What this course is about

A globalized world: trade, money and migration tie distant places together.

The week

A second course, about evidence

Every day, the same habit: ask what the evidence actually shows.

• Monday: a map reveals what a national average hides
• Tuesday: one wave of populism, four competing explanations
• Wednesday: one event, two economists, opposite conclusions
• Thursday: change the unit you measure, and the answer changes

What you will be able to do

• Describe where globalization’s costs landed, and why averages mislead
• Explain economic and cultural accounts of the populist backlash
• Explain why people migrate, and what the evidence says about wages
• Measure how segregated a city is, and why the unit matters
• Read a claim critically — rival explanations, and what data cannot show
• Produce a short report with maps and graphs, in any tool

Logistics

• When: Monday to Friday, 13:10–15:50 · Where: Aulas 11201
• Instructor: Bogdan Popescu · bgpopescu@tec.mx · office hours by appointment
• Course site: bgpopescu.net/teaching/diversity, one page per session
• Reading: one article per session, read together in class
• Software: Excel, Tableau or, ideally, R

How you are graded

Contributions to class — the average of:

• Attendance
• Participation in the in-class exercises
• Progress on your report

The team report

Teams of three. One report, made with any tool, containing:

• Two to three maps
• One simple graph and one more complex graph
• An interpretation in your own words
• A table declaring your use of AI

Where the report marks are

Worked examples

Three finished reports built from this course’s data, made in R. Copy the structure in any tool.

• The China Shock — three maps and two graphs; the cleanest template
• Whose jobs? — the same data, a question the authors did not ask
• Ask or delegate? — global AI use, and who the data describes

The final presentation

Friday. Ten minutes per team, plus questions. Follow your report:

1. The question — one sentence
2. The maps — where the dark areas are, and what that means
3. The graph — the relationship you found
4. What you cannot say — the limitation that most threatens your conclusion
5. One thing that surprised you

Where the presentation marks are

Use of AI

You may use AI, such as ChatGPT. You must declare it.

So what?

• The report is 40% of your grade, due Friday at 13:10
• Today’s graphs and maps use the data you will receive

Two numbers

• Trade with China made almost everything in this room cheaper
• It also destroyed manufacturing jobs in particular towns

Today’s objectives

• Describe where the China shock landed and how concentrated it was
• Explain what happened to workers in the places it hit
• Trace how the same shock reached Mexico, through the US market
• Compare a shock that follows local industry with one that does not
• Say what each dataset can and cannot show

What happened

• China joined the World Trade Organization in 2001
• Its manufactured exports rose faster than any country’s in history
• American factories competed with them, and many lost

The scale of it

Yangshan deep-water port, Shanghai. Source: Wikimedia Commons, public domain.

How we measure the shock?

The basic idea is simple:

Which places were most exposed to competition from Chinese imports?

How we measure the shock?

The measure asks:

How much did Chinese imports rise in the industries that workers in this local area depended on?

It then divides that increase by the number of workers in the area.

How we measure the shock?

• The United States is divided into 722 commuting zones — local labour markets where people live and work.
• Each zone gets its own import-exposure score.
• The study compares two periods: 1990–2000 and 2000–2007.

What the score actually is

That score has a unit: dollars of Chinese imports per local worker.

What the score is not

• It is not a pay cut. Nobody handed over $2,640
• It is not money the zone lost
• It is the size of the wave arriving, divided by the number of people standing in it

Where the China shock hit hardest

The shock was not evenly spread across the United States.

Where it landed

Figure 1

Where it landed

So when you look at the map, read it like this:

• Darker / higher-exposure areas = places whose existing industries faced a larger increase in Chinese import competition.
• Lighter / lower-exposure areas = places whose local industries faced much less of that competition.

A few places took most of it

Extra Chinese imports arriving each year by 2007, per local worker:

• The average zone: $2,640
• The hardest-hit zone: $43,000 — more than 16 times the average
• The middle zone: $1,900 — almost two in three zones came in below the average

A few places took most of it

Figure 2

Those 72 red bars absorbed a third of the country’s entire shock.

A few places took most of it

Where the red bars are:

• Two-thirds of the worst-hit zones are in the South
• Most of the rest are in the Midwest
• Only 2 of 72 are in the West

So what?

• A national average hides where a shock actually lands
• The unit you choose decides what you can see
• Commuting zones show it; states would not

What decline looks like

The abandoned Packard plant, Detroit. Source: Wikimedia Commons, CC BY-SA 3.0, Albert duce.

More exposure, fewer factory jobs

Figure 3

The relationship got stronger

Figure 4

Two ways to read the same loss

Figure 5

Why the two panels disagree

• Men started with 14.9% of the population in manufacturing; women 7.2%
• Both lost roughly the same number of percentage points
• Relative to what each group had, women lost twice as much

Exercise 1 — Make a prediction

Your town’s only factory closes. You are forty, and you worked there.

Rank these four, from most to least common:

1. Move to another city
2. Find other work nearby
3. Become unemployed
4. Stop looking for work

If your ranking is right, what should the data show?

What happened in more-exposed places

Figure 6

The mechanism

%%{init: {"flowchart": {"useMaxWidth": true, "htmlLabels": true, "padding": 24,
                        "nodeSpacing": 22, "rankSpacing": 90,
                        "width": 1150, "height": 640},
          "themeVariables": {"fontSize": "21px", "fontFamily": "Arial, Helvetica, sans-serif"}} }%%
flowchart LR
  A["Imports<br/>arrive"] --> B["The plant<br/>closes"]
  B --> C["A new job<br/>nearby"]
  B --> D["Move to<br/>another city"]
  B --> E["Stop looking<br/>for work"]
  C --> F["Adjustment<br/>works"]
  D --> F
  E --> G["Adjustment<br/>fails"]

So what?

• Unemployment and labour-force exit rose about equally in hit places
• Neither moving away nor new local jobs absorbed the loss
• A place can lose its economy and keep its population

Mexico was selling to the same customer

• NAFTA (1994) tied Mexican factories to the US market
• Maquiladoras assembled clothes, televisions and car parts for export
• After 2001, Chinese factories sold into that same market

2003: the year China passed Mexico

Figure 8

What happened inside Mexico

• Maquiladora employment and growth fell (Utar & Torres Ruiz 2013)
• The least skilled plants were hit hardest
• Exposed regions lost manufacturing employment (Méndez 2015)
• The loss was larger where they competed with China in the US

Where manufacturing employment changed in Mexico

This is a descriptive approximation of the pattern discussed by Méndez (2015).

Where did manufacturing employment actually grow or shrink between 1998 and 2008?

Where manufacturing employment changed in Mexico

Figure 9

Where manufacturing employment changed in Mexico

The map should be read carefully:

• Red states lost manufacturing jobs between 1998 and 2008.
• Green states gained manufacturing jobs.
• Only four states lost jobs; the country gained 12%.
• This does not by itself show that Chinese competition caused the change.

Growth can slow without reversing

Mexico gained manufacturing jobs overall. Exposed regions still lost out.

So what?

• The same measure travels: exposure follows what a place produced
• Mexico was hit through a third market, not at home
• And workers moved — remember that on Wednesday

AI is not a factory

The question changes

For the China shock we asked: where did it land?

The data

Most research on AI guesses which jobs a machine could do. This data records what people actually asked it for.

Where the world’s AI use comes from

Figure 10

Where it is used for schoolwork

Figure 11

Almost the opposite map

• The United States is 20% of world use; 9% of it is coursework
• Tunisia is 0.3% of world use; 52% of it is coursework
• Mexico: 1.3% of world use, coursework 21%

Exercise 2 — Can you say this?

52% of Tunisian AI use is coursework — the highest share in the world.

Which of these sentences does that number support?

1. Most Tunisians use AI for schoolwork.
2. Most Tunisian conversations with Claude are schoolwork.
3. Tunisians use AI for school more than Americans do.

Mark each one yes or no. Rewrite the ones that fail.

The trap in that number

• 1. No — it counts conversations, not people, and only Claude’s
• 2. Yes — that is exactly what the 52% says
• 3. Careful — a bigger share of a much smaller amount

The trap in that number

• Claude users in Tunisia are young, urban, connected — not a cross-section
• The China shock counted every worker in a place
• Here the denominator is whoever showed up

So what?

• Trade exposure was fixed by a place’s existing industries
• AI exposure looks more diffuse — harder to pin to a map
• Use still varies by place, and how it is used varies more
• And we see users, not populations

They are not the same shape

%%{init: {"flowchart": {"useMaxWidth": true, "htmlLabels": true, "padding": 24,
                        "nodeSpacing": 26, "rankSpacing": 95,
                        "width": 1150, "height": 640},
          "themeVariables": {"fontSize": "21px", "fontFamily": "Arial, Helvetica, sans-serif"}} }%%
flowchart LR
  A["Trade<br/>shock"] --> B["Follows local<br/>industry"]
  B --> C["You can<br/>map exposure"]
  D["AI"] --> E["Not tied to<br/>local industry"]
  E --> F["You must ask<br/>who uses it, and how"]

What each dataset can show you

What we can say

• Import exposure and factory job loss move together, strongly
• The relationship is far stronger after 2001 than before
• Hard-hit places lost workers from the labour force, not to other cities

What we cannot say

• That trade alone caused it — automation was happening too
• That the same towns would have thrived otherwise
• That AI will repeat it: different shape, different data, no outcomes yet

Takeaways

1. Globalization’s gains were spread; its losses were concentrated
2. The unit you measure decides what you can see
3. Denominators decide headlines — ask who is counted
4. Hard-hit places did not empty out; they stopped working
5. AI is a different shape of shock, and it is early