Key Takeaways
Define sex before counting it, or the headline can change overnight
A definition changes the answer. In 1991, 599 US students were asked whether particular acts counted as having sex. About 40% said oral sex did; 60% said it did not. The distinction became unexpectedly consequential when President Clinton denied having sexual relations with Monica Lewinsky. For public health research, Natsal used a practical definition based on oral, anal, or vaginal intercourse, since those activities can transmit infections. That definition counts Clinton and Lewinsky as sexual partners, whatever people choose to call the encounter.
Numbers need a boundary. A survey cannot answer a question reliably until it specifies what counts as sex, a partner, or a sexual experience. Definitions are not technical housekeeping: they determine who appears in the data and what the results can be used to explain.
What's striking is that the dispute is not simply about honesty. It is a problem of construct validity: researchers must translate a flexible human concept into a rule that can be measured consistently. Public health definitions are useful because they align measurement with a practical risk, but they do not settle the cultural or personal meaning of sex. A responsible statistic therefore needs its definition beside it, not buried in a methods appendix. The same principle applies to terms such as violence, addiction, and poverty, where a boundary can shape both public debate and access to services.
Rate the survey design, not the number of responses
Big samples can mislead. Spiegelhalter rates evidence from 0 stars, meaning unsupported, to 4 stars, meaning official records such as births. Well-designed random surveys such as Britain’s Natsal receive 3 stars: they cannot eliminate reporting errors, but their sampling and privacy procedures make results reasonably credible. A survey with thousands of volunteers may be much weaker. Shere Hite’s reports drew thousands of replies from people who chose to return questionnaires, while a condom company’s online poll claimed single Americans averaged 130 sexual encounters a year. Neither sample represented the general population simply because it was large.
Bias beats chance. A random sample of 1,000 people might estimate a population proportion within a few percentage points. A self-selected sample can be wrong by much more, because people who respond may differ systematically from those who do not. The book applies the same skepticism to the ONS estimate of prostitution’s economic contribution, built from uncertain assumptions about worker numbers, prices, and workload.
The star scale is a useful antidote to false precision because it separates two questions: how many people were surveyed, and whether they resemble the population being described. Statistical confidence intervals address random sampling variation, not systematic selection bias. That distinction is easy to lose when a headline presents a striking number without its recruitment method. Still, a low-quality survey can document experiences or suggest questions worth investigating. Its proper status is exploratory, not representative. Good statistical literacy does not require discarding imperfect evidence; it requires matching the strength of a claim to the strength of its design.
Use the median to describe a person, not the mean
Typical is not total. Among sexually active Britons aged 16 to 74, the median reported frequency of heterosexual sex was three times in the previous four weeks. The mean was higher because a small number of very high reports pulled it upward. The median describes the person in the middle; the mean describes total activity divided across everyone. They answer different questions.
The partner puzzle. Natsal-3 found that men reported a mean of 14 lifetime opposite-sex partners, twice the seven reported by women. In a closed group of equal numbers of men and women, however, the mean number of partners must be the same: every partnership connects one person of each sex. Differences in reporting, survey coverage, and definitions may help explain the mismatch. Lifetime counts are also vulnerable to vague recall, especially when people estimate rather than remember each partner.
The median is often the better measure of a typical individual when behavior has a long tail, but it should not replace the mean in every context. Public health planners may care about the total number of encounters because that affects service demand or infection spread. The apparent contradiction between men’s and women’s reported partner counts is especially instructive: a mathematical identity exposes a measurement problem that a headline average might conceal. Rather than treating the discrepancy as proof that one sex is lying, the analysis asks how memory, survey coverage, and the meaning of partner shape the answers.
Measure identity, attraction, and behavior as separate things
One label is not enough. Sexual identity is how a person describes themself; attraction concerns whom they desire; behavior records what they have done. These dimensions overlap, but they are not interchangeable. Kinsey’s 0-to-6 scale tried to represent a continuum of heterosexual and homosexual experience and response. Its flexibility challenged the idea that everyone belongs in one fixed category, though it could also blur distinctions when researchers used it as a single measure.
The totals depend on the question. A US survey found that about 10% of men reported at least one of same-sex identity, attraction, or behavior, while 2.4% reported all three. Many people identifying as heterosexual also reported same-sex contact. The practical lesson is to ask which dimension matters. For studying infection transmission, behavior may be relevant; for understanding self-concept, identity matters more.
This distinction improves both measurement and respect. Identity categories describe how people understand themselves, while behavioral categories can capture exposures that identity-based surveys miss. Treating the two as substitutes can make a population appear smaller or larger depending on the question. The framework also resists the assumption that one experience dictates a permanent identity. Its limit is that separate measures still cannot fully express fluidity, context, or personal meaning. Precision in a questionnaire is a starting point, not a complete account of how someone experiences their sexuality.
Read sexual change through generations, relationships, and history
Behavior has a context. In Natsal, the median frequency of heterosexual sex among sexually active 16-to-44-year-olds fell from five times in the previous four weeks in 1990 to three times in 2010. Frequency also declined with relationship length: younger couples together less than two years reported a median of seven occasions, compared with four among those together at least five years. Work, children, health, and changing partnership patterns may matter alongside age.
Three clocks overlap. Researchers distinguish age effects, changes in the period people live through, and differences between cohorts born in different eras. These are difficult to separate because age, calendar year, and birth year are mathematically linked. The rise in anal sex among 30-year-olds across successive surveys, for example, shows a trend, but cannot by itself identify its cause.
The age-period-cohort problem is a compact lesson in the limits of observational data. A trend can be real while its explanation remains uncertain. That matters because public arguments often turn a pattern into a story about technology, morality, or biology before the evidence distinguishes among them. Relationship duration adds another layer: a decline with age may partly reflect years spent together rather than aging itself. The most defensible interpretation is therefore plural. Sexual behavior responds to bodies, relationships, and social conditions, but these influences are entangled rather than cleanly separable.
Teach consent and competence alongside contraception and anatomy
How sex begins matters. In Britain, the median age at first sex was 16 in Natsal-3. The survey also assessed whether first sex involved unequal willingness, regret about timing, pressure or intoxication, or no reliable contraception. By these criteria, 56% of women and 46% of men aged 16 to 24 were not considered competent at their first experience. The proportion was higher among those whose first sex occurred at 13 or 14.
Circumstances shape risk. Teenage pregnancy rates varied greatly across Britain, tracking deprivation in many areas, while countries with accessible contraception and sex education, such as the Netherlands, had far lower teenage birth rates than the UK. These comparisons do not prove that one policy alone caused the difference, but they challenge explanations that blame young people’s behavior while ignoring services, opportunity, and social conditions.
The competence measure broadens sex education beyond anatomy and pregnancy prevention. It asks whether a young person could participate willingly, make choices without pressure, and use protection. That is a more useful model of readiness than age alone. The measure also has limits: one adverse answer classifies an experience as not competent, even though experiences differ in severity and meaning. Still, the framework makes an important policy point. Teaching consent is not an optional supplement to technical information; it is part of the conditions that make sexual health possible.
Treat contraception as a behavior system, not a perfect shield
Use depends on real life. In Britain, an estimated 999 out of every 1,000 heterosexual sexual acts did not result in conception. Yet contraception is not infallible. The book compares annual failure estimates under perfect and typical use: condoms rise from about 2% to 18%, the combined pill from 0.3% to 9%, and withdrawal from an estimated 4% to 22%. Typical use includes mistakes, inconsistent use, and ordinary human circumstances, not simply a method’s theoretical performance.
Risk is not uniform. A single average failure rate may not describe any particular couple, because age, fertility, consistency, and circumstances differ. Natsal-3 classified 16% of recent pregnancies as unplanned and 29% as ambivalent. The contraceptive pill scare of 1995, which overstated a small absolute risk, was followed by a sharp rise in conceptions and abortions. Risk communication can itself change reproductive outcomes.
The distinction between perfect and typical use is a useful example of how effectiveness depends on a person-method interaction. A device can work extremely well under controlled conditions and less well when its use requires repeated decisions. At the same time, averages should not be converted into personal predictions without caution: the book’s example of different withdrawal risks among different men shows why population rates can mislead individuals. Risk communication also has consequences. Clear numbers should distinguish relative risk from absolute risk, especially when public warnings may prompt people to abandon a protective method.
Do not confuse sexual frequency with sexual well-being
Quantity is not health. The World Health Organization’s definition of sexual health includes physical, emotional, mental, and social well-being, not merely the absence of disease or dysfunction. Natsal-3 found that about one in seven sexually active people were dissatisfied with their sex lives, and roughly one in ten felt distressed or worried. Among people without a partner in the previous year, dissatisfaction increased, but distress remained comparatively low. A person can be sexually inactive without feeling that something is wrong.
Problems are common, not destiny. About a third of women reported a lack of interest lasting at least three months in the previous year, and many men reported difficulties such as early ejaculation or trouble maintaining an erection. These figures describe reported problems, not a universal standard for a normal sex life. Natsal also found that two-thirds of men and one-third of women reported masturbating in the previous four weeks, a reminder that sexual expression is broader than partnered intercourse.
This view challenges a culture that can turn both abstinence and sexual difficulty into personal failure. Frequency alone is a poor proxy for well-being because desire, relationships, health, and choice vary. A useful distinction is between a symptom and distress about that symptom: people may experience a difficulty without wanting treatment, while others may benefit from help. The survey figures cannot decide which response is right for an individual. Their value is to show that variation is common and to make room for care without imposing a single required level of sexual activity.
Ask what else could explain a media-behavior link
Correlation leaves a gap. Studies have linked exposure to sexual media with adolescent behavior, but young people who seek such media may already differ in interest, maturity, or social circumstances. One widely cited study found an association between sexual media exposure and later activity among white adolescents, but not black adolescents. A later re-analysis used propensity score matching, comparing people with similar measured characteristics, and found that the apparent effect disappeared.
The evidence has boundaries. Matching can reduce some confounding, but it cannot imitate random assignment perfectly or remove factors that were not measured. Surveys of pornography use and sexting also produce widely varying estimates because they recruit different people and define behavior differently. The book’s argument is not that media never matter. It is that strong causal claims require more than a headline association, especially when the behaviors and technologies change quickly.
The contrast between association and cause is central to interpreting behavior research. Propensity matching helps by comparing more similar groups, but it cannot guarantee that the groups are alike in every relevant way. This is a valuable middle ground between accepting every alarming claim and dismissing every possible influence. Media can shape norms, supply scripts, or provide opportunities without producing one uniform effect in all viewers. The best public discussion therefore asks which people, which content, and which outcomes are under consideration, rather than treating exposure as a single dose with a single consequence.
Use records to infer hidden behavior, but label the limits
Consequences leave traces. When private behavior cannot be observed directly, official records can reveal part of its history. By linking marriage and birth records, researchers estimated that 38% of brides in 1830 were pregnant when they married. In 1938, 51% of brides under 20 were pregnant at marriage. These figures provide a lower bound on sex before marriage, not a complete count: couples who did not conceive leave no trace in this measure.
The category keeps changing. In 2012, 48% of births in England and Wales were outside marriage, but most were registered by two parents living at the same address. The formal category of illegitimacy therefore no longer meant what it had in earlier periods. Historical records can expose broad shifts, but they cannot recover the feelings, agreements, or circumstances behind each event.
This is a form of indirect measurement: rather than asking people to recall private behavior, researchers infer it from consequences that were recorded for other reasons. It can open a window onto periods with little personal testimony, but the window has a frame. The method misses non-conceptions and depends on reliable record linkage; its meaning also changes as social categories change. The strongest use of such evidence is comparative and modest. It can show that behavior or family formation changed substantially, while leaving individual motives to historical interpretation rather than pretending the register contains them.
Make consent and coercion central to sexual health statistics
Counting harm is difficult. Natsal-3 found that one in ten women and 1.4% of men reported having been made to have sex against their will. Only 13% of women and 8% of men said they told the police. Definitions matter: the survey used the term non-volitional sex, while police and crime surveys use other categories, so their totals cannot be treated as interchangeable. Official police records also undercounted the scale of rape; estimates suggested only about one conviction for every 75 rapes.
Health includes safety. The book argues that sexual health cannot be measured only through pregnancy, infection, or frequency. Consent, coercion, and the ability to seek care belong in the picture. The same caution applies to prostitution and trafficking statistics, which are politically contested and often built from uncertain counts or assumptions. Large numbers in this area may carry moral force, but that does not make them reliable.
The section’s most important implication is methodological and ethical: what gets counted shapes what institutions recognize. Under-reporting, stigma, and inconsistent definitions can make violence appear smaller or can make international comparisons meaningless. Better measurement cannot substitute for survivor support, but it can reveal where systems fail. There is also a danger in treating a single survey estimate as definitive, because disclosure depends on wording, privacy, and trust. The responsible response is neither to inflate figures for impact nor to dismiss uncertain data. It is to improve measurement while keeping the human stakes visible.
Analysis
The book’s central achievement is not a grand theory of sexual behavior. It is a disciplined demonstration that numbers about sex are produced through choices about definitions, sampling, memory, and social permission. The Clinton-Lewinsky example makes the point vividly: a public argument appeared to hinge on a fact, but first required agreement about what the word sex meant. That same issue runs through estimates of partners, sexual identity, consent, and pregnancy.
The recurring star ratings make uncertainty legible rather than hiding it behind precise percentages. This is especially useful because sexual statistics are unusually vulnerable to both missing data and self-presentation. A huge online poll may be less informative than a smaller probability sample, while official records may count outcomes accurately yet say little about motives. The book’s strongest habit is to ask what evidence could have generated a number before accepting what the number seems to say.
Its historical material also complicates easy stories of continuous liberation. Marriage, contraception, law, health services, and economic conditions alter the setting in which people act. Trends such as lower reported frequency or rising same-sex experience are measurable, but the causal story cannot be read directly from a graph. The age-period-cohort problem is a sharp reminder that even careful data can leave important explanations entangled.
The book is less a guide to what people should do than a guide to what claims deserve belief. Its chief limitation is structural: broad survey measures inevitably flatten intimate differences, and commentary on weak evidence can sometimes leave readers with more uncertainty than resolution. Yet that uncertainty is productive. It discourages moral panic, simplistic biological explanations, and universal expectations about frequency or identity. The most durable lesson is that sexual behavior is both private experience and public evidence, and responsible interpretation must protect the complexity of both.
Review Summary
Sex by Numbers receives mostly positive reviews, with readers praising its statistical approach to sexology. Reviewers appreciate the author's grading system for data reliability and the book's ability to debunk myths. Many find it educational and engaging, noting its historical context and balanced views. Some criticism includes overwhelming numbers and potential trivialization of serious topics. Readers highlight interesting facts and the book's exploration of research difficulties in studying sexual behavior. Overall, it's seen as informative and thought-provoking, albeit occasionally dry.
People Also Read
Glossary
Natsal
British sexual behavior surveyThe National Survey of Sexual Attitudes and Lifestyles is a series of large British probability surveys conducted roughly every ten years. Researchers interview randomly selected adults about sexual behavior, relationships, health, and attitudes. The survey uses private computer responses for sensitive questions and weighting to make its results better represent the population.
star rating
Evidence reliability scaleSpiegelhalter’s informal scale ranks sexual statistics from 0 to 4 stars. Zero-star figures are unsupported or invented; one-star results are severely biased; two-star results may offer rough estimates; three-star results come from reasonably careful surveys; and four-star figures are official records, such as registered births, that are highly accurate for the outcomes they count.
social-desirability bias
Answers shaped by approvalA response pattern in which people report behavior in ways they believe will be socially acceptable or protect their self-image. It need not involve deliberate lying. Respondents may forget, reinterpret, exaggerate, or conceal experiences, especially when questions concern stigmatized or private behavior. Confidentiality and survey format can affect how much this bias influences answers.
Kinsey scale
Continuum of sexual experienceA seven-point scale developed by Alfred Kinsey, ranging from exclusively heterosexual at 0 to exclusively homosexual at 6. Kinsey intended it to reflect psychological reactions as well as overt experience, allowing intermediate positions and possible change over a lifetime. Later uses sometimes treated it as a measure of behavior alone, which can obscure the distinction between identity, attraction, and experience.
age-period-cohort explanations
Three sources of changeA way of describing possible influences on behavior: age-related change within a life, period effects from the historical moment, and cohort effects associated with the generation into which someone was born. These influences are difficult to distinguish statistically because age, calendar period, and birth year are mathematically connected.
fecundability
Monthly chance of conceptionThe probability of becoming pregnant during one menstrual cycle when contraception is not used. It is usually discussed as a population average, but actual chances vary substantially with the couple’s fertility, age, timing, and frequency of sex. A single average therefore cannot predict an individual couple’s experience with precision.
London Measure of Unplanned Pregnancy
Pregnancy intention questionnaireA six-question scale used to assess how planned a pregnancy was. Questions cover contraception, timing, intentions, feelings about having a baby, discussion with a partner, and health preparations. Responses are added to produce a score from 0 to 12, which can classify a pregnancy as unplanned, ambivalent, or planned.
non-volitional sex
Sex without free agreementThe term used in Natsal-3 questions about someone trying to make a respondent have sex against their will, or succeeding in doing so. It was chosen to capture experiences that may not fit neatly into a respondent’s understanding of a criminal label such as rape. It is a survey category, not a legal judgment.
Relative Age Effect
Advantage from birth timingThe tendency for children born earlier within a school-year cohort to perform better initially than classmates born near the cutoff. A few months of age difference can influence achievement measures and access to advanced opportunities. The book notes that later selection can reverse some patterns among those who reach university or elite sport.
capture-recapture methods
Population size from overlapA method for estimating a population that is difficult to count directly. Researchers compare two lists or samples and examine how many individuals appear in both. If overlap is large, the total population is estimated to be relatively small; if overlap is small, the estimate is larger. The method depends on assumptions about how individuals enter each list.
FAQ
What’s [Sex by Numbers: What Statistics Can Tell Us About Sexual Behaviour] by David Spiegelhalter about?
- Statistical exploration of sex: The book investigates human sexual behavior using statistics, drawing on historical and modern surveys to reveal patterns, myths, and realities.
- Bridging private and public: Spiegelhalter discusses the challenges of studying a topic that sits between private life and public interest, using data to illuminate this complex area.
- Interdisciplinary approach: The author integrates insights from history, biology, psychology, and sociology to provide a nuanced understanding of sexuality.
- Focus on variability: The book emphasizes the diversity and variability of sexual behaviors and outcomes, cautioning against simplistic interpretations of averages.
Why should I read [Sex by Numbers] by David Spiegelhalter?
- Demystifies sexual statistics: The book helps readers critically evaluate common claims and media reports about sex, contraception, and sexual health.
- Balanced perspective: It offers a comprehensive view that acknowledges both biological and social influences on sexual behavior.
- Practical insights: Readers gain knowledge about contraception effectiveness, sexual satisfaction, and societal changes, supported by robust data.
- Historical and modern context: The book combines historical trends with contemporary research, making it relevant for understanding changes over time.
What are the key takeaways from [Sex by Numbers] by David Spiegelhalter?
- Complexity of sexual behavior: Sexuality is influenced by a mix of biology, society, and chance, and cannot be reduced to simple averages.
- Methodological challenges: The book highlights the difficulties in collecting reliable sex statistics, including bias, sampling, and question wording.
- Changing trends: There are notable shifts in sexual activity, attitudes, and behaviors over time, such as declining frequency and increased acceptance of diverse practices.
- Critical use of data: Spiegelhalter encourages readers to question the reliability of statistics and understand their limitations.
How does [Sex by Numbers] by David Spiegelhalter define “having sex,” and why does this matter?
- Varied definitions: The book shows that people differ in what they consider “having sex,” ranging from intercourse to oral sex and other activities.
- Legal and health implications: Clear definitions are crucial for legal cases and health research, especially regarding infection risks.
- Survey consistency: Modern surveys like Natsal define “sex” as activities that can transmit infections, ensuring clarity in data collection.
- Impact on statistics: How sex is defined affects the accuracy and comparability of sexual behavior data.
What methods and challenges does [Sex by Numbers] by David Spiegelhalter highlight in collecting reliable sex statistics?
- Sampling and bias: Obtaining accurate data is difficult due to privacy concerns, social desirability bias, and non-random sampling.
- Reliability rating system: Spiegelhalter uses a star system to rate the trustworthiness of different statistics, from official records to anecdotal numbers.
- Best practices: Large, well-designed surveys use random sampling and confidential interviews to minimize bias.
- Technological improvements: Computer-assisted self-interviews have increased disclosure of sensitive behaviors.
What does [Sex by Numbers] by David Spiegelhalter reveal about the frequency and patterns of sexual activity?
- Average frequency: Sexually active adults in Britain report about three sexual encounters in the last four weeks, with frequency declining with age.
- Declining trend: Sexual activity frequency has decreased over recent decades, especially among younger adults.
- Influencing factors: Partnership status, health, relationship length, and lifestyle pressures all affect sexual frequency.
- Persistence in older age: Sexual activity remains significant even among older adults.
How does [Sex by Numbers] by David Spiegelhalter address the number of sexual partners and gender differences in reporting?
- Reported numbers: Men report more lifetime opposite-sex partners than women, though median numbers are closer than means suggest.
- Mathematical paradox: In a closed population, average numbers should be equal; discrepancies arise from sampling biases and reporting differences.
- Social desirability and recall: Men may exaggerate, women may under-report, and definitions of “partner” can vary.
- Role of sex work: Inclusion of sex workers and partners outside the surveyed population can skew results.
What insights does [Sex by Numbers] by David Spiegelhalter provide about sexual activities beyond intercourse?
- Varied repertoire: Activities like oral sex, anal sex, and BDSM are increasingly common and accepted.
- Historical context: Early research by Kinsey revealed a wide range of behaviors, challenging social taboos.
- Changing trends: Oral and anal sex have become more widespread, especially among younger cohorts.
- Normalization: Practices like BDSM are increasingly seen as mainstream leisure activities.
How does [Sex by Numbers] by David Spiegelhalter explore same-sex sexual behavior and identity?
- Historical pioneers: The book discusses early researchers like Magnus Hirschfeld and the development of the Kinsey scale.
- Complexity of identity: Many people with same-sex experiences identify as heterosexual, highlighting the fluidity of sexual orientation.
- Prevalence and acceptance: Around 1–1.5% of UK adults identify as gay or lesbian, with higher bisexual identification among younger women.
- Attitudinal shifts: Societal acceptance of same-sex relationships has increased significantly in recent decades.
What does [Sex by Numbers] by David Spiegelhalter say about masturbation and solo sexual activity?
- Historical stigma: Masturbation was heavily stigmatized for centuries, with persistent myths about health risks.
- Modern prevalence: About two-thirds of men and one-third of women report recent masturbation, with rates declining with age.
- Changing attitudes: Masturbation is now seen as a normal and healthy part of sexuality, though some stigma remains.
- Concerns about compulsion: The book notes ongoing debates about compulsive solo sexual behavior.
How does [Sex by Numbers] by David Spiegelhalter address contraception effectiveness and the chance of pregnancy?
- Variability in failure rates: Typical-use failure rates for contraception methods vary widely among individuals and can be misleading.
- Fecundability explained: The chance of pregnancy from a single act of sex depends on timing relative to ovulation and individual fertility.
- Limitations of statistics: Published rates are based on assumptions and may not apply to everyone; they serve as guidelines, not predictions.
- Behavioral factors: Success with contraception depends on competence, age, and personal circumstances.
What does [Sex by Numbers] by David Spiegelhalter reveal about the influence of media, technology, and prostitution on sexual behavior?
- Media sexualization: The book documents increased sexual content in advertising and media, especially in recent decades.
- Limited behavioral impact: Studies show little evidence that exposure to sexual media directly causes riskier sexual behavior.
- Technological change: Internet pornography, sexting, and dating apps are rapidly changing sexual behavior, outpacing research.
- Prostitution statistics: The book discusses the challenges of estimating the size and economic impact of the sex industry, highlighting contested figures and methodological difficulties.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.