the education premium

[nsacpi]

what has happened to the education premium since 1970

The education premium (usually called the college wage premium)—the wage advantage of people with a college degree over those with only a high-school diploma—has gone through three distinct phases since about 1970.

---

1. 1970–1980: The premium was relatively modest​

Around 1970, the wage advantage of a college degree was much smaller than today.

• College graduates earned roughly 40–60% more than high-school graduates.
• In some years during the late 1960s and early 1970s the premium even declined slightly, partly because the supply of college graduates was growing quickly.

This period was characterized by:

• Strong unions
• A compressed wage structure
• Manufacturing jobs that paid relatively well even without college

So the U.S. economy was less stratified by education.

---

2. 1980–2000: The premium rose dramatically​

This is the big structural change.
From about 1980 onward:

• The college wage premium rose rapidly.
• It increased from about 50% in 1979 to roughly 90–100% by 2000.

In other words:

• By 2000, a typical college graduate earned about double the wages of a high-school graduate.

Economists attribute this to several forces:
Skill-biased technological change

• Computers increased demand for analytical and technical workers.

Globalization

• Manufacturing jobs moved abroad or faced wage pressure.

Institutional changes

• Decline of unions
• Falling real minimum wage

Slower growth in college-educated labor supply

• Demand for educated workers grew faster than supply.

This surge in the education premium was a major driver of rising U.S. wage inequality.

---

3. 2000–present: The premium stabilized (very high but no longer rising quickly)​

Since around 2000:

• The premium has leveled off at historically high levels.
• It generally fluctuates around 80–90% higher wages for college graduates.

Important nuance:

• Wage inequality within education groups (especially among college graduates) has increased.
• So the college vs. non-college gap is no longer the main driver of inequality growth.

Recent data also suggests some short-term narrowing, as wages for skilled trades rise and some white-collar sectors soften.

---

The long-run picture (very simple summary)​

Period	College wage premium	Trend
1970	~40–60%	Moderate
1980–2000	~50% → ~100%	Rapid increase
2000–2025	~80–90%	High but stable

---

The key historical insight:
The modern U.S. class structure—where a bachelor’s degree strongly separates income groups—is largely a post-1980 phenomenon.

---

If you want, I can also show you something interesting economists discovered: the education premium used to be very high around 1910, then fell for 70 years before rising again after 1980. That long cycle explains a lot about U.S. inequality history.

 

[nsacpi]

is part of it perhaps that those people not getting a college education today are different from those who didn't back in 1970...how do researchers try to control for that

Yes—this is a major issue in the literature, and economists spend a lot of effort trying to address it. The concern is called selection bias or ability bias.


Your intuition is exactly right:

• In 1970, a relatively small and highly selected group went to college.
• Today, college attendance is much more common, meaning the average ability and background of the non-college group may be different than it used to be.

So if we simply compare wages of college vs. non-college workers, we might attribute too much of the difference to education itself rather than to who goes to college.


Researchers use several methods to try to control for this.

---

1. Controlling for observable characteristics​

The most basic approach is regression analysis.


Researchers compare workers while controlling for variables like:

• test scores or cognitive ability
• parental education
• family income
• race and gender
• region
• high-school GPA

This approach is used in large surveys such as:

• U.S. Census Bureau datasets
• National Center for Education Statistics longitudinal surveys

Examples include:

• NLSY79 (people born 1957–1964)
• NLSY97 (people born 1980–1984)

These datasets include AFQT test scores, which allow economists to partially control for ability.


But this still leaves unobserved differences.

---

2. Sibling and family comparisons​

Another method is family fixed-effects models.


Researchers compare:

• siblings where one went to college and the other did not.

This helps control for:

• family background
• neighborhood
• parental education
• some genetic traits

These studies typically find the college premium falls somewhat but remains large.

---

3. Natural experiments (instrumental variables)​

A powerful strategy uses quasi-random variation in college attendance.


Economists exploit factors that influence education but are not directly related to earnings ability, such as:

• distance to a college
• changes in tuition
• GI Bill eligibility
• draft deferments during the Vietnam War

These methods were pioneered by economists like:

• David Card
• Alan Krueger

Their research generally finds that the causal return to college is still very large, though slightly smaller than naive comparisons.

---

4. Quantile and distributional analysis​

Another issue related to your point:


As college expanded, the composition of the non-college group changed.


Researchers study:

• wage differences across the entire distribution
• changes within education groups

Work by economists such as:

• Claudia Goldin
• Lawrence Katz

shows that much of the rising premium reflects demand for skill increasing faster than supply, not just selection.

---

5. Structural models of selection​

More advanced research builds models that explicitly account for who chooses college.


These models estimate:

• ability
• preferences
• expected wages

They then simulate what wages would look like if the same people attended college as in earlier decades.


These studies typically conclude:


Selection explains some of the premium, but not most of it.

---

The bottom-line consensus​

Most studies find roughly:

• 10–30% of the college premium may reflect selection
• 70–90% reflects the causal return to education and skill demand

So your intuition is correct—but the evidence suggests the large rise since 1980 is not mainly due to changing composition.

 

[nsacpi]

A deeper (and fascinating) question related to your observation is this:


The non-college group today has become increasingly male, less urban, and more regionally concentrated.


This compositional change may explain a lot of the political and cultural polarization between college and non-college America.


If you'd like, I can also show you a striking fact economists discovered:
The average cognitive test scores of the non-college population have fallen sharply since the 1960s because college attendance expanded so much.

 

[nsacpi]

ok tell me more about the average cognitive test scores of non-college people...who documented this

Yes—this phenomenon has been carefully documented, and it’s one of the most important compositional changes in the U.S. labor force since 1970.
The key point is:

As college attendance expanded, higher-ability individuals increasingly sorted into the college group, which mechanically lowered the average measured ability of the non-college group.

Researchers demonstrated this using cognitive test scores in longitudinal surveys, especially the AFQT (Armed Forces Qualification Test).

---

The Key Evidence​

1. Claudia Goldin and Lawrence Katz​

The most influential documentation comes from work by

• Claudia Goldin
• Lawrence Katz

Their book:

• The Race between Education and Technology (2008)

uses data from the NLSY cohorts and other historical sources.
Their main finding:

• In the 1960s, many people in the top half of the cognitive distribution did not attend college.
• By the 2000s, the majority of high-scoring individuals go to college.

So the ability composition of the non-college group changed dramatically.

Simplified illustration​

Year	Share of high-ability youth going to college
~1960	~30–40%
~1980	~50–60%
~2000	~70–80%
today	often >80%

This means the non-college population today is much more negatively selected on ability than in 1970.

---

Direct Evidence Using AFQT Scores​

Another influential study uses the National Longitudinal Survey of Youth run by the

• U.S. Bureau of Labor Statistics.

Economists compared two cohorts:

• NLSY79 (youth in the late 1970s)
• NLSY97 (youth in the late 1990s)

They found:

Average AFQT percentile of groups​

Group	1979 cohort	1997 cohort
College graduates	~70th percentile	~75–80th
Non-college workers	~45th percentile	~35th percentile

So the average ability of the non-college group fell noticeably.
This shift is sometimes called:
“cognitive sorting by education.”

---

Research by Charles Murray​

The sociologist

• Charles Murray

popularized this pattern in:

• Coming Apart: The State of White America, 1960–2010

He showed that:

• College attendance increasingly concentrates people from the top of the cognitive distribution in metropolitan professional classes.

While Murray’s broader social interpretations are controversial, the sorting fact itself is widely accepted in labor economics.

---

Research by David Autor and coauthors​

Economist

• David Autor

has also emphasized this compositional shift.
His work on labor market polarization shows:

• The non-college workforce increasingly consists of people from the lower half of the skill distribution, while the college workforce increasingly captures the upper half.

---

Why This Matters for the Education Premium​

This sorting creates a statistical exaggeration of the college wage premium.
Part of the wage gap reflects:

1. Education itself
2. Ability differences between groups

Because ability differences between the groups grew larger over time, simple comparisons overstate how much the return to education increased.
But researchers find:

• The sorting explains some but not most of the rising premium.

---

A Striking Historical Comparison​

In 1970:

• Many bright students from working-class backgrounds did not attend college.
• They often went into unionized manufacturing, trades, or the military.

By 2020:

• Most of those individuals do attend college.

So the non-college group lost a large share of its high-ability members.

---

One striking statistic from the literature:
Among people in the top quartile of test scores, college completion rates roughly:

• doubled or tripled between 1960 and today.

That change dramatically reshaped the composition of the workforce.

---

If you want, I can also show you something very interesting that many economists think is even more important than cognitive sorting:
the geographic sorting of college graduates into a small number of “superstar cities.”
That change interacts strongly with the education premium and may explain a lot of modern U.S. inequality.

 

[nsacpi]

ok tell me more about the geographic sorting

Geographic sorting of educated workers is one of the big structural changes in the U.S. economy since about 1980. The basic pattern is:

College-educated workers have become increasingly concentrated in a relatively small number of metropolitan areas, while non-college workers are more geographically dispersed.

This has major implications for wage inequality, regional divergence, and politics.

---

1. The Core Fact: College graduates cluster together​

Research by economists such as

• Enrico Moretti
• Edward Glaeser
• David Autor

shows that the share of college graduates varies enormously across cities, and the gap has widened since 1980.
Examples (approximate recent shares of adults with bachelor’s degrees):

Metro area	College share
San Jose	~55–60%
San Francisco	~50%
Boston	~50%
Washington	~60%
many small metros	~20–25%

So the U.S. workforce has become highly stratified geographically by education.

---

2. Why this clustering happens​

A. Knowledge spillovers​

Educated workers benefit from being near other skilled workers.
Economists call this agglomeration effects.
Evidence shows:

• wages rise when skilled workers cluster
• innovation rises sharply in dense knowledge hubs

This is central in Moretti’s book:

• The New Geography of Jobs

---

B. Superstar tech and knowledge industries​

Certain industries concentrate geographically.
Examples:

• tech in San Jose / Silicon Valley
• biotech in Boston
• finance in New York
• government/policy in Washington

These industries disproportionately hire college graduates.

---

C. Migration patterns of graduates​

College graduates are far more geographically mobile.
Young graduates disproportionately move to:

• large metros
• innovation hubs
• places with strong labor markets

Meanwhile, non-college workers move much less.

---

3. The divergence accelerated after 1980​

Before 1980:

• regional income levels in the U.S. were converging.

Poor regions were catching up.
After 1980:

• convergence largely stopped or reversed.

High-skill cities pulled ahead dramatically.
Research shows that cities with more educated populations in 1980 grew faster afterward.

---

4. The multiplier effect​

One of the most striking findings from
Enrico Moretti:

Each high-skill job in a tech hub creates about 3–5 additional local service jobs.

Examples:

• doctors
• lawyers
• restaurant workers
• construction workers

But these service jobs often pay much less, which increases within-city inequality.

---

5. Housing constraints intensified the divide​

Many high-skill cities severely restrict housing construction.
Examples include:

• San Francisco
• Boston
• Seattle

This creates two effects:

1. Rising housing prices
2. Limited migration into high-productivity regions

Economists estimate U.S. GDP could be several percent higher if housing supply in these cities were less constrained.

---

6. Political consequences​

Because education is geographically sorted, voting patterns also sort.
Highly educated metro areas vote very differently from less-educated regions.
This helps explain the political geography seen in recent elections, especially the divide between:

• major metro counties
• smaller cities and rural areas

---

7. An especially interesting pattern​

Economists have discovered that college graduates increasingly marry other college graduates and move to the same cities.
This phenomenon is called:
assortative mating + spatial sorting
It further concentrates:

• education
• income
• professional networks

in a small number of metropolitan regions.

---

The big picture
Since about 1980 the U.S. has developed:

• educational stratification (college vs non-college)
• geographic stratification (high-skill cities vs everywhere else)

And the two reinforce each other.

---

If you want, I can also show you something fascinating that economists discovered recently:
the number of U.S. counties producing large numbers of patents has actually shrunk since 1980—even though total innovation has risen.
That tells us just how concentrated the knowledge economy has become.

 

[nsacpi]

are these trends even more pronounced for those workers with graduate degrees

Yes—the geographic sorting is even more pronounced for workers with graduate and professional degrees. In fact, the higher the level of education, the more geographically concentrated the workforce tends to be.
Economists studying spatial inequality have repeatedly found a “skill gradient in agglomeration”: the most highly educated workers cluster the most.
Researchers such as

• Enrico Moretti
• Edward Glaeser
• David Autor

have documented this pattern using census and labor-market data.

---

1. Concentration rises with education level​

A stylized version of the pattern looks like this:

Education level	Geographic concentration
High school or less	widely dispersed
Bachelor’s degree	moderately concentrated in large metros
Master’s degree	strongly concentrated
PhD / professional degree	extremely concentrated

Workers with PhDs, MDs, and advanced technical degrees are disproportionately located in a relatively small set of metro areas.
Examples include:

• San Jose
• San Francisco
• Boston
• Seattle
• Washington

These places have very high shares of residents with graduate degrees.

---

2. Why advanced degrees cluster even more​

A. Specialized labor markets​

Highly specialized fields require thick labor markets.
Example:

• an AI researcher
• a biotech scientist
• a quantitative finance specialist

There may be dozens of relevant employers in a few cities, but none in most places.
This pushes those workers toward clusters.

---

B. Research universities​

Many graduate-degree jobs are linked to major research universities.
For example:

• Massachusetts Institute of Technology
• Stanford University
• University of California, Berkeley

Universities create ecosystems of:

• startups
• labs
• venture capital
• research hospitals

That ecosystem attracts highly educated workers.

---

C. Innovation spillovers​

Research shows that innovation productivity rises when researchers cluster.
For example, patent and publication output rises sharply in dense research clusters.
This effect is especially strong in fields like:

• AI
• biotech
• semiconductors

---

3. Evidence from patenting and research activity​

Studies show that innovation has become increasingly concentrated geographically.
A relatively small number of metro areas now produce a large share of patents.
Research by economists including
Nicholas Bloom
and
Scott Stern
documents the rise of “innovation hubs.”
These hubs disproportionately employ workers with PhDs and advanced STEM degrees.

---

4. The “superstar city” effect​

Economist
Enrico Moretti
describes a reinforcing dynamic:

1. A city gains a cluster of highly educated workers.
2. Firms move there to access talent.
3. More educated workers move there for jobs.
4. Innovation increases.
5. Wages rise further.

This feedback loop creates superstar cities.

---

5. Graduate-degree sorting has accelerated recently​

The concentration of advanced-degree holders increased especially after 1990 with the rise of:

• the internet
• biotech
• venture capital ecosystems
• AI research clusters

For example, the AI research workforce is heavily concentrated around:

• Silicon Valley
• Seattle
• Toronto
• London

---

6. The extreme tail: PhDs​

For PhD holders, concentration is extremely high.
A relatively small number of metropolitan areas account for a very large share of all U.S. PhD employment.
This is especially true in fields like:

• computer science
• biotech
• economics
• physics

---

Bottom line
The higher the level of education, the stronger the geographic clustering:

• BA holders cluster somewhat
• MA holders cluster more
• PhD and professional degree holders cluster the most

This means the knowledge economy is even more geographically concentrated than the college premium alone would suggest.

---

If you're interested, there’s an especially striking statistic economists discovered: roughly half of U.S. innovation (measured by patents) comes from about 20 metropolitan areas. That concentration has increased sharply since 1980.

 

[nsacpi]

what role does Venture Capital play in this geographic and social sorting

Venture capital (VC) plays a very large role in reinforcing both the geographic clustering of talent and the social sorting of highly educated workers. In many ways, VC acts as a concentration mechanism in the innovation economy.

Economists studying innovation and regional growth—such as

• Enrico Moretti
• Josh Lerner
• Scott Stern

have shown that venture capital tends to amplify existing innovation hubs rather than create new ones.

---

1. Venture capital is extremely geographically concentrated​

A striking fact: the vast majority of venture capital investment happens in a small number of regions.

The biggest hubs include:

• Silicon Valley
• San Francisco
• New York
• Boston

Together these areas account for a very large share of U.S. VC funding.

Even within Silicon Valley, activity is concentrated around places like:

• Palo Alto
• Menlo Park
• Mountain View

---

2. Why venture capital clusters geographically​

A. High-information investing​

Venture investing relies heavily on:

• informal networks
• reputation
• rapid information flow
• repeated interactions

Historically, investors strongly preferred founders they could meet frequently in person.

Research shows that VCs strongly prefer startups located within a short travel distance of their offices.

This creates strong local investment ecosystems.

---

B. Networks of founders and investors​

Venture capital is embedded in dense social networks of:

• entrepreneurs
• engineers
• universities
• lawyers
• early employees

For example, many startups are founded by alumni of companies like:

• Google
• Facebook
• PayPal

This creates repeated founder–investor relationships that reinforce clustering.

---

C. Serial entrepreneurship​

Successful founders often become investors.

Examples:

• founders become angel investors
• angel investors start venture funds

This creates self-reinforcing innovation ecosystems.

---

3. Venture capital amplifies talent migration​

Because funding is concentrated geographically, entrepreneurs and technical workers move to where the money is.

For example, many founders relocate startups to:

• Silicon Valley
• San Francisco

even if the company originated elsewhere.

This further concentrates:

• engineers
• data scientists
• product managers
• AI researchers

---

4. Universities + VC ecosystems​

Venture capital strongly interacts with major research universities.

Examples include ecosystems around:

• Stanford University
• Massachusetts Institute of Technology
• University of California, Berkeley

These universities produce:

• graduate students
• spinout startups
• patentable research

VC firms finance the commercialization.

This further concentrates PhD-level talent geographically.

---

5. Venture capital also reinforces social sorting​

VC-backed firms disproportionately recruit from:

• elite universities
• elite tech companies
• elite professional networks

This creates social clustering among people with:

• advanced degrees
• high technical skills
• similar educational backgrounds

Economists sometimes describe this as “elite network concentration.”

---

6. Winner-take-most outcomes​

VC funding tends to produce extreme success for a small number of firms.

That dynamic favors regions with:

• deep talent pools
• large VC ecosystems
• experienced founders

Because investors want the highest probability of very large outcomes, they tend to fund startups in already successful ecosystems.

---

7. Attempts to spread venture capital geographically​

Governments and universities have tried to replicate Silicon Valley through:

• startup incubators
• public venture funds
• technology parks

Examples include efforts in cities like:

• Austin
• Pittsburgh
• Denver

Some have succeeded partially, but very few ecosystems approach Silicon Valley’s scale.

---

The big picture​

Venture capital acts as a powerful reinforcing mechanism in the knowledge economy:

1. Talent attracts venture capital.
2. Venture capital attracts more startups.
3. Startups attract more talent.
4. The ecosystem becomes self-sustaining.

The result is strong geographic and social clustering of highly educated workers.

---

One especially striking statistic:
A handful of U.S. metro areas account for the majority of venture capital investment and startup exits.

That concentration helps explain why innovation—and many of the highest-paying jobs—are clustered in relatively few places.

---

If you want, I can also explain a less obvious effect: how venture capital has changed the career paths of top STEM graduates—especially shifting them away from corporate R&D and toward startups. That shift has quietly reshaped the American innovation system.

 

[nsacpi]

How does policy at the state and local level affect this sorting and stratification?

State and local policy affects geographic and social sorting mainly through housing rules, education systems, taxation, infrastructure, and local economic development policy. These policies shape who can move where, who stays, and which industries grow in a region.


Economists studying regional divergence—such as Enrico Moretti, Edward Glaeser, and Chang-Tai Hsieh—often argue that local policy is one of the biggest drivers of the modern U.S. regional divide.


Below are the main mechanisms.

---

1. Zoning and housing supply (the most powerful local policy)​

Local zoning rules determine how much housing can be built and what types.


Many high-productivity regions have restrictive zoning such as:

• single-family-only neighborhoods
• height limits
• parking requirements
• long permitting processes

Cities often cited in the research include:

• San Francisco
• Boston
• San Jose

Effects:

• housing supply grows slowly
• prices rise dramatically
• only high-income workers can move in

This produces economic sorting by income and education.


Research suggests many workers who would benefit from moving to high-productivity regions simply cannot afford to live there.

---

2. Local public schools and residential sorting​

Public school quality is tied to school districts, which strongly influence where families live.


High-performing districts tend to have:

• higher property values
• more college-educated parents
• stronger local tax bases.

Examples of metro areas where school-district sorting is well documented include:

• Boston suburbs
• New York suburbs
• San Francisco Bay Area suburbs.

Because housing prices determine access to schools, school districts become a mechanism of income and education sorting across neighborhoods.

---

3. State university systems​

State governments strongly influence where highly educated workers are produced and retained.


Examples:

• University of California helped anchor the California tech ecosystem.
• University of Texas at Austin helped build the tech cluster in Austin.
• Carnegie Mellon University has been central to robotics and AI growth in Pittsburgh.

State funding for research universities influences:

• startup formation
• patent activity
• local venture capital ecosystems.

---

4. State tax policy​

States vary widely in their tax structures.


Examples:

• high-income taxes in states like California
• lower taxes in states like Texas or Florida.

However, research suggests taxes matter less than agglomeration effects. Highly skilled workers often remain in high-tax regions because those regions offer:

• better job opportunities
• stronger innovation networks.

Still, tax policy can affect firm location decisions and migration patterns at the margin.

---

5. Economic development incentives​

States and cities often try to attract firms through subsidies and tax incentives.


Examples include bids to attract large employers such as:

• Amazon during the search for its HQ2 location.

Cities competing included:

• New York
• Arlington
• Austin.

These incentives can shift where new high-skill jobs appear, though economists debate how effective they are.

---

6. Transportation and urban planning​

Local infrastructure decisions affect urban density and labor markets.


Examples:

• transit systems in New York and Boston support dense knowledge economies.
• car-dependent metro areas may spread workers over larger regions.

Transportation policy affects how easily workers can access high-skill job clusters.

---

7. Licensing and occupational regulation​

States regulate many professions through licensing requirements.


These rules affect labor mobility in occupations such as:

• healthcare
• law
• education
• construction trades.

Licensing differences between states can discourage workers from moving, reinforcing regional labor market differences.

---

8. Local quality-of-life policy​

Policies affecting amenities can attract or repel skilled workers.


Examples include:

• environmental regulations
• parks and public spaces
• cultural institutions
• crime and policing policy.

Highly educated workers often prioritize amenities and lifestyle, which influences city growth.

---

Big picture​

State and local policies influence sorting through three main channels:

1. Housing constraints → who can afford to live in productive regions
2. Education systems → where skilled workers are produced
3. Local economic ecosystems → where high-skill jobs cluster

Together these policies help shape the modern U.S. pattern where highly educated workers cluster in a relatively small number of metropolitan areas while other regions fall behind economically.

---

If you're interested, there is a particularly important policy debate right now among economists: whether housing reform in high-productivity cities could significantly reduce U.S. inequality and increase national productivity. The estimates of the potential gains are surprisingly large.

 

[Tapate50]

Premium is right

Student loans go to pay bloated staff and excess curriculum

Good Ponzi scheme if you can get on board