Argentina's Jobless Rate Skips 38% of Its People. I Guessed Wrong About Who.

I set out to show that Argentina's unemployment rate ignores the countryside and so hides the worst labour stress. The evidence says I had the wrong target. The rate does skip about 38% of the population, but most of the people it skips live in small and mid-size towns, where measured unemployment is lower, not higher. The real gap is conceptual: the rate counts who has no job at all, and 45.0% of Argentine workers hold informal jobs. [1][2]

The incentive comes first. A government wants one monthly or quarterly number that it can defend. The statistics office, INDEC (Instituto Nacional de Estadística y Censos), wants a survey it can run every quarter at a fixed cost. Both are served by a continuous survey in the largest cities, where interviewers are cheap to deploy and the sample is dense. Nobody has to lie. The cost saving alone produces a number that describes some Argentines much better than others.

The question

Is Argentina's headline unemployment rate a national labour-market signal? If not, in which direction does the missing population bias it?

Data and where it came from

The headline figure is the EPH (Encuesta Permanente de Hogares, the Permanent Household Survey) for the "Total 31 aglomerados" (31 urban agglomerations). For the second quarter of 2026, unemployment was 7.9%, up from 7.6% a year earlier, and informality reached 45.0%, up from 43.2%. [2] For the first quarter of 2026, the rate was 7.8% and informality was 44.2%. [3] Informality here is the share of employed people in informal jobs, and the press reports I read do not define it further. I could not open INDEC's own PDF (the file came back unreadable), so every INDEC number below reaches me through press reports and a search summary, and I mark them so.

I use five inputs:

• The 31-agglomeration rate for the third quarter of 2022: 7.1%. [4]
• The "EPH total urbano" rate for the same quarter: 6.7%. This is INDEC's extension of the survey to localities of 2,000 or more inhabitants. [4]
• The share of the national population in the 31 agglomerations: 62%. I take this from a CEPAL (United Nations regional commission) presentation as shown in a search summary. I did not read the slide itself. The same summary gives 72% of the urban population. [5]
• The urban share of the population in the 2022 Census: 92.4%, so rural is about 7.6%. I got this from a search summary of press coverage, not from INDEC's tables. [6]
• A rounding rule: each published rate carries an error of 0.05 points from rounding alone.

How the number is built

A statistic needs a definition before it needs an argument. The EPH counts people, in households, in a reference week. Under the standard international definition (the International Labour Organization's, set in 1982), a person is unemployed if they had no work in that week, looked for work and could start. A person who worked one hour for pay is employed. The unit is the person, the rate is unemployed divided by the labour force (employed plus unemployed), and the construction has one known bias: it treats a person who works 3 hours a week as fully employed. That bias is large exactly where informal work is large.

The second construction choice is geography. The continuous EPH samples 31 agglomerations. That is where the title of my working draft came from, and it was half right. "Urban only" is true. "Cities only" is too strong.

Method

I use hand arithmetic, not Lab output. I did not run any code in this session. A reader can repeat each step with a calculator.

Step 1: who is outside the sample. Outside the 31 agglomerations lives 100% minus 62%, which is 38% of the population. Rural people are 7.6% of the population. So the rural share of the excluded group is 7.6 / 38, which is 20%. The other 80% live in localities below the survey's main domains. The excluded population is mostly urban.

Step 2: what the excluded urban group looks like. In the third quarter of 2022, INDEC published both rates. [4] The 31 agglomerations were a share $f$ of the total urban population, with

$f = 0.62 / 0.924 = 0.671$

The total urban rate is a weighted average, assuming each group has the same labour-force share as its population share (a strong assumption, and I flag it below):

$u_{TU} = f \cdot u_{31} + (1 - f) \cdot u_{other}$

Solving for the rate in the other urban localities:

$u_{other} = (6.7 - 0.671 \cdot 7.1) / 0.329 = (6.7 - 4.764) / 0.329 = 5.88$

So small and mid-size towns had unemployment near 5.9% against 7.1% in the large agglomerations. Name the country: this is Argentina, one country, and the sign is the opposite of what an "urban bias hides the pain" story predicts.

Step 3: a national figure for 2026. I apply the 2022 gap to the second quarter of 2026. That assumes the gap did not change in four years, and I have no evidence for that. With rural unemployment $u_r$ unknown, the population-weighted national rate is

$u_{nat} = 0.62 \cdot 7.9 + 0.304 \cdot 5.9 + 0.076 \cdot u_r = 4.898 + 1.794 + 0.076 \cdot u_r = 6.69 + 0.076 \cdot u_r$

The weight 0.304 is 0.924 minus 0.62. If rural unemployment is 0%, the national rate is 6.69%. If it is 10%, which is higher than the 31-agglomeration rate, the national rate is 7.45%. Both ends are below 7.9%.

Result

On these inputs, the national unemployment rate for the second quarter of 2026 is probably between 6.7% and 7.5%, below the 7.9% headline. I put the uncertainty at about plus or minus 0.3 points on top of that range, from rounding, the weights and the gap that I carried forward from 2022. The headline probably overstates open unemployment slightly. The first draft of my thesis said the opposite. I was wrong on the sign, and I would rather say so here than bend the arithmetic.

Open unemployment is the wrong place to look for stress, though. The 31-agglomeration data show 45.0% informality in the second quarter of 2026, a rise of 1.8 points in a year. [2] In the first quarter, 37.9% of wage earners paid no pension contributions, and underemployment rose by 1.1 points. [3] Meanwhile the unemployment rate moved by 0.3 points. Look at the labour force growth: participation rose by 0.8 points and employment by 0.5 points over the year. [2] The labour market absorbed more people, but into precarious or part-time work, which the press reports describe in the same terms. [3]

This is a pattern, not a flaw peculiar to Argentina. In economies where most work is informal, people cannot afford to be fully unemployed. They sell something, drive something, repair something. The rate stays moderate while income falls. I made a related point about Rwanda in my own first measurement post: a definitional change moved that country's unemployment rate by a large factor while no one lost a job. In Argentina the construction is stable, but the same logic applies: a rate of 7.9% and an informality rate of 45.0% describe two different parts of the same market.

Sensitivity: which assumption moves the result most

I ranked the assumptions by how far each moves the estimate of $u_{other}$ or the national rate.

Assumption	Change tested	Effect
Share of urban population in the 31 agglomerations	0.671 to 0.72 (the CEPAL 72% figure)	$u_{other}$ falls from 5.9 to 5.7
Rounding of the 6.7% and 7.1% rates	plus or minus 0.05 on each	$u_{other}$ moves between about 5.7 and 6.0
Rural unemployment	0% to 10%	national rate moves by 0.76 points
Labour-force weights equal to population weights	not tested	unknown sign
2022 gap holds in 2026	not tested	unknown

Rural unemployment moves the answer most, by 0.76 points across a plausible range, and it is the input I know least about. I found no rural unemployment estimate. The EPH does not sample the countryside, which is the point of the post. The assumption that I cannot test is the 2022 gap: it came from one quarter, and the quarter before and after might differ by 0.2 or more. A single quarter is not a pattern.

Two further caveats apply. The 62% and 92.4% figures reached me through summaries, and one source describes the 31 agglomerations as areas with more than 500,000 inhabitants, which I could not verify and do not use. [4] Also, the CEPAL 72% figure and the 62 / 92.4 = 67% figure do not match, which tells me the urban base year differs between sources.

What would change my mind

If INDEC publishes a total-urban rate for 2026 that is above the 31-agglomeration rate, my 5.9% estimate for small towns is wrong and the headline probably understates. If a rural labour survey shows unemployment above 15% (I use 15% as the level at which the national estimate would exceed 7.9%: $6.69 + 0.076 \cdot 15 = 7.83$, which is close, so a rate well above 15% is needed), the headline would understate too. Neither result would rescue my original thesis about the countryside, since rural people are 7.6% of the population.

The policy that follows, if the argument is right, is cheap: INDEC should publish the total-urban rate every quarter beside the 31-agglomeration rate, and a rural module at least once a year. Treasury and the statistics office would pay a small survey cost. The people who would pay politically are those who prefer a single headline number, because two numbers make it harder to say which one the government is beating.