Indeed, I alluded to this when I first started an investigation into trends in unintended pregnancy, writing,

As with all things reproductive, these results are heavily age-covariated. What age groups and birth cohorts are experiencing these trends in the unintended pregnancy rate?

What age groups are most at risk for unintended pregnancy? Has this changed over time?

Furthermore, it is possible that the surge in the rate of unintended pregnancy per capita in the 1970s is an illusion. It could be an artifact of age composition effects. Sometimes demographic changes within a population create the illusion of a trend that disappears when analysis is broken down by age.

This age composition hypothesis is plausible. The 1970s followed the Baby Boom. As the Baby Boom generation has moved through various age groups, they have dramatically changed the age distribution in the United States.

Could the supposed surge in the unintended pregnancy rate be an illusion created by there being more women in the age groups with the highest risk for unintended pregnancy?

What Age Groups Are Most at Risk?

Figure 1 depicts trends in the rate of unintended pregnancy per woman per year in the United States from 1974 to 2014, broken down by age group.

The 40+ age group is not included due to insufficient sample size.

We can see in Figure 1 that the 20-24 year old age group has by far the highest risk of unintended pregnancy. More generally, by comparing the individual trendlines with the dashed grey line, groups 15-19, 20-24, and 25-29 have higher-than-average risk, and ages 30 and over have lower-than-average risk.

Furthermore, we can immediately reject the age-composition hypothesis. The 1970s saw an increase in the rate of unintended pregnancy across almost all the age groups: 15-19, 20-24, 25-29, and 30-34. The surge was real.

While the 1970s surge was across all age groups under 30, the decline after the 1990s occurred primarily among teens aged 15-19. Indeed, this age group had the greatest changes in risk of unintended pregnancy.

For ages 20 and higher, there is a consistent ordering of risk from younger age groups (with the greatest risk) to older age groups (with the least risk), as seen in Figure 2.

While the overall ordering remained, the spread started to narrow slightly in the 21st century, due to a decline in the unintended pregnancy rate among the highest-risk 20-24 group and, to a lesser extent, a smaller increase in risk among the 35-39 and 40-44 groups.

The Decline in Unintended Pregnancy among Teens

The 15-19 age group is the only one that breaks this younger-implies-higher-risk ordering and is the only age group that changed its ordering relative to the others during this time frame, as depicted in Figure 3.

In 1974, the 15-19 age group had a risk for unintended pregnancy greater than the 25-29 age group and less than the 20-24 age group. However, in the 1990s, the unintended pregnancy rate began to decline among 15-19 year olds. By the 21st century, the unintended pregnancy rate of the 15-19 group was consistently lower than the 25-29 group.

This makes sense because reducing teen pregnancy was top of mind in the United States in the 1990s. The nonprofit National Campaign to Prevent Teen Pregnancy was founded in 1996. It was unusual (by 21st-century standards) in that it worked across ideological lines and included advocates of both sex education and abstinence.

Furthermore, these initiatives were not confined to the nonprofit sector; they were also incorporated into public policy. For instance, the 1996 Temporary Assistance for Needy Families (TANF) welfare reform legislation included financial incentives for states to decrease the out-of-wedlock birth rate without increasing the abortion rate, especially among teenagers.

It makes sense that teen pregnancies would get the most attention.

Teens consistently have the highest proportion of pregnancies that are unintended, as seen in Figure 4. Thus, a pregnant teen is — at least three-quarters of the time — a visible sign of social disorder.

However, we should note that a simple majority of pregnancies in the 20-24 age group are unintended. This age group has the highest rate of unintended pregnancy per woman, and this rate has decreased less than the rate for the 15-19 age group. (Figure 3)

The rate for the 15-19 group decreased by about 50% from its peak, whereas the rate for the 20-24 group decreased by about 25%.

Age Composition Changes

If we look at the age composition of women in the 1970s, we can further reject the hypothesis that the surge in unintended pregnancies was an illusion. (Figure 5) The percentage of the population in their teens — the second-most-at-risk age group — was actually decreasing. The percentage in their 20s was mostly the same. And the percentage in their 30s — when the risk for unintended pregnancy was less than half of that of the younger groups (Figure 2) — was actually increasing.

We can further reject the age composition hypothesis by computing an age-standardized rate. For this, you pick an age composition and calculate what the rate would have been at various times if the age composition remained the same. The trends in the unintended rate remain the same, no matter how you do this. (Not pictured.)

Period Effects, Not Cohort Effects

Finally, whenever investigating age, it is worth checking for cohort effects.

As seen in Figure 6, there are no cohort effects. The trends for each cohort have the same shape and largely overlap. The major differences come from different cohorts being at certain ages when the unintended pregnancy rate was relatively low or high, not from persistent between-cohort differences.

Perhaps the cohort born in 1990-1999, who grew up after the teen pregnancy prevention initiatives of the 1990s, will be the first to have a consistent between-cohort difference, but that remains to be seen.

Conclusion

The 20-24 age group has consistently had the highest rate of unintended pregnancy. Risk for unintended pregnancy declines with each older age group, and this has been relatively stable.

The most dramatic change in our analysis period has been among the 15-19 age group, which went from a higher unintended pregnancy rate than the 25-29 group to a consistently lower one.

The surge in the per capita rate of unintended pregnancy in the 1970s was real, not a demographic illusion. The increase occurred across nearly all age groups, and the changes in age composition of women actually worked against the surge.

In previous analyses, I quantified the surge in unintended pregnancies that occurred from 1974 to 1981 in the wake of the so-called “sexual revolution.”

However, that analysis did not resolve the role of birth control in the surge. The question remained, “To what extent does the increase in the unintended pregnancy rate occur among those using contraception versus among those who are not?”

It turns out, all of it.

I did not expect such a clean answer. Most empirical results at the population level are messy and multifactorial.

Contraceptive Users Drove the Entire Surge

The average number of unintended pregnancies per year went from approximately 2.49 million per year to 3.06 million per year, a difference1 of about 566 thousand pregnancies. About 475 thousand of this increase was among contraceptive users, versus about 90 thousand among those not using a birth control method.

However, the 90 thousand increase among non-contraceptive users is within its margin of error of 0, so this supposed difference could be due to sampling variation.

Furthermore, remember that the population was growing during this time frame, as well.

I therefore divided the estimates by the median number of women of reproductive age during each period.

The estimate for the rate of unintended pregnancies among women not using a contraceptive method actually decreased slightly during the surge from 33.1 unintended pregnancies per 1,000 women per year to 32.1, though zero is well within the difference’s margin of error. Thus, we can conclude that the rate of unintended pregnancy among non-users of birth control was basically unchanged.

All of the increase in the unintended pregnancy rate was among contraceptive users, going from 18.6 unintended pregnancies per 1,000 women per year to 26.1.

There were still more unintended pregnancies among non-users than among contraceptive users, but the fact that the surge occurred entirely among users of birth control narrowed this gap quite a bit.

The Pill's Outsize Role

This raises the question, “What methods were being used?”

Cycle 3 of the National Survey of Family Growth (NSFG) asked respondents to report on the contraceptive methods they used in the time interval leading up to the pregnancy.

Respondents could have up to four answers if they changed contraceptive regimes during the period, and each regime could be either a single birth control method or a combination of methods.

I first estimated the percentages of unintended pregnancies for which the various methods were mentioned at least once in the lead-up period, as shown in Table 3.

I found no statistically significant difference in the percentages of pregnancies associated with each method between 1974-1977 and 1978-1981, with only a couple of exceptions,2 so Table 3 includes only the estimates for the whole analysis period.

In the lead-up to unintended pregnancies during contraceptive use, about half of women at least tried the oral contraceptive pill, a greater proportion than any other method.

What about the last method regime mentioned, which is presumably the one used during the conception of the unintended pregnancies?

Looking at Table 4, again, the oral contraceptive pill accounts for the greatest proportion. Indeed, the older methods based on the use of condoms, withdrawal, or calendar timing combined account for fewer unintended pregnancies than the pill.

This is remarkable because hormonal contraceptive pills were the “modern” birth control method in the sexual revolution. Yet the subpopulation experiencing the surge in unintended pregnancies relied on the pill more than any other method.

Discussion

The surge in the rate of unintended pregnancies per woman of reproductive age that occurred in the 1970s occurred entirely among respondents who reported using contraception. Furthermore, the most common method of birth control among those experiencing unintended pregnancies was the oral contraceptive pill.

We, of course, do not know how well respondents were using their birth control methods. Some contraception users could have been using their methods inconsistently. Perhaps some respondents claimed to have been using contraception when they actually were not, due to social desirability bias.

Furthermore, since a much greater proportion of women experiencing unintended pregnancies tried the pill than were using the pill at the time of conception, many women quit the pill, likely due to side effects.

However, as we will see in subsequent analysis, “modern” contraceptive methods are just not as effective as many think, and this gap between expectations and reality likely explains a lot of the surge in unintended pregnancies.

Regardless of the more proximal causes of the contraceptive failures, one striking conclusion emerges from this analysis: the birth control movement was a massive failure.

A Movement That Failed Its Own Goals

The birth control movement promised voluntary motherhood and an end to unintended pregnancies. It also promised — as the modern “family planning” establishment would like you to forget — to put the abortionists out of business. It promised that,

Birth Control will prevent abortion. It will do away with the practice of taking drugs and poisonous nostrums to end undesired pregnancies. It will put an end to the tens of thousands of illegal operations to which women resort in despair. Mothers will not submit to the murder of unborn children when they can control conception. (Sanger 1923)

Those promises did not come to fruition. The oral contraceptive pill — and hormonal birth control more generally — was the last “hurrah” of the birth control movement. The movement's now-maligned founder, Margaret Sanger, even emerged from her retirement to raise funds for the development of the pill. (Eig, 2016)

However, once the pill was actually used by the general population, it led to more unintended pregnancies, not fewer. Rather than bringing induced abortion to an end, it contributed to the surge in the rate of induced abortions in the 1970s, which was driven by the surge in the rate of unintended pregnancies.

Perhaps things could have been different. There were many cultural changes in the 1960s. Perhaps if the pill had been introduced in 1915 instead of 1960, it might have been used more responsibly. This was, at least technically, possible, since a basic understanding of ovulation suppression was already known at the time. (Loeb, 1911)

Alternatively, perhaps the culture has always been such that a technology as flawed as hormonal contraception was going to backfire. These are speculations for which we will never know the truth.

The birth control movement would largely die after the invention of the pill. Alan Guttmacher became president of the Planned Parenthood Federation of America (PPFA) in 1962 and converted the PPFA from a birth control organization to an abortion advocacy organization. Since 1960, no truly new method of birth control has been introduced to the market.3

Sanger witnessed how the birth control movement led by Maria Stritt in Germany in the 1910s went “out of existence” and was replaced by the abortion movement of 1920s Germany. So, too, did the birth control movement in the United States that she founded go out of existence in the 1960s, replaced by the Guttmacher vision of unrestricted abortion. Since then, we have had technological stagnation in contraception.

The birth control movement’s crowning achievement backfired, and the movement failed to achieve its stated goals.

Methodological Issues

Timing of Surge in Unintended Pregnancies

As discussed earlier, my estimates necessarily begin in 1974 due to limitations in data availability. The surge, when adjusted for population growth, plateaus between 1980 and 1981, as seen in Figure 2. I therefore considered the surge period for this analysis to be 1974 to 1981.

Conveniently, this falls entirely within the analysis period for Cycle 3 of the NSFG, so I did not need to harmonize data across surveys.

Contraceptive Use Assessment

To determine whether contraception was used during the time frame in which a pregnancy was conceived, I used the variables C4, C6, C17, and C23 from Cycle 3 of the NSFG.

Both C4 and C17 encode answers to the same question,4

Between (A and B), did you or your partner ever use any method of birth control or family planning for one month or more?

Likewise, C6 and C23 encodes answers to very similar questions,

At the time you became pregnant with (B), had you stopped using all methods of birth control?

and

“Had you stopped using all methods before you became pregnant?

respectively.

I considered a pregnancy to have been conceived during contraception use if the respondent answered “yes” to C4 / C17 and “no” to C6 / C23.

Contraceptive Method Assessment

The NSFG Cycle 3 pregnancy file records the contraceptive methods used during each pre-pregnancy interval through a series of “mention” variables (C5_1M, C5_2M, C5_3M, C5_LM for one interview loop and C18_1M, C18_2M, C18_3M, C18_LM for the other).

These variables encode responses to the question,

Looking at the card and starting with first method you used between (A and B), please tell me the letter for each method used for one month or more in the order you used them.

The variables record methods in chronological order, with C5_1M / C18_1M being the first method used and C5_LM / C18_LM being the last.

Each variable can encode either a single method (codes 1–15, such as pill, condom, or IUD) or a simultaneous pair of methods (codes 16–120, such as “condom + foam”).

Intent Assessment and Pooling for Estimation

A pregnancy was considered unintended exactly as before, and estimates were pooled across four-year intervals due to limitations in sample size. Thus, I compared an average year in 1974-1977 with an average year in 1978-1981.

Citations

Eig, Jonathan. 2016. The Birth of the Pill: How Four Pioneers Reinvented Sex and Launched a Revolution. Pan Macmillan.

Loeb, L. (1911). The Cyclic Changes in the Mammalian Ovary. Proceedings of the American Philosophical Society, 50(199), 228–234.

Sanger, Margaret. 1923. “A Better Race Through Birth Control.” The Thinker, November 1923.

At first blush, it may look like this did not affect the number of live births, which also surged in the 1970s.

However, this is confounded by the fact that the number of women of reproductive age increased throughout the 20th century.

This analysis follows the latest iterations of the National Survey of Family Growth (NSFG) and takes 15 to 44 years as “reproductive age.” (Henceforth, plots that say “per woman” mean per woman of reproductive age.)

Even adjusting for population growth, the rate of induced abortions per woman surged in the 1970s post-Roe and Doe.

However, after the population adjustment, we can see that the picture is more complicated for live births.

The surge in abortions may have indeed altered the trajectory of the overall birth rate in the 1970s and 1980s.

Here is where the analysis of pregnancy intent versus outcome adds an interesting dimension.

The number of live births that began as unintended pregnancies continued to increase throughout the 1970s and into the early 1980s. Again, let us adjust to get per capita estimates.

The increase in live births from unintended pregnancies in the 1970s and early 1980s is ever so slight when seen per capita.1

We learned in the previous analysis that most induced abortions (~90%) end unintended pregnancies, so whatever depression in the birth rate we see due to the post-Roe/Doe surge in abortions would show up especially among unintended pregnancies.

However, the rate of unintended pregnancy per woman continued to increase during the abortion surge, not decrease.

How can the abortion rate surge and the rate of births from unintended pregnancies continue to increase?

This is because the surge in the abortion rate in the 1970s and the slight increase in the rate of births from unintended pregnancies were both driven by the surge in the rate of unintended pregnancies.

Even the surge in induced abortions after Roe v. Wade and Doe v. Bolton could not stop the sexual revolution’s legacy from increasing the rate of births from unintended pregnancies. The birth rate would have likely been even higher without the abortion rate surge, but the birth rate nonetheless continued to increase.

The surge in the rate of unintended pregnancies during the sexual revolution led to many millions of excess unintended pregnancies and millions of excess births from unintended pregnancies.

That analysis required my weight adjustment for abortion under-reporting in the National Survey of Family Growth (NSFG) data.

The analysis and the adjustment together surface two variables: pregnancy outcome and pregnancy intent. Comparing those two variables with each other illuminates both the dramatic surge in the number of abortions in the 1970s and the long, gradual decline in abortions after 1990.

Intent and Outcome

Figure 1 illustrates that unintended pregnancies account for a far greater proportion of induced abortions than live births.

Of live births, about 30% to 40% began as unintended pregnancies over four decades.

Among induced abortions, more than 90% started as unintended pregnancies in the 1970s and 1980s, after which a slight declining trend may be present, but this is obscured by sampling error.

Figure 2 looks only at pregnancies that end in induced abortion.

This illustrates why the abortion under-reporting adjustment is especially needed for trend analysis of unintended pregnancies. The adjustment has a relatively small impact on estimates of intended pregnancies because relatively few intended pregnancies are ended by induced abortion.

However, the difference in the rate of induced abortion between unintended and intended pregnancies might be narrowing ever so slightly.

Intended Pregnancies

Let us look at the converse comparison. Figure 3 illustrates trends in pregnancy outcomes for intended pregnancies.

The proportion of intended pregnancies ended by induced abortion is small (less than 5%) and flat (no obvious increasing or decreasing trend). The proportion ending with live birth appears to be declining slightly, but that is attributable to an increasing trend in miscarriages, etc.

Why the Miscarriage Trend Is Misleading

We should not put too much stock in supposed trends in the miscarriage rate. These are trends in the noticed miscarriage rate. Most miscarriages go unnoticed, especially those that occur early in pregnancy.

The best evidence of what proportion of miscarriages go unnoticed comes from studies in which women are given repeated human chorionic gonadotropin (hCG) assays. (Jarvis, 2017)

Most of us are familiar with hCG assays because home pregnancy tests are simple versions of them. Studies using more sensitive assays find that the majority of miscarriages are subclinical and would likely have gone unnoticed without testing. (Wilcox et al., 1988; Wang et al., 2003)

However, these studies use convenience samples, which are inherently unrepresentative of larger populations. Therefore, the results do not generalize, and we should not put too much stock in the specific numbers found.

The important takeaway is that the apparent increase in miscarriage rate is likely just an increase in miscarriage detection. More and more hCG testing options, including those that integrate with apps that monitor menstrual cycles, are becoming available to consumers. As these options become more sophisticated and are used more often, more miscarriages will be detected.

Figure 4 repeats Figure 3, but excludes miscarriages, ectopic pregnancies, and stillbirths.

We can see that there is no trend in the outcomes for intended pregnancies if we accept the hypothesis that miscarriage detection is increasing.

Unintended Pregnancies

Now, let us take a look at unintended pregnancies.

There does appear to be a trend here in Figure 5, unlike for intended pregnancies. However, we once again see the apparent increasing trend in miscarriages, likely due to increased detection of miscarriage. Let us consider just pregnancies that end in live birth or are ended by induced abortion in Figure 6.

During the surge in unintended pregnancies in the wake of the sexual revolution, unintended pregnancies were split almost evenly in outcome between live birth and induced abortion. In the late 1980s and early 1990s, unintended pregnancies were slightly more likely to be ended by induced abortion than to end in live birth.

However, since then, there has been a reversal of the trend such that unintended pregnancies are becoming less likely to be ended by induced abortion and more likely to result in live birth. In the most recent NSFG complete data set,1 the split is 60/40 in favor of live birth.

Explaining Both Ends of the Abortion Arc

After this and the preceding analysis, we now have a much better understanding of the components of the trends in total induced abortions in the United States after Roe v. Wade and Doe v. Bolton, as seen in the Guttmacher Institute’s Abortion Provider Census (APC) data. These trends are summarized in Figure 7.

The number of induced abortions per year in the United States surged in the 1970s, crested in the 1980s, and started to decline gradually in the 1990s. (Panel 1) This is the titular “arc.”

The trend in the total number of unintended pregnancies per year is similar at first, surging in the 1970s and cresting in the 1980s. However, this trend does not decline as dramatically as the number of induced abortions did. In fact, the total number of unintended pregnancies is more or less level (within sampling error) from 1987 on. (Panel 2)

So what is going on here?

There are two contradictory trends. Throughout this period, the number of women of reproductive age in the United States increased. (Panel 3) From the mid-1980s on, the rate of unintended pregnancy per woman per year decreased. (Panel 4) These two trends combine to produce a roughly constant number of total unintended pregnancies per year from about 1987 on.

So if the total number of unintended pregnancies per year is more or less constant past 1987, how do we see the declining trend in induced abortions per year from 1990 on? (Panel 1)

There are two drivers. First, the rate of unintended pregnancy per woman continued to decline past 1987. (Panel 4) This decline is not visible in total unintended pregnancies (Panel 2) because it is offset by the continued growth in the number of women of reproductive age. (Panel 3) Second, the rate at which unintended pregnancies were ended by induced abortion also declined from about 1990 on. (Panel 5) Together, fewer unintended pregnancies per woman and a lower propensity to abort them combined to produce the declining trend in total induced abortions.

The surge in the number of induced abortions per year in the 1970s was dramatic. In 1974, there were 898,570 induced abortions compared with 1,577,340 in 1981, according to APC data.

During this time, the increase in the rate at which unintended pregnancies were ended by induced abortion was relatively small and within sampling error, going from 48.6% (±2.1%) in 1974-1977 to 50.7% (±2.1%) in 1982-1985.

However, the increase in the unintended pregnancy rate was more dramatic, going from 51.6 (±3.3) per 1,000 women in 1974-1977 to 59.5 (±4.0) in 1982-1985. The increase in the total number of women of reproductive age was also dramatic, going from 46.6 million in 1974 to 53.9 million in 1981.

The increase in the rate of induced abortion per unintended pregnancy was small and may just be sampling variation. Most of the surge in abortions in the 1970s came from the increase in the rate of unintended pregnancy per woman and the increase in the number of women of reproductive age.

The abortion surge followed from the surge in unintended pregnancies. The abortion decline required both fewer unintended pregnancies and a lower propensity to abort them.

Methodological Issues

Abortion Under-reporting

These results depend on the assumption that the induced abortions reported on the NSFG are representative of the abortions that are not reported.

As I discussed in the previous analysis, the abortion under-reporting in the NSFG is so severe that less than half of pregnancies ended by induced abortion are reported in the NSFG. Because of this, the official documentation advises us not to use the NSFG for substantive analysis of covariates of induced abortion.

This is because the assumption that the reported abortions are representative of the unreported ones is almost certainly false for most covariates. The subpopulation of women reporting abortions is likely different from the subpopulation not reporting abortions in important ways.

However, for this specific analysis, the relevant question is not whether non-reporters differ from reporters in general; it is whether they differ in one specific respect: the proportion of their abortions that ended unintended pregnancies versus intended pregnancies. This is a narrower assumption than representativeness in general, and it is the only dimension of unrepresentativeness that can affect these results.

This is actually a much safer assumption, since the vast majority of induced abortions end unintended pregnancies. As we saw in Figure 2, based on abortions reported in the NSFG, ~90% of induced abortions end unintended pregnancies, whereas only ~10% end intended pregnancies. Even if the proportions differ among the unreported abortions, it is likely that the proportion that end unintended pregnancies is much greater than the proportion that end intended pregnancies.

Because of this, the estimates in this analysis might be biased (in the statistical sense) due to abortion under-reporting, but the magnitude of this bias is unlikely to be sufficient to change the conclusions.

Furthermore, the trend analysis in this article is on even better ground. A fixed bias in the composition of unreported abortions would shift the estimated levels up or down but would leave the trend intact. For the observed decline in abortion per unintended pregnancy to be untrue, the bias would have to change across survey cycles. The unreported fraction would need to shift systematically toward or away from unintended pregnancies over time in a way that mimics a real trend.

Therefore, in this specific case, I think it is sound to use NSFG for data to analyze trends in pregnancy intent versus outcomes, including induced abortion.

Nonetheless, because of the abortion under-reporting problem, this is likely the last analysis of abortion covariates based on NSFG data that you will find on Born Curious.

Pooling Estimates Across Years

In the previous analysis, I generated estimates for the various quantities per year. This was more feasible because the NSFG samples were being narrowed to subsamples based on a single variable: pregnancy intent.

The perceptive reader will notice that the estimates in this analysis are not generated on a per-year basis. This is because the subsamples used to estimate them were restricted by two variables simultaneously, pregnancy intent and pregnancy outcome, thereby reducing their size considerably. Annual estimates led to confidence intervals that were too wide to support reliable inference.

To address this, estimates were pooled across multi-year periods. For Figures 1 through 6, the pooling follows the “analysis periods” that correspond to the time periods assigned to each NSFG cycle. For Figure 7, to make the estimates comparable to the annual quantities, the estimates were pooled over periods of no more than 4 years.

Citations

Jarvis, G. E. (2017). Early embryo mortality in natural human reproduction: What the data say. F1000Research, 5, 2765. https://doi.org/10.12688/f1000research.8937.2

Wang, X., Chen, C., Wang, L., Chen, D., Guang, W., & French, J. (2003). Conception, early pregnancy loss, and time to clinical pregnancy: A population-based prospective study. Fertility and Sterility, 79(3), 577–584. https://doi.org/10.1016/S0015-0282(02)04694-0

Wilcox, A. J., Weinberg, C. R., O’Connor, J. F., Baird, D. D., Schlatterer, J. P., Canfield, R. E., Armstrong, E. G., & Nisula, B. C. (1988). Incidence of Early Loss of Pregnancy. New England Journal of Medicine, 319(4), 189–194. https://doi.org/10.1056/NEJM198807283190401

As I have written previously, much criticism of the sexual revolution rests on a false premise: that we live in a society with effective contraception. In reality, ours is a society that thinks it has effective contraception, but does not.

These two ideas are related. A society with genuinely effective contraception and the culture to use it consistently would not have experienced the surge that occurred from the 1970s into the 1980s.

A First Look at the Data

As in previous analyses, I used data from the National Survey of Family Growth (NSFG).

Totals vs. Rates

The naive approach would be to estimate the number of unintended pregnancies each year, as illustrated in Figure 1 below.

Even in the naive approach, we can see a spike in unintended pregnancies in the 1970s in the wake of the sexual revolution. However, this naive approach is incorrect for several reasons.

For one, the United States' population was increasing during this period. Figure 2 is an equivalent plot, but for intended pregnancies.

It appears in Figure 2 that intended pregnancies had an increasing trend for the last quarter of the 20th century. However, when we divide by the number of women of reproductive age1 in the United States in Figure 3, we see something else.

The increase in intended pregnancies can be explained almost entirely by population growth. In Figure 3, any increase is less than the sampling margin of error.

Thus, we are more interested in the rate of unintended pregnancies per woman of reproductive age than the total number of pregnancies.

Figure 4 shows unintended pregnancies per 1,000 women of reproductive age per year. It still appears as though there was a spike in unintended pregnancies in the 1970s, followed by a decline in the 1980s back to a steady rate. However, this is also incorrect.

Abortion Under-reporting

An issue known to affect surveys such as the NSFG is the fact that fewer than half of pregnancies that ended in induced abortion are reported. This abortion under-reporting is so bad that the official NSFG documentation instructs us not to attempt to analyze abortion phenomena with the NSFG data.

Most pregnancies that end in induced abortion started as unintended pregnancies, so the bias in estimates due to abortion under-reporting is particularly bad for unintended pregnancies.

Figure 5 illustrates the effect of abortion under-reporting by comparing the NSFG’s estimates of the number of induced abortions per year with the number of induced abortions per year from an external source, namely, the Guttmacher Institute’s Abortion Provider Census (APC).

This is an issue just for induced abortions. Live births do not have this problem, as illustrated by Figure 6.

Adjusted Estimates and Their Implications

Fortunately, I have already addressed this by adjusting the NSFG’s weights to compensate for abortion under-reporting.

Figure 7 repeats the comparison from Figure 5, but now uses my adjusted weights. Note that estimates of the number of induced abortions per year from the NSFG now match the number of induced abortions per year from the APC.

This adjustment changes the estimates of unintended pregnancy, as illustrated in Figure 8.

The unadjusted estimates peak at 47.6 (± 8.9) unintended pregnancies per 1,000 women of reproductive age per year in 1979. However, the adjusted — and more accurate — estimates peak at 60.8 (± 9.8) unintended pregnancies per 1,000 women in 1983. Indeed, the adjusted estimates of the unintended pregnancy rate are roughly 30-60% higher in the earlier cycles.

Also, the timing of the trend is different in the more accurate estimates. The rate of unintended pregnancy increases throughout the second half of the 1970s, crests in the 1980s, but remains high until about 1985, and then slowly starts to decrease with a few interruptions thereafter. (The discontinuity in 1986 is likely due to sampling variation.)

Because our estimates begin in 1974, which is late into the sexual revolution, we cannot estimate how much of an increase in the unintended pregnancy rate occurred during the sexual revolution. We can estimate something of a minimum, though.

The estimated 1975 rate is lower than the 1974 rate. Since we lack estimates before 1974, it is not clear which of these is more of a discontinuity, so let us be conservative and use the higher 1974 rate as a baseline.

Even using the higher baseline rate, there were approximately 2,730,000 excess unintended pregnancies between 1975 and 1985. This is the difference between the actual estimates of unintended pregnancies each year and the number that would have occurred if the rate stayed at the 1974 level.

This 2,730,000 figure is conservative. The unintended pregnancy rate may have been even lower in the 1960s. It may, of course, have been higher in the 1960s, as well. We do not know whether the 1970s trend of increasing unintended pregnancy rate started in the 1960s. But considering the unintended pregnancy rate was increasing rapidly in the 1970s and decreased to even lower rates in the 2010s than the lowest rates measured in the 20th century, it is not an unreasonable conjecture that the unintended pregnancy rate was lower in the 1960s than in the 1970s.

This raises the question of why our estimates begin in 1974. This is addressed in “Methodological Issues” below.

Regardless, the 2,730,000 figure is not intended to be a precise estimate but rather to give us an idea of the order of magnitude of the increase in the unintended pregnancy rate.

The sexual revolution likely led to additional unintended pregnancies in the multiple millions.

Further Analysis

Of course, this analysis is preliminary.

For one, it only implicates the sexual revolution in the rise in unintended pregnancies. The extent to which contraception in particular is implicated requires further analysis.

• To what extent does the increase in the unintended pregnancy rate occur among those using contraception versus among those who are not?

This is actually a surprisingly complicated question, but it would shed light on how much of the degradation of reproductive responsibility in the wake of the sexual revolution was contraception-driven.

This analysis also raises other questions.

• As with all things reproductive, these results are heavily age-covariated. What age groups and birth cohorts are experiencing these trends in the unintended pregnancy rate?

• Is the rate of induced abortion per unintended pregnancy increasing or decreasing?

Methodological Issues

Several methodological constraints shaped this analysis. The abortion under-reporting adjustment is described above; two additional issues deserve elaboration here.

Truncation Effect

Because of truncation, survey data are typically accurate only for pregnancies that occurred within a limited period before the survey. This truncation effect occurs because after any given amount of time, some percentage of women who had pregnancies will age out of the target population, die, emigrate, or otherwise leave the survey population.

For instance, a survey done in 1990 will miss any pregnancies from 1970 among women who died between 1970 and 1990, since they are no longer around to report on their pregnancies.

I have found that pregnancy estimates are accurate for about 5 to 10 years before a survey, due to the truncation effect.

The NSFG, administered by the National Center for Health Statistics, is considered the premier source of population-level data on pregnancy intent and outcomes in the United States. Its first cycle ran in 1973. Before then, there were other surveys, such as the National Fertility Survey (NFS) administered by Princeton’s Office of Population Research.

You would think, then, that we would have estimates going back to about 1963.

Unfortunately, these earlier surveys (including the first two cycles of the NSFG) surveyed only married women,2 reflecting the sexual mores of their day. Needless to say, unintended pregnancy is a phenomenon that is experienced by unmarried women, as well. Therefore, these earlier surveys are biased — in the statistical sense — rendering them useless for estimates of unintended pregnancies.

The first cycle of the NSFG to survey all women, regardless of marital status, was Cycle 3, conducted in 1982, giving a range for first estimates of 1972-1977.

My analysis begins in 1974 because that is when the Guttmacher Institute first began reporting APC counts, which are needed for the weight adjustment.

Pregnancy Intent

Another major issue is that the concept of pregnancy intent is itself an oversimplification, though one that is still a useful abstraction.

For most of human existence, for most of the population, pregnancies were just something that happened in life, and people dealt with them. The idea of consciously planning a pregnancy has historically been relevant for a relatively privileged minority. It has become a concept for the general population only in the mid-20th century. (Campbell & Mosher, 2000)

Furthermore, pregnancy intent is complicated to measure cleanly. For instance, it is surprisingly common for people to be using contraception and yet not classify the resulting pregnancy as unintended. Indeed, many such women report being happy about the pregnancy. (Trussell et al., 1999)

Because of this complexity, Barrett et al. (2004) suggest using multiple dimensions and treating pregnancy intent more as a spectrum, rather than a binary intended vs. unintended classification.

I agree with this approach, and I even tried a limited version of it in this analysis. Unfortunately, the data just are not there, especially in the earlier cycles of the NSFG, and because this is a trend analysis, I needed to harmonize the data across all cycles of the NSFG I used. This harmonization required me to take the lowest common denominator across all cycles.

What I was left with was just the classification of the NSFG encoded in the WANTWIFE and OLDWANTR variables. These encode answers to questions such as

Right before you became pregnant, did you yourself want to have a(nother) baby at any time in the future?

and

So would you say you became pregnant too soon, at about the right time, or later than you wanted?

If the respondent indicates she did not want to have the pregnancy at any time in the future, the pregnancy gets classified as unwanted. If the respondent answers that she became pregnant too soon, the pregnancy gets classified as mistimed.

In my scheme, I classified any unwanted or mistimed pregnancies as unintended.

Bibliography

Barrett, G., Smith, S. C., & Wellings, K. (2004). Conceptualisation, development, and evaluation of a measure of unplanned pregnancy. Journal of Epidemiology & Community Health, 58(5), 426–433.

Campbell, A. A., & Mosher, W. D. (2000). A History of the Measurement of Unintended Pregnancies and Births. Maternal and Child Health Journal, 4(3), 163–169. https://doi.org/10.1023/A:1009519329226

Trussell, J., Vaughan, B., & Stanford, J. (1999). Are all contraceptive failures unintended pregnancies? Evidence from the 1995 National Survey of Family Growth. Family Planning Perspectives, 31(5), 246–247, 260.

However, this is the wrong premise.

We are actually living through a time in which a society thinks it has effective contraception, but does not. One of the major goals of my empirical analysis is to correct this mistaken belief.

But before we get into pregnancies conceived during contraceptive use, let us examine a related topic: “modern” contraceptives available today come with side effects — side effects so severe that women oftentimes stop using them.

Measuring Dissatisfaction

The National Survey of Family Growth (NSFG), the preeminent fertility survey in the United States, has asked female respondents ever since its Cycle 6, which was done in 2002, questions such as:

Some people try a method and then don’t use it again, or stop using it, because they are not satisfied with the method. Did you ever stop using a method because you were not satisfied with it in some way?

Then, for select methods, the NSFG asks a follow-up question such as:

What was the reason or reasons you were not satisfied with [method name]?

Together, these give us a picture of how frequently side effects make a contraceptive method unbearable.

Figure 1 gives us a baseline for the prevalence of the most common methods of contraception in the United States. I use the 2015-2019 data set because, as I have written about, there are issues with the 2022-2023 data (and, indeed, survey data more generally after the COVID-19 pandemic).

The numbers from Figure 1 serve as the denominators for the rest of the analysis, and the important takeaway is that the stakes are large. For instance, a one percentage point difference in discontinuation rate for pill-based methods represents about half a million women.

Discontinuation Rates by Method

The raw discontinuation rates (which include discontinuations for reasons other than side effects) are mostly consistent over time, except for the 2022-2023 data set. The estimated discontinuation rates all decrease in 2022-2023. However, due to methodological changes for the 2022-2023 NSFG, it should not be compared with earlier data, and the earlier data are generally more reliable.

Generally, there are three tiers of discontinuation rate, in order of lowest to highest:

1. withdrawal and condom

2. pill, implant, and intrauterine device (IUD)

3. patch, ring, and Depo-Provera

Withdrawal

Withdrawal is a natural comparison baseline for contraceptive methods, since it predates all “modern” methods by millennia. The discontinuation rate for withdrawal is in the range of 5% to 13%. (Figure 3) This leads to 2,729,000 (±514,000) total discontinuers based on 2015-2019 data.

Condoms

Condoms have a discontinuation rate similar to that of withdrawal. (Figure 3) However, because more people have tried condoms than any other method, this low rate still represents a lot of women in total terms. There were 5,102,000 (± 783,000) women who discontinued condom use due to dissatisfaction, according to 2015-2019 data.

Hormonal Methods

All the remaining methods (with one major exception) are essentially the same method in different modalities: exogenous hormones. They all involve different ways of delivering a cocktail of progestins and estrogens.

Pill

The discontinuation rate for pill methods hovers around 30% to 35%. (Figure 3) Pills are the most popular method after condoms. With their higher discontinuation rate, 17,910,000 (± 1,760,000) women have discontinued them due to dissatisfaction, based on 2015-2019 data.

To put that in perspective, according to the U.S. Census Bureau, there were only 74,394,051 women of reproductive age in the United States eligible for the 2015-2019 NSFG. About a quarter of women in the United States have quit the pill due to dissatisfaction!

There are over a hundred and fifty different brand names of contraceptive pills available in the United States. Even though some are basically the same as others, and some are in practice not manufactured, there are still dozens of discrete formulations of oral contraceptives.

Implant (Nexplanon)

In contrast, there has basically been one implant-based contraceptive available at a time in the United States. Norplant had Food and Drug Administration (FDA) approval starting in 1990, but was withdrawn in the United States in 2002 amid massive litigation. Implanon was introduced in the United States in 2006, but was discontinued in 2012. Nexplanon started replacing Implanon in 2010, and is the only implant-based contraceptive currently available in the United States.

Implants have a discontinuation rate right around ~30%, similar to contraceptive pills. (Figure 3) The 2015-2019 NSFG did not ask about implants, leaving a data gap because implant use has increased substantially in recent decades. According to the 2022-2023 data, 1,727,000 (± 406,000) women have discontinued implants due to dissatisfaction.

IUD

The discontinuation rate for intrauterine devices (IUDs) appears to be the least consistent based on NSFG data. (Figure 3) However, several things explain this. The NSFG changed its handling of IUDs midway through the run of the discontinuation questions, separating copper IUDs from hormonal IUDs.

Furthermore, the IUD market in the United States was effectively destroyed by the Dalkon Shield scandal of the 1970s. It did not recover until the 2000s — right when the NSFG began asking these questions. The small number of IUD users in early survey cycles, combined with a rapidly changing user population as the method regained acceptance, makes for noisy estimates.

Broadly speaking, IUD discontinuation rates stabilized in the same tier as pill methods, hovering around 25% to 35%.

Injectable (Depo-Provera)

Methods in the high-discontinuation tier — Depo-Provera, patch, and ring — tend to have rates of discontinuation due to dissatisfaction in the 40% to 50% range. (Figures 2 and 3) However, patch and ring methods are less common than other methods.

Depo-Provera is notable both for being the only injectable contraceptive method available in the United States and for having a very high discontinuation rate. Around 7,382,000 (± 824,000) women have discontinued Depo-Provera due to dissatisfaction. Additionally, there is active litigation over allegations that Depo-Provera increases the risk for brain tumors.

Side Effects Prevalence

The gap between the low-discontinuation-rate tier of methods and the two higher-discontinuation-rate tiers widens when we look at why women discontinue due to dissatisfaction.

Looking at the answers to follow-up questions asking the reason for discontinuation of condoms, most discontinue either because of the effects on the pleasure felt by the respondent or her partner. (Figure 4)

Looking at the answers for the pill, by comparison, the reasons cited for discontinuation are overwhelmingly side effects. (Figure 5)

This allows us to calculate a discontinuation rate due to side effects for the four methods that have follow-up questions in the NSFG asking about reasons. For these purposes, I took “you had side effects,” “did not like menstrual cycle changes,” and “doctor told you not to use” as discontinuation due to side effects.

Discontinuation rate due to side effects for condoms is consistently less than 2.5%. For pills, it is in the 20% to 25% range, and for Depo-Provera, it is around 40%. (Figure 6)

This leads naturally to the question, what exactly are these side effects?

What Are the Side Effects?

This is easy enough to look up because official FDA-approved documentation must list the adverse effects of a drug. The prevalence of side-effects in the documentation tends to be underestimated because the estimates come from clinical trials, which overstate drug efficacy and underestimate side-effects — a topic worthy of another article of its own.

Nonetheless, the FDA documentation at least gives a list of what the side effects are.

Figure 7 comes from the tables on the official Depo-Provera label. “Amenorrhea” refers to the absence of regular menstruation. The percentages are the number of women with each adverse reaction, expressed as a proportion of the total number participating in the clinical trial.

As discussed above, while there is one injectable (Depo-Provera) and one implant (Nexplanon), there are many different brands of oral contraceptive pills. I just pulled the label for one of them to get a list of side effects in Figure 8, so the incidence percentages shouldn’t be overinterpreted. “Menorrhagia” refers to abnormally heavy or prolonged menstrual bleeding, and “dysmenorrhea” refers to painful menstruation.

Generally, the side effects of hormonal contraception include absence of menstruation, excessive bleeding during menstruation, headache, abdominal pain and cramping, nausea, weight gain, acne, dizziness, and loss of sexual desire.

Conclusion

Data from the National Survey of Family Growth reveal that contraceptive discontinuation due to dissatisfaction is widespread, particularly for hormonal methods.

Discontinuation rates fall into three tiers: withdrawal and condoms cluster around 5–13%, hormonal pills, implants, and IUDs hover between 25–35%, and high-discontinuation methods like Depo-Provera, the patch, and the ring reach 40–50%.

The reasons for discontinuation differ dramatically by method type. Women who stop using condoms cite effects on sexual pleasure; women who stop using hormonal methods overwhelmingly cite side effects.

This is the second in a series of empirical analyses that scrutinize the culture that emerged from the sexual revolution. The previous was “Intercourse Is Remarkably Bad at Making Women Orgasm.” Upcoming analyses will investigate the increase in the incidence of unintended pregnancies after the sexual revolution and estimate the rates at which contraceptive methods fail in real life.

Some would dismiss—without reasoning or argument—the criticism contained in this series of articles as “right-wing.” This is itself a symptom of cultural hegemony.

The Rhetoric of Progress

One way to manipulate common sense is to portray certain viewpoints as inevitable, when in reality they become popular because of hegemonic control over culture. Thus, a cultural hegemony will portray its preferred viewpoints as “progress” and appeal to the masses’ desire to be on the “right side of history.”

In reality, no change in public opinion is inevitable. Everybody thinks that their vision for society is “better” than the current state of society. If public opinion changed to match your views, this would be “better” to you and so constitute “progress.” A hegemony manipulates the rest of us into thinking that the hegemony’s viewpoints have a monopoly on “progress.”

Granted, a cultural hegemony will tend to be anti-traditionalist and anti-religious. Traditional values and religion are refuges of the old common sense. A cultural hegemony successfully achieves submission by replacing an older common sense with a new one that conforms to its desires. Therefore, a cultural hegemony views traditional cultural institutions as enemies that must be overcome.

Thus, the cultural hegemony of the 21st century in the United States is sometimes described as “progressive” or “left-wing,” and counter-hegemonic criticism is sometimes described as “conservative” or “right-wing.” However, this is itself a tactic used by the cultural hegemony.

The term left-wing emerged from the French Revolution to refer to those sitting on the left side of the National Assembly. It does not describe a distinct political philosophy. However, certain historical tendencies are associated with the term. It is used to describe concern about economic inequality, a desire for redistribution of wealth from the rich to the poor, workers’ movements, and a class-based analysis of society.

None of this is aligned with cultural hegemony. In fact, this original definition of “left wing” and our definition of cultural hegemony are directly opposed. The left wing is historically aligned with working-class interests against capitalist elites; however, cultural hegemony is staffed by the professional-managerial class (PMC), highly credentialed, and aligned with wealth—either directly from wealthy capitalists or from activist groups funded by wealthy capitalists.

Why, then, do the hegemonized claim the title of “left-wing?” We live in a stupidly partisan time, and invoking partisanship is a technique to shut down thinking and debate, because many of our contemporaries appear to have an us-versus-them mentality as an overriding motive.

The PMC of the United States consolidated into the Democratic Party in the 21st century (but not the 20th). The Democratic Party is considered the “left-wing” party of the United States. By labeling anything counter-hegemonic as “right-wing” or “conservative,” the hegemony leverages this base partisan mentality to dismiss a viewpoint without serious consideration simply because it comes from “the other side.”1

Why Cultural Hegemony Theory Matters

This brings us to why our theory of cultural hegemony is useful. My hypothesis is that we are living through a period of peak cultural hegemony. The hypothesis explains several things:

1. the current political realignment happening in American society, and

2. practical implications for dissidents.

Political Realignment

The Democratic Party of the United States was dominant in the 20th century. With brief interruptions, it controlled the U.S. Congress from 1933 to 1994, including a remarkable 40 consecutive years with a majority in the House of Representatives.

This party had a broad coalition centered on working-class voters. This broad coalition included culturally “conservative,” “moderate,” and “liberal”2 wings, brought together by support for “left-wing” social democracy described earlier, such as the New Deal and Great Society platforms.

In the 21st century, the Democratic Party has a narrower coalition centered on PMC voters. Its “conservative” and “moderate” wings are basically nonexistent. When it wins a majority in Congress, it does so narrowly and for short periods of time. It is the dominant political alignment among the wealthiest Americans. It has, with one brief respite, been losing working-class voters throughout the 21st century.

What happened? The Democratic Party’s constituency got hegemonized.

The Republican Party, on the other hand, has been affected more indirectly.3

Trust in mass media has declined dramatically since the 1970s.

And this decline has particularly affected Republican-leaning voters.

A similar phenomenon has occurred regarding trust in academia: a general decline that is particularly pronounced among Republican-leaning voters.

The decline in trust in the organs of cultural hegemony is, in large part, a logical reaction to their complicity in manipulating common sense.

While Americans—and Republican-leaning voters in particular—might be justified in their distrust of mass media and academia, this does leave a void in information dissemination. The organs of cultural hegemony are also the vehicles that, in a more honest context, would spread knowledge. Into this vacuum of knowledge-spreading institutions, therefore, comes some amount of conspiracy theories, etc.

This creates an environment ripe for capturing by a demagogue, which is precisely what Donald Trump has done with the Republican Party in the past decade. This regime often employs the rhetoric of counter-hegemonic populism, which appeals to many in the 21st-century Republican constituency, but only occasionally acts as a counter-hegemonic force.

Living as a Dissident

In addition to being a tool useful for analyzing the demographics of political realignment, the theory of cultural hegemony serves as a warning for those of us who are dissidents, i.e., those who disagree with the hegemony’s orthodoxy on one or more issues.

All dissidents deal with cultural hegemony, as it operates in part through mass media. However, this is particularly acute for those dissidents who are themselves part of the PMC or for those dissidents who aspire to a life of the mind.

For PMC dissidents, you will constantly encounter people in your life—your coworkers, your neighbors—who are hostile to your views, prejudicially assume you agree with their opinions, do not understand your views, have false and defamatory beliefs about your reasons for having your views, expect their opinions to be enforced in the workplace, and are either unwilling or unable even to have a conversation with you about your disagreement.

This is because the hegemonized adopted their opinions as common sense, in the Gramscian sense. Their opinions are the axiomatic premises—not the conclusions—of their thinking. These opinions were formed by a profound incentive structure prepared for them long before they started their journey up the ladder of college-educated strivers.

For dissidents who aspire to a life of the mind, you are doubly at a disadvantage. We live in an era of elite overproduction, when it is extremely unlikely that any one of us can secure a job in intellectual work.

On top of this, you must contend with a modified version of O’Sullivan’s First Law: any institution not overtly “conservative” or “right-wing” will gradually become more and more hegemonized.

Therefore, if you disagree with self-styled “progressive” opinion on any culture war issue, the faster you accept that you must seek out alternative spaces, the better off you are.

This is not something a young person wants to hear. If elite overproduction has already made it hard to get an intellectual job in academia or “mainstream” media, then trying to target the even smaller number of jobs in “conservative” or “heterodox” spaces looks even worse.

Yet this counsel of realism should not become counsel of despair. Counter-hegemonic spaces exist precisely because others before you made the same calculation and built institutions outside the hegemonized mainstream.

Conclusion

Cultural hegemony in the 21st century explains why the Democratic Party abandoned its working-class coalition for a narrower PMC base, why trust in major institutions has collapsed—particularly among Republican-leaning voters—and why political discourse has become so difficult, especially among the educated who should be capable of reasoned debate.

For dissidents, understanding this dynamic matters less for changing the hegemony than for navigating it. Institutions will not become more open to heterodox views. The incentive structures that produce submission remain intact.

The theory of cultural hegemony is, ultimately, a tool for seeing clearly. It explains why certain views backed by wealth and status dominate society, and why the political landscape has shifted in ways that confuse those using 20th-century analytical frameworks.

The decline in trust in hegemonized institutions—mass media, academia, credentialed expertise—suggests we may be witnessing not the permanent triumph of cultural hegemony, but rather its peak.

Putting Theory to Work

That is enough theory. What empirical analyses does this lead to?

• PMC Consolidation — the demographic shift in the Democratic Party from a broad coalition to a narrow, PMC-dominated constituency

• “The Big Lurch” — the dramatic change in opinion in the Democratic Party constituency on culture war issues in the 21st century from the 20th century

• The Packaging of Opinions — the increase in the probability of having any one opinion promoted by the cultural hegemony, given that you hold any other opinion promoted by the cultural hegemony

The first two analyses are straightforward trend analyses that can be done using data sources such as the General Social Survey and the American National Election Studies.

The third analysis would involve the use of a statistical model. However, the modeling is not the challenging part. The hard part is finding a public use data set that asks each respondent about enough culture war issues and has been done so repeatedly over time. This may not exist, so I will have to see what I can come up with instead.

As we briefly mentioned in the previous article, cultural hegemony arises from wealth, social prestige, and institutional control.

Activist Groups, Mass Media, and Education

The main institutions that the 21st-century cultural hegemony controls are activist groups, mass media, and educational institutions.1 These types of organizations are heavily dependent on wealthy donors or owners.

Billionaire philanthropists typically finance prominent activist groups. John D. Rockefeller III—one of the heirs to the Rockefeller oil fortune—funded the movement to legalize abortion-on-demand by way of the Population Council and the Association for the Study of Abortion.

Mass media corporations are often privately owned by wealthy capitalists: for example, Arthur Ochs Sulzberger Jr. personally owns and chairs The New York Times, while Jeff Bezos, one of the wealthiest individuals in the world, personally owns The Washington Post.

Educational institutions, particularly elite ones, are responsive to the direction of their largest donors. This is not hidden, since large enough donations to academia come with the status symbol of naming rights.

For instance, the Weill Cornell Medical College is named after Citigroup chairman Sanford Weill and his wife, who donated over $600 million to Cornell and directly shaped curriculum reforms, research priorities, and hospital expansions.

On the other coast, the Knight-Hennessy Scholars program at Stanford University—created with a $400 million gift from Nike co-founder Phil Knight—established a new admissions track and campus facilities explicitly tailored to his vision of cultivating “global leaders.”

Credentialism and the Professional-Managerial Class (PMC)

While wealthy capitalists have their names emblazoned on these institutions, the institutions themselves are staffed by members of the professional-managerial class (PMC).2

The PMC comprises individuals who earn a living not through physical labor, but through abstract, symbolic skills. A rough way to define the PMC is to include all jobs for which a four-year college degree is required for entry.

This is a broad definition. The CEO of a billion-dollar corporation or the president of an Ivy League university wields far more social capital than an accountant or a computer programmer. Still, they are all part of the PMC.3

The PMC is usually contrasted with two other classes.

The capitalist class comprises less than 1% of the United States population and includes individuals who own companies or have inherited wealth from such ownership, thereby becoming independently wealthy.

The working class comprises individuals who earn a living through physical labor. The converse of the above definition of the PMC is that the working class includes all jobs that do not require a four-year college degree as a prerequisite.

One of the defining traits of our society is the explosive growth of the PMC. According to the Education Data Initiative, only 2.7% of adults over 25 years of age even had a college degree in 1910. According to the Bureau of Labor Statistics' estimates for 2021-22, 41.6% of workers aged 25 or older had a bachelor’s, master’s, doctoral, or professional degree.

Thus, the PMC has experienced unprecedented expansion relative to the population as a whole, though it remains a minority. On the other hand, the working class has shrunk proportionally, but remains the majority of the country.

While one might argue that modern professions genuinely require more education, anyone with a college degree knows how little their coursework prepared them for their actual job.

The main driver of PMC growth is credentialism, the gatekeeping of roles in society by requiring a college degree.

Americans have strong incentives to attain college degrees. According to the U.S. Bureau of Labor Statistics’ The Economics Daily of April 29, 2025, high school graduates with no college education had median weekly earnings of $953. Those holding at least a bachelor’s degree had median weekly earnings of $1,754, almost twice as much.

In addition to the economic incentive, nearly all positions of leadership or cultural influence in modern society require a college degree to occupy them. This was not always the case. For instance, journalism was once viewed as a “blue-collar” job for which a college degree was not necessary; however, it is now a heavily credentialed profession.

Representative data are only available from 1971 on, when the American Journalist surveys began. However, even as late as 1971, only 58.2% of American journalists had college degrees. Furthermore, there was an age cohort effect: only 45.5% of journalists over 55 years old had college degrees, compared with 64% of journalists aged 25-35 years old. (Weaver et al., 2009, pp. 36-38)

Now, according to the most recent survey from 2022, a whopping 96.4% of American journalists have college degrees.

The rise of credentialism is closely tied to that of cultural hegemony. The culture of the university—particularly elite universities—and the actual culture promoted by today’s hegemony are virtually the same.

Thus, incentivizing people to attend university ensures that more people—including most affluent and influential members of society—will be exposed to hegemonic culture for four years or more. During these years, young PMC aspirants learn to emulate certain viewpoints and signal particular virtues to advance their status and careers.

Historically, wealthy members of the capitalist class have financed cultural hegemony. This is still often the case, but due to the rise of managerialism, wealth can sometimes separate from capitalists, allowing members of the PMC to direct a hegemony.

While John D. Rockefeller III personally directed the funds he inherited toward population control and abortion advocacy, Henry Ford II was infamously so distraught over the Ford Foundation’s use of his father’s money that he disavowed himself from the organization. In the case of the Ford Foundation, high-status members of the PMC were the hegemons instead of wealthy capitalists.

Deplatforming

Credentialism, in addition to inculcating conformity with cultural hegemony, also provides the hegemony with a robust selection mechanism to control what opinions the public encounters.

Processing people through the university system—where taking a “wrong” viewpoint becomes conspicuous—allows selection processes to admit, hire, and promote only those who sufficiently conform to the hegemony’s orthodoxies.

An aspiring journalist who deviates from the hegemonic viewpoint on race, gender, immigration, abortion, foreign policy, firearms policy, etc., will be passed over for “mainstream” opportunities and left to try to find work in alternative or “right-wing” media.

Aspiring scholars who apply to graduate school can be weeded out by their political viewpoints through perusal of their mandatory DEI statements. This requirement makes formal what was done informally in past generations.

A similar process applies to professional organizations in law, medicine, public health, and social work. Thus, official-sounding pronouncements from professional organizations tout hegemonic viewpoints.

The result is that viewpoints that differ from cultural hegemony are deplatformed from “mainstream” society.

Much has been made of very public “cancellations,” in which a celebrity says something opposed by hegemonic forces and loses a job and/or a place in public life. These cancellations are relatively rare.4

However, what is more profound than cancel culture is the daily advancement of certain people in society over others, again and again, until dissenting voices are deplatformed.

Language Control

Today’s cultural hegemony is very concerned with controlling language. This is partly an extension of the professional instincts of the hegemony's foot soldiers. Members of the PMC often have vocations concerned with symbolic capital—to borrow a phrase from —so it follows they would be concerned about what symbols are allowed and not allowed.

However, controlling language is also a deliberate way to control how people think. This tactic is used prolifically by today’s cultural hegemony.

For instance, the phrase “affirmative action” was originally coined in President John F. Kennedy’s Executive Order 10925, Part III. “Obligations of Government Contractors and Subcontractors”:

The contractor will not discriminate against any employee or applicant for employment because of race, creed, color, or national origin. The contractor will take affirmative action to ensure that applicants are employed, and that employees are treated during employment, without regard to their race, creed, color, or national origin.

Thus, it originally meant affirming that there was no racial discrimination occurring. Over the years, hegemonic forces have reinterpreted the phrase to mean the exact opposite: the presence of race-based discrimination (favoring certain “underrepresented” groups). This inversion in meaning can confuse public opinion.

For instance, a Pew Research Center poll of Asian Americans (2023) found that 53% of respondents who have heard of affirmative action said it is a good thing. The same poll also found that only 21% of respondents said that race or ethnicity should be considered in college admissions decisions.

Thus, about 32% of respondents believed “affirmative action” to be a good thing while simultaneously opposing race-based college admissions. This is more than the 21% who actually believed that college admissions should be based on race or ethnicity.

A similar phenomenon can be found in the way that mass media has slowly purged the word “illegal” from coverage of illegal immigration in the United States.

The United States has been unusually welcoming of immigrants over its history, and contrary to popular perception, it remains so. A Pew poll of U.S. adults (2024) found that 46% of Americans think legal immigration levels should stay at present levels, and 30% say it should be increased. Only 22% of U.S. adults believe it should be decreased.

A separate Pew poll (2024) found that 79% of Americans are concerned about the amount of illegal immigration to the United States, with 38% "extremely," 18% "very," and 23% "somewhat" concerned.

In summary, 86% of Americans approve of legal immigration at or above current levels, and 79% of Americans are concerned about illegal immigration. Americans' feelings about immigration depend heavily on whether it is done legally or not.

During this time, mass media have shifted from reporting on “illegal immigrants” to reporting headlines about “immigrants” without any qualifier whatsoever. This obscures from the headlines the very piece of information that public opinion cares about, i.e., whether immigration is done legally or illegally.

Looking Ahead

We have now examined both the definition of cultural hegemony and the methods through which it operates: institutional control, credentialism, deplatforming, and manipulation of language. But understanding these mechanisms raises further questions.

If cultural hegemony operates through these methods, what are its effects on our political landscape? The professional-managerial class people who staff these institutions are not randomly distributed across our political parties. How has the consolidation of the PMC into one party, and the resulting distrust of hegemonized institutions among the other party’s voters, reshaped American politics?

And more personally: if you find yourself in disagreement with the hegemony’s orthodoxy, what does that mean for your life? How do you navigate a world where the institutions that credential expertise, disseminate information, and provide platforms for public discourse are all aligned against your position?

Part 3 will explore both the analytical and practical implications of living under cultural hegemony—how it explains our current political realignment, and what it means for those of us who dissent from it.

Concepts

Sometimes, the ruling classes enforce their will on others through the coercive power of the state. However, the ruling classes can also dominate culture through soft power, leveraging wealth, prestige, and social position. This is cultural hegemony.

Common Sense

The set of values and beliefs that the people absorb uncritically in a given society is that society’s common sense. It is driven not by active inquiry, but by passive conformity to what people perceive as the settled, default, logical, neutral, or obvious viewpoint.

Common sense, when it contradicts the views of the ruling classes, can be a barrier to their domination. It is also the primary means by which the ruling classes accomplish cultural hegemony. The ruling classes achieve cultural hegemony by manipulating what people perceive to be common sense.

Of course, you think this. How could you not? This is sane and sensible.

You don’t want to be “ideological,” “crazy,” or “extremist.”

This is the language of common sense.

A hegemony subverts common sense only when common sense disagrees with it. Thus, hegemonic tactics are typically employed only when the public disagrees with a viewpoint. Consequently, when common sense language is invoked, it is usually the hegemony that holds the “extreme” or minority viewpoint.1

Submission

Once a cultural hegemony has succeeded in rewriting common sense, then submission2 has occurred. Submission is popular support made to look “spontaneous,” but in reality, it has been engineered by cultural hegemony.3 It is the result of controlling the narrative and subverting popular opinion to the point that popular opinion itself changes, becoming what the cultural hegemony wants.

Cultural Hegemony without Submission

We can see the phenomenon of cultural hegemony empirically.

Cultural hegemony is most apparent when submission has not occurred. In this case, there are typically bipartisan4 majorities that oppose a viewpoint; however, elite culture—mass media, academia, and activist organizations—frames the viewpoint as the only acceptable one.

In other words, when cultural hegemony is being exercised without the public’s submission, the commanding heights of culture act as though their preferred viewpoint is common sense. At the same time, the public has a consensus that disagrees with the viewpoint.

Cultural Hegemony with Submission

Cultural hegemony is less apparent when submission has occurred. However, this is only because the exact same conditions of cultural hegemony without submission had happened in the past.

Even after public opinion is in submission to the cultural hegemony, it leaves its distinctive trace. There was a point in recent history when elite opinion-makers pretended that their preferred viewpoint was the only acceptable one, even though a majority opposed it.

Packaging of Viewpoints

When people conform most or all of their viewpoints to a cultural hegemony, they are hegemonized. The existence of hegemonized people leads to a correlation of viewpoints together into a package.

This correlation arises when hegemonized people share the same viewpoints on separate issues because they were influenced by the same hegemony. The farther along a hegemony is, the more hegemonized people there are, and the more unrelated viewpoints are packaged together.

Examples

Cultural Hegemony without Submission

I have already discussed an example of hegemony without submission in Born Curious. In “Gender Critics Shouldn’t Take a Victory Lap Just Yet,” I commented on a New York Times / Ipsos poll published in January 2025 that found that majorities of both Democratic-leaning and Republican-leaning voters opposed transgender athletes competing in women’s sports and puberty blockers or cross-sex hormones for gender dysphoric minors.

This is remarkable because people had been getting deplatformed or ostracized over the previous several years for expressing viewpoints that, it turns out, most Americans shared.

For instance, on February 17, 2019, the Sunday Times published an op-ed by tennis champion Martina Navratilova arguing that transgender athletes should not compete in women’s sports. Just two days later, Athlete Ally published a statement denouncing her comments as “transphobic,” removed Navratilova from its Advisory Board, and dropped her as an Athlete Ally Ambassador.

Another example occurred the following year, when Abigail Shrier’s book, Irreversible Damage, was set to be published on June 30, 2020. The book argued that puberty blockers and cross-sex hormones were overprescribed and often led to permanent health and fertility consequences. Campaigns launched to have the book removed from major retailers, including Amazon and Target.

Amazon temporarily suspended a paid advertising campaign for Irreversible Damage one week before its release, but went on to sell the book and continues to sell it to this day. Target, on the other hand, removed the book from its online catalog on November 16, 2020, after complaints on social media. This caused a backlash to the cancellation, prompting Target to re-list the book on November 17, 2020. However, on February 24, 2021, Abigail Shrier reported that Target had once again removed her book from its website, quietly and without explanation.

What is remarkable about these cases is the disconnect between hegemonic elites and public opinion. The authors expressed viewpoints held by an overwhelming majority of the public.5 Still, these viewpoints were treated as beyond the pale, equivalent to racism, and in need of deplatforming and censorship.

Cultural Hegemony with Submission

I have also already discussed an example of hegemony with submission in my analysis of public opinion about the legalization of abortion-on-demand (in which abortion is legal for any reason).

A majority of Americans were opposed to legalized abortion-on-demand from 1965 to 2018, according to the General Social Survey (GSS) and its preceding “amalgam” survey.6

In the year 2000, majorities of both Democratic-leaning voters and Republican-leaning voters opposed legal abortion-on-demand.

However, also in the year 2000, the Associated Press published the 35th edition of its Stylebook, which introduced,

Use anti-abortion instead of pro-life and abortion rights instead of pro-abortion or pro-choice.

At a time when a bipartisan majority of Americans opposed legal abortion-on-demand, the AP Stylebook recommended oppositional framing (“anti-abortion”) for this majority opinion and reverential framing (“abortion rights”) in accordance with elite, minority opinion.

Media bias extends far beyond the AP Stylebook. The Stylebook just wrote down the bias down in so many words.

Since 2000, the AP Stylebook has continued to pile on. It strongly prefers “fetus” over “unborn child” or “unborn baby” in most contexts, and it recommends “anti-abortion centers” to “pregnancy resource centers.”

Indeed, the AP Stylebook now discourages the very terms “abortion-on-demand” and “elective abortion.” But as the GSS and other data show, public opinion on abortion turns heavily on whether an abortion is elective, versus done because of risk of death or serious bodily injury, because of rape, or because of severe fetal deformity.

Packaging of Viewpoints

The phenomenon of viewpoint packaging is something I alluded to in my article that was inspired by lawn signs of the “In This House” or “In Our America” variety. In it, I noted that the probability that someone would have uniformly partisan-conforming views on eight separate, unrelated issues is infinitesimal. Nonetheless, these lawn signs exist, proudly proclaiming the home's occupants’ conformity.

This is a phenomenon that the writer labeled the “Omnicause” in her article “Why is Greta wearing a Keffiyeh?”

…all contemporary radical causes seem somehow to have been absorbed into one. A protean animating energy seems to ingest every progressive issue it encounters, to create a kind of ever-spreading, all-encompassing omnicause.

It is highly unlikely issues as unrelated as anthropogenic climate change and the conflict in the Middle East are “profoundly interrelated” as some hegemonized people would assert. Instead, the issues are united only by being championed by the same social forces. The hegemonized simply reflect back all the causes promoted by the hegemony that dominates them.

Looking Ahead

This is the first article in a multi-part series.

We now have the conceptual tools to identify cultural hegemony empirically. When elite opinion-makers portray their viewpoint as common sense while a majority of the public holds the opposite view, cultural hegemony is being exercised.

If public opinion eventually matches the elite opinion-makers, then submission has occurred. In this case, we need to look for evidence of hegemony in the past.

When submission has not happened, hegemony is more evident due to the ongoing disconnect between majority opinion and elite opinion-makers.

Finally, the increase in individuals holding unrelated viewpoints packaged together with remarkable consistency is evidence of hegemony in action.

Understanding what cultural hegemony looks like is only the beginning. The next question is how it works. How do wealthy elites and their professional allies manipulate culture so effectively? What are the mechanisms by which a minority viewpoint becomes institutionalized as common sense despite majority opposition?

Part 2 will examine the methods of cultural hegemony: the role of wealth in financing activist organizations, the credentialing systems that select for conformity, the deplatforming mechanisms that exclude dissenting voices from mainstream discourse, and the deliberate control of language that shapes how people think about contested issues.

Understanding these methods reveals cultural hegemony not as a vague theory, but as a concrete set of institutional practices with observable patterns.

They only lived in Old Town Alexandria for 2 years. During that time, between the two of them, they called 911, he guessed, about 50 times. Most of the time, it was for property crimes—people breaking into cars, etc. But on 4 or 5 occasions, it was because of violence with firearms.

When his wife first called 911 in response to gunshots, the 911 operator dissuaded her. “You are probably just hearing fireworks.” At first, his wife believed the 911 operators. But on a subsequent incident, his wife found spent shell casings on the street the next day, and she realized the 911 operator was gaslighting her.

Three Ways Crime Statistics Fall Short

The incidence of actual crime is much greater than the incidence reported in crime statistics. However, crime statistics have improved remarkably in recent decades.

Crime statistics undercount crime for three major reasons:

1. Most crimes are not reported to the authorities.

2. The authorities, when crimes are reported to them, sometimes do not process the crimes.

3. The authorities, when processing some crimes, downgrade the severity of the crimes.

The Stark Gap Between Real and Reported Crime

We can measure the extent to which 1 and 2 occur. For crimes in the United States at the national level, we have two main instruments:

• The Uniform Crime Reporting (UCR) Program of the Federal Bureau of Investigation (FBI)

• The National Crime Victimization Survey (NCVS) of the Bureau of Justice Statistics (BJS)

The UCR collects statistics from the approximately 18,000 police agencies in the United States. Participation in the program is voluntary, with approximately 95% of police agencies participating.

Each year, the BJS selects a representative sample of approximately 150,000 households in the United States and administers the NCVS to all members of the chosen households aged 12 or older. The NCVS asks each respondent about criminal victimizations experienced in the preceding 6 months.1

Figure 1 shows the estimated number of violent crimes in the United States based on UCR and NCVS data. The UCR data were taken from the FBI’s Crime Data Explorer (CDE), and the NCVS data were taken from the NCVS Dashboard (N-DASH).2

It is well known that crime peaked in the 1990s. You can sort of see this in the UCR data. According to the UCR, there were 1,376,611 violent crimes in 1988, and this increased year by year until 1992, when it peaked at 1,867,287.

However, at the same time that the UCR reported 1,822,382 violent crimes in 1993, the NCVS estimated about 6.1 million violent crimes. Indeed, next to the plot of the NCVS data, the UCR plot almost looks flat.

What is going on here?

Most Victims Never Call the Police

The NCVS asks its respondents whether they reported each victimization to the police. Most are not reported. This underreporting is even worse for property crimes than for violent crimes, but even with violent crimes, only between 40% and half of the victims report the crime to the authorities.

Figure 2 plots the same UCR data, but against NCVS estimates of reported violent victimizations reported to the authorities.

There is still a large gap in 1993 between 3.1 million reported violent victimizations from the NCVS and the UCR’s 1,822,382 violent crimes. It seems like the authorities neglected 1.3 million violent crimes reported to them.3 Still, these 1.3 million crimes that the authorities neglected are less than the approximately 3 million crimes not reported to the authorities in the first place.

Closing the Gap: Police Reporting Gets Better

Something interesting happened in the year 2000. The gap between victimizations reported to the police from the NCVS and police agencies’ reporting of crimes to the UCR went away. The difference between the two lies within the sampling error of the NCVS, as indicated by the 95% confidence interval band plotted around the NCVS estimates.

The United States, as a nation, has largely solved the problem of authorities underreporting crime in their statistics in the 21st century. Of course, this is only the case when crime is aggregated at the national level; any specific locality might still be underrepresenting crime occurring within its jurisdiction.

The problem of victims not reporting crimes to the authorities remains, however. In the 21st century, this accounts for nearly all of the crime statistics underreporting, at least at the national level.

Why It Got Better

The fact that UCR crime statistics plot as a nearly flat horizontal line hints at one reason for the authorities’ underreporting of crime in the 20th century. The UCR likely measures the amount of throughput police agencies have for processing crime. The high crime rate at the end of the 20th century may have saturated the system.

During the 1990s, in response to the large discrepancy between real and reported crime, there were several political initiatives to improve crime reporting. For instance, the New York Police Department launched CompStat in 1994, which led to greater transparency and accountability and was imitated by many police agencies across the country.

The convergence of the two lines in Figure 2 thus represents the confluence of two trends:

1. Actual crime decreased, reaching more manageable levels.

2. Police reporting of crime improved.

A Classification Problem

You may notice that the NCVS estimates of violent victimizations reported to police are actually lower than the UCR crime statistics in some years, e.g., 2016, 2019, and 2020. How can this be?

This is almost certainly a classification problem.

Figures 3 and 4 plot robberies and sexual assaults reported to the police from the NCVS against robberies and rapes reported in the UCR, respectively. As can be seen, the two estimates overlap in the 21st century.

The subtitle to Figure 2 is “excluding simple assault.” This is to make UCR and NCVS comparable, since the UCR does not include simple assault counts in its count of violent crimes.

As we can see in Figure 5, aggravated assaults—unlike robberies or rapes—do appear to be subject to the phenomenon in which the victimizations reported to the police from the NCVS are lower than the UCR counts.

Many things can escalate a simple assault to an aggravated assault, e.g., use of a weapon, severity of injury, the assault occurring during the commission of another crime, whether the victim is a child, elderly, or a police officer, etc. Police know these legal distinctions better than most people do, so police are more likely to classify an assault as “aggravated” (in the UCR) than a victim describing the same incident to a survey interviewer (in the NCVS).

The D.C. Crime Statistics Controversy

Classification of crime is an issue not just for the comparison of violent crime as reported by the NCVS versus the UCR, but for an ongoing controversy about the accuracy of crime statistics in Washington, D.C., the national capital of the United States.

As previously mentioned, the agreement between the UCR and the NCVS in the 21st century does not imply that all jurisdictions in the United States are now accurately reporting all the violent crimes they are aware of. The NCVS estimates of violent crime can have a sampling error of about half a million crimes, which means there is wiggle room for individual jurisdictions to go unnoticed in a large country with 18,000 police agencies.

On July 18 of this year, local news broke a story that Michael Pulliam, a district commander for the Metropolitan Police Department (MPD) in Washington, D.C., had been suspended earlier in May. MPD leadership alleges that Pulliam manipulated crime statistics to make crime appear lower in his district.

This brought to public attention the tension that had been ongoing for years between MPD patrol officers and MPD and D.C. management.

The local Fraternal Order of Police alleged that this type of crime minimization—in which a supervisor arrives after a crime has occurred and downgrades the initial classification of the crime made by responding officers—is rampant throughout the MPD. It further alleges that Michael Pulliam is being made a scapegoat for the MPD’s misconduct, noting that he filed a complaint against an assistant chief just a week before he was suspended.

This has come at a bad time for Washington, D.C., as crime in the nation’s capital is under scrutiny due to the Trump Administration’s deployment of federal police and National Guard troops in Washington, D.C.

In August, the U.S. Congress and the Justice Department both launched investigations into the allegations of crime statistic minimization by MPD leadership.

A Washington Post article was published this past Sunday that summarizes the current state of allegations and investigations. Approximately three dozen MPD officers are voluntarily cooperating with the investigations, providing information about how supervisors routinely reclassify crimes.

The article noted that these allegations are not particularly new. When Lisa Burton, a former MPD officer, sued D.C. for alleged discrimination, the court case brought to the public the allegations that high-ranking officials were minimizing crime statistics to boost their careers.

The Scale of the Drop in Dispute

Despite these investigations, the nature of the dispute is nuanced.

No one seems adamant that crime in Washington, D.C., is increasing. Even Greggory Pemberton, the Chairman of the Fraternal Order of Police, has said, “Crime might have ticked down from 2023. But I certainly don’t think it’s down in the amounts that the department is claiming it is.”

This would be consistent with national trends. Both the UCR and the NCVS show a spike in violent crime in 2022 after the pandemic, followed by a gradual decrease in violent crime each year after that.

What makes Washington, D.C. particularly conspicuous is that authorities are claiming that violent crime is plummeting, with consecutive year-over-year decreases of 34% and 28%.

An Unfunded Answer

One potential improvement to this situation would be victimization survey data for Washington, D.C. (The NCVS is designed to provide estimates for the United States as a whole, rather than specific jurisdictions.) It turns out that Washington, D.C., passed a bill entitled the “Crime Victimization Survey Amendment Act of 2023,” which was ratified into law on April 4, 2024.

Unfortunately, the bill remains unfunded, and no victimization survey has taken place. This leaves Washington, D.C., without an independent mechanism to verify its crime statistics even as federal investigators probe whether those statistics have been manipulated.

Conclusion

Crime statistics in the United States have improved substantially over the past three decades. By 2000, the gap between crimes reported to police and crimes recorded in official statistics had essentially closed at the national level—a genuine achievement in government transparency and accountability.

However, we should remain skeptical of crime statistics. Most crimes are never reported to police in the first place. Official statistics can only reflect what victims choose to report and should always be analyzed in conjunction with victimization surveys.

Moreover, the incentives for authorities to minimize crime in their jurisdictions remain. The allegations currently under investigation in Washington, D.C.—where officers claim supervisors routinely downgrade crime classifications—remind us that even improved systems remain vulnerable to manipulation.

Crime statistics are better than ever, but that bar was remarkably low. We have made real progress, yet we remain far from having a complete picture of crime in America.

Recently, I’ve seen a new type of meme circulating in online comment sections: claims that “right-wingers” are disproportionately guilty of sex crimes against minors.

The most elaborate example I came across showed 42 mugshots, labeled as “right-wingers” arrested for sex crimes against children in September 2025 alone. The images were pulled from news reports, social media, and arrest records.

But what does this actually prove? Not that “right-wing” people are uniquely dangerous. It only demonstrates the statistical illiteracy of whoever compiled the list.

Why 42 Seems Like a Lot

Why does 42 seem striking? Because of the base rate fallacy—our minds are not adapted to think in terms of thousands of crimes per month. Forty-two faces on a meme look overwhelming until you realize they represent a tiny fraction of the actual base rate.

The Actual Numbers

The National Crime Victimization Survey estimates that in 2024 there were about 92,000 sexual assaults of victims aged 15–17 and about 29,000 of victims aged 12–14. On average, that means roughly 10,000 victimizations of minors every month.

The FBI’s Uniform Crime Reporting tells a similar story: between 9,800 and 11,000 reported rapes per month across all age groups.1

Defining “Right-Winger”

In competitive presidential elections, the Republican candidate typically wins 46–48% of the vote. By that measure, nearly half the country could be placed in the “right-wing” category. If there are 10,000 monthly sex crimes against minors, then about 5,000 of those offenders would be “right-wing” if there is no correlation whatsoever.

The “right-winger” label can be broadened further—for example, to include anyone who believes biological males should not compete in women’s sports, a view held by about 80% of Americans—and the expected number rises to 8,000 monthly cases.

Selection Bias, Not Sampling

Finally, the creation of the meme is an exercise in selection bias. Whoever made the meme went out looking specifically for “right-wing” offenders. Out of the many thousands who statistically exist, they spotlighted 42. That is not a representative sample—it’s cherry-picking.

Motivations for the Meme

Why produce such a meme at all? Likely for two reasons. One is to keep the Epstein–Trump association alive in the news cycle. The other is to counter criticism of transgender policies that allow male-bodied individuals into female spaces, by implying that “right-wingers” are just as dangerous.

However, this overlooks the point of the criticism it is attempting to counter. The point is not to count the number of transgender offenders. The point of the criticism is that a specific policy (gender self-identification) has a specific consequence (exploitation by sex offenders to access vulnerable women and girls). By contrast, no policy says, “If you identify as right-wing, you go into the girls’ locker room.”

Conclusion

Memes like this exemplify what has gone wrong with public discourse. Like calling someone a “fascist,” accusing someone of being a pedophile without evidence is the lowest form of argument. Extending that accusation to tens of millions of people is worse.

Even if such a correlation could be proven, it would have no practical value. If people in Group A have a 0.5% chance of being a pedophile and people in Group B have a 0.8% chance of being a pedophile, we don’t treat people in the latter group differently. Doing so would be prejudice.

In the end, these memes do not tell us anything meaningful about sex crime and partisan affiliations. It is the behavior of those who have no argument to make, no information to share, and nothing to say.

Further Reading

I have written in more detail about this very fallacy in my past article “The Fallacy of Listing Examples.”

While these documents detail the failures of Arlington County to protect its girls and women from a registered sex offender, the conduct of the authorities in Fairfax County—where I live—was even worse. Had Fairfax County acted decisively when Cox’s pattern of behavior first emerged, he would never have had the opportunity to offend in Arlington County.

This case reveals how ideologically-driven prosecution policies and “inclusive” gender self-identification policies can compromise public safety.

Basic Facts of the Case

Richard K. Cox is a Tier III registered sex offender. Tier III is the most severe and requires lifetime registration.

In June of 2024, Cox identified as a transgender woman and entered the women’s locker rooms at several Planet Fitness locations.

After the episodes in Planet Fitness gyms, Cox proceeded to loiter in the women’s locker rooms at various Fairfax County and Arlington County recreational centers and high schools — the latter when their facilities were open to the public.

His modus operandi for these excursions was to loiter in the women’s locker room with his penis exposed and a towel draped over his head and shoulders.

Many complaints were made by women and by concerned parents of girls to the operators of the various establishments and to the police. At least one witness alleged having seen Cox masturbating in the locker room.

These encounters continued until December 6, 2024, when Cox was booked in Arlington County. Since then, he has been in the custody of Arlington County as his criminal charges go through trial.

Fairfax vs. Arlington County

The first complaints were made about Cox’s 2024 behavior due to his loitering in Planet Fitness gyms in June—in Fairfax County. The first complaint about Cox’s visits to a rec center again occurred in Fairfax County, in response to his visit to Franconia Rec Center on August 24, 2024.

According to local news, the first complaints about Cox’s behavior in Arlington County did not occur until October 2024. Furthermore, complaints about Cox leering at women and girls in Fairfax County rec centers’ locker rooms continued through November 2024, until he was eventually administratively banned from Fairfax County rec centers.

Throughout all of 2024, Fairfax County never put Cox behind bars. It could have done so when Cox first began exploiting the self-identification policy to prey on women and girls. If Fairfax County had acted, it would have prevented Cox from going on to offend in Arlington County.

The behavior of Arlington County officials in this case is indeed despicable. Still, at least Arlington County had the decency to eventually lock up Cox and prevent him from continuing to victimize women and girls. Fairfax County never did.

Context: The Commonwealth’s Attorney

This is particularly sensitive because Fairfax County experienced an abrupt change in law enforcement leadership in 2020.

The chief law enforcement officer of any county in Virginia is an elected official: the Commonwealth’s Attorney (CA). This is similar to the District Attorney in other states, but differs substantially from the systems in other countries. For instance, in the United Kingdom, the Crown Prosecutor serves a similar role, but is typically an appointed career civil servant rather than an elected official.

Fairfax County had over 50 years of continuity in the CA position. Bob Horan was CA from 1967 to 2007. His successor, Ray Morrogh, served from 2008 to 2019. Morrogh had been Horan’s deputy and largely continued Horan’s approach to law enforcement.

Horan and Morrogh were classic suburban prosecutors, characterized by professional, cautious, and tough-on-crime attitudes.

Then 2019 happened. Approximately $1,000,000 in out-of-county money was poured into the 2019 Democratic primary election for CA of Fairfax County from the “Justice & Public Safety” Political Action Committee (PAC), a George Soros-funded activist organization. (Fairfax County has never had a Republican serve as CA, so the primary election for the Democratic nominee for CA is typically the “real” election.) This was part of a campaign to install “progressive” prosecutors in districts across the United States.

By all accounts, Morrough, who was used to running for reelection with minimal opposition, was blindsided. He narrowly lost the primary to Steve Descano, who had received the lion's share (about $628,000) of the Justice & Public Safety PAC money. Morrogh would go on to endorse independent Jonathan Fahey in the general election instead of Descano, despite being ostensibly part of the same political party as Descano.

It seems that the major difference between having a “progressive” prosecutor such as Descano versus the previous prosecutors is that the role becomes politicized, and actual criminal prosecution decreases. Descano wrote an op-ed in The New York Times asserting he would not enforce any abortion restrictions in Fairfax County. Descano’s office dropped charges against a local brothel just down the street from me that had set up in a residential neighborhood across the street from a school bus stop.

After getting our “progressive” prosecutor, the official website of the Office of the Commonwealth’s Attorney (OCA) tells us, “The OCA is committed to promoting the long-term safety of the community while reforming Fairfax County’s justice system to make it more equitable.”

In its biography of Descano, the website states,

Since taking office, Steve [Descano] has undertaken a reform agenda aimed at mitigating racial and socioeconomic disparities and countering mass incarceration in Fairfax County’s Justice System — while keeping our communities safe. Under his leadership, the Office of the Commonwealth’s Attorney has stopped requesting cash bail, implemented a new sentencing policy to address mass incarceration, and restructured the office to produce accountability around reforms, among other initiatives.

Basically, the OCA is now awash in the jargon and rhetoric of “equity” and “mass incarceration.”

It should be noted that — even before the rise of “progressive” prosecutors — a minority of crimes were ever punished, and a tiny percentage of offenders were ever sent to prison for their crimes. But this is a topic worthy of its own article.

What are these reforms that the above blurb alludes to? From the OCA’s “Common Misconceptions” website,

However, felony charges carry serious, life-altering consequences. Someone convicted of felony charges will have future barriers to finding a job, getting an education, or securing housing. We don’t want that — because it can lead to them committing more crime in the future. Instead, our office charges some conduct as misdemeanors — not felonies — when appropriate, to ensure accountability without sending someone up the ladder of criminal activity.

Thus, one “reform” appears to be to charge people with misdemeanors instead of felonies.

The practical impact of this philosophical shift became clear when Cox’s case reached Descano’s office. Local news reports,

Court records show before Cox visited Planet Fitness in Fairfax County, he was charged with three felonies for sex offender registry violations. But Descano’s office agreed to drop two of the felony charges and reduce the third felony charge to a misdemeanor, according to court records.

Court records also show Cox’s sentence ended up being 90 days in jail when he originally faced up to 15 years behind bars.

Again, from the “Common Misconceptions" website,

Prisons teach individuals how to get better at crime, and oftentimes push them to commit more crime once released. To avoid that, we don’t always want to send low-level, first time, and nonviolent offenders to jail. Instead, in certain situations we can use diversion, treatment, and suspended sentences to reduce the likelihood of these individuals further victimizing their neighborhoods — getting better safety outcomes without contributing to mass incarceration.

This is the theory behind non-punishment. These theoretical approaches to criminal justice had real-world consequences in the Cox case.

Cox was a registered Tier III sex offender who was not even allowed to be within 100 feet of children, let alone loitering naked in the same room with girls and women while they changed their clothes. The Fairfax County OCA could have prevented this from occurring.

Cox was the opposite of a “low-level, first time, and nonviolent offender.” Cox has spent a good chunk of his life in and out of prison from 1988 to 2018, having been charged with crimes such as exposing his genitals to a child under 14 years old and possessing obscene materials with a minor.

As later came to light, according to the charges filed in July of 2025 against him, Cox is alleged to have sexually assaulted a minor in Fairfax County way back in 1988. This case came about only because Cox was made infamous by his 2024 actions, and the victim recognized him in media coverage.

He was and is a predator that needs to be sequestered from the rest of the community. The Fairfax OCA could have and should have protected its citizens from him, but instead fell for his ruse of transgender identification. Cox spoke their ideological language and exploited this to evade serious prosecution.

Unsurprisingly, Jason Miyares, the Republican Attorney General of Virginia, publicly admonished CA Descano for his handling of the case. What is a bit more telling is that Dave Marsden, State Senator from Fairfax County and a Democrat, also called out CA Descano for this matter.

The Police’s Hands Are Tied

A lot of people are understandably angry about the way the Cox case was handled, and some of this anger is directed toward the Fairfax County Police Department (FCPD).

However, it is important to remember that in the United States, police departments have a lot of limits on what they can and cannot do.

For instance, in the brothel case, the day after the OCA pleaded the brothel owner down to a single charge of operating a massage parlor without a license, the FCPD parked a police cruiser directly across the street from the brothel in question, eventually leading to the brothel’s closure.

Timothy Carney, who lives in the area and so took an interest in the case, reports that

The police wanted to shut the place down, but they told neighbors, “Our hands are tied,” presumably a reference to Descano’s tolerance of the place.

With regard to the Richard Cox case, FCPD Chief of Police Kevin Davis, in public statements, sounds frustrated by the situation. When asked about the Cox case by Nick Minock of WJLA news, Chief Davis replied,

Nick, I am going to go through this, and this is being recorded, so please take the time to go back and listen to my response to your very specific question about Richard Cox.

Richard Cox on June the 10th, 2024 was arrested by Fairfax County police officers. That case made its way into a courtroom was dismissed because the victim in that case did not show up. That’s the first encounter.

The next encounter we had with Richard Cox on November 3rd, 2024, and he was in a shower in a locker room taking a shower, and the policy Parks and Rec allowed him and other persons to — and I am reading this verbatim — “to use the locker of room of the gender that they identify with.” So a crime was not committed. So he was not arrested by Fairfax County police officers.

The next time we had contact with Richard Cox was on November 15th, 2024. He was also in the women’s locker room. He was allowed to be in there — didn’t commit a crime where we had probable cause to arrest him. So we did not arrest him.

The next encounter we had with him — same month — November 16th, 2024, he was banned for the first time from Fairfax County Parks and Rec facilities, so if Richard Cox returns — and I don’t know where he is right now, don’t know if he is still being held in Arlington — I am sure you are going to share that with me. If Richard Cox returns to any Fairfax County Park and Rec facility, we will arrest him for tresspass.

I will arrest him myself. Okay?

If the Fairfax County OCA had proceeded with the felony sex offender registry violations, none of this would have mattered, including whether the victim appeared to testify in the misdemeanor case against Cox in June. Later, as Cox’s behavior continued, the FCPD was frustrated in their efforts by the policies of Fairfax County Parks.

The FCPD has to maintain an active working relationship with the OCA to do their job, so it is unlikely that any police officer will throw the OCA under the bus publicly. Indeed, Chief Davis’s insistence that he only answer “specific questions” and his word-for-word quoting of policy are indicative of the kind of pressure on the public statements by police on this matter.

While police might not say much publicly, reporter Nick Minock also interviewed Bryan Hill, the Fairfax County Executive. (The County Executive is a position appointed by the Board of Supervisors.)

Reporter Minock asked,

A lot of people are wondering why Fairfax County police have not charged Richard Cox for things like indecent exposure and indecent liberties with children. Those are some of the charges he’s facing in Arlington County, but not in Fairfax County. Why is that?

County Executive Hill replied,

Unfortunately, it’s a situation we are all trying to resolve, obviously. That’s why you are here. When we asked the Commonwealth’s Attorney if there were criminal charges that we could press, they stated, no, I can’t press a charge. I can’t charge the person.

It is clear that the failures in the handling of the Richard Cox case were due to the policies of the OCA and Fairfax County Parks, not the FCPD.

The Problem with Self-Identification

In court Thursday morning, Cox repeatedly insisted he is a woman, and insisted the judge order the prosecutor to stop misgendering him and to use female pronouns.

The judge declined.

One thing that is remarkable about this case is that no one seems to be taking Richard Cox’s claims to transgender status seriously. This makes sense. Richard Cox is a very unsympathetic figure, a repeat sexual offender and pedophile. He has never, by any account, ever made any sincere attempt to transition, even socially. Indeed, as several victims have personally witnessed during his criminal acts, he is quite fond of his male anatomy.

This skepticism about Cox’s claims reveals a fundamental problem with self-identification policies. According to the logic of self-identification, someone merely says, “I am gender such-and-such,” and that person must be treated as being “really” gender such-and-such. If Richard Cox says he is a woman, then he is “really” a woman.

In short, the exploit that Richard Cox found is a feature, not a bug, of gender self-identification.

Gender self-identification is not a merely theoretical matter; it is the official policy of Fairfax County. Supervisor Pat Herrity testified to this before the Board of Supervisors,

On July 1, 2020, without Board input, review or approval, the Office of Human Rights and Equity Programs issued Memorandum Number 39-08 entitled “Employment Policies Related to Facility Access for Transgendered Workers” which states that, “all persons should have access to the facilities corresponding to their gender identity.”

The policy goes on further to state “No individual shall be asked to provide any medical or legal documentation of their gender identity in order to have access to gender-appropriate restrooms, sleeping quarters, locker rooms, or changing room facilities.”

No one seconded Supervisor Herrity’s motion to review this policy, so the motion failed. The Board seems content to treat the Cox case as a “one off” event.

County Executive Hill asserts,

So, I will say this to you again, we have had one situation. We have given a one-trespass, no trespass policy. That person comes on our property, he will be arrested. We have trans, we have bathrooms, all-gender bathrooms associated throughout the facilities, and people will use those bathrooms. This issue is one issue, one off.

Self-identification ideology demands that any declaration of gender identity be accepted without question or verification. There is no way to invalidate self-identification. But when faced with Cox, whose identification was clearly predatory, authorities implicitly acknowledged it was fraudulent and created a one-time exception just for him. This is contradictory. Can gender self-identification be invalidated or not?

When a policy is so easily exploited, the logical thing to do is to revisit the policy. Incidents such as these should call into question the ideology of gender self-identification itself.

Electoral Remedies

What can be done about this? Someone might say, “Vote.” Couldn’t Fairfax County get a better Commonwealth’s Attorney again?

Neighboring Loudoun County recently did just that. Bob Anderson, who was previously CA from 1996 to 2003, narrowly won an election against incumbent CA and “progressive” prosecutor Buta Biberaj with a margin of only 68,068 to 67,768 votes. This came after Biberaj had prosecuted Scott Smith, the father of a high school girl who had been raped in the bathroom of a Loudoun County school, for an incident that occurred while Superintendent Scott Ziegler made false statements that “we don’t have any record of assaults occurring in our bathrooms.”

However, Loudoun County could utilize this remedy because it occasionally holds competitive two-party elections. (Bob Anderson ran as a Republican.) This is unlikely to occur in Fairfax County, which does not elect Republicans for the Commonwealth’s Attorney. Steve Descano has already survived one primary challenge in 2023 from a more law-and-order Democrat in Ed Nuttall, and Descano is now entrenched in the Virginia Democratic Party establishment.

This kind of lax law enforcement appears to be what Fairfax County can expect for the foreseeable future.

Conclusion

The Richard K. Cox case highlights two unfortunate trends in criminal justice that intersected to create an environment in Fairfax County that Cox could exploit for months:

1. The installation of “progressive” prosecutors, supported by outsider money, who use the rhetoric of “mass incarceration” to minimize the punishment of criminal offenders

2. The rise of policies of gender self-identification, which positively assert that someone’s gender identity should have no criteria whatsoever other than a person’s own declaration

The Fairfax County Police Department attempted to protect the citizens of Fairfax County, but the police were hindered for months by either or both of these factors.

This is a cautionary tale for other jurisdictions. This could happen in many counties in the United States. A well-funded “progressive” prosecutor can narrowly win an election against an incumbent. An unelected bureaucracy can institute a gender self-identification policy with little public scrutiny. Your daughters, too, can walk into their locker rooms to find a sexual predator masturbating in the shower.

However, I do occasionally nerd out about the history of ideas. And to be a “serious thinker,” you are typically expected to have a “big idea.”

My big idea is cultural hegemony. It is a concept that explains a lot for me. I believe it explains why the people of the United States are so divided in the 21st century. It explains why my neighbors are so boring to me. And it explains why I was foolish to try to pursue a career in scholarship.

Like everybody else, I did not invent my big idea. Therefore, I thought it was a good idea to understand what earlier thinkers have thought about cultural hegemony before taking the concept in my own direction.

Antonio Gramsci was an Italian communist who was imprisoned by Mussolini’s Fascist government on November 8, 1926. He was given a 20-year sentence, but his health deteriorated to the point that he was released to medical custody on August 19, 1935. He was formally released before dying on April 27, 1937.

I am not interested in an international workers’ revolution or a dictatorship of the proletariat. Reading a Marxist typically results in me rolling my eyes. However, Gramsci is generally credited with coining the term “cultural hegemony” and providing the concept with its earliest known serious treatment, so I attempted a reading.

I use the phrase “attempted a reading” deliberately. Most of Gramsci’s writings are from his “prison notebooks,” fragmentary writings he cobbled together while imprisoned. He would write a paragraph or two about one topic, then write a few paragraphs about another topic, then another. In time, he would eventually return to write more about the first topic. This went on for dozens of notebooks.

Gramsci’s prison notebooks were rediscovered in the 1950s and 1960s, becoming influential in the “New Left” movement.

I tried various approaches to reading Gramsci’s prison notebooks, including consulting what would have been an authoritative academic translation by Joseph Buttigieg,1 only to discover that Buttigieg died before completing his translation, leaving notebooks that contained passages about the concept of cultural hegemony untranslated.

Ultimately, I adopted a straightforward approach. I obtained a copy of Quintin Hoare and Geoffrey Nowell Smith’s Selections from the Prison Notebooks (1971), reviewed the topic index, and read any section in the book index under “hegemony” or related concepts such as “common sense.”

Civil Society vs. Political Society

A lot of political theory is concerned with the coercive power of the state. Gramsci noted that there is another component that is just as important, if not more so.

What we can do, for the moment, is to fix two major superstructural “levels”: the one that can be called “civil society”, that is the ensemble of organisms commonly called “private”, and that of “political society” or “the State”. These two levels correspond on the one hand to the function of “hegemony” which the dominant group exercises throughout society and on the other hand to that of “direct domination” or command exercised through the State and “juridical” government. The functions in question are precisely organisational and connective. The intellectuals are the dominant group’s “deputies” exercising the subaltern functions of social hegemony and political government. These comprise:

1. The “spontaneous” consent given by the great masses of the population to the general direction imposed on social life by the dominant fundamental group; this consent is “historically” caused by the prestige (and consequent confidence) which the dominant group enjoys because of its position and function in the world of production.

2. The apparatus of state coercive power which “legally” enforces discipline on those groups who do not “consent” either actively or passively. This apparatus is, however, constituted for the whole of society in anticipation of moments of crisis of command and direction when spontaneous consent has failed.

This way of posing the problem has as a result a considerable extension of the concept of intellectual, but it is the only way which enables one to reach a concrete approximation of reality.

Political society refers to the state's coercive power. Think police, courts, and prisons.

Cultural (or “social”) hegemony occurs in civil society and is the way the ruling classes dominate society not by force, but by prestige. Think Harvard or the New York Times.

Mechanisms of Cultural Hegemony

Gramsci puts his quotation marks to work in the preceding passage. The way cultural hegemony works is to create the appearance of “consent” from the masses and to appear to be “spontaneous.” In reality, consent and spontaneity are engineered by the ruling classes and their deputies.

This engineering can be done by controlling the mass media.

The ‘normal’ exercise of hegemony on the now classical terrain of the parliamentary regime is characterised by the combination of force and consent, which balance each other reciprocally, without force predominating excessively over consent. Indeed, the attempt is always made to ensure that force will appear to be based on the consent of the majority, expressed by the so-called organs of public opinion—newspapers and associations—which, therefore, in certain situations, are artificially multiplied. Between consent and force stands corruption/fraud (which is characteristic of certain situations when it is hard to exercise the hegemonic function, and when the use of force is too risky).

It can also be done by controlling education.

In my opinion, the most reasonable and concrete thing that can be said about the ethical State, the cultural State, is this: every State is ethical in as much as one of its most important functions is to raise the great mass of the population to a particular cultural and moral level, a level (or type) which corresponds to the needs of the productive forces for development, and hence to the interests of the ruling classes. The school as a positive educative function, and the courts as a repressive and negative educative function, are the most important State activities in this sense: but, in reality, a multitude of other so-called private initiatives and activities tend to the same end—initiatives and activities which form the apparatus of the political and cultural hegemony of the ruling classes.

Regardless of how hegemony is practiced, the hegemonic power of culture and the coercive power of law complement each other.

When Gramsci refers to the “ethical State” or the “cultural State” in the above, he is referring to civil society, as his definition of the state encompasses civil society.

But what does that signify if not that by “State” should be understood not only the apparatus of government, but also the “private” apparatus of “hegemony” or civil society?

State = political society + civil society, in other words hegemony protected by the armour of coercion.

Intellectuals, aka the Professional-Managerial Class

Gramsci refers to “intellectuals” as the ruling class’s deputies. This is an early reference to what James Burnham would call the “managerial class,” what John and Barbara Ehrenreich would call the “professional-managerial class (PMC),” what Daniel Patrick Moynihan would call “The New Class,” and more recently what Musa al-Gharbi calls “symbolic capitalists.”2

The thinkers in this genre necessarily grapple with the rise of the PMC, which came with the growth of credentialism and university degrees in the 20th century. The PMC comprises individuals who hold college degrees and earn an income through occupations that do not involve manual labor. I would hazard a guess that most of the people reading this article are in the PMC.

Gramsci further distinguishes between “traditional” and “organic” intellectuals.

“Traditional” intellectuals are those who have been around for so long that they are no longer associated with a particular social class and so are effectively a separate social class themselves. Gramsci’s go-to example appears to be the clergy of the Catholic church.

Organic intellectuals, on the other hand, represent a specific social class. When a social class becomes the new ruling class (or tries to be), it brings along its own organic intellectuals. The rise of entrepreneurial capitalism brought with it its own organic intellectuals.

If not all entrepreneurs, at least an elite amongst them must have the capacity to be an organiser of society in general, including all its complex organism of services, right up to the state organism, because of the need to create the conditions most favourable to the expansion of their own class; or at the least they must possess the capacity to choose the deputies (specialised employees) to whom to entrust this activity of organising the general system of relationships external to the business itself. It can be observed that the “organic” intellectuals which every new class creates alongside itself and elaborates in the course of its development, are for the most part “specialisations” of partial aspects of the primitive activity of the new social type which the new class has brought into prominence.

Gramsci paints “intellectuals” with a broad brush, much like how later authors describe the PMC.

The function of organising social hegemony and state domination certainly gives rise to a particular division of labour and therefore to a whole hierarchy of qualifications.… at the highest level would be the creators of the various sciences, philosophy, art, etc., at the lowest the most humble “administrators” and divulgators of pre-existing, traditional, accumulated intellectual wealth.

As early as the 1930s, Gramsci wrote that the PMC was undergoing “unprecedented” growth.

In the modern world the category of intellectuals, understood in this sense, has undergone an unprecedented expansion. The democratic-bureaucratic system has given rise to a great mass of functions which are not all justified by the social necessities of production, though they are justified by the political necessities of the dominant fundamental group. Hence Loria’s conception of the unproductive “worker” … a conception which could in part be justified if one takes account of the fact that these masses exploit their position to take for themselves a large cut out of the national income. Mass formation has standardised individuals both psychologically and in terms of individual qualification and has produced the same phenomena as with other standardised masses: competition which makes necessary organisations for the defence of professions, unemployment, over-production in the schools, emigration, etc.

Symptoms of Cultural Hegemony

Hegemony is not as simple as autocracy. It is not simply the ruling class bossing everyone else around. In a hegemony, the ruling classes must at least consider the interests of those they subordinate.

Undoubtedly the fact of hegemony presupposes that account be taken of the interests and the tendencies of the groups over which hegemony is to be exercised, and that a certain com­promise equilibrium should be formed— in other words, that the leading group should make sacrifices of an economic-corporate kind.

Cultural hegemony tends to be anti-conservative and anti-religious because the new ruling classes will need to replace traditional intellectuals with their own organic intellectuals.

…the content of the political hegemony of the new social group which has founded the new type of State must be predominantly of an economic order: what is involved is the reorganisation of the structure and the real relations between men on the one hand and the world of the economy or of production on the other. …Cultural policy will above all be negative, a critique of the past; it will be aimed at erasing from the memory and at destroying. The lines of construction will as yet be “broad lines”, sketches, which might (and should) be changed at all times, so as to be consistent with the new structure as it is formed. This precisely did not happen in the period of the mediaeval communes; for culture, which remained a function of the Church, was precisely anti-economic in character (i.e. against the nascent capitalist economy); it was not directed towards giving hegemony to the new class, but rather to preventing the latter from acquiring it.

Common Sense

Gramsci spends a considerable amount of time ruminating about “common sense,” the received wisdom of the masses.

A work like the Popular Manual,3 which is essentially destined for a community of readers who are not professional intellectuals, should have taken as its starting point a critical analysis of the philosophy of common sense, which is the “philosophy of non-philosophers”, or in other words the conception of the world which is uncritically absorbed by the various social and cultural environments in which the moral individuality of the average man is developed. Common sense is not a single unique conception, identical in time and space. It is the “folklore” of philosophy, and, like folklore, it takes countless different forms. Its most fundamental characteristic is that it is a conception which, even in the brain of one individual, is fragmentary, incoherent and inconsequential, in conformity with the social and cultural position of those masses whose philosophy it is.

Common sense can be a barrier to the hegemony of ruling classes. Therefore, aspiring hegemons often try to rewrite common sense to suit their own worldviews. This can be observed in the experiences of France, the United States, and the United Kingdom.

The attitude of French philosophical culture towards common sense can indeed offer a model of hegemonic ideological construction. American and English culture can also offer some suggestions, but not in such an organic and complete way as the French. “Common sense” has been treated in various ways. Sometimes it has even been taken as the base of philosophy itself. Alternatively it has been criticised from the point of view of another philosophy. In reality, in either case, the result was to transcend a particular form of common sense and to create another which was closer to the conception of the world of the leading group.

Rewriting common sense is also what many Marxists try to do.

References to common sense and to the solidity of its beliefs are frequent in Marx. But Marx is referring not to the validity of the content of these beliefs but rather to their formal solidity and to the consequent imperative character they have when they produce norms of conduct. There is, further, implicit in these references an assertion of the necessity for new popular beliefs, that is to say a new common sense and with it a new culture and a new philosophy which will be rooted in the popular consciousness with the same solidity and imperative quality as traditional beliefs.

Conclusion

In Gramsci’s work, you can clearly see the beginnings of much of the criticism I have regarding cultural hegemony in the United States in the 21st century. You can also see the outlines of a whole genre of criticism of elites that followed. However, he left a large number of scattered writings, so he addressed these points in part due to sheer volume.

Gramsci is a much more nuanced and interesting thinker than the archetypal blowhard in the Marxist-Leninist tradition. Despite his nuance, he is not very readable because of the fragmentary nature of his work. Undoubtedly, both of these were caused, at least in part, by his imprisonment at the time of his writings.

Furthermore, most of his thinking is not timeless, but particular to his historical context. He spends a lot of time thinking about the French Revolution, the Risorgimento (the 19th-century movement that reconstituted Italy as a nation), the Russian Revolution, and why Marxism failed to materialize in Italy and the West. He goes over every minute difference between his philosophy and that of Benedetto Croce, the liberal Italian statesman, and Giovanni Gentile, the philosopher of Fascism.

Am I glad I took the time to read some of Antonio Gramsci’s writings? Yes. Am I blown away? No. Would I recommend others read Gramsci? Probably not.

Gramsci is of historical interest to those of us who labor intellectually in a tradition that he, in some sense, started. He gave us some labels, such as “cultural hegemony,” that endured. He started other labels, such as “organic intellectual,” that were quickly replaced. It is interesting to see many of the concepts we use today expressed as early as the 1930s. Still, anyone who wants to learn about them is better off learning from a contemporary writer who uses modern examples.

Reference

Gramsci, A. (1971). Selections from the Prison Notebooks of Antonio Gramsci. New York: International Publishers.

Additionally, for the past several months, at the advice of a friend, I have been using a pseudonym.

Both of these facts had a common cause: since February, I have been a full-time job-seeker. As it turns out, finding employment in 2025 is quite an ordeal. The job market is terrible. I applied to 328 job postings before becoming gainfully employed again.

I want to keep my research work on Born Curious separate from my other career, so I will just leave it at that.

Now that I am gainfully employed, I see no good reason to continue using the pseudonym. I will take this opportunity to discuss a prospectus of work that I intend to publish here throughout the remainder of 2025 and to explain the reasons for using a real name in a Substack newsletter.

Prospectus

In 2025, I intend to

1. wrap up my (first-pass on) studies in reproductive responsibility and

2. start an inquiry on “Why Americans Are So Divided.”

#1 entails two analyses that leverage my weights for the National Survey of Family Growth (NSFG) adjusted for abortion under-reporting. The first will be on trends in unintended pregnancy from 1973 to the present. The second will be on real-world contraceptive failure rates.1 This will be capped off by an editorial (which I alluded to in my article on what makes women orgasm) about how the sexual revolution of the 1960s and 70s failed us, especially women, who bear the brunt of unintended pregnancy and whose pleasure the rewrite of our sexual culture seems to have ignored.

#2 sounds played out, I concede. However, I believe I have a couple of interesting new dimensions: my analysis is grounded in a Gramscian framework of cultural hegemony, and it will be empirically driven, utilizing data from the General Social Survey (GSS), American National Election Studies (ANES), and other sources.

My working hypothesis2 for #2 is that the division we see in the 21st-century United States is the result of a peak in the cultural hegemony of the ruling class mediated by the professional-managerial class. This has led to populist backlashes against the “elite,” such as the Occupy Wall Street movement, the Tea Party movement, and Trumpian populism.

To understand how we arrived at this point, we will take a data-driven tour through the partisan sorting of the 1990s, the unilateral lurch among Democratic-leaning voters in the 21st century, and the emergence of a “package” of viewpoints that nowadays often come together.

Along the way, we might explore the tension between the ruling class of wealthy capitalists and their functionaries in the professional-managerial class. Sometimes the ruling class seems to “win” (e.g., John D. Rockefeller III), and sometimes the professional-managerial class runs away with things (e.g., John Ford II).

Cadence

I still intend to publish weekly. I realize my work is not the typical Substack (or, indeed, blogosphere) fare of hot takes on the new cycle or running commentary on politics, so my audience will always be limited. Still, everything I have read suggests that to create a successful Substack newsletter, you must publish regularly.

It takes me about a month, working nights and weekends, to complete an analysis. (The actual writing part goes by pretty quick, but there are many more hours of exploring the data, figuring out where I want the analysis to go, etc.) My plan, therefore, is to take each analysis I complete, break it up into four or more chunks, and release it in pieces while I work on the subsequent analysis.

Thus, there will still be a lull in output for about a month, and then, after that, I promise myself that I will release an article per week going forward.

I may also try to work in some lower-level-of-effort conceptual articles from time to time to keep the cadence up, but we shall see.

To Use Your Real Name or a Pseudonym?

Many authors who use their real name on Substack already had a career in journalism, and their name brings in subscribers. It definitely makes sense for these people to use their real names.

But what about the rest of us normies who have never been published? It might make more sense to just be PlumpTurnip877 or whatever in our online presence. The major pro to this approach is obvious: You can write what you really think without fear of repercussions in real life.

In my case, I started this newsletter using my real name, so my attempt to use a pseudonym was kind of pointless. If you searched for my name and “blog” in any major search engine over the past few months, you would get Born Curious pretty quickly. Therefore, I have pretty much self-doxed, and there is no going back.

Here are my arguments for self-doxing your writing, which might just be me post hoc rationalizing.

• There is a slight chance you will get to the point that someone will want you to write something in their periodical or appear on their podcast. If you are already using your real name, then this process would be smoother than figuring out who PlumpTurnip877 is. I admit, for us nobodies, this is a very slight chance, indeed.

• Using a real name and face serves as an extra reminder not to engage in negative online interactions. The pro to using a pseudonym is also a con in this regard. One reason people can be so rude online is that anonymity often encourages rudeness. At any given point in time, I see dozens of people online who I think are stupid and have nothing to contribute. The Internet lowers the barrier to scratching the itch to actually tell someone this. Using a real name and face raises the barrier a little.

• Much of my writing is about conformity and the social penalties for dissent. Perhaps, then, I should criticize this phenomenon without insulation from the social penalties of dissenting, especially when the risk of any penalties is so small. My work is long and empirical. I have minimal readership. And this is Substack, which is supposed to be spicy.

The 2024 data for the General Social Survey were recently released, though they have yet to be included in the GSS Data Explorer. Therefore, I had to download the cross-sectional file and merge it with the data from my previous analysis.

This is a quick update of my article on trends in attitudes toward abortion1 in the United States to extend the analysis to include the 2024 data.

For background, you may have seen chatter in the mass media about support for legalized abortion-on-demand being especially popular among women. However, women were actually more likely than men to oppose the legalization of abortion-on-demand in the run-up to Roe v. Wade and Doe v. Bolton back in the 1960s and early 1970s.

Furthermore, when I went to find out when support by gender flipped, I was surprised. It never did. Women have never supported legalized abortion-on-demand at a higher rate than men.

The General Social Survey is a big project that lags behind telephone and Internet polling. I hypothesized that this alleged phenomenon would take longer to appear in the GSS data, so I waited for it to appear to analyze it. (Telephone and Internet polls typically do not release public use data files like the GSS.)

With that, let us dive into the new data.

New Data

Reviewing Figure 1, opposition to the legalization of abortion-on-demand was the majority viewpoint in the United States up until 2016. Sometime around 2018, this reversed. From 2021 on, support for legal abortion-on-demand has been the majority viewpoint.

The 2024 data continue the status quo. If you look at those approving of legal abortion-on-demand, the proportion is about the same as in 2022. If you look at those disapproving of legal abortion-on-demand, this appears to decrease in 2024 to just under 40%.

This happened because more respondents answered “don’t know” or didn’t answer the question in 2024.

This is all well within sampling error. Thus, we won’t know if the increasing-approval/decreasing-disapproval trend is continuing or if viewpoints have stayed roughly the same between 2022 and 2024 until we’ve seen a few more iterations of the GSS.

The gap in viewpoints by party lean appears to have narrowed a bit between 2022 and 2024, with a decrease in approval among Democrat-leaners and an increase in approval among Republican-leaners. However, this is again well within sampling error and may be nothing, so I caution against interpreting this as an actual change.

Much to my chagrin, the proportion of women approving of legal abortion-on-demand remained lower than the proportion of men approving of legal abortion-on-demand in 2024 as in 2022. I fit a logistic regression model to this result. The gender difference was within sampling error. Thus, either the proportions of men and women who approve of legal abortion-on-demand are the same, or approval is slightly lower among women and the GSS does not have enough power to detect the difference.

At this point, I think the supposed gap in which women approve of legal abortion-on-demand at a higher rate than men does not exist. If the gap did exist, we would have seen it on the GSS by now. At the very least, the estimate of the rate of approval would be higher for women than for men, instead of the reverse.

As I have written previously, the other polls that supposedly find such a gap are methodologically inferior to the GSS. The GSS has been asking the same questions since 1965.2 These questions are about specific policies and not vague “a lot,” “somewhat,” or “a little” type questions. Furthermore, the GSS uses an address-based sampling frame with probability-based sampling to generate a representative sample of the United States. These other polls you see — whether the telephone polling done by Gallup or the one-offs you see reported in the media — are just not as robust.

Ramifications

The cultural hegemony of the professional-managerial class (PMC) likely hurt itself in 2024 by believing its own false narratives.3

The Democratic Party was once dominant among the working class. Now it’s squarely the party of the PMC. The 2024 Democratic presidential campaign centered on support for legal abortion-on-demand, a core PMC position. Indeed, the 2024 Harris-Walz campaign often struggled to speak concretely about any issue except abortion.

However, Harris-Walz actually did worse in 2024 among women than did the Democratic ticket in 2020. Democrats running for President have generally done better among women voters than Republicans, but that advantage actually narrowed between 2020 and 2024, rather than increasing. Shortly after the election, reporting in The Hill noted:

The CNN exit poll indicated Harris’s advantage with women was 8 points, rather than Biden’s 15 points in 2020. The AP/Fox survey found a 7 point edge for Harris rather than Biden’s 12 point advantage.

This shouldn’t be surprising given what we know from the GSS data. Women do not support legal abortion-on-demand more than do men (Figure 3). Furthermore, given the polarization by party in viewpoints on abortion (Figure 2), most people who would change their voting preference based on abortion have already done so.

Furthermore, support for legal abortion-on-demand is ~55% vs. ~40% in the United States. (Figure 1) This is far less of a difference in viewpoints than there are for other issues, for instance, transgender athletes in women’s sports, which is opposed by ~79% vs. ~18% of Americans.

Abortion is a rally-the-base get-out-the-vote issue for Democrats, not an issue with which they can win persuadable voters—or, indeed, Presidential elections.

Men are more likely to experience orgasms during sexual encounters than women. At the same time, there are common misperceptions about what is likely to make a woman orgasm.

In particular, the vast majority of the population considers intercourse to be the definitive sexual act, but intercourse is one of the least effective ways to make a woman orgasm.

We now have three large surveys confirming the fact that intercourse is poor at causing orgasms in women. They are:

• Australian Study of Health and Relationships (Richters et al., 2006)

• NBC News’ Website Survey (Frederick et al., 2018)

• KnowledgePanel / OMGYes.com Survey (Herbenick et al., 2018)

I discuss the quality of each survey in the appendix.

Orgasm Gender Gap

In a heterosexual context, the Australian Study of Health and Relationships (ASHR) found that it is about six times more common for women to go without orgasm than men. Ritchers et al. report:

It is unusual for a man not to have an orgasm when he has sex with a woman: only 5.2% of men did not reach orgasm at their last heterosexual encounter. It is much more common for women, 31.1% of whom did not reach orgasm.

In homosexual relations, women fare better, but still do not experience orgasms at the same rate as men in either heterosexual or homosexual relations. Frederick et al. tell us:

Heterosexual men were more likely than heterosexual women to always orgasm (75% HM vs. 33% HW; p < .001) and usually-always orgasm (95% HM vs. 65% HW; p < .001) when ‘‘sexually intimate’’ during the past month. Lesbian women were less likely than heterosexual men to always orgasm (59% LW vs. 75% HM; p < .001) or usually-always orgasm (86% LW vs. 95% HM; p < .001). Lesbian women were, however, more likely than heterosexual women to always orgasm (59% LW vs. 33% HW; p < .001) or usually-always orgasm (86% LW vs. 65% HW; p < .001). The patterns for bisexual women were similar to heterosexual women, and patterns for gay and bisexual men were similar to those of heterosexual men, except that they were slightly less likely to always orgasm when sexually intimate.

What Is “Having Sex?”

The fact that the majority of the population equates “having sex” with intercourse, at least in a heterosexual context, complicates the data. There is much more evidence for sexual encounters involving intercourse than for sexual encounters that do not include intercourse.

In the ASHR, ~95% of most recent heterosexual encounters included vaginal intercourse, consistent with previous studies.

There are two ways in which vaginal intercourse can account for so much of heterosexual encounter data in surveys.

Firstly, many couples treat it as obligatory. When such couples decide to “have sex,” they include intercourse in their sexual encounter because that is what “having sex” means. This phenomenon, though worthy of criticism in other regards, does not compromise the data.

Secondly, survey respondents whose most recent sexual encounter did not include intercourse might report some other encounter instead because they do not consider their actual most recent encounter to count as “having sex.”

This latter phenomenon creates an issue for analysis because data are lost. Unfortunately, there isn’t anything we can do about respondents who fail to report such encounters, but it is worth acknowledging that this is a potential issue.1

Which Acts Are Better at Making Women Orgasm Than Intercourse?

Intercourse itself either has a negligible impact on the rate of orgasm in women or actually decreases the rate of orgasm.

Ritchers et al. looked into the most recent heterosexual encounter reported on the ASHR and tabulated whether the respondents experienced an orgasm, categorized by the sex acts they practiced.

Men, by and large, reliably had an orgasm. Men had an orgasm at least 81% of the time, regardless of the specific sexual acts involved. However, there was an increase in the orgasm rate when vaginal intercourse was involved. Men whose last encounter included only manual and oral stimulation experienced an orgasm at a rate of 87.3%. In contrast, those who engaged in vaginal intercourse in addition to receiving manual and oral stimulation experienced an orgasm at a rate of 97.7%. Thus, there was about a 10 percentage point “bonus” from intercourse for men.

On the other hand, only about half women (49.6%) experienced an orgasm when their last encounter consisted only of vaginal intercourse. When the women were manually stimulated, this rate increased to 70.9%. When oral and manual stimulation were included, the rate further increased to 85.6%.

Moreover, there appeared to be a “penalty” rather than a “bonus” from vaginal intercourse. Women who had manual stimulation without intercourse had an orgasm at a rate of 78.7%, versus women who had manual stimulation plus intercourse, who had an orgasm at a rate of 70.9%. Women who had manual and oral stimulation without intercourse had an orgasm at a rate of 90.0%, versus women who had manual and oral stimulation plus intercourse, who had an orgasm at a rate of 85.6%. Thus, including intercourse in a sexual encounter actually decreased the rate of orgasm.2

This phenomenon was partially corroborated by the NBC News website’s survey.

Frederick et al. (2018) did something odd with their analysis of the NBS News website’s survey data. Instead of using whether or not the respondent had an orgasm in the most recent encounter as the response variable, they used whether the respondent “usually” or “always” had an orgasm versus “rarely” or “never” had an orgasm over the past month as the response variable. However, they still used the practices in the most recent encounter as the explanatory variable for categorization.

Frederick et al. justify this analysis with this discussion:

We … examined event-level data, which may be less prone to recall biases. Participants were asked about their activities during the last time they had sex. If these behaviors are representative of what couples typically do, they may provide information about which practices are linked to greater orgasm frequency. Consistent with the proposal that event-level data are informative about general practices, 87% of women and 92% of men in this data set who received oral sex during their last sexual encounter reported usually-always receiving oral sex in the past month. Also consistent with this proposal, reports of oral sex during last encounter were lower if they reported oral sex half of the time (50% women; 60% men) and lowest if they reported oral sex never-rarely (10% women; 12% men) during the past month.

Thus, their justification for this arrangement was that survey respondents are better at remembering a specific event than what they did over a whole month, and the practices during the most recent event appear to be a good proxy for what people do over the past month.

The lowest rate of “usually” or “always” orgasm frequency (35%) among heterosexual women occurred with those who had only vaginal intercourse in their last sexual encounter. The highest rate among heterosexual women occurred among those who experienced all the practices except intercourse — 80% among women who received oral sex, genital stimulation, and deep kissing, but not intercourse.

Whether or not there is an intercourse penalty is more equivocal in the NBC News data. For instance, women who who received oral sex and genital stimulation without intercourse reported usually-always orgasming at a higher rate (73%) than did those who also had vaginal intercourse (69%). However, there are some counterexamples. For instance, women who received oral sex and deep kissing without intercourse reported usually-always orgasming at a lower rate (69%) than did those who also had vaginal intercourse (71%).

I suspect these differences are within sampling error of zero, though neither Frederick et al. (2018) nor Richters et al. (2006) quantified sampling error in their analyses.

Generally, in the NBC News survey data, the rate of usually-always orgasm frequency was a function of practices other than intercourse, and the presence or absence of intercourse had a negligible effect on the usually-always orgasm rate.

Things do not look good for intercourse with regard to women’s orgasms. Why might this be the case?

Physiological Reasons

One reason might be purely physiological. Intercourse directly stimulates the penis, but does not directly stimulate the clitoris, which is the homologue of the penis in female genitalia.

The KnowledgePanel survey commissioned by OMGYes.com addressed this directly. Herbenick et al. report:

In this sample, 36.6% (n = 347) of intercourse-experienced women reported that they needed clitoral stimulation in order to orgasm during intercourse; 36.0% (n = 341) reported that although they did not require clitoral stimulation for orgasm during intercourse, adding it enhanced orgasm; and 18.4% (n = 174) reported that vaginal penetration alone during intercourse was sufficient for orgasm. The remaining 9.0% of women reported they did not have orgasms during intercourse (7.5%, n = 71) or described other patterns or routines (e.g., clitoral stimulation prior to intercourse, orgasm via cunnilingus followed by intercourse, etc.).

Thus, stimulation of the clitoris is either required for orgasm or enhances orgasm for the vast majority of women (72.6%).3

Furthermore, clitoral stimulation appears to increase the frequency of orgasm during intercourse among women. Herbenick et al. (2018) found that 22.3% of respondents reported an orgasm frequency of “always” when receiving clitoral stimulation during intercourse, versus 13.5% of respondents during intercourse without clitoral stimulation. This trend is present for all possible orgasm frequency responses, as seen in the bar plot above.

Motivations for Sexual Encounter

Richters et al. propose the hypothesis that the sex acts involved in an encounter might correlate with the motivations for the sexual encounter.

First, they observe that there is an asymmetry in heterosexual couples in the desire for sexual activity:

When asked how often they would ideally like to have sex, men’s stated ideal frequency was somewhat higher than women’s. The median response from men was four to six times a week, whereas the median response from women was in the “two or three times a week” category.

It follows that there might be encounters in which one partner desires sexual activity, but the other does not, and the undesirous partner engages in the encounter merely for the benefit of the desirous partner. Couples likely engage in sex acts agreeable to the desirous partner in these encounters.

Thus, intercourse-only sexual encounters might comprise a greater proportion of sexual encounters done when the man desires sexual activity and the woman is participating for the man’s sake. Similarly, encounters involving manual or oral stimulation of the woman but not intercourse might involve a greater proportion of sexual encounters done when the woman desires them and the man is participating for the woman’s sake.

When an encounter occurs just for one partner’s benefit, the undesirous partner might be less likely to orgasm due to a lack of motivation. This could explain some of the differences in the rate of orgasm among women between categories of sexual encounter by sex act.

Orgasm Is More Likely with a Regular Partner

If the presence or absence of intercourse has little effect on the rate of orgasm for women, then what does?

The factor that stood out in the ASHR was having a regular sexual partner.

Ritchers et al. tabulated whether respondents had an orgasm during their last heterosexual encounter by several variables from the respondents’ sexual histories. Men reliably had an orgasm regardless of their backgrounds.

Women experienced an orgasm at lower rates overall. The differences between the various levels of the variables were minor for the most part (less than 10 percentage points), with one major exception: women who did not have a regular partner experienced an orgasm at a rate of 48.7%, but the rate for those who did have a regular partner was 70.0% — a whopping 21 percentage point difference!

Having a regular partner benefits the rate of orgasm among men, as well, but the difference is less dramatic since men tend to reliably orgasm. Men who did not have a regular partner experienced an orgasm at a rate of 87.6%, but the rate for those who did have a regular partner was 95.5% — an 8 percentage point difference.

Unlike in their sex act analysis, Richters et al. gave us 95% confidence intervals for the odds ratios of the rates of orgasm. Because the confidence intervals for the odds ratios by partner type “regular” versus “other” for both women and men do not include 1.00, we can infer that the results are statistically significant at the α = 0.05 error rate.

The NBC News website survey only targeted respondents with regular sex partners, so Frederick et al. do not shed light on whether women orgasm at a higher rate with regular partners due to a lack of comparisons.

The KnowledgePanel / OMGYes.com survey did not specifically ask what kind of relationship respondents were in. However, the survey did have a more open-ended question about what enhanced orgasms among women.

The second and third most popular answers of “having a partner who knows what I like” and “emotional intimacy” are synonymous with having a regular sex partner and are only behind “spending time to build-up arousal” in popularity among women for orgasm enhancement. This is consistent with the findings from the ASHR that having a regular sexual partner increases the likelihood of orgasm among women.

Conclusion

Overall, women experience orgasm during a sexual encounter at a lower rate than men. This is true for both same-sex and opposite-sex sexual encounters, but the difference between men and women is greater in opposite-sex encounters.

Intercourse is considered the definitive sexual act in heterosexual encounters. Simultaneously, the inclusion of intercourse in a sexual encounter has either no effect or an adverse effect on the rate of orgasm among women. The inclusion of intercourse does have a small positive impact on the rate of orgasm among men. Thus, the focus on intercourse in sexual encounters exacerbates the orgasm gender gap.

Sex acts that are more adept at making women orgasm include manual stimulation of the clitoris, cunnilingus, and kissing. Furthermore, having a regular sexual partner who knows what a particular woman likes and with whom she has emotional intimacy increases the rate of orgasm among women and enhances orgasms for most women.

There is an obvious editorial to be written based on these data. However, I will save that editorial article until after I discuss two more analyses: one on the trend in rates of unintended pregnancy after the “sexual revolution” of the 1960s and 70s, and another on the ineffectiveness of contraception from then until this very day.

Appendix: Data Sources

Throughout the article, I have written about “women” and “men.” This should more precisely be “women and men in Australia, the United States, and possibly other countries with similar sexual cultures.”

The ASHR surveyed a nationally representative sample from Australia, and the KnowledgePanel / OMGYes.com survey used a representative sample from the United States. All three surveys reinforce similar conclusions for the results discussed in this article, so it is safe to conclude that the populations of Australia and the United States are similar enough with regard to these specific conclusions that the results apply to at least these two countries.

It is quite possible that results would differ in other countries. This is less likely for results based on physiological causes, assuming small differences in female physiology from country to country. However, the size of the orgasm gender gap, for instance, presumably differs from place to place based on the sexual cultures there.

The article was already long, and general readers seem to care little about the scope of inference of studies, so I have relegated this discussion to this appendix.

I discuss the specifics of each of the three surveys below.

Australian Study of Health and Relationships (Richters et al., 2006)

The ASHR used random-digit dialing (RDD) via telephone, targeting all men and women in Australia aged 16 to 59 years. RDD is not as good as address-based sampling (ABS), but RDD is cheaper and is still used by Gallup and other pollsters.

Ritchers et al. report “The overall response rate was 73.1% (69.4% among men and 77.6% among women).” If this survey were done in 2025, this would be an outstanding response rate, but even by 2006 standards, it was pretty good.

The overall sample size was effectively n = 5,118 for the results discussed in this article, because the survey instrument about recent sexual behavior was given to a subsample of 7,653 people, and 5,118 people from this subsample were sexually active and reported a recent sexual encounter with an opposite-sex partner.

The raw sample would almost certainly be biased (e.g., some people have multiple phone numbers by which they can be reached, while others have none; some people don’t pick up the phone if they don’t recognize the number). Additionally, some oversampling was done (by region and of men) to reach the target sample sizes. It was, therefore, important to adjust the sample weights to make the sample match the target population.

Fortunately, the ASHR did such a weight adjustment. Ritchters et al. describe:

Data were weighted to adjust for the probability of household selection (households with more phone lines were more likely to be contacted) and to adjust for the probability of selection within a household (individuals living in households with more eligible people were less likely to be selected). Further weighting on the basis of age, gender, and area of residence ensured that both the full sample and the subsample matched the Australian population as reported in the 2001 Census.

NBC News (Frederick et al., 2018)

The survey from the NBC News website was the weakest of the three. It was basically an opportunity sample of people who happened to browse the NBC News website. Thus, there was no guarantee that the sample was representative of any population whatsoever. This, coupled with the strange way Frederick et al. framed their Table 4 results, almost made me not include the paper. However, the results largely corroborate the ASHR study and extend the results to same-sex couples, so I opted to include it.

The sample size was n = 52,588. The criteria for inclusion were:

…aged 18–65 years; completed the full survey via the NBC News entry portal; indicated they were married, remarried, cohabiting, or dating/seeing one person; and reported being intimate in the past month in response to the question about orgasm frequency over the last month.

Frederick et al. note that NBC News might have a biased television audience, but that:

Market research on NBCNews.com (formerly msnbc.com) shows that, at the time of the surveys, it routinely ranked among one of the most popular Web sites in the U.S. Its 58 million unique monthly visitors included a broad diversity of people in terms of age, income, and political orientation. It is important to note that msnbc.com, the general news Website, was a different entity than MSNBC TV and had substantially different demographics, including approximately equal numbers of Democrat and Republican visitors.

Frederick et al. make no mention of weight adjustment to try to match the target population’s known demographic characteristics. The lack of such discussion, coupled with the fact that they had an opportunity sample, not a probability-based one, leads me to believe that no weighting was done at all.

Thus, the results of the NBC News survey shouldn’t be generalized to any population other than the NBC News users who felt like filling out a survey during that specific ten-day period. Still, this limited population seems to have had similar experiences concerning women’s orgasms as the target population of the ASHR.

KnowledgePanel / OMGYes.com (Herbenick et al., 2018)

Despite being commissioned by a website with a whimsical name, the KnowledgePanel / OMGYes.com survey seems to have been taken seriously. Herbenick et al. tell us:

The study was conducted during June 2015 with GfK Research (Menlo Park, CA), who sampled adult (ages 18+) female U.S. residents from its KnowledgePanel®, a probability-based web panel demonstrated to be representative of the non-institutionalized U.S. population and frequently utilized by scientists to collect U.S. probability data on a range of topics. Probability-based sampling techniques (address-based sampling [ABS] using the U.S. Postal Service’s Delivery Sequence File and random digit dialing [RDD] methods) have been used to recruit panel members. Households without Internet access are provided with hardware and/or Internet access as needed, in order to minimize the risk that only higher-resource or regular Internet-using households are included.

These hybrid online panels have become popular because they try to get the cost savings of an online survey while maintaining the sampling rigor of traditional surveys. They accomplish this by recruiting people into their online panel using traditional probability-based sampling methods and running many online surveys using the already recruited panel.

The sample size was n = 524, and the response rate was 43.7%. (I count the response rate as the percentage of people who received the invitation to complete the survey who actually did.)

GfK Research did a weight adjustment to try to make the sample match the target population:

Although the recruitment process is intended to represent the U.S. adult population, GfK then prepared post-stratification statistical weights to correct for possible non-response or any under- or over-coverage based on demographic variables from the Current Population Survey (e.g., gender by age, race/ethnicity, education, census region, household income, Internet access).

It’s odd to use the Current Population Survey (CPS) instead of the American Community Survey (ACS) and decennial Census data since the ACS was invented for such a purpose. (The CPS is commissioned by the Bureau of Labor Statistics and is used primarily for unemployment statistics.) However, this is just a methodological curiosity; I doubt this would change the results substantially.

This is about as robust a survey as you can get nowadays without resorting to the kind of budget that the General Social Survey or large government bureaucracies have.

References

Frederick, D. A., John, H. K. St., Garcia, J. R., & Lloyd, E. A. (2018). Differences in Orgasm Frequency Among Gay, Lesbian, Bisexual, and Heterosexual Men and Women in a U.S. National Sample. Archives of Sexual Behavior, 47(1), 273–288. https://doi.org/10.1007/s10508-017-0939-z

Herbenick, D., Fu, T.-C. J., Arter, J., Sanders, S. A., & Dodge, B. (2018). Women’s Experiences With Genital Touching, Sexual Pleasure, and Orgasm: Results From a U.S. Probability Sample of Women Ages 18 to 94. Journal of Sex & Marital Therapy, 44(2), 201–212. https://doi.org/10.1080/0092623X.2017.1346530

Richters, J., de Visser, R., Rissel, C., & Smith, A. (2006). Sexual practices at last heterosexual encounter and occurrence of orgasm in a national survey. The Journal of Sex Research, 43(3), 217–226. https://doi.org/10.1080/00224490609552320

For over half a century, the “gold standard” for learning about real-world social phenomena has been a representative survey with an address-based sampling frame, probability-based sampling, and face-to-face (FTF) interviews.

Unfortunately, response rates to these “gold standard” surveys have steadily declined. Then, in 2020, the COVID-19 pandemic happened, and thereafter, response rates took a nosedive.

For instance, let’s look at what happened with the National Survey of Family Growth (NSFG).

Response rates declined from ~90% in 1973 to ~63% in 2017-2019, but this was followed by a sudden drop of about 40 percentage points from a ~63% response rate in 2017-2019 to a ~23% response rate in 2022-2023.

It hasn’t just been the NSFG. The 2022-2023 NSFG User’s Guide states:

The NSFG response rates for 2022-2023 are markedly lower that the response rates for the last 2-year data collection period of 2017-2019. Some of the sharper decline in participation can be attributed to the effects on society of the COVID-19 pandemic, from respondents’ willingness to participate to the ability to recruit and retain interviewers.

Complex and long FTF surveys such as the National Health and Nutrition Examination Survey (NHANES) that did not undergo major design changes from before the pandemic, also experienced drops in participation. For example, the NHANES unweighted interview response rate fell from 51.0% for the 2017-2020 data collection to 34.5% for 2021-2023. The examination response rates fell from 46.9% to 25.6%.

Other similarly long and sensitive household surveys to the NSFG, such as the National Survey of Drug Use and Health (NSDUH) that underwent a redesign from only FTF to a web and FTF design, like NSFG, have also observed substantial drops in survey participation. The weighted overall survey response rate for NSDUH dropped from 45.8% in 2019 (pre-pandemic) to 12.3% in 2023.

As this passage alludes, there was a change in methodology between the 2017-2019 and 2022-2023 iterations of the NSFG when this dropoff in response rate occurred. The 2017-2019 NSFG used exclusively face-to-face interviews, but the 2022-2023 NSFG used a multimode sampling design that included both a Web-based form and face-to-face interviews.1

Plummeting response rates cause problems for surveys. They make surveys more expensive because larger samples must be drawn to get the target number of respondents needed for analysis. They also introduce more opportunity for response bias, the phenomenon in which respondents to the survey differ substantially from the survey’s target population.

Issue in 2022-2023 NFSG

When doing a recent analysis with NSFG data, I noticed an anomaly with the 2022-2023 NSFG public use data.

Since the NSFG is a fertility survey, it asks respondents whether they have ever had heterosexual, vaginal intercourse.2

If you look at people grouped by their birth cohort, you would expect the percentage of people who answer yes to this question to be strictly nondecreasing in subsequent surveys. The percentage of people born in certain years who have had intercourse should not decrease when asked in 2011 from what it was in 2006. You cannot undo having had intercourse.3

This is precisely what we see for three birth cohorts of women surveyed by the NSFG between 2006 and 2019.

Estimates of the percentage of women who have had sex by birth cohort increased for all three birth cohorts between the 2006-2010 and the 2011-2015 surveys. Point estimates appear to have declined between the 2011-2015 and 2015-2019 surveys for women born in the [1970,1975) interval,4 but this is well within fluctuations due to sampling error.

Now, let’s look at the trend including 2022-2023.

The 2022-2023 estimate for the percentage of women who have had sex appears much lower than the 2015-2019 estimate. Still, this strangeness could plausibly be explained by sampling error.

Looking at later birth cohorts, it becomes clear that mere sampling error does not explain this apparent and nonsensical decline.

So what is going on?

Part of the story is that the Web respondents to the 2022-2023 NSFG differ in their answers from face-to-face respondents.

In particular, women born after 1980 who responded to the 2022-2023 NSFG with a face-to-face interview reported having had sex a higher rate than women who responded via the online Web form.

If we exclude the Web respondents from our estimates for the [1980,1985) birth cohort, then the trends appear more plausible.

Using only face-to-face respondents, the dropoff in the percent of women who have had sex can plausibly be explained by sampling error for women born in [1980,1985) as seen in Figure 7, like it was for women born in [1975,1980) in Figure 4.

This phenomenon also exists among male respondents but is less dramatic. Curiously, it does not occur among cohorts born before 1980.

Response Bias or Measurement Error?

There are two plausible hypotheses for why Web respondents to the 2022-2023 NSFG would report a lower rate of sexual activity than face-to-face respondents among cohorts born after 1980.

1. Measurement Error: Under this hypothesis, people respond differently to an online survey than a face-to-face survey. In particular, responses are, on average, more honest in one mode or another.

2. Response Bias: Under this hypothesis, the people who respond to a survey online tend to differ from those who respond to a face-to-face survey.

Social Desirability

here on Substack took a look at the 2022-2023 NSFG data earlier this year and similarly noted anomalies with the online respondent portion of the data:

This could indicate two possibilities. One is the presence of social desirability bias, which might influence FTF responses more than the relatively anonymous online responses; the other is differences in characteristics between those who participated in the traditional FTF interview and those who completed the survey online.

Under this hypothesis, having had sexual intercourse is socially desirable, and people are less likely to admit that they have never had sexual intercourse when asked by a face-to-face interviewer.

The Nuance Pill seems to think this hypothesis is unlikely:

When it comes to social desirability bias, a few studies have found that online surveys are more susceptible to this than FTF interviews (Heerwegh, 2009; Berzelak & Vehovar, 2018). Interestingly, if it is having a large impact on the results, then women must be at least as ashamed to admit to being sexless as men, as women who completed the survey online had a similarly higher sexlessness rate to offline respondents as did men.

Social Aversion

A different hypothesis considered by Nuance Pill is that people who answer a survey via the Web tend to be more socially averse and so actually do have intercourse at a lower rate:

I found that online respondents were more educated and white on average, as were respondents in general compared to an earlier survey. One study found that online survey respondents were lower in openness to experience than FTF respondents, and this trait correlates with sexual activity.

While weighting can do a good job of accounting for these demographic characteristics, differences in personality/behaviour will likely remain, and there may be proportionally more people of a certain type than there were in previous samples…

False Negative Answers

I add a third hypothesis to Nuance Pill’s two hypotheses: people who answer a survey online might be more likely to lie. The NSFG offers a small monetary incentive, and it is a long survey designed to take 50 to 75 minutes to complete. Some people might just want the cash without having to spend an hour answering questions. If you say you haven’t had heterosexual, vaginal intercourse, you skip over most of the questions in the NSFG.

A little insider knowledge: having worked on the NSFG, I know that the survey administrators look for this and try to purge obviously fraudulent responses. However, nothing is perfect, and there may still be fake respondents who just want the $40 cash incentive slipping through.

This would also explain why the discrepancies are seen in younger cohorts, but not older ones. It is easier for younger people to pretend to have never had intercourse than for someone who is older, since by age 30, about 98% of people have had sexual intercourse.

Something about facing another human being who can read your demeanor and body language may make people less prone to lie. Also, people who intend to lie to get the $40 incentive with minimal effort might opt out of a face-to-face interview at a higher rate than an online survey, since letting someone into your home to interview you is more effort than using a Web browser.5

Conclusion

Surveys are moving away from the “gold standard” approach of face-to-face interviews to multi-modal approaches that primarily use online Web forms with face-to-face interviews as a backup. This is being done out of necessity due to rising costs from plummeting response rates and increasing difficulty hiring face-to-face interviewers.

We have reason to believe that online survey respondents either differ substantially from face-to-face respondents or are more apt to give false responses. If either is true, this compromises the veracity of comparing newer multi-modal iterations of surveys with previous face-to-face-only iterations. If the latter is true, this questions the veracity of the online responses themselves.

Going forward, we should exercise caution when doing any trend analysis that includes the time when a survey transitions from face-to-face-only to multi-modal, and we should express skepticism about online respondent data generally.

For instance, Jean Twenge — psychologist and author of the book Generations — was writing about this topic as early as 2015, finding that,

15 percent of 20-to-24-year-olds born in the 1990s have not had sex since turning 18, compared to only 6 percent of comparable young adults born in the late 1960s—more than twice as many.

Numerous articles have been written on this topic for the Institute for Family Studies by authors such as and Nicholas H. Wolfinger.

It remains to be seen if this phenomenon is substantial enough to have an effect on rates of unintended pregnancy, etc. Still, it hasn’t affected one aspect of Americans’ sexual behavior: when people have sexual intercourse for the first time.

Results

I used the National Survey of Family Growth (NSFG) public use files (PUFs) for this analysis. Because the NSFG is a fertility survey, most questions about timing of respondents’ first sexual activity ask specifically about heterosexual, vaginal intercourse. “Sexual intercourse” in this analysis refers to this.1

I divided respondents into cohorts by the half-decade of their birth. As you can see in Figure 1, there appears to be a trend in which subsequent generations, on average, first had sexual intercourse at gradually younger ages until the cohorts born in the 1960s.

For cohorts born in the 1960s and onward, the mean age at first sexual intercourse has stayed right around 17.5 years. The apparent deviation from this for the [1990,1995) cohorts is almost certainly due to sampling error, as alluded to by the exceptionally large confidence interval brackets around the estimates.

Of course, the mean of a distribution is only part of the story.

Figure 2 gives a more complete illustration of ages at first sexual intercourse for the most recent cohort of women for which we have good data. About 20% had sexual intercourse before age 14.5. About 60% had sexual intercourse for the first time between ages 14.5 and 19. About 18% had sexual intercourse for the first time between the ages of 18 and 30. About 1% of women in this cohort never had vaginal sexual intercourse with a man as of 2019.

The median for this distribution appears to be right around 16.5. Because the median is lower than the mean (~17.5), it is a right-skewed distribution. The distribution is dense between ages 14 and 18, but with a long tail of individuals who have intercourse for the first time in their late teens or twenties.

Distributions for other cohorts look remarkably similar. They are included in the appendix for the sake of completeness.

Let’s compare the [1945,1950) and [1960,1965) cohorts.

The “S” shapes of the curves for the two cohorts are similar, with the [1960,1965) curve just shifted to the left. Generally, it looks like women gradually started having sexual intercourse at a younger average age during the Baby Boom generation, but with the same overall pattern.

Comparison plots for [1960,1965) and [1980,1985) — the last cohort for which we have good data — show a lot of overlap.

Generally, the distribution of ages at first sexual intercourse has been stable for generations of Americans born after 1960.

Discussion

The age at which people start engaging in sexual activity can potentially be a factor in rates of unintended pregnancy because people are at their most fecund when they are young — at the same time that they tend to be their most reckless.

The average age of first sexual intercourse reached a minimum for women born in the 1960s. This generation lived through their teens when unintended rates in the 1970s and early 1980s were soaring. Perhaps the trend toward younger ages at sexual intercourse affected this increase in unintended pregnancy.

One thing that can be certain, however, is that trends in age at first sexual intercourse did not have anything to do with the increase and subsequent decline in teen pregnancies in the 1990s. We know this because there was virtually no change in age at first sexual intercourse among cohorts who reached their teens before, during, and after the 1990s.

Data Sources and Methods

Cycles 1 and 2 of the NSFG surveyed women who had ever been married, which would give biased results for this analysis, so I did not include them. Cycle 3 of the NSFG surveyed women regardless of marital status, but it had a very high rate of missing answers to the question about age at first sexual intercourse,2 so I elected not to use it. Therefore, I use Cycle 4 and onward for this analysis.

Men were first surveyed in Cycle 5 of the NSFG, which is why I have results for more cohorts of women than men.

Because this analysis does not specifically pertain to pregnancy outcomes, I used the analysis weights provided with the PUFs instead of weights adjusted for abortion under-reporting.

Later iterations of the NSFG bottom-code the answers for age at first sexual intercourse to 10 years old for disclosure risk reduction.3 For consistency, I performed this bottom-coding retroactively on earlier iterations of the NSFG in this analysis.4 However, the number of people who have sexual intercourse before age 10 is minimal, so this should not change the substance of the results.

The month and year of a respondent’s birth were available in the PUF for earlier iterations of the NSFG, but this information was removed from the PUF for disclosure risk reduction starting in the 2011-2013 iteration of the NSFG. For 2011 and onward, I calculated a presumptive birth year for respondents based on their age at the time of their interview and the month that their interview occurred.

There are issues with the 2022-2023 iteration of the NSFG, specifically with the responses that used the new online mode instead of traditional face-to-face interviewing, so I only use face-to-face responses from the 2022-2023 NSFG. I will address the issues I found with the 2022-2023 NSFG in a separate article.

The distributions of ages at first sexual intercourse for face-to-face respondents to the 2022-2023 NFSG look very similar to the distributions for respondents to the 2015-2019 NSFG, so the face-to-face responses, at least, should be fine.

In the two plots above, the [1980,1985) cohort estimates are taken from the 2015-2019 NSFG, and the [1985,1990) estimates are from just the face-to-face responses to the 2022-2023 NSFG.

Conclusion

The average age that American women first had vaginal sexual intercourse with a man gradually got younger for women born from 1945 to 1960. For generations of both women and men born from 1960 onward, the average age of first heterosexual, vaginal intercourse has remained the same (mean ~17.5 years, median ~16.5 years).

The shapes of the distributions of ages at first sexual intercourse have been consistent throughout this whole period. Generally, most Americans engage in sexual intercourse for the first time in their teens, but a substantial minority (~20%) do not do so until their twenties.

Trends in age at first sexual intercourse might have somewhat affected the soaring of unintended pregnancy rates in the 1970s and early 1980s, but had no effect on changes in unintended pregnancy rates thereafter because there has been virtually no change in age at first sexual intercourse among recent generations.

Appendix: Plots for All Birth Cohorts

Women

Men

In the previous article on U.S. public opinion about abortion1 in the 1960s, we saw that attitudes toward the legalization of abortion were divided into three tiers depending on the circumstances leading to abortion.

Over 70% of Americans supported the continued legalization of abortion when it was necessary to preserve the health of the mother against serious injury. However, over 70% of Americans rejected the legalization of abortion in elective cases such as a couple not having enough money, a woman being unmarried, or a couple simply not wanting more children. In the middle, a simple majority of Americans supported the legalization of abortion when the child was likely to be severely deformed and when the pregnancy was a result of rape.

Public opinion about abortion in the United States has changed since the 1960s. It did so during two discrete periods of time. The first occurred during the late 1960s and early 1970s. The second occurred from the turn of the millennium to the present, but only in a subset of the population.

Survey Quality

In my previous article on public opinion during the 1960s, I lamented the lack of data on the subject. After the 1960s, we have the opposite problem. From the 1970s on, there have been many polls about public opinion on abortion in the United States. For a comprehensive summary, see the review by Bowman and Goldstein (2021).

Most of these polls aren’t worth much. The fundamental problem is that survey responses to questions about abortion — like a lot of other controversial issues — are very sensitive to how the questions are framed and worded. Indeed, as we shall see, results are sensitive even to the order in which questions are asked. Therefore, results vary considerably from one bad poll to the next.

For a survey on public opinion about abortion to be useful for analysis, it needs a few characteristics:

• The survey must be repeated over an extended period of time using the same questions.

  The first time a survey is administered, it only establishes a baseline. It tells us what results we get from the particular wording used in the survey, which is not comparable to other surveys because even slightly different wording can result in large differences.

• Questions must be specific.

  This dismisses another swath of bad polling. For instance, asking people whether they consider themselves “pro-choice” or “pro-life” is pointless because respondents have wildly different ideas of what it means to be “pro-choice” or “pro-life.”

  Another genre of non-specific questions involves questions that presume a linear scale. These are of the “a lot more,” “a little more,” “a little less,” or “a lot less” form. It is dumb to ask respondents “how much” abortion restriction they support because people have strong opinions on the validity of abortion based on the specific circumstances.

  For instance, if the modal American respondent of the 1960s was living in the modal American state of the 1960s, that respondent would want restrictions of abortion loosened or lifted in the cases of severe fetal deformity and pregnancy from rape simultaneously with wanting restrictions maintained or even strengthed in elective cases.2

• Samples must be representative.

  Since public opinion analysis is concerned with what proportion of a population gives a specific response to a question, it is important for the sample to accurately reflect the population. It is well-known how to do this. You need a sampling frame that enumerates the target population (and only the target population). Then, you need to randomly select from that frame with a known probability of selection for each respondent.

  While this is a solved problem in theory, actually doing it is another thing. In practice, it is very expensive and will become even more expensive as survey response rates continue to plummet.

  Because of low response rates, care must be taken to adjust results for response bias. Typically, this involves weighting adjustments to make estimates match known population totals for key variables. However, this isn’t always done.3

General Social Survey

The only data source for analyzing trends in public opinion about abortion in the United States that meets all these criteria is the General Social Survey (GSS), which is run by the National Opinion Research Center (NORC).4 NORC has run the GSS since 1972, and the GSS includes questions about abortion attitudes in its core module. Therefore, the GSS has repeatedly asked the same questions about abortion attitudes every other year or so since 1972. In fact, NORC actually did a survey with the same questions about abortion attitudes in 1965, so we have data going back to before the GSS started.

The GSS is by no means perfect. For instance, its core module only asks questions about the legalization of abortion by circumstance. Another large factor in public opinion on abortion is the timing of the abortion (e.g., first versus second versus third trimester). Such questions have only been rarely asked on the GSS.

Another issue is that the GSS core only asks about abortion policy. People have dramatically diverse views about abortion outside of the public policy context, e.g., what they think about abortion from a moral standpoint, whether or not they would themselves seek an abortion, whether they would help others procure an abortion, etc.

The biggest issue with the GSS for this analysis, however, is that it doesn’t cover most of the time period when the first change in public opinion about abortion occurred. The 1965 survey by NORC gives us a snapshot of public opinion before this change, and the GSS catches the tail end of this change, but that is it.

Another issue is that NORC changed from in-person interviews to online questionnaires for the 2021 GSS and then to a mixture of in-person interviews and online questionnaires for the 2022 GSS. NORC advises:

Changes in opinions, attitudes, and behaviors observed in 2021 and 2022 relative to historical trends may be due to actual change in concept over time and/or may have resulted from methodological changes made to the survey methodology during the COVID-19 global pandemic. Research and interpretation done using the 2021 and 2022 GSS data should take extra care to ensure the analysis reflects actual changes in public opinion and is not unduly influenced by the change in data collection methods. For more information on the 2021 and 2022 GSS methodology and its implications, please visit https://gss.norc.org/Get-The-Data.

In this article, I do not investigate whether the methodological changes affected the abortion attitudes results from the 2021 GSS and the 2022 GSS. It isn’t critical to this article’s analysis, but it is good to be aware of.

Gallup

This analysis uses Gallup polling to supplement the GSS. Gallup’s polling is not as good as the GSS for various reasons.5 However, Gallup has asked more questions about abortion attitudes repeatedly than the GSS.

In addition, Gallup polling about abortion took place during the critical 1968-1975 time period. The results of these polls, however, are not available on Gallup’s public website. Sociologist Judith Blake commissioned the 1968-1975 polls, and it appears the only publicly available results from these polls are whatever she happened to publish in academic journals.

National Fertility Survey

The National Fertility Survey (NFS) was administered in 1965, 1970, and 1975, and it included questions about attitudes toward abortion. Unlike the GSS and Gallup polls, the NFS target population comprised women who had ever been married rather than all adult Americans.

In fact, the 1975 NFS target population is even stranger. The survey tried to create a 1975 sample that revisited women from its 1970 sample. Thus, the 1975 target population became continuously married White women whose duration of marriage was less than 25 years, who married at age 25 years or less, and whose husbands had only been married once.

Therefore, the 1975 NFS does not reflect a cross-section of American ever-married women, and we should view any results from it with skepticism.

1968 to 1973

Gallup/Blake

The Gallup polls from 1962 to 1977 that were done at the request of Judith Blake (henceforth “Gallup/Blake” polling)6 started off asking three questions:

Do you think abortion operations should or should not be legal in the following cases:

a. Where the health of the mother is in danger?

b. Where the child may be born deformed?

c. Where the family does not have enough money to support another child?

From 1968 on, a fourth question was added.

d. Where the parents simply have all the children they want although there would be no major health or financial problems involved in having another child?

Blake’s 1971 paper only reported on the opinions of White people. (See previous discussion about this.) However, her 1977 paper expanded results to include all Americans, and it is unlikely the racial exclusion in the 1971 paper changed the results substantially.

The 1971 paper only reported disapproval rates, whereas the 1977 paper reported both disapproval and approval. However, the 1977 paper only reported on the responses to the question about when “parents simply have all the children they want.”

There appears to be a downward trend in disapproval from 1968 on for the “no more children” scenario. However, the question was first asked in 1968, so we can’t be sure if this downward trend started earlier. There might be a smaller downward trend in the disapproval rate for the other questions between 1965 and 1968.

Therefore, the change in public opinion toward elective abortion started at least by 1968 and perhaps as early as 1965.

The change in public opinion appears to level off by 1974. In 1974, 31% of Americans approved of the legalization of abortion in the “no more children” case, and in 1977, 30% did.

Even after this change in public opinion, in 1974, 63% of Americans disapproved of the legalization of abortion when parents simply had all the children they wanted, and in 1977, 63% did again. When the question was first asked in 1968, the disapproval rate was 85%.

Therefore, according to Gallup/Blake polling, there was a change of about 22 percentage points in American attitudes toward elective abortion during the 1968-1974 time frame before leveling off. Even with this change, the legalization viewpoint was the minority position.

National Fertility Survey

The NFS confirms the Gallup/Blake polling. For all six questions7 about abortion attitudes asked in the three iterations of the NFS, there was a trend of increasing approval and decreasing disapproval.

Since the 1975 NFS is not a representative cross-sectional sample, its results should not be taken as reflecting the proportion of attitudes in the population. That being said, it is directionally consistent with the Blake/Gallup polling as approval of abortion in elective cases remains a minority position.

The timing of the NFS in 1965 and 1975 means that these results do not narrow the time frame for the change in public opinion. We already knew from the Blake/Gallup polling that it started as early as 1965, had started no later than 1968, and ended by 1974.

1965 NORC Survey and the General Social Survey

If we take the results of the survey done by NORC in 1965 together with the GSS during the 1970s, the combined results confirm there was a trend of increasing approval of legalized abortion in response to all six questions.8

Again, these results do not help identify the beginning of the trend. We already knew it might have been as early as 1965.

The GSS shows a distinct leveling off of the trend by 1973, which is only slightly earlier than the 1974 inflection point in the Blake/Gallup results.

Approval of elective abortion appears to plateau at a much higher rate in the GSS results than in the Blake/Gallup results. In 1973, a peak of 47% of Americans approved of legalized abortion in the “married and does not want any more children” case on the GSS, which is the closest to the “parents simply have all the children they want” case in the Blake/Gallup results where approval peaked at 31% in 1974.

The 1965 NORC survey found that 16% approved of the legalization of abortion in the “parents simply have all the children they want” case. The 16% to 47% change in approval in the NORC/GSS results would represent a 31 percentage point change, larger than the 22 percentage point change found in the Blake/Gallup polling.

Blake (1977) explains the difference between the NORC/GSS and the Blake/Gallup results as due to differences in wording and question order. The Blake/Gallup question explicitly stated that there were “no major health or financial problems involved in having another child,” and it was asked after three questions about health and financial issues.

In contrast, the “married and does not want any more children” question in the NORC/GSS surveys was the second question, after only the “serious defect in the baby” question and before questions about health and financial issues.

Thus, respondents to NORC/GSS surveys might think of health or financial issues when answering the “married and does not want any more children” question, but respondents to the Blake/Gallup “no more children” question almost certainly did not think of these extenuating circumstances.

The NORC/GSS results, therefore, overestimate support for the legalization of elective abortion somewhat. Exactly how much, we cannot know.

Like the other sources, support for the legalization of elective abortion plateaus in the GSS results as minority viewpoints, except for the low-income case, which plateaus at just above 50%.

Summary

There was a broad increase in public approval of the legalization of abortion that began at least by 1968 — perhaps as early as 1965 — and ended in 1973 or 1974. For elective abortion, this change was somewhere around a 22 to 31 percentage point increase.

Even with this trend, the viewpoints that were in the majority remained the same. A majority of Americans approved of the (continued) legalization of abortion in cases of serious health risks to the mother, high likelihood of severe fetal deformity, and pregnancy resulting from rape. A majority of Americans disapproved of the legalization when a woman was unmarried or a couple simply did not want another child.9

Despite not resulting in a change in which viewpoints were in the majority, the 1968-1973 change made abortion politics much more contentious. In the 1960s, the legalization of abortion in elective cases was a political non-starter. In the 1970s, a sizable minority supported the legalization of elective abortion.

Women’s Opinions

The 1970s continued the phenomenon from the 1960s that women were more opposed to the removal of abortion restrictions than men in the United States. The overall difference in both decades was small, but it could be more substantial in certain subdomains or on certain questions.

In the Gallup survey of January 1973, more women than men believed that abortion should be allowable only if the pregnancy was at 3 months or less. Indeed, more women volunteered that abortion should never be performed than selected the response option that there should be no legal restriction on the timing of abortion.

This result was confirmed by the Gallup survey of April 1975, which found that more women (73%) than men (58%) disapproved of a law that would permit women to have an abortion even if they were more than three months pregnant.

When asked “when human life begins” in the January 1973 and April 1975 surveys, more women than men answered “at conception,” whereas more men than women answered “at birth.” (Overall, pluralities of both women and men replied “at conception.”)

Roe v. Wade and Doe v. Bolton

Public opinion was overruled in the 1973 U.S. Supreme Court opinions Roe v. Wade and Doe v. Bolton, which removed the issue of abortion from the normal political process. Roe v. Wade overturned all laws prohibiting abortion for any reason up until fetal viability, which was stipulated to occur at some point during the seventh month (24 to 28 weeks) of pregnancy.

Roe and Doe occurred at a local maximum of public approval of elective abortion cases and right when the change in public opinion plateaued. Asking respondents about the circumstances under which they approved of abortion was suddenly a lot less relevant. Roe had legalized abortion-on-demand, i.e., abortion whenever it was requested, without any criteria about the circumstances that led to the request.

The GSS added a question asking specifically about abortion-on-demand in its 1977 iteration, a few years after Roe v. Wade. The question reads:

Please tell me whether or not you think it should be possible for a pregnant woman to obtain a legal abortion if the woman wants it for any reason?

The 1977 GSS found that the legalization of abortion-on-demand had only 37% approval.

Judging by how flat the support for legalized abortion in the elective cases was between 1973 and 1977, it is unlikely that support for abortion-on-demand was much higher in 1973 than in 1977. Therefore, when the Roe v. Wade opinion was issued, only about 37% of Americans approved its policy implications. This 37% estimate is closer to the 31% approval Blake/Gallup found for the “no more children” case than other GSS results from the same time frame.

In contrast, about 61% of Americans rejected the legalization of abortion-on-demand.

Public opinion was actually worse for Roe v. Wade’s policy implications than this 61%-to-37% split because, as we have seen, most Americans rejected the legalization of abortion after the first trimester, as well, and Roe legalized abortion-on-demand until the early third trimester.

1977 to 2000

We see in Figure 7 that, in the 1977 GSS, when the abortion-on-demand question was first asked, there was a clear ordering of support for legalized abortion in elective cases. Abortion in the case of low income had the highest approval, followed by the case of unmarried women, followed by married couples who do not want more children, and finally, abortion-on-demand.

From 1977 into the mid-1990s, approval of all of the elective abortion cases consolidated in such a way that the approval rate for any one case became similar to the approval rate for any other.

Approval of elective abortion reached a local maximum in the 1994 GSS in the high 40 percents. After this, there was a decline in approval of elective abortion to a nadir of around 37% in the year 2000.

2000 to 2022

According to GSS results, support for the legalization of elective abortion climbed steadily throughout the Twenty-first Century. It finally became a majority viewpoint sometime between 2018 and 2021.10

Overall, approval for the legalization of abortion-on-demand went from 36% in 2000 to 55% in 2022. This 19 percentage point change is smaller in magnitude than the 22 to 31 percentage point change that occurred in the 1968-1973 period, and it occurred over a longer period of time.

However, the 2000-2022 change has more implications. Support for the legalization of elective abortion went from a minority viewpoint to a majority viewpoint. Furthermore, the 2000-2022 change has been very partisan.

Partisanship

In the mid-1970s, Republican-leaners were slightly more likely to approve of the legalization of elective abortion than Democratic-leaners,11 but the difference was small. From the late 1970s through the 1980s, there was little to no partisan difference in attitudes toward elective abortion.

Starting in the 1990s, Democrats and Republicans separated in a process of partisan sorting, with Democrats approving elective abortion at a higher rate than Republicans. However, throughout the 1990s there was a decline in approval of elective abortion in both parties until the nadir of approval of elective abortion in 2000.

From the year 2000 on, Republicans have approved of elective abortion at more or less the same rate. However, approval of elective abortion by Democrats has skyrocketed during this same time period.

In 2000, about 29.8% of Republicans approved of the legalization of abortion-on-demand, and in 2022, it was about 29.4%. However, Democrats went from 41.3% approval of the legalization of abortion-on-demand in 2000 to a whopping 82.2% in 2022 — a difference of about 41 percentage points!12

In terms of magnitude, the 41 percentage point change among Democrats is substantially larger than the 22 to 31 percentage point change among all Americans from 1968 to 1973.

The effect on the Democratic Party cannot be understated. A position that was mildly unpopular among Democrats in 2000 has become extremely popular among Democrats in 2022.

Women’s Opinions

After I learned from reading Judith Blake that women in the 1960s and 1970s disapproved of legalized abortion at a higher rate than men, I asked myself, when did this change?

It turns out it didn’t. Women have never approved of legalized elective abortion at a higher rate than men.

The difference in approval of legalized abortion in the four elective cases between men and women is small, but generally, men approve of abortion at a higher rate than women.

There have been 33 iterations of the GSS that asked about abortion attitudes. The first 5 iterations asked 3 questions about elective abortion, and the remaining 28 iterations asked 4 questions. This means there are 127 comparisons in Figure 10.

In these 127 comparisons, the estimate of approval for legalized elective abortion was higher for men in 104 cases (82%) and women in 23 cases (18%).

However, I fit simple logistic regression models with abortion attitude as the response variable and sex as the explanatory variable for each of these cases. Even before adjusting for multiple comparisons, no comparison with a higher estimate for women than men was statistically significant with α = 0.05. Therefore, these cases are likely due to statistical noise.

Approval of legalized elective abortion has either been greater among men than among women or has been the same (within the statistical precision of the survey).

Gestational Age

How about the result found by Judith Blake that most Americans disapprove of laws permitting abortion after the first trimester?

This also hasn’t changed.

Gallup polling from 1996 to 2023 has found that a majority of Americans reply that “abortion should generally be legal” during the first trimester, and majorities reply that “abortion should be generally illegal” during the second and the third trimester. The gap has narrowed since 2011, but which response is in the majority has not changed.

Popularity of Roe v. Wade

At the same time that most Americans have not supported the actual policy implications of Roe v. Wade, most Americans have opposed repealing Roe v. Wade.

From 1989 to 2022, a majority of Americans have indicated on Gallup polling that they would not like to see Roe v. Wade overturned.

Discussion

The 1968-1973 increase in approval of the legalization of abortion follows an injection of funding for abortion advocacy by billionaire benefactors. In 1966, John D. Rockefeller III began funding Robert Hall’s Association for the Study of Abortion and Roy Lucas’ James Madison Institute, which lobbied for the legalization of abortion. During this time period, wealthy men such as Warren Buffet and Richard Mellon Scaife funded litigation efforts to legalize abortion. (Forsythe 2013, pp. 59-60)

The 1968-1973 increase in approval for legalized abortion thus may have been the consequence of a well-funded advocacy effort without an organized counteracting lobby. This hypothesis is consistent with the change in public opinion leveling off in 1973, when Roe and Doe provoked the creation of organized anti-abortion lobbying at a national level.

The effect of partisan sorting on public opinion about legalized abortion in the United States cannot be understated. The changes in public opinion after the year 2000 are extremely partisan.

During the 1990s, Democrats had higher approval rates for the legalization of elective abortion than Republicans for the first time. However, during partisan sorting in the 1990s, support for legalized abortion actually declined.

In the Twenty-first Century, it has been trendy for Democrats to accuse Republicans of “asymmetric polarization” and of becoming more “extreme.” It is clear that on abortion, the opposite is the case. Republicans are basically where they were after the partisan sorting in the year 2000, whereas Democrats have undergone a massive 41 percentage point change. Republicans in 2022 are closer to where Democrats in 2000 were on abortion than are Democrats in 2025.

The partisan sorting was likely a prerequisite for this. Grouping most of the abortion advocacy into one party created an echo chamber that allowed viewpoints to snowball. There has been a bias in mass media and academia since at least the 1970s in favor of abortion advocacy. Mainstream media and academia today are heavily aligned with the Democratic Party. The partisan sorting allowed a confluence between these phenomena.

Women Have Never Approved of Abortion Legalization at a Higher Rate than Men

When I went looking for the year that women started to support legalized elective abortion at a higher rate than men, I was genuinely surprised to find this never occurred. I was likely influenced by the mainstream media narrative that constantly repeats the refrain that “women want abortion.”

This result is even more surprising considering the gender polarization going on in the United States — in which women are more likely to vote Democratic and men more likely to vote Republican — and the polarization between parties on the issue of abortion. There is likely something going on here that is worth investigating in a subsequent article.

Democrats’ belief in their own false narrative likely contributed to one failure of the Harris-Walz campaign for U.S. President in 2024. The Harris-Walz campaign was criticized for not saying anything concrete about many issues. However, this criticism did not apply to abortion, as the Harris-Walz ticket prominently emphasized the issue of abortion throughout the campaign.

However, the Harris-Walz campaign performed worse among women voters in the 2024 election than the Biden-Harris campaign did in the 2020 election.

I suspect there is a minority of women for whom opposition to restrictions on abortion is their one main issue. As a result of this being their issue, this subset of women is quite loud and vociferous about their views on abortion. Furthermore, because of the partisan sorting, this subset of women is now virtually all in Democratic-leaning spaces.

Thus, in Democratic-leaning spaces, this subset of women is prominent and interpreted as representing “women” writ large. However, given how extensive party polarization is on abortion, most of the voters — women or men — who would vote for a Democrat because of abortion have already been Democratic-leaning voters for years now.13 Therefore, running on abortion is unlikely to get you additional votes in high-turnout elections like for the U.S. President.

Looking back at the first paragraph in this Discussion section, there are a lot of male names who funded the abortion lobby and who led the organizations lobbying for abortion in the 1960s and 1970s. Roe v. Wade and Doe v. Bolton — the great win of the abortion lobby — were the opinions of seven men. Maybe I shouldn’t have been so surprised that legalized abortion has been more popular among men than women.

Repudiation of Gallup’s “Circumstances” Question

I partially retract an analysis from an earlier article in which I wrote:

Democrats as a party have been captured by what is a minority viewpoint among all Americans. Approximately 60% of Democrats think that abortion should be legal in all circumstances, compared with only 34% of Americans who think that abortion should be legal in all circumstances.

This is based on a Gallup polling question about whether abortion should be “legal in all circumstances,” “legal in some circumstances,” or “illegal in all circumstances.”

I now think it was a mistake for me to even care about the results of this question. As I discussed in the “Survey Quality” section of this article, these kinds of questions are too non-specific to shed much light on public opinion.

I think the 55%-42% split in favor of legalized abortion-on-demand in the 2022 GSS results is more relevant because it is based on a specific policy question. Indeed, this 55%-42% split is closer to the voting results on several state abortion referenda, such as in Ohio and Florida.

I suspect a lot of the people who pick the “some circumstances” option in Gallup polling are thinking of limiting abortion to the first trimester of pregnancy since, as we have seen, majorities in the same Gallup polling think that abortion in the second or third trimesters should generally be illegal.

Of course, the abortion referenda that recreate Roe and Doe legalize abortion-on-demand throughout the second trimester and into the early third trimester. It may be the case that some voters simply value the legality of first-trimester abortions over the illegality of second-trimester and third-trimester abortions.

Indeed, the state of Nebraska defeated its referendum recreating Roe and Doe and passed a counter-referendum in its 2024 election. Nebraska currently has generally legal abortion in the first trimester, which may have undercut support for the referendum to recreate Roe and Doe.

This being said, my main thesis from the earlier article is confirmed by the analysis of this article: Abortion is a get-out-the-vote issue for Democrats, not an issue Democrats can use to win persuadable voters, because party sorting by views on abortion is already nearly complete.14

The Policies of Roe v. Wade Never Had Majority Support

A majority of Americans finally supported the legalization of abortion-on-demand shortly before the Dobbs v. Jackson Women’s Health Organization decision of 2022 overturned Roe v. Wade, which is what had legalized abortion-on-demand (in 47 states) in the first place.

Furthermore, during the entirety of the time between Roe and Dobbs, a majority of Americans opposed the legalization of abortion in the second trimester and the third trimester, and Roe legalized abortion-on-demand in the second trimester and part of the third trimester.

Thus, Roe and Doe were a political coup d'état without which the abortion lobby could not have been so successful. The legalization of abortion-on-demand was supported by only 37% of Americans in 1977 and, again, by only 37% of Americans in 2000. In 2022, with abortion policy finally becoming a subject of legislation again, support had risen to 55% of Americans.

The strategy that Judith Blake outlined and that the abortion lobby took in the 1970s was not to make public policy match public opinion on abortion, but to make public opinion match public policy — after changing public policy by way of judicial activism. Thus, politics has been a tool of the abortion lobby to change public opinion.

This has worked. The referenda enshrining abortion in the various state constitutions have been retreads of Roe and Doe, sometimes using the exact same wording. Unlike Roe and Doe, which were opinions of seven unelected judges, these referenda have passed by popular vote. Roe and Doe gave 45 years of cover to the abortion lobby to make this work.

If Planned Parenthood v. Casey in 1992 had overturned Roe and Doe instead of reaffirming their core holdings, the political landscape would have been very different. Casey occurred right before a downward trend in approval of abortion-on-demand that only started to reverse in 2000.

It is curious that a majority of Americans simultaneously opposed Roe’s policies and opposed its overturning. I speculated in the past that this was in part due to many Americans not understanding what policies Roe actually created.

I have seen numerous people, including book authors and journalists, wrongly state that Roe legalized abortion-on-demand in just the first trimester when it actually legalized abortion-on-demand throughout the first and second trimesters and into the beginning of the third trimester. Indeed, sometimes, polling questions themselves misrepresent Roe in this way. (Bowman and Goldstein, 2021)

Some of this false belief might be due to Roe legalizing abortion-on-demand until “viability.” Some people do not know when viability occurs.15

Another major reason for the seemingly contradictory simultaneous opposition to Roe’s policies and to overturning Roe is that some people prefer to maintain the status quo. I have seen numerous people assert that Roe and Doe were “long-standing” or “established” and so should continue.

Gallup didn’t start asking about overturning Roe until 1989, more than 16 years after Roe, by which time it was presumably “established.” In 1989, 57% of Americans were against the legalization of abortion-on-demand, but 58% of Americans were also against overturning Roe.

Regardless of what accounts for it, the difference between support for legalized abortion-on-demand and support for overturning Roe has been critical in the abortion lobby’s program of changing public opinion through policy.

Conclusion

In the 1960s, a majority of Americans approved of the legalization of abortion when it was necessary to preserve the health of the mother, when a pregnancy was the result of rape, or when the child was likely to be severely deformed. At the same time, most Americans rejected the legalization of abortion in elective cases, such as when a woman was unmarried or when a couple didn’t want another child.

Starting sometime between 1965 and 1968, public opinion began to change toward greater approval of legalized abortion. By 1973, this trend plateaued, and approval remained at the same levels in subsequent years before dropping again. During this plateau, a greater proportion of Americans approved of the legalization of elective abortion than during the 1960s, but this remained a minority viewpoint.

In the 1990s, there began to be a difference in viewpoints on the legalization of elective abortion based on political party affiliation, with Democratic-leaning voters approving of elective abortion legalization at a higher rate than Republican-leaning voters. Still, approval of elective abortion legalization declined throughout the 1990s, reaching a nadir in the year 2000.

From the year 2000 on, the rate of approval of the legalization of elective abortion among those who tend to vote Republican has remained more or less the same. At the same time, the rate at which Democratic-leaning voters approve of the legalization of elective abortion has skyrocketed. Approval of legalized abortion-on-demand was a minority position among Democrats in 2000. By 2022, it was the view of over 80% of Democrats.

Because of this dramatic change in viewpoints among those who tend to vote Democrat, and because of the lack of a change in viewpoints among those who tend to vote Republican, support for the legalization of abortion-on-demand went from a minority viewpoint to a majority viewpoint sometime between 2018 and 2021, shortly before Dobbs overturned Roe v. Wade.

References

Blake, J. (1971). Abortion and Public Opinion: The 1960-1970 Decade. Science, 171(3971), 540–549. https://doi.org/10.1126/science.171.3971.540

Blake, J. (1977). The Supreme Court’s Abortion Decisions and Public Opinion in the United States. Population and Development Review, 3(1/2), 45–62. https://doi.org/10.2307/1971759

Bowman, K., & Goldstein, S. (2021). Attitudes about Abortion: A Comprehensive Review of Polls from the 1970s to Today. https://www.aei.org/wp-content/uploads/2021/10/Attitudes-About-Abortion.pdf?x91208

Evers, M., & McGee, J. (1980). The trend and pattern in attitudes toward abortion in the United States, 1965–1977. Social Indicators Research, 7(1), 251–267. https://doi.org/10.1007/BF00305601

Forsythe, C. D. (2013). Abuse of discretion: The inside story of Roe v. Wade. Encounter Books.

Those who scratch this itch are prone to protest, “Why not? Why is it out of bounds to use the word ‘fascist?’ Why can’t I say ______ is a ‘fascist’ when he/she is?”

It is not offensive to use the word “fascist.” It does not violate some arbitrary rule of social decorum to use the word “fascist.”

Describing your political enemies as “fascist” signifies stupidity. That is the problem. The problem is not that it offends the delicate sensibilities of those around you. The problem is that it makes you look like an idiot, and no one wants to be associated with idiots.

To understand this, we need to take a walk through history through the period when the use of the “fascist” slur arose in the 1920s.

Many — if not most — of those who enjoy calling their political enemies “fascist” do so out of the belief that they are guarding against the next Adolph Hitler. However, the inventors of the “fascist” slur failed to identify the literal Adolph Hitler as a threat.

Marxism, a Very Brief Review

Since this article is written for a general audience, here is a speedy summary of some concepts so that the quotes in the rest of the article will make sense.

If you are old enough to remember communism, you might want to skip this section. Otherwise, bear with me.

Marxism is a kind of socialism named after Karl Marx. Typically, “Marxism” and “communism” are synonyms, though some Marxists would say that organized communism deviated from Marxism at some point.

Marxists believe a class struggle exists between the proletariat (the workers) and the bourgeoisie (the capitalists). According to this ideology, capitalism is a system of oppression in which the bourgeoisie immiserate the proletariat. The bourgeois class does this by owning the means of production (e.g., factories) while not doing anything valuable.

As this thinking goes, capitalism is inherently unstable because of this, and eventually, capitalism will collapse. This leads to a revolution in which workers will seize the means of production, eliminate the useless bourgeois class, and establish a dictatorship of the proletariat.

Even though Marxism is just one kind of socialism, it came to dominate the socialist scene of the early twentieth century. Marxism believes that it is the “true” socialism and that other kinds of socialism are imperfect ideologies that would become Marxism if they were developed further.

The main takeaways for interpreting the quotes in this article are:

• Whenever a communist refers to “workers” or “proletariat,” those are the good guys.

• Whenever a communist refers to “capitalists” or “bourgeoisie,” those are the bad guys.

The major players in our story representing communism are:

• the Communist International (Comintern), an international association of political parties that existed from 1919 to 1943

• the Communist Party of Germany (KPD), which also started in 1919 and technically existed until 1946

Social Democracy, an Even Briefer Review

The Social Democratic Party of Germany (SPD) was founded in 1875 as a merger of various smaller socialist parties. It had a strong Marxist influence throughout its early decades but gradually moderated its platform. It is still a popular party in Germany today.

Unlike communism, which seeks a violent cataclysm leading to a dictatorship of the proletariat, the term “social democracy” became associated with moderate attempts to achieve some of the same ends as Marxism.

Social democracy is typically willing to use the existing system of constitutional norms and democratic processes to redistribute wealth and alleviate workers’ immiseration.

The SPD became the dominant political party of the Weimar Republic — the government of Germany between the end of World War I and the rise of the National Socialist German Workers’ Party (NSDAP) in Germany. Because of the SPD’s moderation away from revolutionary Marxism, the SPD’s platform during the Weimar Republic was more akin to the Labour Party’s in the United Kingdom or the New Deal in the United States.

Capital-“F” Fascism

Before “fascist” became a slur, “Fascist” described a political movement in Italy.

The word “fascio” had been used in Italian politics since the 1890s non-specifically for various radical or revolutionary organizations. In English, it literally translates to “bundle” and thus means a political “group” figuratively.

With a capital “F,” it became associated with Benito Mussolini’s political career and Giovanni Gentile’s political philosophy.

Before his ascent in politics, Mussolini was a journalist and a member of the Italian Socialist Party. He was editor of La Lotta di Classe (The Class Struggle) and then Avanti! (After You!)

Italian socialism of the time believed itself to be part of the international workers’ movement. World War I, which was perceived to be a clash of nation-states, was not something good internationalists wanted their country to be part of. Mussolini, however, had a strong sense of Italian nationalism.

This is what caused Mussolini to break with organized Italian socialism. He founded the Fasci d’Azione Rivoluzionaria in 1914 to promote Italian entry into World War I. After World War I, Mussolini founded the Fasci Italiani di Combattimento in 1919 and then the Partito Nazionale Fascista, which took power in Italy during the March on Rome in October 1922 when Mussolini was appointed Prime Minister.

In terms of outlook, capital-“F” Fascism positioned itself as a “third way” in opposition to both liberal democratic capitalism and international communism. It broke with the latter by being overtly nationalist instead of internationalist and sought a peaceful resolution of class conflict in which the workers would be represented in trade unions, typically called “syndicates.”

Thus, capital-“F” Fascism was self-described as “Italian nationalist syndicalism with the philosophy of Actualism.” (This last part was a nod to Giovanni Gentile’s neo-Hegelian philosophy.)

Needless to say, the Comintern was not amused.

Mussolini's success brought the subject of fascism sharply to the attention of the Communist International, which had previously given it little consideration. An Italian Commission was set up at the Comintern's Fourth Congress in November-December 1922, and its resolution referred to the fascists as “the most radical wing” of the bourgeoisie. But the old Italian Socialist party was blamed most for Mussolini's victory. “The real forerunner of fascism was reformism,” the resolution declared. “The treachery of the reformists is primarily responsible for the great sufferings of the Italian proletariat.” (Draper 1969)

As far as the Comintern was concerned, Mussolini was an apostate — like someone who grew up a Christian or a Muslim and then became an atheist or a Buddhist later in life. Like religious fundamentalists who blame watered-down forms of their religion when their children leave their faith, the Comintern blamed moderate forms of socialism — the “reformists” — as leading to this apostasy.

Small-“F” “Fascists”

A year after Mussolini’s successful March on Rome, the KPD staged the German October, an unsuccessful communist uprising in 1923. The KPD sought the help of elements of the SPD, but the SPD did not give such help, and much of the uprising was aborted. In response to what uprisings did occur, the SPD-led government responded with force. Dozens of people were killed.

Grigory Zinoviev, Chairman of the Comintern, deployed the f- word in 1924:

The leading strata of the German Social Democrats are at the present moment nothing else than a fraction of German fascists under a socialist mask . . . It is not just now that the German social democrats have passed over to the side of capitalism. At the bottom, they have always been the class enemies of the proletariat.1

Josef Stalin, General Secretary of the Communist Party of the Soviet Union, added to Zinoviev’s assessment:

Fascism is a fighting organisation of the bourgeoisie dependent upon the active support of Social Democracy. Objectively, Social Democracy is a moderate wing of fascism. They are not poles apart, but immediate neighbours.2

Previously, the KPD and the SPD tried to work together to form a “united front” of socialism in Germany. After the events of the German October, the SPD would be called “social fascists” instead of “social democrats” by the Comintern and the KPD.

The first to put the words “social” and “fascism” together in that order was apparently Heinz Neumann, one of Stalin's early German protégés. In the Comintern's German organ, dated October 7, 1924, Neumann contributed an article subtitled “The Newest Form of Social Fascism in Germany.” (Draper 1969)

This is the origin of “fascist” with a small “f” as a slur.

Once demagoguery starts, it tends to keep spreading. Zinoviev called lots of moderate socialists “fascists.”

What are Pilsudski and the others? Fascist Social-Democrats. Were they this ten years ago? No. It goes without saying that they were already then fascists in nuce. But they have become fascists precisely because we are living in the epoch of revolution. What is Italian Social-Democracy? It is a wing of the fascists; Turati is a fascist Social-Democrat. Could this statement have been made five years ago? No. Think of a group of academicians who gradually developed into a bourgeois force. Italian Social-Democracy is now a fascist Social-Democracy. Take Turati, D'Aragona, or the present Bulgarian government Socialists. They were opportunists, but could one say ten years ago that they were fascist Social-Democrats? No, that would have been stupid then. Now they are that.3

Who is that over there? “Fascist!” How about this guy? “Fascist!” Have you met— “Fascist!” It turns out there were “fascists” suddenly everywhere in 1924.

Józef Piłsudski was the Polish armed forces commander and head of state in Poland when the Polish Socialist Party was in power. Augusto Turati and Ludovico D’Aragona were Italian socialist politicians. Turati would convert to Fascism with a capital “F,” and D’Aragona would not, instead fleeing to Paris. “Bulgarian government Socialists” was conveniently not specific.

The word “fascist” was quickly becoming a blanket term for anyone who was a political enemy of the Comintern. It seemed especially useful for the demagoguery of other socialists. Still, Zinoviev also applied it to Ramsay MacDonald, Prime Minister of the United Kingdom, and Hans von Seeckt, Commander-in-Chief of the German Army.

By applying it to figures as far apart as General von Seeckt, Marshal Pilsudski, Turati, and Ramsay MacDonald, Zinoviev made social-fascism into little more than a catch-all for Communism's enemies and opponents from moderate Left to far Right. (Draper 1969)

Zinoviev’s successor as leader of Comintern was Nikolai Bukharin, who tried to back away from demagoguery of the SPD as “fascist” a little.

The bourgeoisie resorts either to the method of fascism or to the method of coalition with Social Democracy according to the changes in the political situation; while Social Democracy itself often plays a fascist role in periods when the situation is critical for capitalism. In the process of development Social Democracy reveals fascist tendencies which, however, does not prevent it, in other political situations, from acting . . . against the bourgeois government in the capacity of an opposition party.4

Thus, Bukharin’s interpretation was that the SPD could be “fascist” bad guys or “against the bourgeois government” good guys, depending on the situation.

“Class against Class”

Still, the Ninth Plenum of the Comintern in February 1928 adopted a “class against class” policy that declared there were two only classes — the proletariat and the bourgeoisie — and they were in exclusive conflict. You can either be with us or against us.

Communists could no longer support moderate parties. However, this was mainly to the detriment of communist parties like the KPD because they weren’t that popular.

In 1929, the British Communist party claimed no more than 4,000 members against over 3,000,000 for the Labour party. Some British Communist leaders were understandably reluctant to cut themselves off from the Labour party and to pretend that they, not the Labourites, represented the British working class. …

Even in Germany, which had the largest and strongest Communist party outside Russia, “class against class” meant that the Communists consigned to an enemy class the organizations which contained the vast majority of workers. In 1930, for example, the German Communist party reported a membership of 135,808; the German Social-Democratic party, 1,021,777.

The Communist trade-union opposition claimed a following of 136,000. But the “free trade unions,” associated with the Social-Democratic party, contained 4,716,569 members; the so-called salaried employees (angestellten) unions, 1,620,970; the Christian (Catholic) trade unions, 778,863; and another group (Deutsche Gewerkvereine), 163,302.

The Communist vote was about half that of the Social-Democrats—4,592,100 to 8,577,700. Even at the Communist high point and Social-Democratic low point in November 1932, the latter's vote was still considerably larger—7,248,000 to 5,980,200. Many more workers, of course, voted for the Catholic Center and other parties.

The theory of social-fascism, then, put by far the largest number of organized workers into quasi-fascist parties which were so far gone that, for the Communists, there was only one thing to do—destroy them. (Draper 1969)

The conflict between the KPD and SPD worsened after the Blutmai (Bloody May) incident in May 1929. In response to increasing tensions, the police chief of Berlin banned all public gatherings on International Workers’ Day. The KPD went ahead with gatherings anyway, leading to an immediate crackdown by police that led to days of violence. Again, dozens were killed.

Hitler Not the Main Threat?

During this period in which the KPD was fighting the SPD, Hitler and the NSDAP were active. However, according to Comintern, the NSDAP wasn’t the real threat. The SPD was.

It would, however, be incorrect to conclude from this, that Germany is directly faced with the establishment of a fascist government a la Mussolini. Even fascist methods are subject to the changes of time and circumstances, i.e., to the development of capitalism, and are adapted to the economic and political situation of the country in question. The great change that has taken place is the growth of fascism within social-democracy, and in German social-democracy particularly the German capitalists have found a strong support with increasingly definite fascist tendencies.5

In March 1930, the SPD lost control of the German government to the (Catholic) Center Party under Heinrich Brüning. The Brüning government lasted until June 1932.

The Reichstag elections in July of [1932] gave the Nazis 230 seats, a gain of 123, and doubled their popular vote. The Social-Democrats went down to 133 seats, a loss of ten. The Communists gained twelve, from 77 to 89. Without the Nazis or Communists, a majority government had become impossible, and neither could or would take part. Brüning had already given way to [Franz] von Papen who was willing to take in Hitler as vice-chancellor, and Hitler would not settle for anything less than total power. (Draper 1969)

The Comintern made some preliminary admissions that it had made a mistake in concentrating its animosity almost exclusively against the SPD and not the NSDAP. Still, it continued to insist that the SPD was the main enemy.

At this juncture, in September of that cursed year, the Twelfth Plenum met in Moscow. According to the Comintern's spokesman, Kuusinen, the “revolutionary upsurge” had moved on to an even higher stage. The only united front was the “united front from below” which he defined as one between the “Communist vanguard” and the non-revolutionary masses for the purpose of isolating the Social-Democratic “agents of the bourgeoisie.”

He admitted that “for a long time, the Communist party of Germany underrated the National Socialist movement; and in part neglected to struggle against it.” But social-democracy was still “the main social support of the bourgeoisie” and “we ought to direct our main offensive against social-democracy.” The only concession he would make was that this offensive should be waged “in such a way that we may win over the Social-Democratic workers.”

In his concluding remarks, four months before Hitler took power, he reiterated: “The main blow, as I have already stated in my report, must in the present period of preparing for the revolution be directed against social-fascism and the reformist trade-union bureaucracy.” (Draper 1969)

Even with the SPD out of power, the KPD continued to insist that the SPD was a more significant threat than the NSDAP as late as December 1932.

The decline of the Social-Democratic party in no way reduces its role as the main social buttress of the bourgeoisie, but on the contrary, precisely because the Hitler party is at present losing followers from the ranks of the workers, instead of penetrating still more deeply into the proletariat, the importance of the Social-Democratic party for the fascist policy of finance capital increases.6

Franz von Papen would quickly be brushed aside, and the government would be taken over by the NSDAP. Hitler was appointed Reichskanzler on January 30, 1933.

Thus, the “fascist” slur was not used to prevent the rise of Hitler. It was used by communists who were so hostile to a more moderate workers’ party that they concentrated their hostility toward the moderates instead of the literal Nazis.

The outcome of the German Communist’s adoption of the theory of social fascism was catastrophic for the party, as well as the German working class. The increased hostilities between the Social Democrats and Communists enabled the rise of Hitler’s National Socialists. The Nazis made no distinction between the Social Democrats, Communists, and trade unionists when they rounded them up and placed them in concentration camps. The point here is not merely that the KPD failed to assess the dangers of fascism during the Weimar Republic, but rather that the KPD had no intention of creating a theory that reflected the reality of the Weimar period. The theory of social fascism had very little, if anything, to do with fascism. (Haro 2011)

Conclusion

It will be seen that, as used, the word ‘Fascism’ is almost entirely meaningless. In conversation, of course, it is used even more wildly than in print. I have heard it applied to farmers, shopkeepers, Social Credit, corporal punishment, fox-hunting, bull-fighting, the 1922 Committee, the 1941 Committee, Kipling, Gandhi, Chiang Kai-Shek, homosexuality, Priestley's broadcasts, Youth Hostels, astrology, women, dogs and I do not know what else. …

But Fascism is also a political and economic system. Why, then, cannot we have a clear and generally accepted definition of it? Alas! we shall not get one — not yet, anyway. To say why would take too long, but basically it is because it is impossible to define Fascism satisfactorily without making admissions which neither the Fascists themselves, nor the Conservatives, nor Socialists of any colour, are willing to make. All one can do for the moment is to use the word with a certain amount of circumspection and not, as is usually done, degrade it to the level of a swearword.7

Far be it for me to contradict George Orwell writing in 1944, but from my perspective, writing now in 2025, I think we should abandon the use of the word “fascist” entirely.

Over the past eighty years, “fascist” hasn’t been “degraded” to a swear word. A hundred years ago, when the use of “fascist” was first started, it was already used as a swear word.

However, we should abandon the word “fascist” not just because its use associates us with 1920s communism. If you think this article is just an anti-communist screed, take a look at the references to this article.

Thomas Draper was a communist and renowned historian of communism. Even in his old age, he left the Hoover Institution because it was too “conservative.”

The full title of the journal Critique is “Critique: Journal of Socialist Theory.” It was founded as “Critique: Journal of Marxist Theory and Soviet Studies,” but the title was changed after the end of the Soviet Union.

Maybe you are a college student with a Che Guevara t-shirt and a hammer-and-sickle sticker on your MacBook, so you are inclined to like the association with communism that the use of the “fascist” slur brings you. However, when you use the “fascist” slur, you do not just associate yourself with communists; you associate yourself with the stupidest kind of communists that other communists criticize as a stain on the history of communism.

We should abandon the word “fascist” because it is now and has always been used for demagoguery. The word “fascist” is too flawed to be useful for serious analysis. It is the adult equivalent of calling someone “poopy pants.”

As soon as someone uses the word “fascist,” I stop paying attention. It is a sign of a flawed intellect, a lack of self-awareness, and a lack of substance. There are many better criticisms that anyone can make about any political figure. I welcome those criticisms, and I reject childish name-calling.

References

Draper, T. (1969, February 1). The Ghost of Social-Fascism. Commentary Magazine. https://www.commentary.org/articles/theodore-draper/the-ghost-of-social-fascism/

Haro, L. (2011). Entering a Theoretical Void: The Theory of Social Fascism and Stalinism in the German Communist Party. Critique, 39(4), 563–582. https://doi.org/10.1080/03017605.2011.621248