The most cited credit indicator, and the history you cannot download

Charts of high yield credit spreads across 2008 and 2020 are everywhere, and on 13 August 2026 the public file behind them served three years and no more. Asked explicitly for 1996 onward, it returned 786 daily values beginning in August 2023, and the page that would explain why refused two different clients. So this page does something narrower and checkable: it measures the window that was served, in full, and then names three long, fully documented series that measure related things and can be read back for decades.

What was asked for, and what came back

On 13 August 2026 the CSV endpoint of the Federal Reserve Bank of St. Louis was asked for series BAMLH0A0HYM2 with an explicit start date of 1 January 1996. It returned 786 daily values, the first dated 14 August 2023— three years to the day before the request. The investment-grade series BAMLC0A0CM returned 785 values from the same start date. Requesting them without a start date gave exactly the same result.

For comparison, requested in the same session and from the same endpoint: the ten-year minus two-year Treasury spread returned 13,097 rows back to 1976, the unemployment rate 942 rows back to 1948, and the consumer price index 1,362 rows back to 1913. The truncation is specific to these two series, not a property of the service.

The obvious next step is to read the series page, which would state the observation range and any redistribution condition. It could not be read. A request with a browser user agent over HTTP/1.1 timed out; a second client returned 403. So this page states the finding and stops there: the file served three years, and the explanation printed on the page was not obtained. What it does not do is guess at a reason, or fetch the missing history from a secondary site that republishes it. A gap declared is information; a gap filled from an unnamed source is a risk dressed as completeness.

The three years that were served, measured in full

Within its own window the file is complete and consistent, and there is no reason not to measure it exactly. Every figure below is computed over all 786 and 785 observations, not a sample.

The whole three-year range of the high yield series is 2.02 points, from 2.59 to 4.61. Its mean of 3.157 sits above its median of 3.090, which is what a series spends most of its time low and occasionally spikes looks like: only 50 of the 786 days, or 6.4%, printed at or above 4.00 points. The 5th, 25th, 50th, 75th and 95th percentiles of the window are 2.67, 2.83, 3.09, 3.32 and 4.04.

The relationship between the two series is more informative than either alone, because it isolates the part that depends on credit quality rather than on the general level of yields. Over the 785 days present in both files:

The two do not move by the same amount. Between the narrowest and widest days the high yield series moved 2.02 points while the investment grade series moved 0.39 — a ratio of about five to one. And the two extremes of the difference are less than three months apart, on 22 January and 7 April 2025. Whatever else these series measure, they can travel most of their observed range inside a quarter.

Three long series that are fully readable

The honest response to a truncated file is not to substitute a similar-looking one and pretend it is the same measurement. It is to say what else exists, what each one actually measures, and how far back it goes. These three are published by US public bodies, carry decades of history, and were served in full on the same afternoon.

None of these is a credit spread and none is offered as a substitute for one. Two are composite indices constructed so that zero is an average condition, which means a reading is only interpretable against the distribution of the series itself — hence the percentiles above rather than a verdict. The third is a delinquency rate, an actual count of loans in arrears, and its range says more about credit conditions than any index does: it ran from 1.41% at the end of 2004 to 11.48% in the first quarter of 2010, a factor of eight, and stood at 1.89% in the first quarter of 2026.

Neither composite index carries a published threshold in anything read here. A reading of −0.549 is not “loose” and −0.7709 is not “calm”; they are numbers whose meaning comes from where they sit in their own history, which is why this page gives the percentile and the extremes rather than an adjective.

What can and cannot be said about spreads and recessions

The claim that credit spreads widen ahead of recessions is testable in principle and was not testable here. The National Bureau of Economic Research chronology has no peak after February 2020, and the file served begins in August 2023. There is no recession inside the window, so there is nothing to measure a lead against.

It is worth being precise about what that does and does not imply. It does not mean the claim is false. It means that this page cannot check it from the sources it could open, and therefore does not assert it. The same discipline applies to every figure that circulates for these series across 2008 and 2020: those numbers may well be right, and they did not come from the file this page could read.

There is also a definitional limit worth stating plainly. The term “option-adjusted spread” appears in the identifiers, and its precise construction — which bonds are in the index, how they are weighted, how the embedded options are valued, how often the constituents change — is set out in a methodology document that was not opened. A page that explains a measure it has not read is describing its own assumptions.

Four rules for using any spread series

• Check the first observation date before the chart. Two series requested from 1996 began in 2023, and nothing in the file said so.
• Never fill a gap from a secondary site. If the primary source will not serve the history, the honest output is a shorter chart with a stated window.
• Give the percentile, not the adjective. A composite index reading means nothing without the distribution it sits in, and neither of the two here publishes a threshold.
• Keep the two grades apart. Over the same 785 days the high yield series moved five times as far as the investment grade one.

What these sources do not say

• The index methodology was not read. No definition of the option adjustment, no constituent rule and no weighting scheme appears on this page.
• The window served contains no recession. Nothing here tests what spreads do in a downturn, and no lead time is claimed.
• Neither financial conditions index publishes a threshold. A reading is reported with its percentile and the observed extremes, and is not translated into a verdict.
• No default rate, recovery rate or issuance figure. None was read for this page.
• Nothing about credit outside the United States. No European, sterling or emerging market index was opened.
• No statement about equities or any other asset. No price series was read for this page.

How to verify this yourself

1. Request fredgraph.csv?id=BAMLH0A0HYM2&cosd=1996-01-01 and look at the first row after the header.
2. Request the same file with no start date. The answer should be identical.
3. Request fredgraph.csv?id=T10Y2Y in the same session and count the rows. The truncation is not a property of the service.
4. Count the rows in the high yield file at or above 4.00. You should get 50 out of 786.
5. Join the two spread files on date, subtract, and find the largest and smallest differences. They fall on 7 April 2025 and 22 January 2025.
6. Download NFCI and count how many of its 2,901 values are below the latest one. That share is the percentile quoted above.

Where to go next on this site

• Recession indicators — the same window problem set beside three indicators that do have a testable history.
• Interest rates and asset prices — the Treasury yields these spreads are measured against, read in full.
• The yield curve — a spread that is published back to 1976, and what fifty years of it show.
• All macroeconomics guides — the area index, one click away.

Sources opened for this page

Each source below was opened and read on 13 August 2026. Verified on 13 August 2026.

• Federal Reserve Bank of St. Louis — FRED series BAMLH0A0HYM2 and BAMLC0A0CM, requested with and without an explicit 1996 start dateUsed for: the 786 and 785 daily values beginning 14 August 2023, the latest readings of 2.72 and 0.79 for 11 August 2026, the extremes of 2.59 and 4.61 and of 0.73 and 1.33, the mean of 3.157 and median of 3.090, the 50 days at or above 4.00, the percentiles 2.67, 2.83, 3.09, 3.32 and 4.04, and the widest and narrowest differences of 3.41 and 1.78. Read 13 August 2026.
• Federal Reserve Bank of St. Louis — FRED series page for BAMLH0A0HYM2, two attemptsUsed for: the record that on 13 August 2026 one client timed out on this page and another received HTTP 403, so the stated observation range, index methodology and any redistribution condition were not read and are not reported. Read 13 August 2026.
• Federal Reserve Bank of St. Louis — FRED series NFCI and STLFSI4Used for: the 2,901 weekly values from 8 January 1971 with a range from minus 1.097 in August 1993 to plus 5.222 in July 1974 and a latest of minus 0.549 at the 32nd percentile, and the 1,702 weekly values from 31 December 1993 with a range to plus 9.6766 on 10 October 2008 and a latest of minus 0.7709 at the 10th percentile. Read 13 August 2026.
• Federal Reserve Bank of St. Louis — FRED series DRSFRMACBS, delinquency rate on single-family residential mortgagesUsed for: the 141 quarterly values from the first quarter of 1991, the range from 1.41% in the fourth quarter of 2004 to 11.48% in the first quarter of 2010, and the reading of 1.89% for the first quarter of 2026. Read 13 August 2026.
• Federal Reserve Bank of St. Louis — FRED series T10Y2Y, UNRATE and CPIAUCNS, requested in the same sessionUsed for: the 13,097, 942 and 1,362 rows returned for those series, used to establish that the three-year truncation is specific to the two spread series and not a property of the endpoint. Read 13 August 2026.

Change log

• 13 August 2026— page rewritten from scratch against the two FRED spread files, two attempts at the FRED series page, and four other FRED series read in the same session, all on that day. Removed from the previous version: typical spread ranges for investment grade and high yield given with no source and no period; named crisis peaks of 2,182, 1,928 and 1,116 basis points that the file available cannot support; thresholds such as 600 or 700 basis points presented as signals, published by nobody; a definition of the option-adjusted spread taken from no document; a description of CDX indices and their daily notional, read nowhere; and explicit instructions to reduce or add risk at given spread levels, which is advice this site does not give.

Vextor Capital is not authorised under MiFID II as an investment firm, is not registered with the SEC, FINRA, the CFTC or the NFA, is not an investment adviser, holds no position in any instrument named on this page, and receives no payment from any bank, broker, exchange, index provider or data vendor. Every outbound link on this page points to a statistical agency, a central bank or a public data series.