Diversity in a Globalized World

Lecture 4: Who Lives Next to Whom

Bogdan G. Popescu
bgpopescu@tec.mx

Tecnológico de Monterrey

Where the week has been

  • Monday: a shock landed on particular places, not on a country
  • Tuesday: those same places voted differently afterwards
  • Wednesday: some people left — and the money came back
  • Today: who lives next to whom — and why

Knowing who lives in a country is not the same as knowing who lives next to whom.

Four numbers

  • 23.2 million identify as indigenous — 19.4% of everyone aged 3 or over
  • 7.4 million speak an indigenous language — 5.8% of the population
  • 2.6 million identify as Afro-Mexican — 2%
  • 371 municipios are majority indigenous-speaking; 1,047 are under 1%

The first two are not the same question. Of the 23.2 million who identify as indigenous, seven in ten do not speak an indigenous language.

The country is diverse. Most of the places in it are not.

Those three numbers, on a map

Figure 1

The dark municipios are the 371. Oaxaca alone holds 203 of them.

Learning objectives

  • Say where difference sits in Mexico, and separate three questions people run together:

    • how much
    • which places
    • how mixed

  • Turn “this city is divided” into a number — and say what that number leaves out

  • Explain how a city can divide itself with nobody intending it, and why the sorting outlives whatever started it

  • Say whether living closer together actually reduces prejudice, and under what conditions

Diversity is not spread evenly

Figure 2

A second map, a different geography

Figure 3

Source: INEGI, Perfil sociodemográfico de la población afrodescendiente en México (Censo 2020).

Same population, different segregation

Both cities have 100 residents: 25 blue and 75 green.

Figure 4

Same composition. Different segregation.

What do our Mexico maps actually tell us?

The question What we learn
Composition Who lives in each place? A municipio is 30% indigenous-speaking
Geographic distribution Where are groups located? Which municipios have the highest shares of indigenous-language speakers
Segregation Do groups live in the same neighbourhoods? Our maps cannot tell us

Our maps show who lives where across municipios.

They do not show whether indigenous-language speakers and non-speakers live together or apart within a municipio.

To answer that question, we need smaller geographical units.

So what?

  • Mexico is diverse as a country
  • 83% of Mexicans live in a municipio that is under 5% indigenous-speaking
  • “How much difference?” and “who meets it?” are different questions

Which raises the question of whether where you live also shapes how you live.

The two maps look alike

Figure 5

What this does and does not say

  • The pattern is strong and it is real
  • It does not tell us why
  • Remote municipios are poorer, and also more indigenous

Geography, history and discrimination all predict the same map. A correlation cannot separate them — which is why the rest of the lecture takes the mechanisms one at a time.

A municipio is not a person

Does this graph prove that indigenous-language speakers are poorer than non-speakers living in the same municipio?

No.

Imagine a municipio where: - 80% speak an indigenous language - 60% live in poverty

The poverty rate could be 60% for both groups.

Our graph tells us which municipios are poor, not which people within them are poor.

This is the ecological fallacy: assuming that a relationship between places also holds between individuals.

Poorer municipios, poorer people?

Hypothetical example: two municipios, 100 residents each.

Who lives where?

Figure 6

Who is poor within each municipio?

Figure 7

Who is actually poor?

Hypothetical example: 100 residents in each municipio.

Municipio A Municipio B
Indigenous-language speakers 20 80
Spanish-only speakers 80 20
Indigenous-language speakers in poverty 7 (35%) 52 (65%)
Spanish-only speakers in poverty 28 (35%) 13 (65%)
Total in poverty 35 (35%) 65 (65%)

Who is actually poor?

Between municipios: More indigenous-language speakers, more poverty.

Within each municipio: Both groups have the same poverty rate.

Ecological fallacy: Assuming that a relationship between places also holds between individuals within those places.

So what?

  • Where a group lives is bound up with how it lives
  • At this scale we compare whole municipios
  • The interesting sorting happens at a much smaller scale

So let us take one city apart.

Monterrey, thirteen municipios

Same question, different cleavage: this metro is only 0.4% to 4% indigenous-speaking, so the difference you can see inside it is class.

Figure 8

The same city, ranked

Figure 9

Putting a number on unevenness

The index of dissimilarity asks one question: what share of one group would have to move for the two groups to be spread evenly across the city?

Figure 10

Both sit near the left end. A five-fold gap between two municipios is not yet a divided city.

Which city is more segregated?

Two cities, four neighbourhoods. Same 200 blue and 200 green in each.

Figure 11

Why City 2?

  • The middle two neighbourhoods are identical in both cities
  • The cities differ only at the ends — and City 2 pushed those all the way
  • One neighbourhood is all blue, the other all green

\[D \;=\; \tfrac{1}{2}\sum_{i}\left|\;\frac{b_i}{B}-\frac{g_i}{G}\;\right|\]

City 1 · ½(0.3 + 0.1 + 0.1 + 0.3) = 0.40 — 40% of blue would have to move

City 2 · ½(0.5 + 0.1 + 0.1 + 0.5) = 0.60 — 60% would

Is Monterrey unusual?

Figure 12

No. Every large metro has a gap like this.

But read it carefully: this is a gap between municipios. It says the groups are sorted. It does not say how mixed any one of them is — that is the third word.

One honest caveat

In the Valle de México the gap runs from 8% to 76%. Monterrey’s runs from 5% to 28%. Tijuana’s is only 10 points wide.

But these compare municipios, and Monterrey municipio alone holds over a million people.

Real segregation happens street by street. Every number here is a floor, not a ceiling. Say so when you report one.

You have met this problem all week: Monday’s commuting zones, Wednesday’s whole-city comparison for Mariel. Each time, the unit decided what you could see.

Exercise 1 — Where would you look?

You want to know whether your own city is segregated.

What is the smallest unit you could get data for, and why does that choice change your answer?

Then: name one thing your measure would still miss entirely.

Five minutes, in pairs.

Exercise 1 — Responses

  • Mexico publishes four levels: estado → municipio → AGEB → manzana, the city block
  • The smaller the unit, the more segregation you find — big units average it away
  • A municipio of a million people can look mixed while every street in it is not

What your measure still misses: where people spend the day. Living apart and meeting are different things — people work, study and shop across the line.

Same city, two answers

Figure 13

Same 400 people. Same city. Drawn one way it is perfectly mixed; drawn another way it is completely divided. Nothing moved.

So what?

  • Sorting is measurable, and the measure depends on the unit
  • Prosperous-looking cities contain enormous internal gaps
  • Nothing so far tells us why the sorting happens

The obvious answer is that people are prejudiced. The obvious answer turns out to be unnecessary.

Schelling’s question

In 1971 Thomas Schelling asked what would happen if people were only mildly concerned about their neighbours.

Not hostile. Not seeking to exclude anyone. Simply preferring not to be an extreme minority on their own street.

Your intuition says mild preferences should produce mild sorting. Hold on to that, and watch what the model does.

What actually happens

Figure 14

Two lessons, pulling opposite ways

At the start, the average person’s neighbours were 50% their own group. At the end: 75%.

  • You do not need bigots. Nobody here is hostile, no law excludes anyone, and nobody coordinated
  • So finding segregation does not prove that anyone is prejudiced
  • But it also does not prove that nobody is

The mechanism

%%{init: {"flowchart": {"useMaxWidth": true, "htmlLabels": true, "padding": 26,
                        "nodeSpacing": 30, "rankSpacing": 70,
                        "width": 1150, "height": 650},
          "themeVariables": {"fontSize": "22px", "fontFamily": "Arial, Helvetica, sans-serif"}} }%%
flowchart LR
  A["A mild<br/>preference"] --> B["A few<br/>people move"]
  B --> C["The street<br/>tips"]
  C --> D["Neighbours<br/>move too"]
  D -->|"and again"| B
  C --> E["Segregation"]
Figure 15

Author’s illustration, after Schelling (1971). The loop is the whole point.

Watch the street tip

An animated grid of red and green squares that starts mixed at random and separates into blocks of one colour over thirteen rounds.

Author’s simulation, after Schelling (1971). Each round, people with fewer than 30% same-colour neighbours swap houses with each other. Nobody’s preference changes.

But is anyone discriminating?

Schelling shows sharp sorting does not require hostility or exclusion. That is not evidence that there is none of either. So stop inferring it from maps and test it directly:

  • Send two applications for the same thing — a job, a flat
  • Make them identical except one detail: a name, a photo
  • Count who replies

Because everything else is held fixed, a gap in replies is hard to explain away. This is the correspondence study design.

What those tests find

Six passport-style photographs of applicants in business dress, arranged as women and men across three phenotypes labelled European, Mestiza or Mestizo, and Indigenous.

Same clothes. Same pose. Identical CV behind each one.

  • Arceo-Gómez and Campos-Vázquez sent these across Mexico
  • Indigenous-looking women were called back less
  • Men showed no penalty

Photographs: Arceo-Gómez & Campos-Vázquez, Appendix A. Bertrand and Mullainathan (2004) found a comparable gap in the United States using names alone.

Mind the domain

  • This is a hiring experiment.
  • It shows discrimination is live in Mexico — it does not show that discrimination produced the sorting you just saw
  • Those are two claims, and only one of them has been tested here

The design travels, though: run it on landlords instead of employers and you have a test of the housing market. That has been done in Sweden and the United States — rarely in Mexico.

Does living closer help reduce discrimination?

Two intuitions, pointing opposite ways:

  • Contact breeds friction — difference itself is a source of conflict
  • Contact breeds familiarity — people who meet stop generalising

Allport’s conditions

%%{init: {"flowchart": {"useMaxWidth": true, "htmlLabels": true, "padding": 22,
                        "nodeSpacing": 12, "rankSpacing": 110,
                        "width": 1150, "height": 620},
          "themeVariables": {"fontSize": "22px", "fontFamily": "Arial, Helvetica, sans-serif"}} }%%
flowchart LR
  A["Equal<br/>standing"] --> X["Intergroup<br/>contact"]
  B["A shared<br/>goal"] --> X
  C["Working<br/>together"] --> X
  D["Institutional<br/>support"] --> X
  X --> E["Less<br/>prejudice"]
Figure 16

Author’s illustration, after Allport (1954). Contact is the mechanism; the four are the conditions Allport said it needs to work. Note what is not here — proximity on its own.

The evidence, fifty years later

Pettigrew and Tropp (2006) pooled 515 studies. Contact reduces prejudice on average — and Allport’s conditions strengthen that effect rather than being required for it.

  • Contact usually helps, conditions or no conditions
  • Contact under Allport’s conditions helps considerably more
  • Contact between unequals, competing, can make things worse

So what?

  • Contact needs opportunities for contact
  • Segregation does not abolish them — people still work, study, commute and shop across the line, as you said in Exercise 1
  • What it thins out is the sustained, equal-footing, repeated kind — exactly the kind the evidence rewards most

So the honest claim is about odds, not impossibility: a divided city makes the good sort of contact rarer, and leaves more of what is left to the kinds that help least.

Takeaways

  1. Mexico is diverse; 83% of Mexicans live somewhere that is not — on this one measure, at this one scale
  2. Where a group lives is bound up with how it lives — but a pattern across places is not a fact about people
  3. Sorting is measurable; the number depends on the unit you pick, and on which dimension you chose to measure
  4. Mild preferences can divide a city — and the division outlives whatever caused it, so repealing a cause does not undo it
  5. So segregation is no proof of discrimination, which is a separate claim, tested one market at a time
  6. Contact usually reduces prejudice, and most when it is sustained and equal — which is what a divided city makes scarce

References

Ahmed, A. M., & Hammarstedt, M. (2008). Discrimination in the rental housing market: A field experiment on the Internet. Journal of Urban Economics, 64(2), 362–372.

Allport, G. W. (1954). The Nature of Prejudice. Addison-Wesley.

Arceo-Gómez, E. O., & Campos-Vázquez, R. M. (2014). Race and marriage in the labor market: A discrimination correspondence study in a developing country. American Economic Review, 104(5), 376–380.

Bertrand, M., & Mullainathan, S. (2004). Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. American Economic Review, 94(4), 991–1013.

CONEVAL. (2021). Medición de la pobreza a nivel municipio 2020. Mexico City: Consejo Nacional de Evaluación de la Política de Desarrollo Social.

References (cont.)

Huntington, S. P. (1996). The Clash of Civilizations and the Remaking of World Order. Simon & Schuster.

Hanson, A., & Hawley, Z. (2011). Do landlords discriminate in the rental housing market? Evidence from an internet field experiment in US cities. Journal of Urban Economics, 70(2–3), 99–114.

INEGI. (2021). Censo de Población y Vivienda 2020. Aguascalientes: Instituto Nacional de Estadística y Geografía.

Massey, D. S., & Denton, N. A. (1988). The dimensions of residential segregation. Social Forces, 67(2), 281–315.

Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90(5), 751–783.

Schelling, T. C. (1971). Dynamic models of segregation. Journal of Mathematical Sociology, 1(2), 143–186.