Labor Market MX — Quarterly Update

ENOE, INEGI · Quarterly, 2005–2026 · National

Updated every time INEGI publishes a new quarter: what changed from the prior quarter and from a year earlier, and how closely this pipeline matches INEGI’s own official bulletin for that same period. Same quality check described on the overview page, laid out here in full instead of a single summary line.

raw = FileAttachment("data/labor-indicators.csv").csv({ typed: true })
rawCortes = FileAttachment("data/labor-indicators-cuts.csv").csv({ typed: true })
validaciones = FileAttachment("data/validations.csv").csv({ typed: true })

datos = raw.map(d => ({ ...d, fecha: new Date(d.anio, (d.trimestre - 1) * 3, 1) }))
datosCortes = rawCortes.map(d => ({ ...d, fecha: new Date(d.anio, (d.trimestre - 1) * 3, 1) }))

nombres = ({
  TPEA: "Labor force participation",
  TD: "Unemployment",
  TIL1: "Labor informality",
  TIL2: "Labor informality (excl. agriculture)",
  TOSI1: "Employment in the informal sector",
  TOSI2: "Informal-sector employment (excl. agriculture)",
  TSUB: "Underemployment",
  TCCO: "Critical employment conditions",
  TDAMPL: "Broader unemployment",
  SUBUTIL: "Labor underutilization"
})

notas = ({
  TCCO: "Likely a rebase of INEGI’s minimum-wage threshold (the bulletin cites a “January 2026 base”); verified correct by independently reconstructing TCCO from raw microdata."
})

// Only these have sex/age/education/state breakdowns computed so far (see
// Participation/Informality/Unemployment) -- TSUB and TCCO stay national-only.
indicadoresConCortes = ["TPEA", "TIL1", "TIL2", "TOSI1", "TOSI2", "TD", "TDAMPL", "SUBUTIL"]

cortes = ({
  sexo: "Sex", edad: "Age", nivel_educativo: "Education", entidad: "State"
})

// categoria_origen labels in the CSV are Spanish -- map to English for
// display, same lookup used on the three theme pages.
etiquetaEn = ({
  "Hombre": "Men", "Mujer": "Women",
  "14 a 19": "14 to 19", "20 a 29": "20 to 29", "30 a 39": "30 to 39",
  "40 a 49": "40 to 49", "50 a 59": "50 to 59", "60 y más": "60 and over",
  "Primaria incompleta": "Incomplete primary", "Primaria completa": "Primary",
  "Secundaria completa": "Secondary", "Medio superior y superior": "High school and beyond",
  "Ciudad de México": "Mexico City", "Coahuila de Zaragoza": "Coahuila",
  "México": "Mexico State", "Michoacán de Ocampo": "Michoacán",
  "Veracruz de Ignacio de la Llave": "Veracruz"
})

serieDe = codigo => datos
  .filter(d => d.indicador === codigo)
  .sort((a, b) => a.fecha - b.fecha)

valorEnPeriodo = (serie, anio, trimestre) => {
  const fila = serie.find(d => d.anio === anio && d.trimestre === trimestre);
  return fila ? fila.valor : null;
}

calcularDeltas = serie => {
  const ultimo = serie[serie.length - 1];
  const trimAnt = ultimo.trimestre === 1
    ? { anio: ultimo.anio - 1, trimestre: 4 }
    : { anio: ultimo.anio, trimestre: ultimo.trimestre - 1 };
  const valorTrimAnt = valorEnPeriodo(serie, trimAnt.anio, trimAnt.trimestre);
  const valorAnioAnt = valorEnPeriodo(serie, ultimo.anio - 1, ultimo.trimestre);
  return {
    deltaTrim: valorTrimAnt != null ? ultimo.valor - valorTrimAnt : null,
    deltaAnual: valorAnioAnt != null ? ultimo.valor - valorAnioAnt : null,
    baseTrim: valorTrimAnt, baseAnual: valorAnioAnt
  };
}

// Same as valorEnPeriodo/calcularDeltas above, but also carries the
// standard error of each period's estimate through, so the by-group
// section below can tell a real move from sampling noise.
valorEnPeriodoConSE = (serie, anio, trimestre) => {
  const fila = serie.find(d => d.anio === anio && d.trimestre === trimestre);
  return fila ? { valor: fila.valor, ee: fila.ee } : null;
}

// SE of a difference between two survey estimates: combine in quadrature,
// assuming independence. ENOE's rotating panel means neighboring quarters
// actually share some of the same respondents, so this slightly overstates
// the true SE of the quarter-over-quarter difference (positive correlation
// between overlapping observations would shrink it) -- the filter below
// errs toward under-claiming significance, not over-claiming it.
calcularDeltasConSE = serie => {
  const ultimo = serie[serie.length - 1];
  const trimAnt = ultimo.trimestre === 1
    ? { anio: ultimo.anio - 1, trimestre: 4 }
    : { anio: ultimo.anio, trimestre: ultimo.trimestre - 1 };
  const trimAntDato = valorEnPeriodoConSE(serie, trimAnt.anio, trimAnt.trimestre);
  const anioAntDato = valorEnPeriodoConSE(serie, ultimo.anio - 1, ultimo.trimestre);
  return {
    deltaTrim: trimAntDato ? ultimo.valor - trimAntDato.valor : null,
    deltaAnual: anioAntDato ? ultimo.valor - anioAntDato.valor : null,
    seTrim: trimAntDato ? Math.sqrt(ultimo.ee ** 2 + trimAntDato.ee ** 2) : null,
    seAnual: anioAntDato ? Math.sqrt(ultimo.ee ** 2 + anioAntDato.ee ** 2) : null,
    baseTrim: trimAntDato ? trimAntDato.valor : null,
    baseAnual: anioAntDato ? anioAntDato.valor : null
  };
}

// A change counts as distinguishable from noise only if it clears 1.96
// combined standard errors -- the same 95% threshold as the confidence
// bands on every chart elsewhere on this site. Without this filter,
// "biggest mover" would almost always just mean "smallest sample": small
// states/categories have the largest standard errors, so their point
// estimates swing the most from quarter to quarter for no real reason.
esSignificativo = (delta, se) => delta != null && se != null && Math.abs(delta) > 1.96 * se

// Unlike the rest of this series (which leaves color neutral on purpose,
// without implying good/bad), this page does color the direction of
// change -- because it's a scan-many-rows-at-once summary, not a
// case-by-case reading of one indicator. Only the arrow carries color,
// and it's kept pale: coloring the whole value read as too loud and too
// normative (good/bad), which this page shouldn't imply.
// base is the prior period's own value (the denominator this change is
// measured against) -- pass it to also show the relative percent change
// alongside the percentage-point change, e.g. "▲ 0.41 pp (0.7%)".
// The pp figure stays unsigned (direction carried by the arrow alone, the
// site's usual convention), but the percent figure keeps its sign -- it's
// also the sort key for these columns, and an unsigned percent next to a
// down arrow read as if sorting ignored direction, which it doesn't.
formatoDelta = (delta, base) => {
  if (delta == null) return html`<span>n/a</span>`;
  if (Math.abs(delta) < 0.005) return html`<span class="delta-flat">no change</span>`;
  const subida = delta > 0;
  const clase = subida ? "delta-arrow-up" : "delta-arrow-down";
  const signo = subida ? "▲" : "▼";
  const relativo = (base != null && base !== 0)
    ? html` <span class="data-table-period">(${(delta / base * 100).toFixed(1)}%)</span>`
    : "";
  return html`<span><span class="${clase}">${signo}</span> ${Math.abs(delta).toFixed(2)} pp${relativo}</span>`;
}

cambiosTrimestre = Object.keys(nombres).map(codigo => {
  const serie = serieDe(codigo);
  if (serie.length === 0) return null;
  const ultimo = serie[serie.length - 1];
  const { deltaTrim, deltaAnual, baseTrim, baseAnual } = calcularDeltas(serie);
  return {
    codigo, nombre: nombres[codigo], valor: ultimo.valor,
    periodo: `${ultimo.anio}-Q${ultimo.trimestre}`,
    deltaTrim, deltaAnual, baseTrim, baseAnual
  };
}).filter(d => d !== null)

categoriasDeCorte = corte => [...new Set(datosCortes.filter(d => d.corte === corte).map(d => d.categoria_origen))]

// One row per category within a cut, latest value plus QoQ/YoY change --
// same calculation as cambiosTrimestre above, just applied per category
// instead of at the national level. Unsorted here -- ordenarFilas below
// applies whichever order the reader picks at render time.
cambiosPorCorte = (indicador, corte) => categoriasDeCorte(corte).map(cat => {
  const serie = datosCortes
    .filter(d => d.indicador === indicador && d.corte === corte && d.categoria_origen === cat)
    .sort((a, b) => a.fecha - b.fecha);
  if (serie.length === 0) return null;
  const ultimo = serie[serie.length - 1];
  const { deltaTrim, deltaAnual, baseTrim, baseAnual } = calcularDeltasConSE(serie);
  return {
    categoria: etiquetaEn[cat] ?? cat, valor: ultimo.valor,
    periodo: `${ultimo.anio}-Q${ultimo.trimestre}`,
    deltaTrim, deltaAnual, baseTrim, baseAnual
  };
}).filter(d => d !== null)

// Sort state for both tables' clickable column headers -- each table gets
// its own independent column/direction pair, so sorting one doesn't
// resort the other. Clicking a header sorts by that column, a fresh
// column starts descending, and a second click on the active column
// toggles descending/ascending. The change columns sort by the relative
// PERCENT change (delta divided by the prior period's own value), not the
// raw percentage-point delta -- pp isn't comparable across rows with very
// different bases (a 0.1 pp move means something different for a 60% rate
// than a 2% one), so sorting on it would rank rows in an order that
// doesn't match what the parenthetical percent figure actually says.
mutable columnaOrdenTrim = "deltaTrim"
mutable direccionOrdenTrim = "desc"
mutable columnaOrdenCorte = "deltaTrim"
mutable direccionOrdenCorte = "desc"

alternarOrdenTrim = columna => {
  if (columnaOrdenTrim === columna) {
    mutable direccionOrdenTrim = direccionOrdenTrim === "desc" ? "asc" : "desc";
  } else {
    mutable columnaOrdenTrim = columna;
    mutable direccionOrdenTrim = "desc";
  }
}

alternarOrdenCorte = columna => {
  if (columnaOrdenCorte === columna) {
    mutable direccionOrdenCorte = direccionOrdenCorte === "desc" ? "asc" : "desc";
  } else {
    mutable columnaOrdenCorte = columna;
    mutable direccionOrdenCorte = "desc";
  }
}

// Every header shows a faint double arrow when sortable but inactive, and
// a bold single arrow matching the current direction when it's the active
// sort column -- so the affordance is visible before a reader clicks
// anything, not just after.
flechaOrden = (columnaActiva, columna, direccion) => columnaActiva === columna
  ? html` <span class="data-table-sort-arrow">${direccion === "desc" ? "▼" : "▲"}</span>`
  : html` <span class="data-table-sort-arrow data-table-sort-arrow-inactive">⇅</span>`

// Setting onclick as an attribute inside the html`` template (rather than
// as a DOM property on the returned node) doesn't wire up as an event
// listener in Quarto's ojs runtime -- it just stringifies the function into
// the cell's text. Building the <th> node and assigning .onclick directly
// is what actually works. subtitulo (e.g. "(t vs t-1)") renders on its own
// line below the label when given.
celdaEncabezado = (etiqueta, subtitulo, columna, columnaActiva, direccion, alClick) => {
  const flecha = flechaOrden(columnaActiva, columna, direccion);
  const celda = subtitulo
    ? html`<th class="data-table-sortable">${etiqueta}${flecha}<br><span class="data-table-header-sub">${subtitulo}</span></th>`
    : html`<th class="data-table-sortable">${etiqueta}${flecha}</th>`;
  celda.onclick = () => alClick(columna);
  return celda;
}

// The relative percent change this sorts by: null when there's no prior
// value to compare against (new category, or a value of exactly zero,
// where "percent change" is undefined) -- pushed to the bottom regardless
// of sort direction via the sentinel below, same as a missing pp delta
// would be.
cambioPct = (delta, base) => (delta != null && base != null && base !== 0) ? delta / base * 100 : null

ordenarFilas = (filas, columnaActiva, direccion) => {
  const copia = [...filas];
  const signo = direccion === "desc" ? -1 : 1;
  const centinela = direccion === "desc" ? -Infinity : Infinity;
  if (columnaActiva === "valor") return copia.sort((a, b) => signo * (a.valor - b.valor));
  if (columnaActiva === "deltaAnual") return copia.sort((a, b) => signo * ((cambioPct(a.deltaAnual, a.baseAnual) ?? centinela) - (cambioPct(b.deltaAnual, b.baseAnual) ?? centinela)));
  return copia.sort((a, b) => signo * ((cambioPct(a.deltaTrim, a.baseTrim) ?? centinela) - (cambioPct(b.deltaTrim, b.baseTrim) ?? centinela)));
}

// Every (indicator, cut, category) combination's change, filtered to only
// the ones that clear the significance bar -- the pool "biggest mover"
// picks from below. Significance is still evaluated on the raw pp delta
// against its standard error (that's the level estimate the SE actually
// describes), but ranked by relative PERCENT change, same as the table
// below it -- ranking by raw pp here would pick, say, a 6.9 pp move on a
// 27% base over a 4.4 pp move on a 14% base, even though the second is
// the larger swing in relative terms, and disagree with how the table
// right below this headline sorts the exact same category-level data.
todosLosCambios = (campoDelta, campoSE, campoBase) => {
  const filas = [];
  for (const indicador of indicadoresConCortes) {
    for (const corte of Object.keys(cortes)) {
      for (const cat of categoriasDeCorte(corte)) {
        const serie = datosCortes
          .filter(d => d.indicador === indicador && d.corte === corte && d.categoria_origen === cat)
          .sort((a, b) => a.fecha - b.fecha);
        if (serie.length === 0) continue;
        const ultimo = serie[serie.length - 1];
        const deltas = calcularDeltasConSE(serie);
        const delta = deltas[campoDelta], se = deltas[campoSE], base = deltas[campoBase];
        if (esSignificativo(delta, se)) {
          filas.push({
            indicador, corte, categoria: etiquetaEn[cat] ?? cat,
            valor: ultimo.valor, delta, base, periodo: `${ultimo.anio}-Q${ultimo.trimestre}`
          });
        }
      }
    }
  }
  return filas.sort((a, b) => Math.abs(cambioPct(b.delta, b.base) ?? 0) - Math.abs(cambioPct(a.delta, a.base) ?? 0));
}

// Split into a biggest increase and a biggest decrease per period, rather
// than one overall winner -- todosLosCambios is already sorted by
// |percent change| descending, so the first row matching each sign is
// that direction's biggest mover.
cambiosTrimOrdenados = todosLosCambios("deltaTrim", "seTrim", "baseTrim")
cambiosAnualOrdenados = todosLosCambios("deltaAnual", "seAnual", "baseAnual")
mayorMovimientoTrimSubida = cambiosTrimOrdenados.find(f => f.delta > 0) ?? null
mayorMovimientoTrimBajada = cambiosTrimOrdenados.find(f => f.delta < 0) ?? null
mayorMovimientoAnualSubida = cambiosAnualOrdenados.find(f => f.delta > 0) ?? null
mayorMovimientoAnualBajada = cambiosAnualOrdenados.find(f => f.delta < 0) ?? null

filaMovimiento = (etiqueta, fila) => fila
  ? html`<tr>
  <td>${etiqueta}</td>
  <td>${nombres[fila.indicador]}<span class="data-table-period"> — ${fila.categoria} (${cortes[fila.corte]})</span></td>
  <td class="data-table-num">${formatoDelta(fila.delta, fila.base)}</td>
  <td class="data-table-num">${fila.valor}%<span class="data-table-period">(${fila.periodo})</span></td>
</tr>`
  : html`<tr>
  <td>${etiqueta}</td>
  <td colspan="3" class="data-table-note">No move clears the 95% confidence threshold this period.</td>
</tr>`

ultimoPeriodoValidacion = validaciones.length
  ? `${validaciones[validaciones.length - 1].anio}-Q${validaciones[validaciones.length - 1].trimestre}`
  : ""

ultimasValidaciones = {
  const ultimoAnio = validaciones[validaciones.length - 1]?.anio;
  const ultimoTrim = validaciones[validaciones.length - 1]?.trimestre;
  return validaciones.filter(v => v.anio === ultimoAnio && v.trimestre === ultimoTrim);
}

Highlights this quarter

Auto-generated from the same by-group data behind Participation, Informality, and Unemployment: the biggest increase and the biggest decrease, both quarter-over-quarter and year-over-year, ranked by percent change. A row only appears if that move clears a 95% confidence threshold — otherwise the quarter’s biggest swing is usually just the smallest sample, not a real move.

html`<table class="data-table">
<thead><tr>
  <th>Move</th><th>Indicator</th><th>Change</th><th>Latest</th>
</tr></thead>
<tbody>
${filaMovimiento("Biggest quarterly increase", mayorMovimientoTrimSubida)}
${filaMovimiento("Biggest quarterly decrease", mayorMovimientoTrimBajada)}
${filaMovimiento("Biggest annual increase", mayorMovimientoAnualSubida)}
${filaMovimiento("Biggest annual decrease", mayorMovimientoAnualBajada)}
</tbody>
</table>`

Quarter-over-quarter changes

Click a column header to sort by it — the change columns sort by the percent change (the figure in parentheses), not the raw percentage-point move, since pp isn’t comparable across indicators sitting on very different bases. Click again to reverse.

html`<table class="data-table">
<thead><tr>
  <th>Indicator</th>
  ${celdaEncabezado("Latest", null, "valor", columnaOrdenTrim, direccionOrdenTrim, alternarOrdenTrim)}
  ${celdaEncabezado("Quarterly change", "(t vs t-1)", "deltaTrim", columnaOrdenTrim, direccionOrdenTrim, alternarOrdenTrim)}
  ${celdaEncabezado("Annual change", "(t vs t-4)", "deltaAnual", columnaOrdenTrim, direccionOrdenTrim, alternarOrdenTrim)}
</tr></thead>
<tbody>
${ordenarFilas(cambiosTrimestre, columnaOrdenTrim, direccionOrdenTrim).map(d => html`<tr>
  <td>${d.nombre}</td>
  <td class="data-table-num">${d.valor}%<span class="data-table-period">(${d.periodo})</span></td>
  <td class="data-table-num">${formatoDelta(d.deltaTrim, d.baseTrim)}</td>
  <td class="data-table-num">${formatoDelta(d.deltaAnual, d.baseAnual)}</td>
</tr>`)}
</tbody>
</table>`

Changes by group

Same quarter-over-quarter and year-over-year changes as above, broken down by the cuts INEGI doesn’t publish on its own site — sex, age, education, and state. Each change shows the move in percentage points, and, in parentheses, the equivalent percent change relative to the prior period. Click a column header below to sort by it — like the table above, the change columns sort by that percent figure, not the pp move. Click again to reverse.

viewof definicion_sel = Inputs.select(indicadoresConCortes, {
  value: "TPEA",
  label: "Indicator",
  format: d => `${d} — ${nombres[d]}`
})

viewof corte_sel = Inputs.select(Object.keys(cortes), {
  value: "sexo",
  label: "Breakdown",
  format: d => cortes[d]
})

html`<table class="data-table">
<thead><tr>
  <th>${cortes[corte_sel]}</th>
  ${celdaEncabezado("Latest", null, "valor", columnaOrdenCorte, direccionOrdenCorte, alternarOrdenCorte)}
  ${celdaEncabezado("Quarterly change", "(t vs t-1)", "deltaTrim", columnaOrdenCorte, direccionOrdenCorte, alternarOrdenCorte)}
  ${celdaEncabezado("Annual change", "(t vs t-4)", "deltaAnual", columnaOrdenCorte, direccionOrdenCorte, alternarOrdenCorte)}
</tr></thead>
<tbody>
${ordenarFilas(cambiosPorCorte(definicion_sel, corte_sel), columnaOrdenCorte, direccionOrdenCorte).map(d => html`<tr>
  <td>${d.categoria}</td>
  <td class="data-table-num">${d.valor}%<span class="data-table-period">(${d.periodo})</span></td>
  <td class="data-table-num">${formatoDelta(d.deltaTrim, d.baseTrim)}</td>
  <td class="data-table-num">${formatoDelta(d.deltaAnual, d.baseAnual)}</td>
</tr>`)}
</tbody>
</table>`

Comparison against the official bulletin

html`<p>Latest validation: <strong>${ultimoPeriodoValidacion}</strong>. Filled in by hand each quarter, comparing this pipeline against INEGI's <strong>quarterly</strong> press release, not the monthly one, which uses a different methodology and isn't comparable (two separate bulletins, with separate release numbers, come out almost the same day).</p>`

html`<table class="data-table">
<thead><tr>
  <th>Indicator</th><th>Official (INEGI)</th><th>Pipeline</th><th>Difference</th><th>Note</th>
</tr></thead>
<tbody>
${ultimasValidaciones.map(v => html`<tr>
  <td>${nombres[v.indicador] ?? v.indicador}</td>
  <td class="data-table-num">${v.oficial}%</td>
  <td class="data-table-num">${v.pipeline}%</td>
  <td class="data-table-num">${v.diferencia_pp.toFixed(2)} pp</td>
  <td class="data-table-note">${notas[v.indicador] || "—"}</td>
</tr>`)}
</tbody>
</table>`

How this is built

Same pipeline and same public repository as the overview page. “Difference” is the absolute gap between this pipeline and the figure INEGI rounds to one decimal in its bulletin — a gap of up to roughly 0.05 pp is rounding, not a sign of error. Quarter-over-quarter and year-over-year changes are computed against the exact matching period, not by row position, so the 2020-Q2 gap (ETOE, not comparable) never produces a change calculated across two quarters of distance. The “biggest mover” highlights use a 95% significance filter (quarter-over-quarter or year-over-year change greater than 1.96 combined standard errors) computed by treating the two periods’ estimates as independent; ENOE’s rotating panel means adjacent quarters actually share some of the same respondents, so this is a conservative approximation — a real quarter-over-quarter move is a little more likely to be flagged as significant than this threshold gives it credit for, not less. The percent figure in parentheses next to each change is that change divided by the prior period’s own value, not a second percentage-point figure — a 0.1 pp move on a 2% base is a much bigger relative swing than the same 0.1 pp on a 60% base, which the parenthetical is there to make visible. TSUB and TCCO don’t appear in the by-group section: no state/demographic breakdown has been computed for those two yet. Source: ENOE microdata, INEGI, 2005–2026.