Chinese Gender Calendar Scientific Evidence: What Peer-Reviewed Studies Actually Show

Before working through the studies themselves, it is worth defining what “scientific evidence” actually means in the context of a predictive method like the Chinese gender calendar. In clinical research, a predictive tool is not judged by how convincing it feels to its users. It is judged by whether, in a large enough sample of independent cases, its predictions match the actual outcomes at a rate meaningfully better than chance — and whether that result holds up across multiple studies done by different research groups.

For a binary outcome like baby sex, “chance” means being right about half the time, because boys and girls are born in nearly equal numbers. That sets the bar a predictive method must clear: simply being right half the time proves nothing, because flipping a coin will produce the same result. To count as real evidence, the calendar would need to consistently beat a coin flip — ideally by a wide margin — in carefully designed studies.

Another piece of the puzzle is peer review. Peer review is the process where a paper is submitted to a recognized academic journal, anonymous expert reviewers in the field critique the methodology and statistics, and the paper is only published if those reviewers find the work sound. A claim in a blog post or a YouTube video is not evidence. A claim published in a journal like the American Journal of Obstetrics and Gynecology — AJOG, one of the most influential obstetrics journals in the world — carries weight precisely because it has cleared that bar. That is the lens we apply to the Chinese gender calendar in the sections below.

The most rigorous test of the Chinese gender calendar comes from Sweden, whose national birth register is one of the most complete population health records in the world. Eduardo Villamor of the University of Michigan and colleagues at the Karolinska Institutet took 2,840,755 singleton births that occurred in Sweden between 1973 and 2006. The abstract is indexed by the U.S. National Library of Medicine onPubMed (PMID 20618730).

The method was straightforward. For each birth they estimated the mother's lunar age and the lunar month of conception, looked up the predicted sex using a published version of the chart, and compared that prediction with the baby's recorded sex. Instead of a raw “percent correct”, they reported Cohen's kappa, which measures agreement beyond what chance alone would produce. A kappa of 0 means no agreement beyond chance; 1 means perfect agreement.

The overall kappa was 0.0002, with a 95% confidence interval from −0.0009 to 0.0014. In plain terms, the chart agreed with reality exactly as often as a coin would. The result did not change with the year of conception, the mother's age, her education, her body mass index, or how many children she already had.

The authors also anticipated the obvious objection that they might have converted the dates differently from the way parents do. They re-ran the test on 1,000 births using the lunar-date algorithm of a popular website (kappa −0.02), and they simulated the common mistake of skipping the lunar conversion altogether. Neither changed the result. Their conclusion: the method “is no better at predicting the sex of a baby than tossing a coin”, and they advised against painting the nursery based on it.

The other study you will see named on almost every gender-prediction website is “Katz, 1999”. The real record is different. Daniel Katz and Blair Wylie presented an abstract titled “The Chinese birth calendar for prediction of gender — fact or fiction?”, published as abstract 568 in the December 2009 supplement of theAmerican Journal of Obstetrics and Gynecology(volume 201, issue 6, page S211). It is a meeting abstract, not a full paper, and its text is not openly available, so we do not quote a sample size or an accuracy figure for it.

We are flagging this because an earlier version of this site repeated the common version of the citation: dated 1999, credited with a sample it does not describe, and linked to a PubMed record that turned out to be an unrelated chemistry paper. In October 2026 we checked every research citation on the site against PubMed and the publishers' own records, and corrected them. The Swedish study above is the one to rely on: its data, methods and numbers are all on the record.

The table below summarizes the published peer-reviewed work most often cited on this question. It is not exhaustive, but it captures the studies that shape the current scientific consensus. The pattern across rows is what matters most: every study with open data finds chance-level accuracy. Sources checked on PubMed and Crossref, October 15, 2026.

Two notes on reading this table. First, “no agreement beyond chance” is not a rhetorical hedge; it is what the kappa statistic in the largest study measured. Second, these are the rigorous studies. The internet abounds with informal “polls” and self-reported testimonials that report much higher accuracy figures; those numbers do not survive selection bias correction and have never been replicated in any peer-reviewed work.

The finding that anchors this entire discussion — no better than chance — is easy to misread, so it is worth pausing on. When statisticians say a predictive method performs at “chance,” they do not mean it is literally useless to use; they mean its observed accuracy is consistent with the simplest baseline possible. For a binary outcome with a roughly even base rate, that baseline is a coin flip.

Concretely, imagine you flipped a fair coin to predict the sex of every baby in a hospital for a year. You would land on the correct answer in approximately half of the cases, simply because two outcomes are roughly equally likely. If the Chinese gender calendar lands at the same accuracy as that imaginary coin, then the calendar is not contributing information; the coin would have done equally well. That is the meaning of the published finding.

A second concept that matters here is the confidence interval. Even a true coin-flip method, tested on a finite sample, will not land on exactly half — it will scatter slightly above or below on any single dataset, simply because finite samples are noisy. A result is considered statistically significant only when the deviation from chance is larger than the noise band that the sample size produces. Villamor et al. reported a kappa of 0.0002 with a 95% interval from −0.0009 to 0.0014, which sits across zero: the small wobble was exactly what would be expected from a method with no real predictive signal. That is why the authors concluded the chart performs at chance.

This is also why anecdotal accuracy rates of “the chart got it right for me!” tell us almost nothing on their own. With a coin-flip base rate, half of any group of parents will have had the chart land correctly purely by luck. That subset will reasonably remember and talk about it. The other half — equally large — will tend not to.

The statistical findings above are reinforced by something even more basic: the biology of sex determination. Whether a baby is born male or female is decided at the precise moment of fertilization by which sperm cell reaches the egg. Sperm carrying a Y chromosome produce males; sperm carrying an X chromosome produce females. The mother contributes only X chromosomes, so the determining input is entirely a property of the single sperm cell that fertilizes the egg. That event takes milliseconds.

The Chinese gender calendar uses two inputs: the mother’s lunar age and the lunar month of conception. Neither input has any known biological mechanism by which it could influence whether an X-bearing or Y-bearing sperm reaches the egg first in any given act of conception. There is no published research demonstrating that the calendar month systematically shifts the X/Y ratio of sperm reaching the egg, and the maternal-age signal that does exist in the literature (a very slight shift in live-birth sex ratio with maternal age) is far too small to drive a calendar-based prediction.

Some of the biological reasons no calendar method can outperform chance are worth listing explicitly:

• Sex is determined by the chromosomal content of the sperm at the instant of fertilization, not by any maternal characteristic.
• The mother contributes only X chromosomes, so her age and the month of conception cannot supply the Y chromosome that determines male sex.
• Sperm motility and viability are not measurably modulated by the lunar calendar in any peer-reviewed reproductive physiology research.
• Boys and girls are born in nearly equal numbers, so over a large enough sample any method that does not contain real biological information will converge toward coin-flip performance.
• Published reproductive endocrinology has identified no maternal-age or seasonal effect large enough to produce above-chance predictive accuracy.

This is the deeper reason the empirical finding (no better than chance) is not surprising to obstetricians. The calendar uses inputs that biology does not use, so its predictions reduce to a guess. For a complementary discussion comparing the Chinese chart to another popular but unvalidated method — the Ramzi theory — see our deep dive on the Chinese gender calendar vs. Ramzi theory.

If the evidence is this consistent, why does belief in the chart persist? The answer is mostly psychological, not statistical. Two well-documented cognitive biases conspire to make a 50/50 method feel uncannily accurate to the people who try it.

The first is confirmation bias: people more readily notice and remember the cases where a belief was confirmed and quietly discount the cases where it was disconfirmed. The second is selection bias in online testimonials: the parents who post “the chart was right!” on forums and TikTok far outnumber those who post “the chart was wrong” (which is rarely entertaining content). The result is that the public conversation systematically over-represents correct predictions, even when the underlying rate of correctness is exactly a coin flip.

None of this means the people who report a correct prediction are lying or foolish. It simply means that the structure of how those reports reach the public is shaped by selection effects that no individual is fully aware of. That structure is exactly what large peer-reviewed studies are designed to cut through — and when they cut through it, the underlying accuracy resolves to chance.

Major obstetrics organizations do not list the Chinese gender calendar among the methods they recognize for fetal sex determination. When the American College of Obstetricians and Gynecologists discusses prenatal screening and diagnostic tests, the named methods for determining fetal sex are ultrasound, non-invasive prenatal testing (NIPT)/cell-free DNA, chorionic villus sampling (CVS), and amniocentesis. Traditional or calendar-based methods do not appear on those lists, because none of them have cleared the peer-reviewed validation bar that those clinical tools have.

The same picture holds in the broader medical literature curated by the National Institutes of Health and indexed in PubMed. A search of PubMed for “Chinese gender calendar” or “Chinese lunar calendar fetal sex” surfaces a handful of papers, of which the recurring conclusion across the high-quality ones is the same as the Villamor work: accuracy is at chance, the method is best understood as a cultural tradition, and it should not be used for clinical decision-making.

This is not a hostile or dismissive position. Medical bodies routinely acknowledge that cultural practices around pregnancy — gender reveal rituals, naming traditions, food and rest customs — play a meaningful role in family life and emotional wellbeing. The narrow point they make is the predictive one: when the question is “what sex is this baby going to be,” the answer should come from a tool with documented accuracy above chance, and the Chinese gender calendar is not such a tool.

The contrast between a method that performs no better than chance and clinically validated modern tools is large. The methods listed below are the ones that obstetricians actually use when an expectant parent wants to know the sex of the baby with confidence.

• Non-invasive prenatal testing (NIPT, cell-free fetal DNA). Available from approximately 10 weeks of pregnancy, NIPT analyzes fragments of fetal DNA circulating in the mother’s bloodstream and can report fetal sex. It is a screening test, mainly for certain chromosomal conditions, so a provider explains the result (MedlinePlus Genetics).
• Mid-pregnancy ultrasound. The anatomy scan typically performed between 18 and 21 weeks of pregnancy (NHS) can often show fetal sex, depending on fetal position and the skill of the sonographer. It is the most common method for parents who want to know during a standard pregnancy.
• Chorionic villus sampling (CVS). An invasive diagnostic test performed around 10–13 weeks, CVS examines the baby’s chromosomes directly, so it shows fetal sex along with chromosomal results, but carries a small procedural miscarriage risk and is usually reserved for cases with a clinical reason to test.
• Amniocentesis. Performed around 15–20 weeks, also examines the baby’s chromosomes directly, so it shows fetal sex. Like CVS, it is an invasive test reserved for clinical indication rather than routine curiosity. The timing for both tests comes from ACOG's Prenatal Genetic Diagnostic Tests FAQ (amniocentesis is usually done at 15–20 weeks but can be done later).

For a broader comparison of these and other approaches in the U.S. context, including how parents typically choose among them, see our overview of U.S. gender prediction methods. The point of including this comparison here is not to argue that the Chinese gender calendar should be replaced by clinical testing for every reader — many parents enjoy the calendar precisely because it is non-clinical. The point is that the gap between a coin flip and a test that reads the baby’s own chromosomes is the entire reason medical bodies recommend the clinical methods for any decision that actually depends on knowing the sex of the baby.

Bringing the threads together: the scientific evidence on the Chinese gender calendar is unusually clean for a question that has been asked this many times. In the largest test ever run, a registry analysis of 2,840,755 Swedish births (Villamor et al., 2010), the calendar predicted baby sex no better than chance, whatever the mother's age, education or number of children. There is no published peer-reviewed evidence to the contrary, and there is a clean biological reason why such evidence would be a surprise: the calendar uses inputs that the actual mechanism of sex determination does not use.

None of this argues against the calendar as a cultural object. It is part of centuries of Chinese pregnancy folklore, it is a charming centerpiece for a gender-reveal party, and it gives families something to talk about with grandparents. The verdict on scientific evidence does not erase any of that. It simply locates the calendar correctly: it is a tradition, not a clinical tool. If you came to this page wanting the academic answer to the question “what does the science say,” the answer is that the science says no better than chance, no statistical significance, no biological mechanism, and no clinical recommendation. For the consumer-facing summary of Chinese gender calendar accuracy, see our companion guide.

To explore the chart on its cultural terms instead — including the underlying logic, the lunar-age inputs, and how to read the grid — return to the Chinese Gender Calendar homepage, or read more from Sukie Chinese, the former Chinese-language teacher who writes the cultural-context guides on this site.