw CSV keeps that full history.

Per-capita measures divide each real aggregate by the BEA mid-period population (all ages, B230RC0Q173SBEA) and are in dollars per person.

Hours, labour productivity and real hourly compensation cover the nonfarm business sector, as published by the Bureau of Labor Statistics. Real hourly compensation is deflated by consumer prices. The participation rate is the labour force as a percentage of the civilian noninstitutional population aged 16 and over.

Real components of consumption and investment

The data problem. BEA publishes the components of consumption (durable goods, nondurable goods, services) and of fixed investment (nonresidential, residential) in chained 2017 dollars only from 2007 onwards. Their chain-type quantity and price indexes, and their current-dollar values, are published from 1947.

Construction. BEA defines a chained-dollar series as its chain-type (Fisher) quantity index multiplied by its current-dollar value in the reference year, divided by 100 (BEA, 2023, chapter 4). The components shown here apply exactly this formula from 1947: Xt = Qt × V2017 / 100, where Qt is the quantity index (2017 = 100) and V2017 the 2017 current-dollar value, the average of the four quarterly values at annual rates. From 2007 the result reproduces BEA’s published chained-dollar series to within rounding (about 0.001 percent); applied to total consumption and total investment, which BEA publishes in chained dollars from 1947, it reproduces the published series over the whole period to within 0.01 percent. Every update checks this again and stops if the difference exceeds 0.05 percent. The real change in private inventories is published by BEA in chained dollars from 1947 and is used as published; it has no quantity index because it can be negative.

Why not deflate? Dividing a nominal series by its own price index gives almost the same numbers, but not exactly, because BEA’s quarterly chain-type price index and the implicit price deflator differ slightly (by up to 0.16 percent for residential investment since 2007). The quantity-index formula is the one BEA itself uses.

Units and interpretation. Chained dollars keep the series in billions of dollars at 2017 prices, so the components and the total can be shown on one chart. They are not additive, however: the components do not sum to the total, and the gap widens with the distance from 2017 because relative prices change (BEA, 2023; Whelan, 2002). Real growth rates are unaffected, as each chained-dollar series grows exactly like its quantity index. Relative sizes in chained dollars are only rough approximations, and only close to 2017; BEA recommends current-dollar shares to measure the relative importance of components (BEA, 2023; Landefeld, Moulton and Vojtech, 2003), which is why the share tables report both and the nominal shares should be read first.

Shares and growth rates

Nominal shares divide each current-dollar component by the current-dollar total (GDP, consumption or investment). They add up to 100 (for GDP, with imports entering negatively) and are the measure of relative importance recommended by BEA, because current-dollar values are the weights of the chain-type indexes (BEA, 2023; Landefeld, Moulton and Vojtech, 2003).

Growth rates are annualised quarter-on-quarter log changes, averaged over each period. Real growth uses the chained 2017-dollar series, so it is not affected by the non-additivity of chained components.

Real ratios divide each chained 2017-dollar component by chained 2017-dollar GDP. A nominal share equals the real ratio times a relative price, PCC / (PYY) = (PC/PY)(C/Y), so the two differ whenever relative prices move — most notably for investment and for durable goods, whose relative prices have fallen for decades. Chained components are not additive, so real ratios do not sum to 100 (the “Sum of components” row of the consumption and investment tables shows by how much) and are most reliable close to the 2017 reference year. Whelan (2002) shows that ratios of chain-aggregated real series can be seriously misleading far from the reference year and recommends nominal shares instead.

Inflation and real interest rates

Inflation is the annualised quarter-on-quarter change, 400 × [ln(Pt) − ln(Pt−1)], in percent. For the consumer price index (CPI, all urban consumers), the monthly index is first averaged over each complete quarter.

The two measures cover different baskets. The GDP deflator is the price of everything produced in the US — consumption, investment, government purchases and exports, but not imports — so it matches the output measure, real GDP. This is why it is the usual inflation measure when estimating macroeconomic models, which are often closed-economy (for example Smets and Wouters, 2007). The CPI is the price of a fixed basket of goods and services bought by urban households, including imported goods; it is the measure most people experience and the basis of many indexation rules.

The federal funds rate is the effective rate, a volume-weighted measure of the overnight rates at which banks lend reserve balances to each other without collateral; the Federal Reserve sets a target for it, so it is the main indicator of the monetary policy stance. It is published from July 1954. The 3-month Treasury bill rate is the secondary-market yield on short-term government debt.

Each ex-post real rate is the quarterly average of the monthly nominal rate minus GDP-deflator inflation realised over the following quarter, both at annual rates and in percentage points. It measures the real return actually earned, not the rate expected when the loan was made or the bill was bought.

Annual or quarterly rates. The inflation and interest-rate charts can be shown at an annual rate (the default) or per quarter. Annual rates are the reporting convention: interest rates are quoted per year, and quarterly inflation is annualised, 400 × [ln(Pt) − ln(Pt−1)], so that it can be compared with them. The quarterly rate is the annual rate divided by four: inflation becomes 100 × [ln(Pt) − ln(Pt−1)], the nominal rate becomes the interest earned over one quarter, and the real rate the nominal quarterly rate minus quarterly inflation over the following quarter.

Which to use in a quarterly model. When a DSGE model is estimated or calibrated on quarterly data, its period is one quarter, so every rate must be measured per quarter: inflation as 100 × Δln Pt and the policy rate as the federal funds rate divided by four, as in Smets and Wouters (2007). Mixing units is a common error: if the interest rate is quarterly but inflation is annualised (or the reverse), the estimated policy response to inflation is off by a factor of four; and steady-state values (for example the discount factor implied by the average real rate) and shock standard deviations are only meaningful at the model’s frequency. Dividing by four is the convention used with log-linearised models; the exact quarterly equivalent of an annual rate i is (1 + i)1/4 − 1, which differs little at usual rates.

Capital

Measure. The capital series is the capital input of the US business sector from John Fernald’s quarterly TFP data (Fernald, 2014): the same capital measure that enters his TFP calculation. Fernald builds perpetual-inventory stocks for each type of asset from disaggregated quarterly national-accounts investment, then aggregates their growth rates with weights based on each asset’s estimated user cost (rental price). This measures the flow of productive services from capital, the input to production, rather than the market value of the stock.

Coverage. Business sector, as in Fernald’s TFP: equipment (information-processing and other), software, research and development, artistic originals, nonresidential structures, residential capital owned by businesses (rented housing), inventories and land. General government, nonprofit institutions and owner-occupied housing are outside the business sector.

Levels. Fernald publishes growth rates only (400 × change in the natural logarithm). The index cumulates them exactly, Kt = Kt−1 × exp(gt / 400), with its 2017 average = 100. There is therefore no level in dollars, and ratios such as capital to output are not available from this series.

Why capital services. BEA’s fixed-asset accounts, the official measure of the US capital stock in constant prices, are annual, year-end values; there is no official quarterly real capital stock. Every capital measure, including BEA’s, is built from past investment and assumed depreciation patterns. Fernald’s series applies these at the level of individual asset types and aggregates them by user cost, so it does not impose a single aggregate law of motion, Kt+1 = (1 − δ)Kt + It, on the data. This matters when the data are used to evaluate models whose own capital-accumulation equation (with or without time to build or adjustment costs) is the object of study.

Capital–output ratio. Because Fernald’s series is an index, the capital–output chart uses a different measure: the current-cost value of BEA’s year-end net stock of fixed assets divided by nominal GDP (annual, BEA via FRED). Ratios of current-dollar values are the usual measure, for the reasons given under Shares and growth rates (see also Gomme and Rupert, 2007).

Alternative: a quarterly capital stock

For readers who need a capital stock in constant dollars, a quarterly series is provided for download and compared with Fernald’s capital input below. It is a construction, not official data, and is not used elsewhere on this page. It covers private nonresidential fixed assets, is built by asset type, as BEA builds its own stocks, and starts in 1950 (from BEA’s end-1949 stocks), the first year shown on this page.

  • Benchmarks. BEA’s real year-end net stocks of private equipment, structures and intellectual property products, in chained 2017 dollars: each type’s chain-type quantity index for the net stock times its 2017 current-cost value, divided by 100 (BEA Fixed Assets tables 1.1 and 1.2, lines 5–7, via the BEA API).
  • Quarters, by type. Within each year y, the end-of-quarter stock of each type follows Kq = (1 − δy) Kq−1 + Iq, where Iq is real investment in that type during the quarter (the annual rate divided by four, in chained 2017 dollars) and δy is the one quarterly depreciation rate for which the stock at the end of the fourth quarter equals BEA’s year-end stock of that type. Each type’s annual path is therefore exactly BEA’s; the movements within the year come from actual investment; and δy is the implied quarterly depreciation rate of that type, published in the download (see the next point for how far it can be read as depreciation).
  • Total. The three stocks are combined with a chained Fisher quantity index, as BEA combines its own, using each type’s stock price: BEA’s year-end implicit price (current cost divided by real stock), with the path within the year following the type’s investment price index. The total is expressed in chained 2017 dollars and differs from BEA’s year-end total only because the chain is linked quarterly rather than annually; the download reports BEA’s year-end total for comparison.
  • Why by type. Building one aggregate stock from total investment fails far from 2017: chained-dollar aggregates are not additive (see Real components of consumption and investment), so the aggregate law of motion does not hold in chained dollars and the implied depreciation rate absorbs the gap. Built that way, it fell to about 0.1 percent a year in the late 1940s. Building by type solves this for structures: their implied rates, about 3 percent a year, match BEA’s own depreciation rates (current-cost depreciation divided by the stock) throughout. Equipment and intellectual property products, however, combine assets with very different price trends (computers, software), so the problem remains within each type: their implied rates are close to BEA’s near 2017 (about 13 and 22 percent a year in the 2010s, against BEA’s 14 and 25) but are too low far from it (about 4 and 8 percent a year in the 1950s, against BEA’s 13 and 16 percent). The stocks still equal BEA’s at every year end; only the path within each year and the implied depreciation rates are affected. A finer split of these two types would reduce the gap.
  • Latest quarters. After BEA’s latest year, each type’s last depreciation rate is kept and prices follow investment prices; these quarters are flagged as extrapolated.
  • Related approach. Gomme and Rupert (2007) also build quarterly stocks by type from quarterly investment and the law of motion, using BEA’s annual depreciation rates; they start from BEA’s 1929 stocks and do not force the series back to BEA’s later year-end values.
  • Limits. Within each year, each type follows the simple law of motion with one depreciation rate and no time to build, so the series should not be used to test a model’s own capital-accumulation equation. Its coverage also differs from Fernald’s: it excludes inventories, land and business-owned housing.
Capital: Fernald’s capital input and the quarterly capital stock

Loading chart…

Both indexed to their 2017 average = 100. Capital services usually grow faster than the stock, because the mix of capital has shifted towards short-lived assets (equipment, software), whose services per dollar of stock are larger.

The file gives, for each quarter and each asset type, the end-of-quarter stock and investment (billions of chained 2017 dollars), the implied quarterly depreciation rate and BEA’s year-end stock (fourth quarters); then the total stock, BEA’s year-end total and the extrapolation flag.

Total factor productivity

Source. John Fernald’s quarterly TFP series for the US business sector, published by the Federal Reserve Bank of San Francisco (Fernald, 2014) and updated each quarter. It is the standard quarterly TFP series in US macroeconomics, used for example to identify technology shocks.

Definitions. Measured TFP growth is output growth minus the contributions of capital and labour, Δtfp = Δy − α Δk − (1 − α)(Δh + Δq), where α is capital’s share of income, k capital input, h hours and q labour composition (quality). Utilization-adjusted TFP subtracts an estimate of changes in the utilization of capital and labour (labour effort and the capital workweek), following Basu, Fernald and Kimball (2006). Measured TFP is procyclical partly because utilization is; the adjusted series is a closer measure of technology.

Levels. Fernald publishes growth rates, gt = 400 × [ln(At) − ln(At−1)], in percent at an annual rate. The levels shown here cumulate them exactly, At = At−1 × exp(gt / 400), and are expressed as an index with its 2017 average = 100. Using (1 + gt/400) instead of the exponential would treat log changes as percentage changes and bias the level over long periods.

Timing. Fernald’s release follows the national accounts and productivity data, so the TFP series usually ends one quarter before the other national-accounts series. The latest quarters are revised with each release.

Trends

On charts with several series, each trend is drawn in its series’ colour (dashed for HP, dotted for Christiano–Fitzgerald). Trends are estimated on each series’ full history, before 1950 included.

Trends are statistical estimates, not observed data. For series measured in levels they are estimated on 100 × the natural logarithm and converted back to the series’ own units; for rates (inflation, unemployment, participation, interest rates) they are estimated on the rate itself.

  • HP: Hodrick–Prescott filter, λ = 1600 (quarterly). It is smooth and two-sided, so its latest values are revised as new data arrive.
  • Christiano–Fitzgerald: the series minus its business-cycle component, defined as fluctuations lasting 6–32 quarters. It still contains fluctuations shorter than the band, so it is less smooth than the HP trend; its latest values are also revised.

Unemployment rate: instead of a statistical trend, the chart shows the Congressional Budget Office’s noncyclical rate of unemployment (NROU), the standard US estimate of the natural rate. Statistical filters applied to the unemployment rate tend to follow its large cyclical swings, whereas the unemployment gap relative to the NROU is the usual measure of labour-market slack (for example in Okun’s law, Phillips-curve and policy-rule analysis). HP and Christiano–Fitzgerald trends of the unemployment rate are still computed and included in the downloads.

The Hamilton filter is not offered as a trend here: its fitted value is a forecast made two years earlier, so it repeats past shocks eight quarters later. Hamilton cycles are shown on the Real Business Cycles page, together with the HP and Christiano–Fitzgerald cycles.

Recession shading and data revisions

Shaded bands show US recessions from the NBER-based quarterly recession indicator published on FRED (USRECQ).

Official statistics are revised after their first release. Each update of this page uses the latest available history, so earlier values can change.

References
  • Basu, S., J. G. Fernald and M. S. Kimball (2006). “Are Technology Improvements Contractionary?” American Economic Review 96(5): 1418–1448.
  • Bureau of Economic Analysis (2023). NIPA Handbook: Concepts and Methods of the U.S. National Income and Product Accounts, chapter 4, “Estimating Methods” (updated December 2023). bea.gov
  • Fernald, J. G. (2014). “A Quarterly, Utilization-Adjusted Series on Total Factor Productivity.” Federal Reserve Bank of San Francisco Working Paper 2012-19 (updated March 2014). Data: frbsf.org
  • Gomme, P. and P. Rupert (2007). “Theory, Measurement and Calibration of Macroeconomic Models.” Journal of Monetary Economics 54(2): 460–497.
  • Landefeld, J. S., B. R. Moulton and C. M. Vojtech (2003). “Chained-Dollar Indexes: Issues, Tips on Their Use, and Upcoming Changes.” Survey of Current Business, November 2003. bea.gov
  • Smets, F. and R. Wouters (2007). “Shocks and Frictions in US Business Cycles: A Bayesian DSGE Approach.” American Economic Review 97(3): 586–606.
  • Whelan, K. (2002). “A Guide to U.S. Chain Aggregated NIPA Data.” Review of Income and Wealth 48(2): 217–233. RePEc

Missing data and estimates

Occasionally a source does not publish a value for a period — for example, the October 2025 household survey was not collected because of the federal government shutdown. The economy did not stop in that month, and gaps break time-series and econometric analysis, so such values are estimated here and flagged rather than left blank.

Method. An isolated gap of up to three consecutive periods is filled by linear interpolation between the published values on either side; for series measured in levels the interpolation is in logs, so growth is constant across the gap. Longer gaps are not filled automatically. Estimates in an input also flag the measures built from it (for example, the GDP deflator enters inflation and the real interest rates).

In the downloads. The raw CSV keeps exactly what the source published, including the gap. The clean CSV and MATLAB files contain the estimate, with an estimated column (1 = estimated) so that it can be excluded or treated differently in your own analysis.

Loading…

Page created on 28/10/2026.

© Arthur Galichère.

  • Errors: Every effort has been made to ensure the data and calculations are accurate; any remaining errors are mine. If you believe you have found a mistake, or have a comment or suggestion, please let me know.
  • Reuse and citation: Data are from the agencies named under each chart: via FRED (Federal Reserve Bank of St. Louis); directly from the Bureau of Economic Analysis (fixed assets, through its API); and, for total factor productivity and capital input, from John Fernald at the Federal Reserve Bank of San Francisco. The constructed series, charts and text are the author’s; the source data remain subject to the terms of their providers. When reusing these figures, please cite the original data providers and this website: Arthur Galichère, Main Economic Indicators.