When I go crawling around in data from the General Social Survey, sooner or later I run into Ryan Burge, who writes Graphs About Religion. In his most recent post, he wrote about changes in interpersonal trust, as measured by this GSS question:
Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?
Among other analyses, he looks at how the responses relate to self-identification as liberal, conservative, or moderate. The results are … complicated. Ryan writes:
Among older Americans who identify as liberal, they are more likely to trust other people. Among Gen Z liberals, it’s exactly the opposite.
To explain why older liberals are more trusting, Ryan suggests:
Liberalism is based on collectivism. It embraces the idea that “together, we can achieve more.” Things like universal healthcare rely on a sense of trust among large groups of people.
But if that’s true, why does the effect go in the opposite direction with Gen Z? Ryan speculates:
Well, maybe liberals feel very slighted by the fact that Donald Trump has been on the ballot for president in every election in which they’ve been eligible to vote, and he’s won twice. Or it could be that they are more impacted by the cynicism of the internet than conservatives? Or maybe it’s because they report higher rates of depression and anxiety than the rest of the ideological spectrum?
As it happened, I looked at the same GSS variable in a recent article. But I didn’t do the breakdown by political ideology, so let’s do that now.
The Story So Far
First, here’s a recap of what we saw in the previous post, with the whole GSS sample. The following figure shows the percentage of respondents who said “most people can be trusted.”
Time series: percent saying most people can be trusted (all respondents)
Interpersonal trust has declined consistently since the survey started in 1972. In the previous post, I decomposed this trend into cohort and period effects. Here’s the estimated cohort effect:
Standardized cohort effect with fixed time mix, percent saying most people can be trusted (all respondents)
The level of trust increased between the cohorts born in the 1900s through the 1940s, and then started a steep decline — more than 30 percentage points over 60 years. And here’s the remaining period effect, after factoring out the cohort effect.
Standardized time trend with fixed cohort mix, percent saying most people can be trusted (all respondents)
There is almost no period effect in the full sample.
Breakdown by Politics
Now let’s get to the question we started with:
Who is more trusting: liberals or conservatives?
Here’s the time series, broken down by ideology.
Time series: percent saying most people can be trusted, by 3-point political views
Until recently, there was not much difference: liberals and conservatives were about equally likely to say people can be trusted — both a little more trusting than moderates (who are, maybe, too distrustful to join a team).
Since the 2000s, liberals and conservatives have diverged, and now liberals are more trusting, by about 5 percentage points. But all three groups are still declining.
Now let’s decompose those trends into cohort and period effects. Here are the cohort effects for the three groups.
Standardized cohort effect for generalized trust by political views (uniform weights on survey years within group)
Here we can see what Ryan reported: in most generations, liberals are more trusting than conservatives; it’s only in the most recent generation that it goes the other way. The crossover happens among people born in the mid-1990s, close to the conventional beginning of Gen Z (born 1997 to 2012).
And here’s the period effect for the three groups.
Standardized period trend for generalized trust by political views (fixed within-group cohort weights)
For conservatives and moderates, the period effect is generally downward, although small. For liberals, it’s the other way around, generally increasing since about 2000.
The cohort-period decomposition provides some hints about what’s going on:
Among conservatives and moderates, the cohort and period effects are both downward, so they contributed to a steeper decline over time.
Among liberals, the cohort effect is also downward, but the period effect is upward, so the period effect mitigates the cohort effect.
To explain the cohort effect, I think Ryan’s suggestions are plausible. People born after 1995 have grown up during a discouraging time to be a liberal. Recent developments contrary to the liberal worldview include
The rise of populist and nationalist movements in several countries, along with democratic backsliding or erosion of liberal norms.
Slow progress or reversal on climate change.
Increasing distrust of experts, journalists, scientists, and public institutions.
Declining confidence in institutions such as Congress, the media, and organized religion.
A series of conservative Supreme Court decisions, most notably the overturning of long-established abortion rights protections in 2022.
Negativity bias in the media — especially social media — is probably a contributing factor. And going to school during the pandemic probably didn’t help.
But what about that period effect — can we think of reasons liberals would be more trusting, starting around 2005? If the recent cohort effect among young liberals is driven by the Trump era (at least in part), maybe the period effect was driven by events and trends of the Obama era that were aligned with the liberal worldview:
Perception that the country was becoming more socially inclusive and diverse.
Greater visibility and acceptance of LGBTQ people, and growing support for same-sex marriage, culminating in nationwide legalization in 2015.
Expansion of health insurance coverage through the Patient Protection and Affordable Care Act.
And falling crime rates from the 1990s through the 2010s might have contributed more directly to increasing interpersonal trust.
Of course these explanations are speculative, so let’s get back to what’s supported by the data:
Since the 1970s, interpersonal trust has declined consistently in all three groups (liberal, conservative, and moderate).
Before 2005, liberals and conservatives were about equally trusting; since then, the decline among liberals has slowed, so they are now the most trusting group.
Among conservatives and moderates, the cohort and period effects are both downward, so they contribute to steeper decline.
Among liberals, the cohort effect is steeply downward in the most recent cohorts, but the decline is mitigated by a positive period effect.
For now, liberals are the most trusting group, but if the cohort effect persists, they might not be for long.
This article is one of a series looking at changes in public opinion over the last 50 years, with a focus on culture war topics. In this installment, we’ll look at responses to four questions in the General Social Survey (GSS) related to sexual activity:
Premarital sex (premarsx): There’s been a lot of discussion about the way morals and attitudes about sex are changing in this country. If a man and woman have sex relations before marriage, do you think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
Teen premarital sex (teensex): What if they are in their early teens, say 14 to 16 years old? In that case, do you think sex relations before marriage are always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
Extramarital sex (xmarsex): What is your opinion about a married person having sexual relations with someone other than the marriage partner—is it always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
Same-sex relations (homosex): What about sexual relations between two adults of the same sex—do you think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
As we’ll see, answers to these questions have diverged in the last 50 years. A large majority answer that extramarital and teen sex are wrong, and that has barely changed (although opposition). At the same time, opposition to premarital sex and same-sex relations has declined substantially.
In this article, we’ll look at these trends and decompose them into cohort and period effects. In the next article, we’ll look at the relationship between these responses and religion, both affiliation and attendance.
We’ll start with the first question, on premarital sex.
Premarital sex
The following figure shows the percentage of respondents saying premarital sex is always or almost always wrong, from 1972 to 2024. The shaded area shows the results from a Bayesian model that estimates the latent trend — that is, a slowly varying underlying level of opposition to premarital sex.
Opposition to premarital sex has declined since 1972, from about 47% to about 20%.
As always, when we see this kind of change over time, it might be caused by cohort or period effects, or a combination of the two. Using a Bayesian model, I estimate a cohort effect for each birth year and a period effect for each survey year. The following figure shows the resulting trajectory for each cohort over time.
Each line represents a single birth year. For example, the yellow line at the top shows the fitted trajectory for people born in 1900, who were 72 when the survey started in 1972 and 90 when they aged out in 1990. The blue line in the bottom right shows responses of people born in 2000, who became eligible to participate in the survey when they turned 18 in 2018.
One pattern is clear: each cohort is less likely than the previous cohort to say that premarital sex is wrong. Among people born in 1900, it was more than 70%. Among people born in the 2000s, it is close to 10%. So that’s a big difference.
From these results, we can estimate the cohort and period effects separately. The following figure shows the cohort effect, standardized to control for the period effect by simulating responses as if every cohort was interviewed during every iteration of the survey.
The decline was steepest between the cohorts born in 1900 and 1950. After that, it leveled off, then declined again among the cohorts born in the 1980s.
Now we can estimate the period effect, standardized to control for the cohort effect, shown in the following figure.
The period effect is smaller than the cohort effect — about 8 percentage points from peak to trough — and less consistent. Opposition to premarital sex increased between 1980 and 2000, which coincides with increasing awareness of AIDS and public messaging about sexual risk, as well as the rise of the Religious Right and “family values” politics.
But before I speculate about the causes of these patterns, it will be useful to look at the responses to the other questions.
When is sex wrong?
The following figure shows the estimated percentage who think sex is wrong (always or almost always) in each of the four scenarios: premarital, teen, extramarital, and same-sex.
Reading from top to bottom:
Nearly everyone thinks “a married person having sexual relations with someone other than the marriage partner” is wrong, and the percentage has barely changed in more than 50 years.
Opposition to teen sex (the question specifies ages 14-16) is nearly as high, although it has declined since 2005 by about 15 percentage points.
Opposition to same-sex relations was high and mostly unchanged between 1972 and 1990. Since then it has decreased by almost 50 percentage points in 30 years, which is an astonishing speed for this kind of social change. Since 2020, opposition has increased a little.
Finally, as we’ve already seen, opposition to premarital sex declined substantially since the beginning of the survey.
For all four scenarios, we can estimate the cohort effect, controlling for the period effect, shown in the following figure.
For all four scenarios, there is a consistent downward trend in the cohort effect, modest for extramarital and teen sex, much steeper for premarital and same-sex relations.
The following figure shows the period effects.
After controlling for the cohort effects, the remaining period effects are more modest.
Opposition to teen sex is mostly unchanged, with some decline since 2010.
Opposition to extramarital sex increased between 1972 and 2010, and decreased since then.
Opposition to same-sex relations shows by far the largest period effect, increasing between 1972 and 1990, and decreasing since then — although increasing again since 2020.
Opposition to premarital sex has increased and decreased modestly.
In most scenarios, the cohort effect accounts for more of the observed change — but for same-sex relations, the period effect also makes a substantial contribution.
Dogma, morality, and health
To make sense of these patterns, let’s think about what people might mean when they say that sex is wrong.
Taking premarital sex as an example, some people consider sex outside marriage to be contrary to spiritual values or religious teachings. Others might be concerned with sexually-transmitted disease or the social consequences of children born outside of stable families.
And for teens specifically, some object because they see adolescence as a period of innocence that should be protected, believe sexual activity compromises purity or chastity, or think sexual restraint reflects virtues like self-discipline and respect for social norms. Also, some might think teenagers don’t have the knowledge, experience, and impulse control to avoid health consequences of sex, especially pregnancy, or the maturity to handle emotional challenges.
Similarly, some people object to same-sex relations because they see them as contrary to religious teachings, inconsistent with traditional ideas about gender and family, or incompatible with social norms about sexuality.
And people might object to extramarital sex because of the emotional harm it causes, because it breaks a vow, because it threatens families and social stability, or because it violates holy matrimony.
In each scenario, objections arise from different concerns: practical risks and harms, social concerns about norms and stability, and moral or religious beliefs. Looking only at multiple-choice responses, we don’t know what respondents had in mind.
But the differences we see — between teen and extramarital sex on one hand, and premarital sex and same-sex relations on the other — suggest a conjecture: objections rooted in immediate harms and concrete consequences might be more stable over time; objections rooted in social norms, morality, and religion might be more historically contingent.
In the next article, we’ll test this conjecture by looking at relationships between religion and attitudes about sex — and how both have changed over time.
This article is one of a series exploring responses to core questions in the General Social Survey (GSS), estimating period and cohort effects, and looking for historical events that might explain the trends we see.
Confidence in American institutions
In this installment, we’ll look at 13 questions related to confidence in institutions. We’ll start with a detailed look at confidence in “the people running Congress”, and then summarize results from the other questions. The complete survey question is:
I am going to name some institutions in this country. As far as the people running these institutions are concerned, would you say you have a great deal of confidence, only some confidence, or hardly any confidence at all in […] Congress.
The following figure shows the fraction of people who answered “a great deal of confidence” or “only some confidence,” and a smooth line fitted to the raw percentages.
The long-term trend is downward, from about 80% during the first iteration of the survey to below 50% in the most recent iterations. But there are ups and downs.
It looks like confidence was increasing during the 1980s before collapsing in the early 1990s. Possible causes of the decline include:
Confidence in Congress recovered between 1995 and 2005, and declined again between 2005 and 2015. A likely contributor is the Great Recession from late 2007 to mid-2009.
This period also saw the rise of anti-establishment politics, including the Tea Party movement and Ron Paul’s presidential campaign.
Now we’ll decompose these changes into period and cohort effects.
Period and cohort effects
Using the Bayesian model described here we can estimate a latent “confidence in Congress” factor for each birth cohort over time. The following figure shows these estimates; each line represents a single birth year.
Those results are easier to interpret if we factor out the cohort effect (keeping the mixture of survey years constant) and the period effect (keeping the mixture of cohorts constant). The following figure shows the standardized cohort effect.
Among people born between 1900 and 1950, there is almost no change. Then starting with people born in the 1960s, confidence in Congress has increased consistently and substantially.
To interpret this result, it is helpful to go back to the previous figure. Starting in the upper left:
Among people born in the 1960s and 1970s, about 80% reported confidence in Congress when they were surveyed as young adults.
Among people born in the 1980s and 1990s, it was closer to 70%.
And among people born in the 2000s it’s below 70%.
The entry point of each cohort is below the entry point of previous cohorts, but because these entry points are above the declining trend of previous generations, this relative optimism is interpreted as an increasing cohort effect.
So we should not conclude that younger generations are more confident in Congress, only that when they are first surveyed, they start out above the trajectory of previous cohorts.
As a simplification, we might imagine an 18-year-old entering adulthood with a relatively idealized understanding of American government — shaped by civics classes and maybe a school trip to Washington D.C. — before later political experiences erode some of that confidence.
Another possibility is that younger generations have grown up with lower expectations of government, so when they say they have “some confidence”, they might be evaluating Congress against a lower standard.
The following figure shows the estimated period effect, with the mix of cohorts held constant.
This decline is steeper than what we saw in the time-only model, because we have factored out the mitigating effect of relative optimism in recent generations.
Government Institutions
Now we’ll apply the same analysis to questions about the executive branch of the federal government, the Supreme Court, and the military (technically part of the executive branch).
The following figure shows the estimated cohort effects for each of these institutions.
And the following figure shows the period effect, after factoring out the cohort effect.
The results for the executive branch are similar to the results for Congress.
The cohort effect is flat between people born in the 1900s and 1950s, and increasing after that.
The period effect declines substantially and consistently — without the ups and downs of confidence in Congress.
When respondents are asked about the “executive branch”, it’s not clear whether they think primarily about the president, federal agencies, or the federal government in general.
The patterns for the military and Supreme Court are different.
Confidence in the military is generally high. The cohort effect declined gradually, and possibly more steeply among people born after 1980. The period effect gradually increased.
Confidence in the Supreme Court is higher than confidence in other branches, although the period effect has dropped steeply since 2015. The cohort effect increased among people born between 1900 and 1960; among more recent generations it is gradually declining.
Historically, the Court cultivated an image of being above politics; that perception weakened substantially in the 2010s. In March 2016, Merrick Garland was nominated to the Supreme Court after the death of Antonin Scalia. The Republican majority in the Senate refused to hold a hearing or vote on his nomination. The seat remained empty until Donald Trump nominated, and the Senate confirmed, Neil Gorsuch, who is considered to be more conservative. To many liberals, the Senate’s 293-day blockade undermined the legitimacy of the Court.
Then when Ruth Bader Ginsburg died in September 2020, Donald Trump nominated Amy Coney Barrett and the Senate confirmed her 8 days before the 2020 presidential election, in a process criticized by Democratic leaders as illegitimate.
These appointments, along with the confirmation of Brett Kavanaugh in 2018, shifted the composition of the Court toward conservatives, forming what is now described as a 6-3 supermajority of conservative justices.
Since then, the Supreme Court has issued several decisions contrary to majority public opinion, most notably the 2022 decision overturning Roe v. Wade. Other unpopular decisions weakened gun control and limited federal regulatory authority, especially over environmental policy. Many of these decisions have been perceived, especially on the left, to be motivated by politics rather than constitutional principles and precedent.
The slope of the cohort effect might reflect a generational change in associations with the Supreme Court. Older cohorts may associate the Supreme Court with landmark decisions like Brown v. Board of Education and the expansion of civil rights. Younger cohorts may instead associate it with partisan conflict, blocked reforms, and ideological polarization. Older generations might also have been more deferential toward the institution itself. Younger generations, exposed to more adversarial and partisan media coverage, might be less inclined to deference.
Economic institutions
The following figures show the cohort and period effects for confidence in economic institutions: banks, major companies, and organized labor.
Confidence in these institutions is generally high. Looking at the cohort effects, the most salient feature is increased confidence in organized labor among people born after 1940.
A possible explanation is that younger cohorts have less exposure to organized labor. Older generations were more likely to be affected by strikes and related economic disruption, and more likely to be aware of corruption in labor unions. As union membership has declined and the gig economy has expanded, younger cohorts are less aware of the negative aspects of organized labor, including dues, and more likely to perceive their lack of negotiating power in the labor market. Also, anti-labor ideology has declined since the end of the Cold War, as the framing has shifted from “labor versus management” to “workers versus corporations”.
By comparison with the cohort effects, the period effects are modest:
The period effect on organized labor is almost unchanged since the 1980s — the change we see over time is almost entirely due to generational replacement.
The period effect on confidence in major companies has declined somewhat.
The trend for banks and financial institutions is more complicated — arguably driven by shorter-term period effects like scandals and financial crises. The first notable downturn, in the 1980s, coincides with the savings and loan crisis, when hundreds of financial institutions failed and taxpayers absorbed large bailout costs, followed by an economic recession from 1990 into 1991. The larger downturn around 2008 coincides with the 2008 global financial crisis, when taxpayers were hit with even larger bailout costs, and public anger at “private gains, public losses” came to a focus in the Occupy Wall Street protests.
Professions, knowledge, and religion
The following figures show estimated cohort and period effects for confidence in education, medicine, the scientific community, and organized religion.
Confidence in these institutions is highest for science and medicine, lower for education and religion.
Looking at the period effects, they are all in decline. Confidence in education declined most steeply; confidence in the scientific community is relatively stable, although it declines after 2020.
Thinking about confidence in education, it might be useful to separate colleges and universities from K-12 schools.
In higher education, the decline might be due to increasing tuition and student debt, credential inflation, and increasing uncertainty about the economic return on a college degree, especially among majors in the arts and humanities. More recently, confidence in universities has decreased steeply, especially among conservatives, due to the perception of ideological bias.
In K-12 education, declining confidence might be related to anxiety about standardized testing, international competition, and especially around the No Child Left Behind Act in 2001, the framing of public schools as underperforming institutions requiring accountability reforms and federal intervention. Also, public schools have increasingly become focal points for political conflicts about curriculum, race, gender, religion, and parental authority.
Most of the cohort effects have increased modestly; in particular confidence in education is higher among people born after 1980, compared to previous generations observed at the same time. But for these institutions, the interaction of the period and cohort effect is similar to what we saw for Congress — it’s not that recent generations have more confidence, it’s just that when they are surveyed as young adults, they come in at entry points above the declining trend of previous generations.
Confidence in organized religion is the exception — the period and cohort effects both trend downward, so the decline is additive. A likely contributor to the period effect is growing public awareness of sexual abuse and institutional coverups in the Catholic Church, which received national attention beginning in the 1980s, escalated after the Boston Globe investigations published in 2002, and continues to the present with additional revelations in the United States and other countries. But the decline is not limited to the Catholic Church, and it began before these scandals were widely known.
The cohort effect likely reflects broader secularization trends, including declining religious affiliation, lower church attendance, and weakening institutional authority among younger generations.
Media
Finally, the following figures show estimated cohort and period effects for confidence in television and the press.
There is almost no cohort effect, although the most recent cohorts might have a little more confidence in television.
The headline here is the period effect, which is consistently downward, and steeper for the press than television. The steepest part of the decline for both media started around 1990, shortly after the 1987 abolition of the fairness doctrine, which required broadcast coverage of controversial topics to be “fair in the sense that it provides an opportunity for the presentation of contrasting points of view,” as described in the 1949 FCC report that established the doctrine.
The end of the fairness doctrine coincided with the rise of talk radio programs with explicit political viewpoints, including The Rush Limbaugh Show, which was nationally syndicated in 1988.
Cable television news followed, including Fox News Channel in 1996, with an explicit conservative orientation, and MSNBC, which developed a more liberal identity in the 2000s.
During this period, more generally, media audiences became more fragmented. Prior to 1980, most Americans were exposed to a small number of shared news sources, notably the three major television networks. Talk radio and cable television offered more options and less common experience.
And then the internet happened, starting in the 1990s with online news and political blogs, including the Drudge Report which started as a weekly email newsletter in 1995, and rose to national prominence when it broke the Clinton-Lewinsky scandal in 1998.
Social media followed. YouTube was founded in 2005; Facebook and Twitter launched in 2006. While these platforms have become important sources of news for many Americans, engagement-driven algorithms often promote emotionally provocative and polarizing content over careful reporting. The rise of the internet contributed to the decline of local newspapers, and eventually national newspapers as well.
Ownership of television stations became increasingly consolidated following the 1996 Telecommunications Act, allowing a small number of national media companies to control larger shares of local news programming. The effect of this consolidation is explained in this Vox article and memorably demonstrated in this Deadspin compilation showing dozens of TV news anchors reading nearly identical scripts provided by the Sinclair Broadcast Group, which requires the channels it owns to air segments called “must-runs” — many of them presenting conservative talking points.
Finally, since the beginning of his presidential campaign in 2015, Donald Trump has repeatedly denigrated television and print media, frequently describing unfavorable coverage as “fake news” and labeling journalists “enemies of the people.” These attacks likely contribute to declining confidence in the press, especially among Republicans.
Negativity Bias
In the previous examples, you might notice that I offer explanations for the downturns, but no explanation for the upturns. That’s because bad things, like scandals and economic crises, often happen quickly and they get a lot of coverage; good things often happen slowly and continue without comment.
For many institutions, no news is good news. When they do their jobs, they don’t get much attention, and public confidence drifts higher, even without specific positive events or coverage.
So I want to end this article by highlighting some of the positive results we see in this data:
Confidence in education, science, and medicine is high and although the period effects are negative, the cohort effects are positive, which bodes well for the future.
Confidence in financial institutions and organized labor is high.
Confidence in government is lower and declining, but as each generation of young adults starts out more optimistic than their elders, there is hope for a turnaround.
But the recent steep decline of confidence in the Supreme Court is a concern, as is the loss of confidence in the media. It’s hard to find a positive take on those trends.
In Graphs About Religion, Ryan Burge recently wrote about changing opinions about assisted suicide and how they relate to religion.
As always, when I see survey responses changing over time, I wonder whether it is driven primarily by period or cohort effects. And if you’ve read my last few posts, you know I’ve been working on a Bayesian model to answer that question.
Ryan’s analysis is based on four questions from the General Social Survey (GSS):
Do you think a person has the right to end his or her own life if this person:
Has an incurable disease? (suicide1)
Has gone bankrupt? (suicide2)
Has dishonored his or her family? (suicide3)
Is tired of living and ready to die? (suicide4)
In addition, we’ll look at results from a related question (letdie1):
When a person has a disease that cannot be cured, do you think doctors should be allowed by law to end the patient’s life by some painless means if the patient and his family request it?
The framing of the questions is different: the first four are about the right to end one’s life and the last is about the legality of doctor-assisted suicide.
Before we look at the breakdown of period and cohort effects, here are the results from a model that estimates latent opposition to each proposition as a smooth function over time.
Opposition to suicide is high in three of the scenarios — bankrupt, dishonored family, and tired of living — and lower in the incurable disease scenarios.
In all five questions, opposition has declined over time, although for the incurable disease scenarios, it might have leveled off after 1990.
Doctor-assisted death
Now let’s see if we can decompose these changes into period and cohort effects. We’ll start with the question about doctor-assisted death when the patient has an incurable disease.
As in the previous posts, I used a Bayesian model to estimate a trajectory over time for each birth cohort, shown in the following figure.
Reading from top to bottom, we can see that opposition has declined from one cohort to the next, and reading from left to right, we can see that opposition has varied over time within each cohort.
The following figure shows the cohort component alone, standardized to factor out the period effect.
Opposition to doctor-assisted suicide has declined from more than 40% in the earliest cohorts to 20% among people born in 2006.
A possible explanation for the cohort pattern is that people anchor their moral judgments to the legal environment they encounter when they are young. During the “impressionable years” of late adolescence and early adulthood, existing laws can establish a moral baseline, so that what is illegal is inferred to be wrong, and therefore should remain illegal. As a result, gradual legalization can generate long-run attitudinal change through cohort replacement: people who grow up after a practice becomes legal are less likely to see it as morally problematic.
The following figure shows the period effect alone, along with the results from the time model (which includes both period and cohort effects).
Comparing the two lines, we can conclude that the decline we see over time is entirely due to the cohort effect — when we control for generational replacement, the estimated period effect has generally increased since 1990.
The increase between 1990 and 2005 might reflect increasing moral concern due to advances in life-sustaining medical technology, high-profile legal disputes like the Terri Schiavo case, and broader discussions of the sanctity of life.
The decline between 2005 to 2015 might reflect normalization of assisted dying following legalization in several states (Oregon in 1997, Washington in 2008, and Montana in 2009, Vermont in 2013), along with a shift in public discourse toward autonomy, dignity, and patient choice, reinforced by high-profile cases like Brittany Maynard.
Other Scenarios
The following figure shows the estimated cohort effects for all five questions.
For the incurable disease scenario, opposition has declined from more than 60% in the earliest cohorts to less than 40% among cohorts born after 1950 — although it might have leveled off since then.
In the other scenarios, opposition has also declined from one cohort to the next, but the size of the effect is smaller.
The following figure shows the estimated period effects, controlling for generational replacement.
Since 1990, most of the period effects are small. The only exception is the “tired of living” scenario, where there is some decline over time, independent of generational replacement.
In the next post, we’ll do the same analysis with questions about abortion and the situations where it should be legal or not.
In a previous article, I claimed that Young adults are not very happy. Now the World Happiness Report 2026 has confirmed that young people in North America and Western Europe are less happy than they were fifteen years ago, and less happy than previous generations.
In this article, we’ll look at results from three related questions in the General Social Survey (GSS):
Trust: “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?”
Fair: “Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?”
Helpful: “Would you say that most of the time people try to be helpful, or that they are mostly just looking out for themselves?”
As we’ll see, young adults in the United States have a more negative outlook than previous generations: they are less likely to say that people can be trusted, that they are fair, or that they are helpful. And we’ll consider connections between this bleak outlook and unhappiness.
Trust
Using the same model from the previous articles, I estimated the percentage who say people can be trusted, following each birth year over time.
Cohort trajectories, percent saying most people can be trusted
With these trajectories, we can decompose the cohort and period effects. The following figure shows the cohort effect, standardized by holding the period effect constant.
Standardized cohort effect with fixed time mix, percent saying most people can be trusted
The level of trust increased between the cohorts born in the 1900s through the 1940s, and then started a steep decline. This is a large cohort effect, dropping about 30 percentage points over 60 years.
The following figure shows the period effect, standardized by holding the cohort mix constant.
Standardized time trend with fixed cohort mix, percent saying most people can be trusted
In contrast, there is almost no period effect.
The conjecture part
About my previous article, one of my former colleagues said he appreciated my attempt to offer explanations, but reminded me that with this kind of data alone, it is hard to say what causes what with any confidence. That’s true, and it’s a good reminder — but we can get some clues:
When we see a strong cohort effect and almost no period effect, that’s evidence that we’re seeing patterns set in childhood.
When we see period effects, we should look for events that affected all cohorts at the same time.
So let’s think about what was happening in the formative years of these cohorts, starting with the 1940 cohort, which was the high point in trust, before the decline:
Cohort 1940 (childhood: 1940–1960): dense local communities, strong civic and religious institutions, frequent face-to-face interaction, and shared media environment.
Cohort 1950 (1950–1970): suburbanization expands, some weakening of community density, television becomes widespread but still shared.
Cohort 1960 (1960–1980): civil rights conflict, Vietnam War, Watergate scandal, rising crime.
Cohort 1970 (1970–1990): reduced civic participation, rising inequality, more cautious parenting, less unstructured social interaction.
Cohort 1980 (1980–2000): increasing inequality, more segregation by class and education, early internet exposure, continued decline in shared institutions.
At this point a multi-generational effect comes into play — the parents of Cohort 1980, born in the 1950s and 1960s, were less trusting than previous generations of parents.
Cohort 1990 (1990–2010): widespread internet use, early social media, more structured childhood, increasing awareness of global risks.
Cohort 2000 (2000–2020): smartphones and social media throughout formative years, algorithmic content, reduced in-person interaction.
If trust is largely set early in life, then differences between cohorts reflect the environments they experienced during their first two decades.
In addition to this question about trust, the GSS includes related questions about fairness and mutual assistance.
Fair
Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair? The following figure shows the percentage who thought people would be fair.
Cohort trajectories, percent saying people would try to be fair
And here’s the cohort effect.
Standardized cohort effect with fixed time mix, percent saying people would try to be fair
And the period effect.
Standardized time trend with fixed cohort mix, percent saying people would try to be fair
The cohort pattern is similar to what we saw in trust: small changes between the 1900s and 1940s cohorts, and then a steep decline — almost 40 percentage points over 60 years.
The period effect is relatively small, varying by only 10 percentage points from lowest to highest point, but it was generally positive until about 2015 (the onset of the Trump Era?).
Helpful
Would you say that most of the time people try to be helpful, or that they are mostly just looking out for themselves?
Here is a period–cohort fingerprint of the responses, showing the percentage who thought people try to be helpful.
Cohort trajectories, percent saying people try to be helpful
Here’s the cohort effect:
Standardized cohort effect with fixed time mix, percent saying people try to be helpful
And the period effect.
Standardized time trend with fixed cohort mix, percent saying people try to be helpful
Again we see the same pattern: little change between the cohorts born between 1900 and 1940, and then a decline of more than 30 percentage points over 60 years.
And again, the period effect is comparatively small and generally increasing — but possibly declining in the most recent cycles of the survey.
Cause and Effect?
It is plausible that the decline in trust is a contributing factor to the decline in happiness. If you believe that people are out to get you, and 80% of your friends agree, that’s not a worldview conducive to a sense of well-being. And generational decline in trust precedes the decline in happiness, so it is at least a potential cause.
The decline in trust-related beliefs also supports the interpretation that recent cohorts are actually unhappy, rather than interpreting the question differently, or being more willing than previous generations to say they are unhappy.
I haven’t done full-on causal modeling to quantify these relationships, but I ran a few regression models to explore. To reduce the number of researcher degrees of freedom, I asked ChatGPT to interpret the results:
Differences in happiness across cohorts appear to be partly explained by differences in social outlook (trust, fairness, helpfulness), and these outlook variables behave like stable, cohort-structured traits rather than period-driven fluctuations.
The AI-generated summary of the experiments follows.
Model 1: Cross-sectional association (complete cases)
Specification:
Outcome: very_happy (binary)
Predictors: trust, fair, helpful (all binary)
Sample: complete cases with all variables observed
Purpose:
Estimate the cross-sectional relationship between social outlook and happiness.
Provides baseline associations without accounting for cohort or period effects.
Interpretation:
Coefficients represent conditional associations among individuals at a point in time.
Answers: Are people with a more positive outlook more likely to be very happy?
Model 2: Outlook + cohort + period (restricted sample)
Specification:
Outcome: very_happy
Predictors:
trust, fair, helpful
cohort_c (mean-centered birth year)
year_c (mean-centered survey year)
Sample: respondents born ≥ 1940 with complete data
Purpose:
Assess whether the outlook–happiness relationship persists after accounting for:
Cohort effects (differences across birth cohorts)
Period effects (changes over survey years)
Interpretation:
Coefficients for outlook variables reflect within-cohort, within-period associations.
Cohort and year coefficients capture linear trends in happiness after controlling for outlook.
Answers:
Are outlook variables still associated with happiness after adjusting for historical context?
Is there an independent cohort or period trend?
Model 3: Cohort + period only (no outlook variables)
Specification:
Outcome: very_happy
Predictors:
cohort_c
year_c
Sample: respondents born > 1940 (larger sample since outlook variables not required)
Purpose:
Estimate total cohort and period effects on happiness without controlling for outlook.
Provides a baseline for comparison with Model 2.
Interpretation:
Cohort and year coefficients reflect combined (direct + indirect) effects.
Comparing to Model 2 shows how much of these effects are accounted for by outlook variables.
Answers:
How does happiness vary across cohorts and over time in aggregate?
How much do these patterns change when outlook is included?
Key Findings
Positive social outlook is associated with higher happiness.
Trust, fairness, and helpfulness all have positive and statistically significant associations with being “very happy.”
Estimated odds ratios:
Trust: ~1.25
Fairness: ~1.36 (strongest)
Helpfulness: ~1.29
These effects are modest in size and explain a small fraction of overall variation (Pseudo R² ≈ 0.016).
These relationships are stable across cohorts and time.
Adding cohort and survey year controls has little effect on the coefficients.
This suggests the outlook–happiness relationship is primarily cross-sectional, not driven by historical shifts.
Cohort and Period Effects
Without controlling for outlook:
Later cohorts are less likely to report being very happy.
There is also a negative period trend (declining happiness over time).
With outlook variables included:
The cohort effect becomes small and statistically insignificant.
The period effect remains negative and significant.
Interpretation
Outlook variables appear to mediate cohort differences in happiness.
Later cohorts tend to report lower trust, fairness, and helpfulness.
These differences account for much of the observed cohort decline in happiness.
Period effects persist independently.
There is a modest downward trend in happiness over time that is not explained by outlook variables.
Data Considerations
Approximately 40% of observations are missing at least one outlook variable, reducing the complete-case sample.
This raises the possibility of selection bias in the estimates.
Bottom Line
A more positive view of others (trust, fairness, helpfulness) is consistently associated with higher happiness.
Differences in these outlook measures help explain why later cohorts report lower happiness.
However, there is also an independent downward trend in happiness over time.
Since 1972, the General Social Survey has asked respondents: “Taken all together, how would you say things are these days—would you say that you are very happy, pretty happy, or not too happy?”
The following figure shows how the responses have changed over time and between birth cohorts. Each line represents one birth year.
People born in 1900 were 72 years old when the survey started; at that point, about 37% said they were very happy. In 1990, the last year they were eligible to participate, a little more than 40% said they were very happy. So it seems like they aged well—or possibly the less happy died earlier.
People born in 1910 were a little less happy when the survey started, but by the time they aged out, they also reached 40%. They were the last generation to reach that mark.
Among people born between 1920 and 1950, each cohort was a little less happy than the one before (or maybe less likely to say they were happy). In these cohorts, we can see a general trend over time: increasing until about 2000, leveling off, and declining after 2010.
The cohorts born in the 1960s and 1970s followed a similar trajectory, with only small differences from one birth year to the next.
And then the bottom fell out. Starting with people born in the 1980s (the earliest Millennials), each successive cohort was substantially less happy than the one before.
When people born in 1990 joined the survey in 2008 (at age 18), only 27% said they were very happy. In the most recent data, from 2024, the number had fallen to 22%.
When people born in 2000 entered in 2018, they set a new record low at 21%, which has now fallen to 18%.
And in the most recent cohort—born in 2006 and interviewed in 2024—only 16% said they were very happy.
These percentages are based on a statistical model that estimates the proportion of “very happy” responses in each group at each point in time. The details of the model and its assumptions are below.
The Time Trend
With an estimated proportion for each cohort and time step, we can compute separate contributions for changes over time and between cohorts.
To characterize the contribution of time, we have to hold the cohort effect constant, which we can do by computing the distribution of birth years across the entire dataset and simulating a population where this distribution does not change over time. The following figure shows the result.
The overall level of happiness increased between 1972 and 2000, leveled off, and then declined after 2010.
Of course it is speculation to say why that happened, but we can think about large-scale economic and social patterns and how they line up with these trends.
Economically, 1980 to 2000 was a period of growth and relative stability. That changed after the end of the Dot-com bubble in 2001 and, more importantly, the Global Financial Crisis in 2008, which had broad and persistent effects on employment, wealth, and economic security.
Geopolitically, the 1970s through the 1990s were relatively quiet compared to what followed. The September 11 attacks in 2001, and the wars in Iraq (2003–2011) and Afghanistan (2001–2021) marked a shift toward a more uncertain and conflict-oriented global environment.
Participation in civic organizations and religious institutions declined over the past several decades. These institutions traditionally provided social support, shared identity, and regular face-to-face interaction. Social isolation is strongly associated with lower well-being.
At the same time, the media environment was transformed. The rise of 24-hour news increased exposure to negative and emotionally salient events, and after 2010 the spread of smartphones and social media made that exposure continuous and personalized.
Finally, measures of trust in institutions and other people have generally declined over this period, while political polarization has increased. These trends may reduce people’s sense of stability and shared purpose.
The COVID-19 pandemic likely contributed to the most recent decline, but the downward trend was already underway before 2020.
The Cohort Effect
Just as we isolated the time trend by simulating a survey with a fixed distribution of cohorts, we can isolate the cohort effect by simulating a survey with a fixed distribution of times. The following figure shows the result.
The cohort effect is larger and more consistent than the time trend: the difference between the happiest and least happy cohorts is more than 20 percentage points.
The decline was relatively slow for cohorts born between 1900 and 1950 and nearly zero for cohorts born in the 1950s, 1960s and 1970s (late Baby Boomers and Gen X). The steep decline begins with the Millennials and continues into Gen Z.
Possible explanations for the recent decline include:
Transformation of childhood: Jonathan Haidt has described childhood in recent cohorts as “overprotected in the real world and underprotected in the online world.” Increased parental monitoring, reduced independent play, and greater time spent online may affect the development of autonomy, risk tolerance, and social skills. If these early-life experiences shape long-term outlook, they could contribute to lower self-reported happiness.
Greater and earlier exposure to media: Younger cohorts were exposed to a media landscape characterized by continuous, personalized, and often negative content. Social media platforms amplify social comparison and negative content, while displacing in-person interaction. Increased awareness of global risks—including climate change—may contribute to a more pessimistic worldview.
Differential impact of economic conditions: Recent cohorts entered the labor market during periods of economic disruption, including the aftermath of the Global Financial Crisis and more recent pandemic-related shocks. These cohorts also face higher housing costs and greater student debt. Economic insecurity during the transition to adulthood may have lasting effects on well-being.
Extension of “liminal” adulthood: Young adults are taking longer to complete education, establish careers, form long-term partnerships, and have children. This extended unsettled period may be associated with lower life satisfaction.
Norms around self-reported well-being. Younger cohorts may also be less likely to say they are “very happy,” either because of changing norms around self-presentation or greater awareness of mental health.
It’s hard to say how much of the recent decline we can attribute to these causes. But the decline is steep, and seems to be ongoing.
How the Model Works
One of the challenges with this kind of survey data is that the sample size is small for each birth year in each iteration of the survey. If we plot raw percentages over time, the result is noisy.
In Probably Overthinking It, I addressed this problem by grouping respondents into decade-of-birth cohorts and smoothing the resulting time series. That approach works, but it has drawbacks: aggregation removes detail, introduces edge effects for the earliest and latest cohorts, and requires an arbitrary choice about the level of smoothing.
The new model takes a more principled approach. Instead of smoothing the observed data, it models an unobserved (latent) propensity to report being “very happy” for each cohort in each year.
We assume that the number of “very happy” responses in each group follows a binomial distribution, where the probability of a “very happy” response depends on this latent propensity. The observed responses provide noisy information about the latent factor; the model combines information across cohorts and years to estimate it.
The latent propensity is modeled as the sum of an intercept, representing the overall level of happiness, a smooth effect of birth cohort, a smooth effect of survey year, and a year-specific random effect that captures short-term fluctuations (overdispersion).
The cohort and period effects are modeled as second-order Gaussian random walks (RW2), which means the model assumes these effects evolve smoothly over time, with a preference for gradual changes in slope rather than abrupt jumps, unless the data provide strong evidence otherwise. The amount of smoothing is not fixed in advance; it is estimated from the data.
The random year effect captures variation from one survey to the next that is not explained by long-term trends, like current events and topics of discussion.
Where we have a lot of data, the estimates track the observed proportions closely. Where data are sparse, the model borrows strength from neighboring cohorts and years, providing principled smoothing and interpolation without arbitrary grouping.
This article is an excerpt from the manuscript of Probably Overthinking It, available from the University of Chicago Press and from Amazon and, if you want to support independent bookstores, from Bookshop.org.
[This excerpt is from a chapter on moral progress. Previous examples explored responses to survey questions related to race and gender.]
The General Social Survey includes four questions related to sexual orientation.
What about sexual relations between two adults of the same sex – do you think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
And what about a man who admits that he is a homosexual? Should such a person be allowed to teach in a college or university, or not?
If some people in your community suggested that a book he wrote in favor of homosexuality should be taken out of your public library, would you favor removing this book, or not?
Suppose this admitted homosexual wanted to make a speech in your community. Should he be allowed to speak, or not?
If the wording of these questions seems dated, remember that they were written around 1970, when one might “admit” to homosexuality, and a large majority thought it was wrong, wrong, or wrong. In general, the GSS avoids changing the wording of questions, because subtle word choices can influence the results. But the price of this consistency is that a phrasing that might have been neutral in 1970 seems loaded today.
Nevertheless, let’s look at the results. The following figure shows the percentage of people who chose a homophobic response to these questions as a function of age.
It comes as no surprise that older people are more likely to hold homophobic beliefs. But that doesn’t mean people adopt these attitudes as they age. In fact, within every birth cohort, they become less homophobic with age.
The following figure show the results from the first question, showing the percentage of respondents who said homosexuality was wrong (with or without an adverb).
There is clearly a cohort effect: each generation is substantially less homophobic than the one before. And in almost every cohort, homophobia declines with age. But that doesn’t mean there is an age effect; if there were, we would expect to see a change in all cohorts at about the same age. And there’s no sign of that.
So let’s see if it might be a period effect. The following figure shows the same results plotted over time rather than age.
If there is a period effect, we expect to see an inflection point in all cohorts at the same point in time. And there is some evidence of that. Reading from top to bottom:
More than 90% of people born in the nineteen-oughts and the teens thought homosexuality was wrong, and they went to their graves without changing their minds.
People born in the 1920s and 1930s might have softened their views, slightly, starting around 1990.
Among people born in the 1940s and 1950s, there is a notable inflection point: before 1990, they were almost unchanged; after 1990, they became more tolerant over time.
In the last four cohorts, there is a clear trend over time, but we did not observe these groups sufficiently before 1990 to identify an inflection point.
On the whole, this looks like a period effect. Also, looking at the overall trend, it declined slowly before 1990 and much more quickly thereafter. So we might wonder what happened in 1990.
What happened in 1990?
In general, questions like this are hard to answer. Societal changes are the result of interactions between many causes and effects. But in this case, I think there is an explanation that is at least plausible: advocacy for acceptance of homosexuality has been successful at changing people’s minds.
In 1989, Marshall Kirk and Hunter Madsen published a book called After the Ball with the prophetic subtitle How America Will Conquer Its Fear and Hatred of Gays in the ’90s. The authors, with backgrounds in psychology and advertising, outlined a strategy for changing beliefs about homosexuality, which I will paraphrase in two parts: make homosexuality visible, and make it boring. Toward the first goal, they encouraged people to come out and acknowledge their sexual orientation publicly. Toward the second, they proposed a media campaign to depict homosexuality as ordinary.
Some conservative opponents of gay rights latched onto this book as a textbook of propaganda and the written form of the “gay agenda”. Of course reality was more complicated than that: social change is the result of many people in many places, not a centrally-organized conspiracy.
It’s not clear whether Kirk and Madsen’s book caused America to conquer its fear in the 1990s, but what they proposed turned out to be a remarkable prediction of what happened. Among many milestones, the first National Coming Out Day was celebrated in 1988; the first Gay Pride Day Parade was in 1994 (although previous similar events had used different names); and in 1999, President Bill Clinton proclaimed June as Gay and Lesbian Pride month.
During this time, the number of people who came out to their friends and family grew exponentially, along with the number of openly gay public figures and the representation of gay characters on television and in movies.
And as surveys by the Pew Research Center have shown repeatedly, “familiarity is closely linked to tolerance”. People who have a gay friend or family member – and know it – are substantially more likely to hold positive attitudes about homosexuality and to support gay rights.
All of this adds up to a large period effect that has changed hearts and minds, especially among the most recent birth cohorts.
Cohort or period effect?
Since 1990, attitudes about homosexuality have changed due to
A cohort effect: As old homophobes die, they are replaced by a more tolerant generation.
A period effect: Within most cohorts, people became more tolerant over time.
These effects are additive, so the overall trend is steeper than the trend within the cohorts – like Simpson’s paradox in reverse. But that raises a question: how much of the overall trend is due to the cohort effect, and how much to the period effect?
To answer that, I used a model that estimates the contributions of the two effects separately (a logistic regression model, if you want the details). Then I used the model to generate predictions for two counterfactual scenarios: what if there had been no cohort effect, and what if there had been no period effect? The following figure shows the results.
The circles show the actual data. The solid line shows the results from the model from 1987 to 2018, including both effects. The model plots a smooth course through the data, which confirms that it captures the overall trend during this interval. The total change is about 46 percentage points.
The dotted line shows what would have happened, according to the model, if there had been no period effect; the total change due to the cohort effect alone would have been about 12 percentage points.
The dashed line shows what would have happened if there had been no cohort effect; the total change due to the period effect alone would have been about 29 percentage points.
You might notice that the sum of 12 and 29 is only 41, not 46. That’s not an error; in a model like this, we don’t expect percentage points to add up (because it’s linear on a logistic scale, not a percentage scale).
Nevertheless, we can conclude that the magnitude of the period effect is about twice the magnitude of the cohort effect. In other words, most of the change we’ve seen since 1987 has been due to changed minds, with the smaller part due to generational replacement.
No one knows that better than the San Francisco Gay Men’s Chorus. In July 2021, they performed a song by Tim Rosser and Charlie Sohne with the title, “A Message From the Gay Community”. It begins:
To those of you out there who are still working against equal rights, we have a message for you […] You think that we’ll corrupt your kids, if our agenda goes unchecked. Funny, just this once, you’re correct. We’ll convert your children, happens bit by bit; Quietly and subtly, and you will barely notice it.
Of course, the reference to the “gay agenda” is tongue-in-cheek, and the threat to “convert your children” is only scary to someone who thinks (wrongly) that gay people can convert straight people to homosexuality, and believes (wrongly) that having a gay child is bad. For everyone else, it is clearly a joke.
Then the refrain delivers the punchline:
We’ll convert your children; we’ll make them tolerant and fair.
For anyone who still doesn’t get it, later verses explain:
Turning your children into accepting, caring people; We’ll convert your children; someone’s gotta teach them not to hate. Your children will care about fairness and justice for others.
And finally,
Your kids will start converting you; the gay agenda is coming home. We’ll convert your children; and make an ally of you yet.
The thesis of the song is that advocacy can change minds, especially among young people. Those changed minds create an environment where the next generation is more likely to be “tolerant and fair”, and where some older people change their minds, too.
The data show that this thesis is, “just this once, correct”.
Sources
The General Social Survey (GSS) is a project of the independent research organization NORC at the University of Chicago, with principal funding from the National Science Foundation. The data is available from the GSS website.
In the third article, I found that generational differences on most questions related to abortion are small and probably not practically or statistically significant.
GSS respondents were asked several questions related to their attitudes about sex:
There’s been a lot of discussion about the way morals and attitudes about sex are changing in this country.
If a man and woman have sex relations before marriage, do you think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
What if they are in their early teens, say 14 to 16 years old? In that case, do you think sex relations before marriage are always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
What about sexual relations between two adults of the same sex–do you think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
What is your opinion about a married person having sexual relations with someone other than the marriage partner–is it always wrong, almost always wrong, wrong only sometimes, or not wrong at all?
For each of these questions, I count the fraction of respondents who reply “always wrong”.
And I looked at responses to one other sex-related question:
Would you be for or against sex education in the public schools?
Here are the results:
Generational changes in attitudes related to sex.
The blue markers are for people whose religious preference is Catholic, Protestant, or Christian; the orange markers are for people with no religious affiliation.
For each group, the circles show estimated percentages for people born in 1968; the arrowheads show percentages for people born in 1993.
For both groups, the estimates are for 2018, when the younger group was 25 and the older group was 50. The brackets show 90% confidence intervals.
In almost every scenario, young Christians are less likely than the previous generation to say that sex is “always wrong”, and in the cases of homosexual and teen sex, the changes are substantial.
Opposition to premarital sex was already low and did not change as much. Support for sex education was already high and is now an overwhelming majority.
The exception is extramarital sex, where there is practically no generational change: more than 80% of both generations think it is always wrong.
Compared to their Christian peers, the non-religious are more sex-positive by 15-30 percentage points. And their generational changes go in the same direction, with young Nones less likely to think sex in these scenarios is wrong.
But again, extramarital sex is the exception; among the Nones, the small generational change is within the margin of error.
This exception suggests that both groups distinguish between actions that harm people and transgressions of divine law.
Then, for a while, the story was that people were leaving organized religion, but they were still religious or at least spiritual; that is, they were “believing without belonging”.
In this series of articles, I have looked at changes among the ones who are left behind; that is, the decreasing fraction who identify as Christian. On many dimensions, the pattern is the same: young Christians are more secular than the previous generation.
Responses that follow this pattern include:
Almost all religious beliefs and activities, except belief in the afterlife.
Opposition to sex and sex education, except extramarital sex.
Matters of public policy including the legalization of marijuana, pornography, and euthanasia; support for affirmative action; and opposition to the death penalty and school prayer.
Many questions related to public spending follow the same pattern, with younger Christians generally moving toward positions held by their secular peers; the only substantial exception is mass transportation, which has less support among young people in both groups [although this result is so surprising to me that I need more evidence to be confident it is correct].
The most notable exceptions are opposition to gun control and abortion, which show almost no generational changes. Maybe not coincidentally, these exceptions are probably the most politicized topics among the questions I explored.
In summary, we can describe secularization in the U.S. as the sum of two trends, changes in affiliation and changes in belief. Both trends are moving fast, and they are moving in the same direction, away from religion.
Generational changes in public spending priorities
In the third article, I found that generational differences on most questions related to abortion are small and probably not practically or statistically significant.
GSS respondents were asked, “We are faced with many problems in this country, none of which can be solved easily or inexpensively. I’m going to name some of these problems, and for each one I’d like you to tell me whether you think we’re spending too much money on it, too little money, or about the right amount.”
Since they asked about 18 areas of public spending, I’ve put them in three categories:
Areas where young people are more inclined to increase spending,
Areas where young people are less inclined to increase spending, and
Areas where generational differences are inconsistent or small.
The following figure shows the first group, areas where young people are more likely to say we spend too little:
Generational changes in views on public spending
The blue markers are for people whose religious preference is Catholic, Protestant, or Christian; the orange markers are for people with no religious affiliation.
For each group, the circles show estimated percentages for people born in 1968; the arrowheads show percentages for people born in 1993.
For both groups, the estimates are for 2018, when the younger group was 25 and the older group was 50. The brackets show 90% confidence intervals.
The way these questions were posed, I suspect that most respondents can’t answer them literally. Few people know how much we spend in each area, what we spend it on, or what effect it would have if we spent more.
So their answers reflect some combination of how important they consider each issue, how much they think we are spending, and how effective they imagine more government spending would be.
With those caveats, we can draw a few conclusions:
On these issues, the priorities of Christians and Nones are generally aligned. The biggest difference between the groups is on spending to protect the environment, but a majority of both groups think we are spending too little.
The biggest generational changes are in foreign aid and protecting the environment; on both issues, young people are substantially more inclined to increase spending. But with respect to foreign aid, it is still a small minority.
The following figure shows areas of public spending where the generational change is generally negative:
Generational changes in views on public spending
Compared to their parents’ generation, young people are substantially less likely to increase spending on the military, transportation infrastructure and Social Security. To me, the direction of those changes is not surprising, but the magnitude is.
The other change I find surprising is in support for mass transportation, which decreased in both groups. I double-checked the data and this result seems to be correct, but it might warrant more investigation.
Finally, the following figure shows areas of public spending where generational changes are small and unlikely to be practically or statistically significant.
Generational changes in views on public spending
On these issues, the spending priorities of Christians and Nones are generally aligned, although Nones are more inclined to increase spending on space exploration, alternative energy, and education.
In the next article, I’ll look at generational changes related to confidence in various government and private institutions.
A large majority of Americans support legal abortion, at least in some circumstances
GSS respondents were asked, “Please tell me whether or not you think it should be possible for a pregnant woman to obtain a legal abortion” under different circumstances.
The following figure shows the results.
Generational changes in beliefs about legal abortion
The blue markers are for people whose religious preference is Catholic, Protestant, or Christian; the orange markers are for people with no religious affiliation.
For each group, the circles show estimated percentages for people born in 1968; the arrowheads show percentages for people born in 1993.
For both groups, the estimates are for 2018, when the younger group was 25 and the older group was 50. The brackets show 90% confidence intervals.
Before we look for generational changes, we should notice the starting point: a large majority of Americans support legal abortion, at least in some circumstances.
In cases of severe birth defects and pregnancy due to rape, the majority is about 70% of Christians and 90% of the nonreligious.
In cases of serious danger to the woman’s health, it’s almost 90% of Christians and nearly all of the nonreligious.
Under other circumstances, opinions are more divided, with support near 40% among Christians and 70% among the Nones.
Looking now at the generational changes, I see only one that is likely to be practically and statistically significant: younger people in both groups are less likely than the previous generation to support legal abortion if there is a chance of serious birth defect.
Even so, there is majority support in both groups, more than 60% among Christians and 80% among Nones at age 25.
In summary:
Beliefs about abortion depend substantially on the circumstances;
In many circumstances, a large majority of Christians and the non-religious support legal abortion;
Even where there is disagreement between the groups, there is substantial diversity of opinion within both groups;
Generational changes in these opinions are generally small and within the statistical margin of error.
