# What does the unemployment rate leave out?

> The unemployment rate counts only people who searched for work in the last four weeks, so people who gave up looking drop out of it entirely.

*EconPlot · https://econplot.com/explainers/what-the-unemployment-rate-misses/ · updated 2026-09-06*

## The answer

The unemployment rate counts a person as unemployed only if they have no job, are available for work, and actively looked for work in the previous four weeks. Anyone without a job who has not searched in that window is outside the labor force and outside the rate entirely, including discouraged workers, who want a job but have stopped looking because they believe none is available. That is why a fall in the rate — from 4.4 percent in February 2026 to 4.1 percent in July 2026, on the Bureau of Labor Statistics series read on 2026-09-03 — can mean people found work, or that people stopped looking.

## The rate counts only people who looked for work in the last four weeks

A person is counted as unemployed only when three things are true at the same time: they have no job, they are available for work, and they actively looked for work in the four weeks ending with the week the survey asks about. That is the definition used by the Bureau of Labor Statistics, the US Department of Labor agency that counts jobs and publishes the monthly Employment Situation report, in its Current Population Survey — the household survey the unemployment rate is calculated from. A person who fails any one of those three conditions is not described by the unemployment rate at all.

The rate is then a fraction with a specific denominator. The labor force is the employed plus the unemployed — everyone working, plus everyone who meets the three-part test above. The unemployment rate is the number of unemployed people divided by that labor force.

That construction rules out several things a reader might expect the number to be. It is not the share of adults without a job, and it is not the share of the population receiving benefits. A person who has no job and has not searched in the last four weeks appears in neither the top nor the bottom of the fraction. Such a person is not counted as unemployed, and is not counted as part of the labor force either. As far as the headline number is concerned, that person is not there.

The timing rule is strict rather than approximate. A search that happened five weeks before the survey week is not a search for this purpose, even though the person doing it would describe themselves, in ordinary language, as looking for work.

## Discouraged workers want a job, and none of them are in the rate

Discouraged workers are the clearest case of what the rate leaves out. The Bureau of Labor Statistics defines them as people not in the labor force who want a job and are available for one, who have looked for work sometime in the past twelve months, but who are not looking now because they believe there are no jobs available or none for which they would qualify. Wanting a job is not enough to be counted. Only the recent search is.

The agency does not ignore these people; it keeps them out of the headline figure and reports them elsewhere. The 4.1 percent published for July 2026 is the measure the agency labels U-3. Discouraged workers are published separately, in the broader U-4 to U-6 alternative measures that appear in table A-15 of the same monthly report. This article does not state a value for U-4, U-5 or U-6, because no such figure is in the record behind this page.

One distinction matters more than it looks. Being outside the labor force is not the same as being discouraged. Retirees, full-time students, people caring for family members and people too ill to work are all outside the labor force too, and none of them are discouraged workers. Retirees, students and carers are not looking because they do not want a job at the moment, not because they have given up on finding one. So a change in the number of people outside the labor force cannot be read straight off as a change in how hard it is to find work: the category holds two kinds of people whose absence means opposite things.

## A falling unemployment rate can mean hiring, or it can mean people quitting the search

A falling unemployment rate has two possible causes, because the rate is the unemployed divided by the labor force and a person who stops searching leaves both the top and the bottom of that fraction. That follows from the arithmetic of the definitions, not from any particular month's data.

- An unemployed person is hired. That person leaves the count of the unemployed, which is the top of the fraction, but stays in the labor force, which is the bottom. The top shrinks, the bottom does not, so the rate falls.
- An unemployed person stops searching for four weeks. That person stops meeting the definition of unemployed and so leaves the top of the fraction — and because the labor force is the employed plus the unemployed, that person leaves the bottom as well. The rate falls again, with nobody hired.

The two cases push the rate in the same direction, so the rate on its own cannot tell them apart. That is the specific blind spot in the number: it is a ratio, and someone leaving the labor force changes both of its parts at once. The agency's own explainer, "How the Government Measures Unemployment", describes this reading of the measure.

It is worth being precise about which direction the ambiguity runs. The rate is a fair description of the people it defines as being in the labor market. What it cannot do is tell a reader how many people the labor market lost — and losing them is one of the ways the number falls.

## Between February and July 2026 the rate fell and participation fell with it

Both series moved down over the same six months. The labor force participation rate, the second series here, is the share of adults who are either working or looking for work, so it falls when people leave the job market altogether. These are the values from the Bureau of Labor Statistics files read on 2026-09-03.

- Unemployment rate: 4.4 percent in February 2026, 4.3 in March, 4.3 in April, 4.3 in May, 4.2 in June, 4.1 in July 2026. No month rose. The total move is 0.3 percentage points down.
- Labor force participation rate: 62.0 percent in February 2026 and 61.4 percent in July 2026, a total move of 0.6 percentage points down, with no month in between higher than the month before it.

Over the same stretch, the count of jobs on employer payrolls barely moved: 158,861 thousand in May 2026, 158,881 thousand in June 2026, and 158,858 thousand in July 2026, so July was slightly below June rather than above it. A burst of hiring large enough to explain a falling unemployment rate would show up as payrolls climbing, and it does not.

So the pattern fits people leaving the search rather than people getting hired. "Fits" is the honest strength of that statement, and it is deliberately weaker than "shows". Participation also declines for reasons that have nothing to do with how hard it is to find work: an ageing population means more people retiring each month, and more people staying in education pushes participation down the same way. Nothing in these two series separates a retirement from a person who gave up, so the size of each contribution is not established here.

## The participation rate is the number that shows whether the labor force shrank

The participation rate shows whether the labor force shrank, because it moves when people enter or leave the labor force — exactly the movement the unemployment rate is blind to. It is the labor force expressed as a percent of the civilian noninstitutional population: people aged sixteen and over who are not living in an institution and are not on active military duty. In plainer terms, it is the share of adults who are either working or looking for work.

That makes the two numbers answer different halves of one question, and reading them next to each other narrows the ambiguity in the previous section:

- A low unemployment rate while participation is rising means people are entering or staying in the labor force and mostly finding work.
- A low unemployment rate while participation is falling is consistent with people giving up the search — but only consistent with it, because retirement and schooling pull participation down too. It can be a weaker labor market than the unemployment rate alone suggests; it does not have to be.

Reading the pair does not turn two ambiguous numbers into one certain answer. It rules out one story rather than proving another: a falling unemployment rate alongside rising participation cannot be a story about people quitting the search, because quitting the search would have pushed participation the other way. When both fall together, as they did between February 2026 and July 2026, the quitting story is still on the table alongside the ordinary demographic ones, and neither series decides between them.

## Every figure on this page

| What it measures | Value | Units | As of | Published by | Series | Data read | Computed from |
|---|---|---|---|---|---|---|---|
| US unemployment rate | 4.1 | percent | 2026-07 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| US unemployment rate | 4.4 | percent | 2026-02 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| US unemployment rate | 4.3 | percent | 2026-03 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| US unemployment rate | 4.3 | percent | 2026-04 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| US unemployment rate | 4.3 | percent | 2026-05 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| US unemployment rate | 4.2 | percent | 2026-06 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| Fall in the unemployment rate from February 2026 to July 2026 | 0.3 | percentage points | 2026-07 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | 4.4 (2026-02) minus 4.1 (2026-07) |
| US unemployment rate a year before the latest reading | 4.3 | percent | 2025-07 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| Lowest unemployment rate in the file, which starts in January 2015 | 3.4 | percent | 2023-04 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| Highest unemployment rate in the file, which starts in January 2015 | 14.8 | percent | 2020-04 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | UNRATE | 2026-09-03 | — |
| Share of adults working or looking for work (labor force participation rate) | 62.0 | percent | 2026-02 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | CIVPART | 2026-09-03 | — |
| Share of adults working or looking for work (labor force participation rate) | 61.4 | percent | 2026-07 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | CIVPART | 2026-09-03 | — |
| Fall in the participation rate from February 2026 to July 2026 | 0.6 | percentage points | 2026-07 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | CIVPART | 2026-09-03 | 62.0 (2026-02) minus 61.4 (2026-07) |
| Jobs on employer payrolls | 158861 | thousands of jobs | 2026-05 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | PAYEMS | 2026-09-03 | — |
| Jobs on employer payrolls | 158881 | thousands of jobs | 2026-06 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | PAYEMS | 2026-09-03 | — |
| Jobs on employer payrolls | 158858 | thousands of jobs | 2026-07 | Bureau of Labor Statistics, retrieved from FRED (Federal Reserve Economic Data) | PAYEMS | 2026-09-03 | — |

## What this page does not establish

- The 4.1 percent figure for July 2026 is the measure the Bureau of Labor Statistics labels U-3, and no count of discouraged workers appears in the record behind this page. A reader can conclude that such people exist and are excluded; a reader cannot conclude from anything here how many of them there are, or whether that group grew between February and July 2026. The broader U-4 to U-6 measures in table A-15 of the monthly report are where those counts live.
- The falling participation rate is consistent with people quitting the search, but a falling participation rate is not proof that people quit. Participation also drops when more people retire or stay in education, and neither the unemployment rate series nor the participation rate series separates a retirement from a discouraged worker. Treating the 0.6 percentage point decline from February 2026 to July 2026 as if it were all people giving up would overstate what these two series establish.
- The payroll figures and the two rates come from different monthly collections, one asking businesses and one asking households, and they count different things — jobs at a business versus people at an address. The flat payroll count and the falling unemployment rate are therefore not two views of the same number, and subtracting one from the other produces nothing meaningful. The payroll figures are used here only as a check on whether a hiring surge is visible, not as a component of the rate.
- These are monthly estimates from a household survey, and the record behind this page contains no margin of error and no revision history for them. A single month's move of one tenth of a percentage point therefore cannot be treated as an established change on the strength of these files alone; the six-month direction, where every month was at or below the one before, is the sturdier reading.
- July 2026 was the latest published month when the data was read on 2026-09-03, because the August 2026 Employment Situation report was not due until the first Friday of September 2026. Anything written here describes the labor market up to July 2026 and no later.

## Questions people also ask

**Does collecting unemployment benefits mean you are counted as unemployed?**

The rate is not built from benefit claims at all. It comes from the Current Population Survey, a household survey, and the test is whether a person had no job, was available for work, and actively looked for work in the previous four weeks. Someone receiving benefits who did not search in that window would fail the test, and someone receiving nothing who did search would pass it.

**What about people working part time who want a full-time job?**

People working part time who want full-time hours are not counted as unemployed in the 4.1 percent figure for July 2026, because that figure counts only people with no job at all. Their situation is picked up instead in the broader alternative measures the Bureau of Labor Statistics publishes in table A-15 of the monthly Employment Situation report, the U-4 to U-6 series. The record behind this page carries no value for those broader measures, so this page does not state one.

**Is 4.1 percent high or low compared with recent history?**

In the file behind this page, which starts in January 2015, the unemployment rate has ranged from a low of 3.4 percent in April 2023 to a high of 14.8 percent in April 2020. The July 2026 reading of 4.1 percent sits near the low end of that range, though above the 2023 low. A year earlier, in July 2025, the rate was 4.3 percent.

**Are people in prison or serving in the military counted?**

No. The participation rate is measured against the civilian noninstitutional population — people aged sixteen and over who are not living in an institution and are not on active military duty. People in prison, in long-term care, and in the armed forces are outside that population, so they are outside both the labor force and the unemployment rate.

**How often does this number come out, and when is the next one?**

The Bureau of Labor Statistics publishes the unemployment rate monthly, in the Employment Situation report, which is released on the first Friday of the month. When the data for this page was read on 2026-09-03, July 2026 was the most recent published month, with the August 2026 report due on the first Friday of September 2026.

**Is someone on temporary layoff who expects to be called back counted as unemployed?**

The three-part test described on this page — no job, available for work, and an active search in the previous four weeks — is the general rule, and the Current Population Survey also has a separate provision covering people on temporary layoff. The evidence record behind this page does not set out the terms of that provision, so this page cannot say how a worker awaiting recall is classified. What it does establish is the general case: with no job and no search in the last four weeks, a person is outside the labor force and outside the rate.

## Sources

- **unemployment rate july 2026** — BLS unemployment rate via FRED, series UNRATE, local file data/UNRATE.csv read 2026-09-03: observation 2026-07-01 = 4.1 percent, the last row in the file (the August 2026 Employment Situation is due the first Friday of September 2026, so July 2026 is the latest published month at render time). Used in the title, cover stamp, stakes narration, scene 3 narration and stamp, and the closing scene ('4.1 alone')
- **unemployment rate february 2026** — Same series UNRATE, observation 2026-02-01 = 4.4 percent. Used in scene 7 narration and headline: 'fell from 4.4 to 4.1 percent'
- **unemployment rate fell** — Same series UNRATE, February 2026 4.4, March 4.3, April 4.3, May 4.3, June 4.2, July 4.1: every month is at or below the month before, so 'fell' over the same months as the participation decline is a direct read of the data. Re-read from data/UNRATE.csv on 2026-09-03
- **four week search definition** — BLS Current Population Survey, 'Concepts and Definitions' (bls.gov/cps/definitions.htm): 'Persons are classified as unemployed if they do not have a job, have actively looked for work in the prior 4 weeks, and are currently available for work.' The narration compresses this to 'counts you as unemployed only if you searched in the last four weeks' and the scene 3 headline to 'Only a job search in the last four weeks counts as unemployed', so the sentence now names what the search counts toward. The 'available for work' condition and the temporary-layoff exception are omitted for length; neither changes the point that no search means not counted as unemployed. Also given in the brief for topic T7
- **bls is the agency that counts jobs** — The Bureau of Labor Statistics is the US Department of Labor agency that publishes the monthly Employment Situation report, which is the source of UNRATE and CIVPART (bls.gov/cps). The one-idea gloss 'the agency that counts jobs' anchors the name for a viewer who has never heard it
- **out of the labor force definition** — BLS Current Population Survey: the labor force is the sum of the employed and the unemployed; a person neither working nor meeting the four-week search test is not in the labor force and so does not appear in the unemployment rate at all.
- **skips people who stopped looking** — Definitional statement, no count claimed: by the definition above, a person with no job who has not searched in the prior 4 weeks is outside the labor force and therefore not in the unemployment rate. The stakes line 'skips people with no job who stopped looking' and the highlight '4.1% skips people who stopped looking' say only that such people exist and are excluded; the earlier word 'many' was removed because no count is recorded here
- **discouraged workers definition** — BLS CPS Concepts and Definitions: discouraged workers are 'persons not in the labor force who want and are available for a job and who have looked for work sometime in the past 12 months ... but who are not currently looking because they believe there are no jobs available or there are none for which they would qualify'. They are counted outside the labor force and therefore outside the unemployment rate; BLS publishes them separately in the U-4 to U-6 alternative measures (table A-15). The narration attaches the term only to the sub-group 'Some of those people [who] want a job but gave up looking', not to everyone outside the labor force (retirees, students and caregivers are outside the labor force but are not discouraged workers). Headline 'Discouraged workers want a job but stopped looking'; stamp 'NOT IN THE RATE'
- **rate can fall for a bad reason** — Arithmetic of the definition: the unemployment rate is unemployed divided by the labor force. A person who stops searching leaves both the numerator and the denominator, which lowers the ratio when that person was unemployed. So the rate can fall with no one hired. This is the standard explanation in BLS's own 'How the Government Measures Unemployment' (bls.gov/cps/cps_htgm.htm) and in the brief for topic T7. Narration keeps it as a mechanism, 'can fall', not a claim about any particular month: 'the rate can fall for a bad reason: fewer people searching, not more hired'
- **participation rate definition** — BLS CPS Concepts and Definitions: the labor force participation rate is 'the labor force as a percent of the civilian noninstitutional population', i.e. people aged 16 and over who are not in institutions or on active military duty. Narration glosses this as 'the share of adults working or looking'; the chart title says the same
- **participation february 2026** — BLS labor force participation rate via FRED, series CIVPART, local file data/CIVPART.csv read 2026-09-03: observation 2026-02-01 = 62.0 percent. Used in scene 6 narration
- **participation july 2026** — Same series CIVPART, observation 2026-07-01 = 61.4 percent, the last row in the file. Used in the cover stamp, chart badge and scene 6 narration
- **participation fell** — Same series CIVPART, February 2026 62.0, March 61.9, April 61.8, May 61.8, June 61.5, July 61.4: no month rose, so 'fell' and 'Participation fell too' are direct reads of the data over the same months as the unemployment decline. Re-read from data/CIVPART.csv on 2026-09-03
- **both fell together** — UNRATE 4.4 to 4.1 and CIVPART 62.0 to 61.4 over the identical span February 2026 to July 2026, both from the local FRED files above. 'Over those months' and 'Both fell' compare these two series over that span and nothing else
- **fits people quitting the search** — Interpretation, stated as 'fits', not as a finding. With payrolls (BLS via FRED PAYEMS, data/PAYEMS.csv read 2026-09-03) at 158,861 thousand in May 2026, 158,881 in June and 158,858 in July, employment was roughly flat and fell by 23 thousand in July, so hiring does not account for the lower rate; a falling participation rate alongside a falling unemployment rate is the textbook signature of people leaving the labor force. The sentence contrasts two explanations, 'people quitting the search rather than people getting hired', so 'getting hired' cannot attach to the people who quit. Caveat kept out of the narration for length: participation also falls as the population ages and people retire, so 'fits' is the honest strength of the claim, and the narration does not say discouraged workers caused all of it. BLS discusses this reading in 'How the Government Measures Unemployment'
- **second number needed** — Stated as a consequence of the arithmetic above ('Hence a second number'), not as a claim about what economists or anyone else does: because a lower unemployment rate can come from people leaving the labor force, the participation rate is the number that shows whether the labor force shrank. The earlier attribution 'economists also watch' was removed so the script makes no claim about other people's practice that would need a named, dated source. The closing headline 'Read the two numbers together' is an editorial instruction to the viewer about reading a statistic, not financial advice
- **strong versus weaker market reading** — Mechanism statement, textbook level, both halves now carry their causal step and their strength. Rising participation means more people entering or staying in the labor force, i.e. looking; a low unemployment rate alongside it means those people are mostly finding work, hence 'more people looking and finding work, a strong market'. A low rate with falling participation is consistent with people giving up, so the narration says 'maybe a weaker market' and the stamp says 'CAN BE WEAKER', not 'means weaker', because participation also falls for reasons unrelated to the job market (retirement, schooling). The comparison is against 'the 4.1 alone', the unemployment rate figure, so the viewer does not hear 'headline' as a news headline. Neither sentence predicts what comes next, asserts the current market is weak, or advises any action
- **chart** — BLS via FRED, series UNRATE, monthly seasonally adjusted, plotted from July 2025 at build time. Values read 2026-09-03: 2026-02 = 4.4, then 4.3, 4.3, 4.3, 4.2, and 2026-07 = 4.1, so 'falling since February 2026' is the file's own path. CORRECTION 2026-09-03: this scene previously plotted CIVPART (participation) under a participation title while the narration talked about the unemployment rate, and an intermediate draft said 'a year earlier the rate was 4.4 percent' — July 2025 was 4.3, so that anchor was wrong and is gone.
- **no advice** — Explanation only: the video explains who is counted and why two numbers are read together; it does not advise on job searching, timing, or any financial decision
- **punchline and length** — Owner rule 2026-09-03 ('we aim for memorable and not for completeness'): the video ends on 'the number can fall because fewer people are looking' and every scene after it was cut. Narration 996 -> 664 characters (77s -> 51s at the measured 12.9 chars/second). No claim was added; cut scenes only removed material.
- **number discipline** — Owner rules 2026-09-03 (at most three numbers, none repeated, at most two significant digits, first slide provokes, last slide is the punch line). Numbers kept: 4.1% (scene 2), 4.4% a year earlier (scene 5, the chart). Dropped the participation series and its 61.4% (three digits); the hook no longer states the rate, and the ending is the reversal the title promises.
- **cover headline length** — Owner catches 2026-09-03: cover headlines were being cut off where they ran under the artwork. The Cover component now sizes the headline to the band above the image, and the headlines themselves were cut to one provoking line.
- **cover is one line** — Owner, 2026-09-03: 'we just need catchy top phrases... nobody can read [the rest] in a fast scrolling environment, so just remove them'. The cover is the headline plus the artwork; the stamps and the bottom kicker bar are no longer rendered at all.
- **no call to action** — Owner catch 2026-09-03: 'Stop looking for work, and you leave the number' reads as an instruction to stop looking for work. The channel explains, it does not advise (STRATEGY.md editorial rules), so the cover and title now describe the mechanism in the third person. The claim is unchanged: someone who has not searched in four weeks is outside the BLS definition of unemployed.
- **two model review round2** — Fixed 2026-09-03 from findings both reviewers raised independently (Claude in the first sweep, DeepSeek in the re-run after its 240s timeouts). 'that number' and 'a number like that' were bare demonstratives pointing across scene cuts; both now name the unemployment rate.
- **doccard retired** — Owner, 2026-09-03: 'why we need doc card? when we already have another card like this'. The doccard scene type rendered a headline on a bare paper card with no artwork at all, which is newscard minus the picture; it was the only text-only type in the renderer and every project had unused card art sitting beside it. DocCard was deleted from remotion/src/MarketShort.tsx and these scenes are newscard now. NewsCard took over its two extras: the kicker renders above the headline, and a highlight that is a whole sentence not lifted from the headline renders as a caption bar over the artwork instead of painting headline words yellow. Not one word of narration or on-screen text changed, so the script fingerprint and its validation still stand.
- **unemployment 4 1 percent july 2026** — BLS via FRED, series UNRATE (civilian unemployment rate, seasonally adjusted, percent), local file data/UNRATE.csv read 2026-09-03: 2026-07-01 = 4.1. Narrated in scene 3 and drawn on the scene 6 chart.
- **unemployment 4 4 percent february 2026** — Same series UNRATE, data/UNRATE.csv read 2026-09-03: 2026-02-01 = 4.4, with 2026-04 = 4.3, 2026-05 = 4.3, 2026-06 = 4.2, 2026-07 = 4.1 — the monthly path behind 'falling since February 2026'. The direction is read straight off those five consecutive prints, not asserted.
- **four week search test** — BLS Current Population Survey definition of unemployment: a person is counted as unemployed only if they were without work, available for work, and had actively searched for work in the four weeks ending with the reference week. Source of the scene 3 and scene 4 statements of the rule.
- **discouraged workers excluded** — BLS Current Population Survey: discouraged workers want a job and are available for work but have stopped searching, so they fall outside the labor force and outside the headline (U-3) unemployment rate. They are counted separately in the broader U-4 to U-6 measures, which this script does not name.
- **rate can fall two ways** — Arithmetic of the definitions above, which is why it is stated as a mechanism rather than a measurement: the rate is unemployed divided by labor force, so it falls either when the numerator falls because people found work, or when both fall because people stopped searching and left the labor force. This is the video's central twist and had no record entry before.
- **deepseek missing source entries** — DeepSeek review 2026-09-03 found seven claims with no matching entry in this record, including the headline 4.1 percent itself and every definitional claim the video rests on — against the channel rule that every claim traces to a source, the largest gap found in any script this round. Claude passed the same check. The six entries above were added; none of them changed a word of the script. Separately, scene 4's 'Not searching for four weeks means out of the labor force' used 'means' to claim an entailment the previous sentence did not supply — it stated only how the BLS counts the unemployed, not what happens to everyone else — so the sentence now states the condition in full.
- **sentence caption removed** — Owner, 2026-09-05, on the staged drafts: 'the label ... should be removed. Way too many words and too little time to read ... they look like a caption, and is redundant with the caption almost. can we just please remove this type of text.' The caption bar was added on 2026-09-03 so the retired doccard's second sentence had somewhere to live; that was the wrong instinct. A short already burns in word-by-word captions, so a second block of prose over the artwork competes with the text the viewer is reading, in a scene lasting a few seconds. Removed from the renderer and from these scenes: scene 3: Not counted as unemployed. Not in the rate.; scene 6: More hired, or fewer looking.
- **chart axes labelled** — validations/chart-axes.md, owner catch 2026-09-05: 'why is there no x-axis and no y-axis labels? add, and add a validation rule. x and y must be clearly labeled.' The renderer drew gridlines with no values on them at all, so every chart in the library showed a shape rather than a measurement - a viewer could not tell a two-point move from a two-hundred-point one. ChartLine now writes the value at the top, quarters and bottom of the plot from yMin/yMax, with a unit from the new yUnit field, and takes an optional xLabel for the horizontal axis. A second defect surfaced with it: xTicks are generated on year changes, so a series that never crosses a year boundary produced NO horizontal labels - which is exactly what happened to the 2026-only jobs chart. build_video.py now falls back to month names in that case.
- **on screen text swept** — validations/on-screen-brevity.md sweep, 2026-09-05. The dominant defect was one habit: the headline restating the scene's own narration, so the card read aloud what the burned-in captions were already spelling out. A headline is the label a viewer takes in at a glance — the angle, not the sentence. Also fixed: highlights appearing nowhere in their headline (they paint nothing), stamps repeating a figure the headline already carries, over-long headlines, and stamps over five words. Scene 3's headline was 62 characters and its second stamp repeated the four-week rule the headline and narration both state; scene 5's second stamp repeated 'not in the rate'.
- **yellow belongs to type** — Owner, 2026-09-06: 'Pull the yellow prompt.' Every image prompt asked for 'one marker-yellow highlight strip'. In black and white that strip WAS the accent and it worked; in colour the generator started using yellow as a compositional block, so the artwork's yellow competed with the yellow our typography uses to point at a word — two things claiming the same signal on one frame. Yellow now belongs to the type alone. The collage keeps its torn-paper colour blocks, which no longer include a deliberate yellow marker.
- **no marks in artwork** — Owner, 2026-09-06: 'costco card usually have costco on it. is it a fair use?' Naming a company in the narration is ordinary editorial use and this channel depends on it; rendering its mark is a different act, and a generated mark is always a wrong mark - an approximation that misrepresents the logo rather than reporting on it, on top of platform brand complaints decided without adjudicating anything. The Costco card was already specified as 'a plain white plastic membership card'. The real exposure was incidental branding nobody had thought of as a brand shot: supermarket shelves, storefronts, packaged goods, and a container port whose boxes carry shipping-line marks. 42 of 63 prompts in the library lacked the clause; every image and video prompt now ends 'no text, no logos, no brand names'.
- **house style 2026 09 06** — Colour magazine collage instead of newspaper halftone (saturation 41 -> 84 measured on a matched card, against Economics Explained's 91); no marker-yellow in any prompt, so yellow belongs to the typography; inner scenes carry the narration captions plus at most one stamp holding a figure, which is what the field's inner slides do. Rebuilt in Fargate, not locally.
- **silent outro card** — Owner, 2026-09-06: 'Can we add a call to action slide at the end ... for users to follow my econplot videos to understand what moves your wallet.' Implemented as the one shape this channel's own rules allow. validations/punchline.md: 'A silent end card (no words) after the punch line is allowed - it is a held frame, not more video', so the punch line still lands last in the NARRATION and nothing follows it aloud. validations/editorial-register.md bans comment-bait specifically ('comment X and I'll send you', 'link in bio for') and says a plain editorial ending is fine - a channel name and what the channel is about is neither bait nor financial advice, which is what STRATEGY.md's 'no call to action' was written to exclude. validations/scene-artwork.md permits shared artwork for 'a channel-level element (a logo, an outro card)', so one card serves every video rather than each generating its own. The line is topic-tied where the topic earns it - 'what moves your mortgage' on the bond video - and the house line otherwise. 1.8 seconds, no narration, no audio.