Unemployment is near its lowest level since 2012. What the rate does not show

In the quarter ending in August 2026, 5.3% of the people in the labor force looked for a job and did not find one. That is the second-lowest figure in the PNAD Contínua, IBGE's continuous household survey, which began in 2012. The lowest, 5.1%, was recorded in December 2025. At the worst moment, in September 2020, the rate hit 14.9%, with 15.0 million unemployed. Today there are 5.8 million.

But the unemployment rate is a narrow yardstick. For IBGE, an unemployed person is someone who did not work, looked for a job and was available to start. Left out are people who work fewer hours than they would like, people who gave up looking, and people who left the labor market. The rate also says nothing about the kind of jobs created or how much they pay. Some of these indicators are at the best point in the series. Others are not.

The numbers that follow are national, month by month. Another Dado Bruto analysis shows how the fall in unemployment was distributed across the states.

The broad measure is also near its floor

IBGE has a wider measure, the composite rate of labor underutilization. It adds two groups to the unemployed: the underemployed due to insufficient hours, who work fewer than 40 hours a week and want and are able to work more, and the potential labor force, people who want to work but did not look or could not start. In August 2026, the rate stood at 13.1%. The series low is 12.9%, in June 2026, and the high, 30.7%, in August 2020. Before 2016, the lowest value had been 14.7%, in July 2014.

By both yardsticks, then, the labor market is near its tightest point since 2012. And the gap between them, which used to be larger than the unemployment rate itself, has narrowed.

Taken separately, though, the layers are not all at the floor. The underemployed due to insufficient hours number 4.3 million, fewer than the 7.0 million of March 2012. The potential labor force, 4.9 million, is above its low of 4.0 million, in August 2014. And discouraged workers, who gave up looking because they cannot find suitable work or because there are no jobs where they live, number 2.4 million. That is less than half the peak of 5.8 million, in March 2021, but still above the series low of 1.4 million, in June 2014, and the figure at the start of the survey, 1.9 million.

Fewer people looking

The unemployment rate can fall in two ways: more people find work or fewer people look. Both happened. The share of the population aged 14 or older that is working, known as the employment-to-population ratio, reached 59.0% in November 2025, the highest in the series, and now stands at 58.9%. The labor force participation rate, which adds those who work and those who are looking, stands at 62.2%. That is practically the same as in March 2012 (62.3%) and below the 2019 average (63.4%).

This difference matters. A simple calculation, not a forecast, gives the measure: if participation were at its 2019 average and employment stayed at 103.5 million, the labor force would have 111.4 million people, not 109.3 million. The unemployed would be 7.9 million, and the rate, 7.1%. The calculation assumes all those people would be looking for work without finding it, which need not be true.

Since 2012, the population aged 14 or older has grown from 153.0 million to 175.6 million, and the number of people outside the labor force from 57.7 million to 66.4 million. This group, in the PNAD, includes students, retirees, people who care for the home or relatives, and people who do not want to work. The data used here do not separate those reasons or age. An aging population is a possible explanation, but it has not been tested.

What jobs were created

From March 2012 to August 2026, the number of employed people went from 87.7 million to 103.5 million: 15.8 million more. Of those, 5.3 million are formal private-sector employees (with a signed work card, the "carteira"). The other 10.5 million came from the self-employed (+6.0 million), informal employees (+2.6 million), the public sector (+2.3 million) and employers (+1.1 million). Domestic work lost 0.5 million, and unpaid family work, 1.0 million.

With a different starting point, the picture changes. Since December 2019, before the pandemic, 9.6 million jobs have been created, and 5.2 million of them are formal private-sector jobs, the majority. The number of formal employees reached 39.5 million, the highest in the series.

By sector, the biggest gains in workers since 2012 were in public administration, education, health and social services (+5.5 million), followed by information, finance and professional services (+3.7 million) and retail and vehicle repair (+2.7 million). Agriculture lost 2.0 million employed people, and domestic services, 0.4 million. Manufacturing employs practically the same number of people as in 2012: 13.3 million, versus 13.0 million.

Informality: slightly lower, and with a different face

IBGE has measured informality since the end of 2015. The calculation includes employees and domestic workers without a signed work card, employers and self-employed workers without a CNPJ (business tax ID), and people who help their families without pay. The rate started at 38.3% in December 2015, reached 41.1% in August 2019 and stands at 37.5%, close to the low of 36.6%. That low, from June 2020, was not an improvement. Between December 2019 and August 2020, informal workers fell from 38.2 million to 30.6 million. They were 41% of the employed but accounted for 59% of the jobs lost. The rate fell because they dropped out of the count. Today they number 38.8 million.

Among the self-employed, the growth came from those with a CNPJ: there were 4.3 million in December 2015 and there are 7.3 million now. Those without a CNPJ went from 18.0 million to 18.9 million. For IBGE, anyone with a CNPJ is formal. The survey does not say whether that CNPJ belongs to a micro-entrepreneur with many clients or to someone who provides services to a single company in place of a formal job.

What it pays

Average real earnings of the employed stand at R$ 3,777, in values adjusted by IBGE to August 2026. The record, R$ 3,804, is from March 2026. That is a rise of 19% since March 2012, 12% since December 2019 and 3.7% over 12 months. The lowest point was December 2021, at R$ 3,068, after the surge in inflation in 2021.

Those who earn less have had the biggest increases since 2012: 35% for informal employees, 27% for the self-employed and 22% for domestic workers. For formal employees, the rise was 12%; in the public sector, 15%. The gap has narrowed: informal employees earn an average of R$ 2,612, 23% less than formal employees (R$ 3,399).

With more people working and earning more, the total earnings bill, the sum of what all employed people receive per month, reached R$ 385.6 billion, the highest value in the series. It was R$ 269.8 billion in March 2012 and R$ 308.7 billion in December 2019. And people are working slightly less: average hours in the main job fell from 40.6 a week in the first quarter of 2012 to 39.1 in the April–June 2026 quarter.

Low unemployment, rising wages?

In theory, when few people are left looking for work, companies compete for workers and wages rise. The chart below tests this: each point is a month, with the unemployment rate on the horizontal axis and the 12-month real change in average earnings on the vertical axis.

The relationship shows up, with varying strength. From 2013 to 2019, it was weak: the correlation between the two series is -0.23. From 2022 to 2026, it became sharper (-0.62): months with lower unemployment tend to have wages rising faster. A negative correlation means that when one series rises, the other tends to fall.

The 2020 and 2021 points break the pattern, and the reason is arithmetic. Between February and August 2020, the country lost 11.9 million jobs, and average earnings rose 6.3%. Those who lost their jobs were mostly informal employees (2.8 million) and domestic workers (1.7 million), who earn less. When the lowest wages leave the calculation, the average rises without anyone getting a raise. The data show the association between unemployment and wages, not how much one causes the other.

The labor market at six moments

The table brings together the main indicators at six moments: the start of the series, the lowest unemployment before the 2014–2016 recession, the last quarter before the pandemic, the worst moment for employment in the pandemic, a year ago and the most recent data.

Limits

The PNAD Contínua is a household sample survey. Each number has a margin of error, and changes of a few tenths from one quarter to the next may not be significant. The data are rolling quarters: the August figure is the average of June, July and August. The series are not seasonally adjusted, and unemployment usually rises at the start of each year. For that reason, comparisons always use the same month in different years, or the entire series.

IBGE revises the survey weighting from time to time, and numbers from earlier years may change. The informality and CNPJ data begin at the end of 2015. And none of the comparisons measures cause and effect.