Series

This page documents seven maintained trends and four commentary ratios. Each builder transforms or combines upstream observations and returns a Series with its source records and display metadata. A build writes series/<id>.json and series/<id>.csv under the output directory. See Output for the file format and Sources and licences for reuse terms.

The trend registry contains sp500-in-gold, btc-in-gold, buffett-indicator, federal-deficit, federal-deficit-gdp, federal-deficit-monthly, and federal-deficit-ttm. The other four ratios belong to commentary article groups and are emitted under dist-commentary/. The daily Treasury commentary pair is documented with its article group.

ID	Measures	Frequency	History starts¹	Scale	Catalogue
sp500-in-gold	S&P Composite index relative to gold	monthly	1871-01	linear	Trends
btc-in-gold	Gold equivalent of one bitcoin	monthly	2010-09	log	Trends
buffett-indicator	Nonfinancial corporate equity value / GDP	quarterly	1947 Q4	linear	Trends
federal-deficit	Federal budget deficit in current dollars	annual (fiscal year)	FY1901	linear	Trends
federal-deficit-gdp	Federal budget deficit / fiscal-year GDP	annual (fiscal year)	FY1930	linear	Trends
federal-deficit-monthly	Actual monthly federal budget deficit in current dollars	monthly	1980-10	linear	Trends
federal-deficit-ttm	Trailing 12-month federal budget deficit in current dollars	monthly	1981-09	linear	Trends
corporate-profit-share	After-tax corporate profits / GDP	quarterly	1947 Q1	linear	Commentary
market-value-per-dollar-of-profit	Nonfinancial corporate equity value / after-tax profits	quarterly	1947 Q4	linear	Commentary
federal-deficit-share	Federal current expenditures minus receipts / GDP	quarterly	1947 Q1	linear	Commentary
effective-tariff-rate	Customs duties / goods imports	quarterly	1992 Q1	linear	Commentary

¹ Start dates reflect the checked-in indexes. Maintained trends use the history available in their inputs; commentary datasets select observations within their configured or overridden date windows. Read firstDate, lastDate, and observationCount from the generated JSON for the actual coverage.

sp500-in-gold

value = monthly S&P Composite index level / gold price in USD per troy ounce

The two inputs come from datasets/s-and-p-500 and datasets/gold-prices. The ratio shows equities moving relative to gold. It uses the index price level; dividends are not reinvested, and it does not represent the return of a purchased share or fund.

The equity package uses Shiller's historical series through June 2023 and extends it with monthly averages of FRED's daily S&P 500 prices. This repository reads the package's SP500 column directly. See the upstream source description and extension script.

Gold is a monthly average from 1960 onward. For 1833–1959, the upstream monthly file repeats each year's annual average in all twelve months. The pre-1960 ratio therefore combines monthly equities with annual gold values; it does not reveal within-year gold movements. See the gold package's construction notes.

Only matching dates with positive prices survive. Values are rounded to four decimal places; precision: 2 requests two decimal places for display. The published unit is ounces of gold, with the index-level interpretation above.

Implementation: src/market_trends/trends/sp500_in_gold.py.

btc-in-gold

value = monthly bitcoin price in USD / monthly gold price in USD per troy ounce

This measures how many troy ounces of gold have the same quoted value as one bitcoin. The source adapter groups blockchain.com timestamps by UTC calendar month, discards nonpositive prices, and takes an arithmetic mean of the remaining readings. A month needs at least 20 readings to qualify. This is a coverage threshold: a month can qualify before it has ended, and the adapter does not check for one reading per distinct day.

Bitcoin monthly averages are rounded to four decimal places before division. The ratio is rounded to six decimal places and carries precision: 2 and scale: log. The extra stored decimals preserve small early values even when fixed two-decimal formatting would show 0.00; consumers may need more digits in tooltips. Only months present in both inputs are emitted.

The generated Bitcoin JSON and CSV are git-ignored because the source has no open licence recorded. A full build still writes them locally. See Bitcoin data before distributing them.

Implementation: src/market_trends/trends/btc_in_gold.py.

buffett-indicator

value = (NCBEILQ027S / 1,000) / GDP × 100

NCBEILQ027S is nonfinancial corporate equity value in millions of dollars; division by 1,000 converts it to the billions used by GDP. The numerator is a quarter-end stock, while GDP is a quarterly flow expressed at a seasonally adjusted annual rate. The result is a percentage of annualised GDP, not of spending during that quarter alone. See the FRED definitions for NCBEILQ027S and GDP.

This construction includes publicly traded and closely held nonfinancial corporations. It excludes financial corporations, so its coverage differs from a total-market stock index. Neither the level nor the path should be assumed identical to another chart labelled “Buffett indicator.” The Federal Reserve's series breakdown describes the equity components.

Implementation: src/market_trends/trends/buffett_indicator.py.

federal-deficit

value = −FYFSD / 1,000

OMB's FYFSD, hosted by FRED, records federal budget surpluses as positive and deficits as negative, in millions of current dollars. Negation makes deficits positive and surpluses negative; division by 1,000 converts to billions. Stored values preserve million-dollar detail with three decimals, and precision: 1 requests one decimal place for display. These are nominal dollars, without inflation adjustment. An annual deficit is a flow, not the stock of federal debt.

The history begins in FY1901. Both annual budget series use historical actuals from their source; neither adds budget projections or a partial current year. Observation dates remain fiscal-year ends: June 30 through FY1976 and September 30 from FY1977. The separate July–September 1976 transition quarter is excluded from these annual inputs. The optional dateBasis: "fiscal-year" metadata tells consumers to label observations as fiscal years. See the OMB Historical Tables introduction for the fiscal-year convention and historical-table definitions.

These unified-budget balances differ from the BEA national-accounts current expenditure/receipt gap in commentary's federal-deficit-share.

Implementation: src/market_trends/trends/federal_deficit.py.

federal-deficit-gdp

value = −FYFSDFYGDP

FYFSDFYGDP is OMB's federal budget surplus or deficit as a percentage of fiscal-year GDP. Negate the published percentage directly, keeping its stored precision, so deficits are positive and surpluses are negative. The result has precision: 1 and begins in FY1930; no values are inferred before the source starts. This is a separate history from the longer dollar series, not a join with quarterly GDP.

Do not substitute the similarly named FRED FYFSGDA188S: its denominator is calendar-year GDP. Fiscal-year dates, actuals coverage, and the omitted 1976 transition quarter follow federal-deficit. Both series use a linear scale so budget surpluses can appear below zero.

Implementation: src/market_trends/trends/federal_deficit.py.

federal-deficit-monthly

value = −MTSDS133FMS / 1,000

Treasury's Monthly Treasury Statement, hosted by FRED as MTSDS133FMS, records the actual federal surplus or deficit for each month in millions of current dollars. Negate and divide by 1,000 to express deficits as positive and surpluses as negative, in billions. Stored values are rounded to six decimals; precision: 1 requests one decimal for display. This is the nominal cash balance of the unified budget, without inflation or seasonal adjustment.

The history begins in October 1980. Dates use the first day of the reference month and have no fiscal-year dateBasis metadata. Each observation is that month's actual balance, not an annualized rate or an interpolated annual value. Missing months are not filled. Completed months remain valid even when the fiscal year is incomplete, so this series can extend beyond the annual actuals. Tax-payment deadlines and payment timing contribute to substantial month-to-month changes, including seasonal surpluses.

Monthly Treasury Statement sums can differ from OMB's revised annual history because source vintages and accounting adjustments differ. For example, the September 2026 snapshot's FY2025 monthly sum is $1,775.711461 billion, while FYFSD reports $1,774.684 billion. Each series preserves its own source; the builder does not force the monthly values to reconcile to the annual total.

Implementation: src/market_trends/trends/federal_deficit.py.

federal-deficit-ttm

value(t) = sum(federal-deficit-monthly for months t−11 through t)

The trailing-year total sums twelve consecutive signed observations from federal-deficit-monthly, including the current reference month. It preserves the nominal billions-of-dollars units and source attribution. math.fsum computes each total, then the builder rounds to six decimals; precision: 1 requests one decimal for display. Deficits remain positive and surpluses negative, on a linear scale.

Each point covers a complete twelve-month span, recalculated monthly. Initial windows with fewer than twelve months are omitted. Windows crossing a missing month are also omitted until twelve consecutive observations are available again; twelve rows spread over thirteen calendar months are not a valid year. The calculation uses full history before any chart range filtering and never uses future months. Dates label the final reference month, with no fiscal-year dateBasis metadata. The first complete window ends in September 1981.

This smooths the repeating calendar cycle by including each calendar month once. It is a trailing-year total, not an official seasonally adjusted monthly estimate. It can lag turning points, and exceptional payment shifts can remain visible. As with the underlying monthly data, source vintages and accounting adjustments mean a twelve-month Treasury sum can differ from OMB's annual history, even when the period ends in September.

Implementation: src/market_trends/trends/federal_deficit.py.

corporate-profit-share

value = CP / GDP × 100

Both inputs are billions of dollars at seasonally adjusted annual rates, so no conversion is needed. CP is after-tax corporate profits without inventory valuation and capital consumption adjustments. Use this exact series when reproducing the calculation; other BEA profit measures give different results. See the FRED definition of CP.

This is the profit component of the decomposition below. Its corporate profit coverage is broader than the nonfinancial equity numerator used by the Buffett indicator; the identity is algebraic, not a claim that their sectors match exactly.

Implementation: src/market_trends/commentary/why_the_buffett_indicator_keeps_rising/corporate_profit_share.py.

market-value-per-dollar-of-profit

value = (NCBEILQ027S / 1,000) / CP

The result is dollars of nonfinancial corporate equity value per dollar of annualised after-tax corporate profit. It is an aggregate valuation ratio, not the P/E ratio of an index or a set of companies with matching earnings. It can change when equity values change, profits change, or both.

On dates shared by all three series, using the same source observations:

buffett-indicator ≈ corporate-profit-share × market-value-per-dollar-of-profit

The equality holds before rounding. Both percentage series use percent units, so no additional factor of 100 is needed in this expression. Independently refreshed inputs can break the comparison; use the same cached responses for all three builders.

Implementation: src/market_trends/commentary/why_the_buffett_indicator_keeps_rising/market_value_per_dollar_of_profit.py.

federal-deficit-share

value = (FGEXPND − FGRECPT) / GDP × 100

All inputs are billions of dollars at seasonally adjusted annual rates. The result is positive for a deficit and negative for a surplus. It compares federal current expenditures and receipts on the National Income and Product Accounts basis. It is not the unified federal budget deficit or a fiscal-year total. Definitions: FGEXPND and FGRECPT.

The shared national-accounts basis makes this useful beside the profit share. The comparison does not establish that a change in the federal deficit caused an equal change in corporate profits. A quarterly annual-rate ratio can also differ substantially from the deficit measured across a full fiscal year.

Implementation: src/market_trends/commentary/why_the_buffett_indicator_keeps_rising/federal_deficit_share.py.

effective-tariff-rate

annualised imports = mean(BOPGIMP for the quarter's 3 months) × 12 / 1,000
value = B235RC1Q027SBEA / annualised imports × 100

B235RC1Q027SBEA is customs duties in billions of dollars at a seasonally adjusted annual rate. BOPGIMP is monthly goods imports in millions of dollars, seasonally adjusted. The conversion puts the denominator on the same annual-rate and dollar-unit basis.

The builder groups import rows into calendar quarters and keeps only groups with exactly three observations. It assumes the upstream supplies one row per month; it does not separately verify three distinct months. A retained quarter must also have a duties observation and a positive imports denominator.

This is an aggregate collections-to-imports ratio. Product mix, exemptions, timing, and sourcing changes can make it differ from announced tariff rates; it is not the rate charged on every shipment.

Implementation: src/market_trends/commentary/analyzing_the_effects_of_tariffs_on_prices_and_inflation.py.

Shared calculation and display conventions

• Dates identify periods. Monthly rows use the first day of the month; quarterly rows use January 1, April 1, July 1, or October 1. A period-start label does not imply a first-day measurement: the equity numerator is a quarter-end value. Annual federal budget observations retain fiscal-year ends.
• Joins use exact dates. After any source aggregation, a builder keeps only dates present in every required input. It drops missing dates and nonpositive denominators rather than filling them. Coverage may have gaps, including early equity data with only annual observations.
• Aggregation follows the inputs. Quarterly GDP-based ratios remain quarterly; daily Bitcoin readings become monthly averages; monthly imports become quarterly annualised values. Historical gold already contains annual values repeated into monthly rows upstream, as described above. Annual federal budget balances and the fiscal-year GDP share remain annual. Monthly budget balances remain actual monthly flows, without seasonal adjustment or annualization. The trailing-year series sums twelve consecutive monthly balances for each point.
• Storage and display differ. Most ratios are rounded to four decimal places; Bitcoin and the monthly and trailing-year federal budget balances use six, while the annual federal budget GDP share preserves the upstream decimals and the annual dollar series preserves million-dollar detail. precision requests display decimals; it does not control calculation rounding. The quarterly series request one display decimal and a linear scale.
• Validation has a limited role. It checks series structure and values, not the economic interpretation, source units, complete periods, or release freshness. See Output for the checks and Development for the tests.