What a country produces determines who stays poor. This briefing maps the evidence at the municipal level.
Prepared for UNICEF · March 2026
The complexity gradient from north to south. The poverty gradient follows.
Northern states have spent decades building industrial supply chains, training workforces, and attracting investment. Southern states have not. The gap in productive capability is structural, not cyclical.
Chiapas: 76.8% poverty. Nuevo Leon: 14.5%. The ranking mirrors the complexity bars above almost exactly. The question is whether this holds at the municipal level, where UNICEF programs.
Mexico's economy grew 76% since 1996. Its billionaires' wealth grew 320%.
For every peso the Mexican economy had in 1996, it had 1.76 pesos in 2025. The richest 10% had 2.14. The richest 1% had 2.38. The 22 billionaires had 4.20.
Growth is real. Its distribution is not.
Each crisis widens the gap. Billionaires recover in months. Wages take decades. The 1982 debt crisis cut real wages in half. They did not return to their 1981 level until 2024. Forty-two years.
1.3 million people capture 35% of national income and hold 40% of national wealth. On the other side: 18.8 million lack food access. 38.5 million face at least one social deprivation.
Mexico's public revenues rank lowest in the OECD. The private sector captures growth. The public sector lacks the fiscal capacity to redistribute it.
One index. Two countries.
Economic Complexity Index (ECI) measures how many different things a place produces, and how few other places produce those same things. High ECI means deep, diversified productive capability. Low ECI means dependence on activities that every municipality can do.
Miguel Hidalgo, Mexico City: ECI 4.55, poverty 10.5%. Santos Reyes Papalo, Oaxaca: ECI -1.53, poverty 98.1%. A 6-point gap in complexity. An 88-point gap in poverty.
5.5 million firms. Most of them in the same places.
Metropolitan areas concentrate businesses at 10 to 50 times the density of rural indigenous municipalities. Nearly all of this activity is private sector. Mexico's public revenues rank lowest in the OECD. Where the private economy is thin, public services are underfunded.
Poverty is not random. It follows the complexity map.
CONEVAL measures poverty across six dimensions: income, education, health, food, housing, and social security. The geographic overlap with the complexity map is near-total.
Each dot is a municipality. As complexity rises, poverty falls. The outliers matter most. Above the line: higher poverty than complexity predicts. Below: outperforming. Chapter 8 explains why.
Economic complexity predicts poverty. But inequality determines how much of that prediction reaches children.
The most unequal states are not the poorest. Some northern industrial states show high Gini despite high complexity. Inequality and poverty are correlated, but they don't travel together. Inequality is an independent force.
Complexity reduces poverty. Inequality amplifies it. Both forces operate simultaneously. Where complexity is high but inequality is also high, welfare gains concentrate at the top.
The deprivations that follow from structural weakness.
CONEVAL tracks educational backwardness, health service access, and food insecurity. Chiapas, Guerrero, and Oaxaca show deprivation rates 2 to 3 times the national average across all three. These deprivations cluster in the same municipalities with the lowest economic complexity.
Homicide concentrates in enclaves: municipalities where complex activity exists but isn't embedded. Domestic violence reporting follows the opposite pattern. Dividend municipalities report 8.3 DV cases per homicide. Traps report 3.8. The gap measures institutional presence, not safety.
The economy barely grew. What growth occurred was captured privately. The state collects less than half the OECD average in tax revenue. This is the fiscal environment UNICEF partners with.
Four structural conditions. Four types of intervention.
Municipalities split into four quadrants based on their ECI and poverty rate relative to the median. Each quadrant implies a different type of intervention.
Low complexity, high poverty
Ocosingo, Chiapas (264K). Predominantly indigenous. Extreme poverty above 50%. Economy: subsistence agriculture, informal commerce, public transfers. Formal firms scarce.
San Felipe del Progreso, Edo. Mex. (156K). Within commuting distance of Mexico City, economically disconnected from it. Poverty above 80%. Illiteracy above 10%.
Low capability constrains employment, income, and fiscal capacity. Service delivery sustains welfare in these municipalities. It does not build it. Violence is lower here than elsewhere: nothing to extract, nothing to fight over. But DV reporting is half the national rate. The absence of violence data is itself a signal of institutional absence.
Higher complexity, persistent poverty
Centla, Tabasco (108K). PEMEX oil extraction and mangrove restoration raise ECI. Employment concentrates in public admin and low-productivity services. Poverty persists.
Silao, Guanajuato (203K). GM assembly. $4.8B in exports (2024). Nearly 50% poverty. Supplier network is geographically dispersed. Local jobs concentrate in assembly and logistics, not higher-value tiers.
Complexity without local linkages does not reduce poverty. The industry is present but not embedded. These municipalities are also 50% more violent than any other quadrant. Concentrated rents and weak governance attract organized crime. Acapulco: 543 homicides. Uruapan: 258. The violence is structural, not incidental.
High complexity, low poverty
Saltillo, Coahuila (metro 979K). Dense automotive and metalworking cluster. Assembly, engineering services, intermediate manufacturing. Diversified supplier networks. Poverty well below national average.
Apodaca, Nuevo Leon (748K). Aerospace, industrial machinery, electronics, logistics. Dense supplier networks and continuous firm entry.
Embedded complexity generates formal employment, municipal revenue, public services, and institutional depth. This is what the mechanism looks like when it works. Absolute homicide counts are high (Tijuana: 1,844, Juarez: 1,034), but these cities have the institutional capacity to report, investigate, and respond. DV reporting is 2x higher per homicide than in traps.
Lower complexity, lower poverty
Coyoacan, Mexico City (602K). Universities, cultural institutions, government. Low poverty through metropolitan labor market integration, not industrial complexity.
Tepic, Nayarit (453K). State capital. Public administration, education, services. Government spending stabilizes income.
Stable, but structurally dependent. If government spending contracts, these municipalities are exposed. Violence is moderate but diffuse. Without a productive base, security depends on the same government spending that funds everything else.
Traps need structural investment alongside service delivery. Enclaves need linkage strategies that connect industry to local welfare. Capability dividend municipalities are models to study and replicate. Overperformers need diversification beyond government spending.
Five findings. One framework. A map for where to intervene.