Steve Levit — Abortion and Crime, Revisited

In one sentence

The dramatic drop in American violent crime starting in the 1990s was significantly driven by the legalization of abortion, which reduced the number of children born into environments where they were not wanted or prepared for.

Why read this (and who it's for)

This summary is for anyone interested in the "hidden" causalities of social policy, criminology, or public health. It solves the mystery of why crime plummeted so unexpectedly in the 1990s by identifying a massive, non-obvious variable: the "unwantedness" effect. You will be able to look at social problems through the lens of multivariate causality—understanding that complex societal shifts are rarely caused by a single factor, but by a web of interacting variables.

The big idea

The central thesis argued by Steve Levit and John Donahue is that the legalization of abortion in the 1970s led to a significant reduction in the number of children born into "unwanted" circumstances. This, in turn, led to a measurable decrease in violent crime rates decades later.

The logic follows a specific causal chain:

1. Access: Legalization allowed women to choose when and if to bring a child into the world.
2. Unwantedness: Research shows that children born to parents who are not ready or do not want them are statistically more likely to have poor health/educational outcomes and higher rates of criminal behavior.
3. Lagged Effect: Because crime is often a result of developmental environments, the impact of a 1970s policy change on the population's behavior became visible in the 1990s as those cohorts reached "crime ages."

Levit argues that while other factors (policing, crack epidemic, economy) played roles, the abortion effect is a massive, independent variable that explains a significant portion of the crime drop.

Key ideas

The "Unwantedness" Effect

• What it is: The social science finding that children born to parents who are unprepared or unwilling to parent them face significantly worse life outcomes.
• Why it matters: It shifts the focus from "bad" individuals to the environmental and social conditions of birth. If you accept this, you recognize that social stability is tied to the readiness of the support system surrounding a child.
• In the book: Donahue and Levit cite research showing that "unwanted" kids are more likely to engage in criminal behaviors as they grow up.

Multivariate Causality

• What it is: The principle that complex social phenomena (like a national crime rate) are never caused by a single factor, but by many overlapping variables.
• Why it matters: It prevents "single-cause" thinking, which often leads to oversimplified policy solutions that fail to address the actual root causes.
• In the book: Levit explains that while the economy, policing, and the crack epidemic all contributed, they were not the only factors; we must look at the "collage of evidence" to see the full picture.

The "Collage of Evidence"

• What it is: Because you cannot run a randomized controlled trial on abortion, researchers must assemble a "collage" of different, imperfect data points to infer causality.
• Why it matters: It teaches you how to think like an economist: look for "natural experiments" and consistent patterns across different datasets.
• In the book: Levit and Donahue compared states with high vs. low abortion rates, found that the high-abortion states saw a much sharper crime drop, and then verified this by looking at specific birth cohorts.

The 15-20 Year Lag

• What it is: The recognition that policy changes affecting birth/childhood (like abortion or lead exposure) won't show results in crime statistics until 15–20 years later.
• Why it matters: It teaches patience in policy evaluation. You cannot judge a 1970s policy by 1980s data.
• In the book: This is why the 1990s crime drop was the perfect window to measure the impact of 1970s abortion legalization.

The Lead Pollution Parallel

• What it is: A secondary finding that the removal of lead from gasoline also contributed to the crime drop.
• Why it matters: It reinforces the idea of multivariate causality—both abortion and lead removal were happening simultaneously and both contributed to the result.
• In the book: Professor Jessica Walpa Rees found that the "abortion effect" remains significant even when the "lead effect" is accounted for.

Frameworks & tools

The "Collage of Evidence" Framework for Causal Inference

When trying to determine if Factor A caused Result B in a complex system, do not look for a "smoking gun." Instead, build a collage:

1. Identify a Natural Experiment: Find a situation where a policy changed in some places but not others (e.g., states legalizing abortion at different times).
2. Compare Groups: Divide the subjects into groups (High, Medium, Low) based on the intensity of the variable (Abortion rates).
3. Track Trends: See if the groups with the highest "treatment" (High abortion rates) showed the most significant change in the outcome (Crime) over time.
4. Cohort Analysis: Verify the data by looking at specific birth years to see if the "treatment" in a specific year correlates with the "result" in that specific age group.

Memorable examples & stories

The 1990s Crime Explosion vs. Collapse

In 1990, New York City had over 2,200 homicides; by 2023, it had fewer than 400. This wasn't a localized fluke; it was a national trend. Lesson: When a massive, unexpected shift occurs, look for a major policy shift that happened 20 years prior.

The "Impossible" Research

Levit and Donahue were criticized by both the "pro-life" and "pro-choice" camps. Pro-life advocates were upset that the study implied a positive outcome from abortion; pro-choice advocates were upset that the study suggested a "moral" benefit to a "tragic" act. Lesson: Data is often stripped of its context by those who only care about the "moral" conclusion.

The "Wrong" Table

Levit admits he and Donahue made a technical "coding error" in their original 2001 paper regarding how they controlled for state effects. However, when they re-ran the data correctly, the results were actually stronger. Lesson: A mistake in execution doesn't mean the underlying hypothesis is wrong; it just means the measurement was flawed.

Key quotes

"The fact that you know... there is an issue here, and it's sort of useful to be able to figure out what the truth is." (Paraphrased: The importance of pursuing "uncomfortable" data to find the truth.) "I think there's actually a lot less disagreement about facts than about the interpretation of the facts." (Paraphrased: Public debate is often a battle of interpretations rather than a battle over the data itself.) "The mechanism by which any effects on crime have to be happening here are the women making good choices." (Levit: Explaining that the "unwantedness" effect is a result of women choosing when they can provide a good environment.)

How to apply it

This week

• Identify a "Single-Cause" Narrative: Find a news story where a complex problem (e.g., housing prices, education) is being blamed on one single factor.
• Apply Multivariate Causality: Write down three other factors that might be contributing to that same problem.
• Audit your "Right/Wrong" Hat: When reading a controversial topic, ask yourself: "Am I looking for a fact here, or am I looking for a reason to justify my pre-existing moral stance?"

Ongoing

• The 20-Year Rule: When evaluating social policy (education, health, housing), remind yourself that the "results" of the policy may not appear for two decades.
• Data over Interpretation: When engaging in debate, try to lead with "Here are the five facts we know" before offering your interpretation of what those facts mean.

Common pitfalls

• The "One Cause" Fallacy: People often want a single "hero" or "villain" (e.g., "It's just the crack epidemic"). This ignores the complexity of human systems.
• Moralizing Data: People often interpret the implications of data as a moral endorsement. Levit argues that a study can show a correlation without endorsing the morality of the act itself.
• Ignoring the Lag: Expecting immediate results from social policy changes often leads to the false conclusion that the policy "didn't work."

Connects to

• The Broken Windows Theory: Challenges the idea that policing alone is the answer, suggesting instead that social conditions (like "unwantedness") are the root.
• Lead Exposure Research: Complements the argument by showing how environmental toxins (Lead) and social policy (Abortion) both independently influenced the crime drop.

If you remember only three things

• Unwantedness: Children born into unprepared environments are significantly more likely to engage in criminal behavior.
• Multivariate Causality: Complex social problems (like crime) are never caused by a single factor; they are a "collage" of many overlapping influences.
• The 20-Year Lag: Significant social changes (like abortion access) may not show their full impact on crime statistics for two decades.

Internalize it

• Reflection questions
  • If I believe a social problem is caused by a single factor, what other three factors am I ignoring to make that story simpler?
  • How does the "unwantedness" concept change how I view the responsibility of the state versus the responsibility of the individual?
  • When I hear a "moral" argument about a policy, can I strip away the moral language and look at the raw data being presented?
• Flashcards
  • Q: What is the "unwantedness" effect? A: The finding that children born to parents who are not ready or do not want them are more likely to have poor outcomes and higher crime rates.
  • Q: Why did the 1990s crime drop provide a perfect window for studying abortion? A: Because there is a 15-20 year lag between the 1970s legalization of abortion and the age at which those children would reach "crime ages."
  • Q: What is "multivariate causality"? A: The principle that complex social phenomena are caused by multiple interacting factors rather than a single cause.
  • Q: How did the "collage of evidence" prove the abortion-crime link? A: By comparing states with high vs. low abortion rates and finding that high-abortion states saw a significantly sharper crime drop.
  • Q: Did the removal of lead from gasoline refute the abortion-crime study? A: No, it reaffirmed it; both factors were operating independently and were of similar magnitude.

Fidelity note

This summary is based strictly on the transcript provided. I am confident in the representation of the "unwantedness" effect and the "collage of evidence" methodology. The summary correctly identifies that the study does not claim a "moral" endorsement of abortion but rather a statistical correlation. The distinction between the "lead" and "abortion" effects as independent variables is a key nuance from the text.