Astrology has a long history, stretching back thousands of years and permeating across numerous ancient civilizations. In modern times, astrology is big business — and it's growing. In 2025, the industry was estimated to be worth around $3 billion.

In this excerpt from "What Science Says About Astrology" (Columbia University Press, 2026), author and science journalist Carlos Orsi looks at a study of 20 million people that sought to test whether star signs have a role in romantic compatibility.

The most robust use of data to test astrology is the study of love signs conducted by David Voas in 2007, involving data from more than 20 million people from the 2001 census from England and Wales. Voas tested the hypothesis that certain sun signs were "more compatible" for romantic relationships.

The use of the supposed romantic compatibility/incompatibility between signs or planetary configurations to test astrology's validity has a long history. This strategy was, for example, employed by Carl Jung (1875–1967) in his work on astrology and synchronicity and in the classic study by Bernie Silverman.

The idea of astrological compatibility or incompatibility in love has strong popular appeal. The book "Love Signs", by Linda Goodman (1925–1995), an almost 1,000-page tome, continues to be reprinted and sold 30 years after the author's death (as of this writing, the most recent edition dates from 2020). In general, signs separated on the zodiac wheel by angles of 60° and 120° are considered favorable for love, while those separated by 180° are seen as extremely incompatible. Right angles also tend to be interpreted as bad omens.

Voas explains the rationale of his study this way: People born during the month-long periods defined by a particular sun sign are supposed to share certain dispositions, for example, to be generous or sensitive or stubborn. These tendencies affect personal relationships.

We know from everyday experience as well as a mass of social scientific data that people who are similar in age, education, social class, religion, ethnicity and so on are far more likely to marry than those who are different in these respects. Couples are regarded as being well or poorly matched on the basis of appearance or personality. If astrological compatibility exists, its effects should be observable.

This last point — that the effects should be observable — is crucial. Astrologers often complain that tests based solely on sun signs are unfair because a sun sign's influence represents only a fraction of an entire birth chart's meaning. However, a sample of 20 million people, like Voas's, neutralizes this objection.

Even if the sun sign accounts for only, say, 0.1% of overall romantic compatibility, in a sample composed of 10 million couples, this should result in an excess of 10,000 formed by people with compatible signs, above and beyond what would be expected if astrology had no effect. Or, as the author states, "With a sufficiently large sample, we should be able to detect any tendency for some signs to attract or repel each other."

The study's initial goal was to find an excess of pairings between signs deemed compatible by the consensus of astrological literature. Unfortunately, Voas writes, such a consensus was hard to find: "There is no great consistency among astrologers, and a survey of books and websites reveals a considerable variety of views concerning propitious pairings." So he opted for the least common denominator, searching for any deviation from basic probability: "In this research I look for evidence that any combination of signs is found more or less often than would be expected to occur by chance."

The results were at least intriguing: The initial analysis indicated an excess of couples where both partners had the same sign or adjacent signs — e.g., more Capricorns with Capricorns or Capricorns with Aquarians than expected. There were about 22,000 extra couples with matching signs beyond what chance would predict and an additional 5,000 couples with adjacent signs. Could this be astrology in action?

Voas dug deeper into the data and discovered more anomalies. For example, the excess of couples born in the same month was even greater (23,000) than that of couples with the same sign, and the proportion of couples with the same birth date was 41% higher than expected by chance. "Now while there may be some people who are drawn to each other because they share a birthday, the excess probably reflects response error for the most part," he wrote. "Census forms are typically completed by one member of the household, and that individual may — through carelessness or forgetfulness — write in his or her birthday when entering details for the spouse."

Other statistical anomalies attributed to errors include an excess of birth dates recorded as January 1 (probably a placeholder when the actual date is unknown), instances of matching days in different months, and matching months with different days. Voas's challenge, then, was distinguishing these potential data entry errors from any real astrological effect — if one existed.

"The partial overlap between astrological signs and months of birth allows a crucial test," he wrote, noting that the first 10 days of the period covered by any sign falls in one month while the other 20 or so fall in the next (for example, Aries runs from March 21 to April 20). So was a person born in the last days of March more likely to be married to someone born in the early weeks of March or perhaps the early weeks of April? In the first case, their spouse would be from the same month but a different sign; in the second, from a different month but the same sign.

"The results were conclusive. The couples whose birthdays belonged to the same sign but fell in different months were no more numerous than chance would dictate. By contrast, there were more combinations of birthdays from different parts of the same month than expected. This excess in shared months of birth is probably the result of response error, but in any event sun sign is not a factor."

The slight excess of couples with adjacent signs was explained by a data-imputation technique used in the British census to fill in missing or illegible data. One partner's birth date was imputed as the first day of a month and the other's as the first of the following month. When these imputed data points were excluded from the sample, the "adjacent sign" effect disappeared. The bottom line is that an analysis of 10 million couples in England and Wales revealed no astrological effect.

But Voas's work illustrates how easy it is to get lost in data or be swayed by enthusiasm. Someone who had stopped at the first step — finding an excess of couples with the same sign — might have mistakenly presented census data as validation for astrology.

This article is excerpted from What Science Says About Astrology by Carlos Orsi. Copyright (c) 2026 Columbia University Press. Used by arrangement with the Publisher. All rights reserved.