Do population studies find a marriage or divorce advantage?

No consistent advantage emerges in the accessible population studies: sign pairs classed as compatible do not reliably appear more often in marriages or show lower divorce risk. These studies assign Western Sun signs from birth dates and examine legal records; they do not compare complete natal charts or interpret what two people’s synastry means. They test recorded marriage formation or divorce, not relationship quality. In Ten million marriages: A test of astrological ‘love signs’, David Voas examined more than 10.3 million married couples in England and Wales. The initial table showed a small same-sign excess, but Voas traced much of the pattern to repeated birthday entries and the handling of missing dates; after adjustments, the remaining differences were essentially random. The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data adds a separate test: it examines marriage pairings and divorce risk across several published compatibility schemes, finding no consistent favorable pattern. Together, these findings weigh against a dependable population-level Sun-sign advantage. They leave a more exact question: what did each study count, and what could that count establish?

Sources: Ten million marriages: A test of astrological ‘love signs’; The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data; The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data

What the census table can—and cannot—count

Voas counted 10,317,649 married couples living together in England and Wales at the 2001 census. The first table showed a small excess of same-sign pairs, along with a smaller pattern among neighboring signs. The apparent clue raised a basic question: compared with what? The paper, “Ten million marriages: A test of astrological ‘love signs’,” did not assume that all 144 sign combinations should occur equally often. Births are not evenly spread through the year, so the signs themselves do not appear equally often in the population. Instead, Voas estimated the expected number for each pair from the separate totals for each partner’s sign. If one sign is common among husbands and another among wives, their pairing should also be common even if signs do not shape partner choice. Voas multiplied each partner’s sign total and divided by the total number of couples. This gave each cell a fairer comparison than dividing the table into 144 equal shares. Against that baseline, the initial table showed about 22,100 more same-sign couples than expected. That was about a quarter of one percent of the total. Yet modest errors in recorded dates could account for it. Voas therefore requested finer-grained counts by day and month, rather than treating the first sign table as the final answer. Those birthday counts exposed a pattern that sign totals alone could hide. Couples had the same recorded birthday more often than expected: roughly 39,800 pairs, compared with an expected 28,300. Voas argued that much of this excess likely came from how census forms were completed. One household member often filled in information for everyone. If that person did not know a spouse’s birthday, or was moving quickly through the form, they could enter their own date by mistake. The paper points to an Office for National Statistics validation exercise in which some respondents also recorded their own sex for a partner. That parallel does not prove every repeated birthday was an error, but it makes response error a plausible explanation. The first day of a month also appeared unusually often. January 1 was recorded 2,560 times, compared with 77 expected under the paper’s baseline; July 1 was the next most common. Some people may have supplied January 1 when they did not know an exact date. Missing dates added another source of clustering. The Office for National Statistics had imputed dates for about 0.5 percent of people with missing or illegible birthdays. Every imputed day was set to the first of a month, while months were assigned in rotation. When both partners’ dates needed imputation, form order could place the wife in the month after the husband. Since signs span parts of two months, this could create apparent spillover between neighboring signs. The correction mattered because it changed what the table appeared to say. Voas first removed couples where either partner was recorded as born on the first of a month, then adjusted same-sign counts for excess same-birthday or same-month entries. In the corrected table, most differences were small and scattered around the expected counts. The paper calls this essentially random variation. The adjustment was approximate, and some neighboring-sign cells remained positive, likely because same-month response patterns were not fully removed. The point is not that every error vanished. The original same-sign surplus weakened after the known recording patterns were examined and corrected. This is why a very large table can still mislead. More than ten million observations make tiny differences visible, but they do not make the dates error-free. A small, systematic recording pattern repeated across many households can create thousands of extra entries in a sign cell. When a headline says one pairing is unusually common, ask how expected counts were built and whether the researchers checked for shared birthdays, default dates, or imputed values. In this census, those checks changed the interpretation from an apparent sign pattern to a small residual distribution with no stable compatibility shape.

Sources: Ten million marriages: A test of astrological ‘love signs’

What do the Swedish registers add?

Helgertz and Scott’s “The validity of astrological predictions on marriage and divorce” adds a population and a way to follow marriages over time. The researchers used Swedish register data from 1968 to 2001. Their sample covered Nordic-born people ages 18 to 70 in heterosexual unions with Nordic-born spouses. About 66,000 marital unions were included in the marriage-pairing analysis. The divorce analysis required unions observed from their beginning, leaving about 46,000 marriages; roughly 15,000 ended in divorce during follow-up. The totals reflect separate questions. The researchers assigned each spouse one of 12 Sun signs from the recorded birth date, creating 144 possible combinations. They did not use birth time and place, or compare Moon, Venus, or Mars placements, houses, or inter-chart aspects. This is a test of date-based Sun-sign pairing claims. Its large register population makes it useful for those claims, while its narrow chart input marks the edge of what it can tell us. The authors tested several compatibility classifications collected from free public astrology websites. A scheme had to define favorable pairings clearly enough to sort them into categories. Some sources divided the signs differently, and two were asymmetric: a site could call one sign a good match for another without making the reverse judgment. The authors tested the rules as published and checked results after adjusting for that asymmetry. These categories are operational rules from selected popular guides, not a universal doctrine of compatibility. For marriage pairing, the authors compared observed unions in each category with expected numbers based on the sign distribution in their study population. Across the schemes, favorable pairs did not show a consistent excess. In the first scheme, 38 of the 144 combinations were classed as “good.” Those pairs averaged 462.47 observed unions against 463.01 expected, a difference that was not statistically significant. Other classifications gave a similar broad result. Favorable labels did not reliably distinguish marriage combinations from the study’s baseline. Divorce is a separate outcome. Starting at each marriage, the authors used Cox proportional hazards models to test whether more favorably classified couples had a lower risk of legal divorce while observed. Death, emigration, or the index person reaching age 70 ended follow-up without counting as a divorce. One estimate suggested about a 2 percent lower risk for a more compatible category, but it was not statistically distinguishable from the comparison group. Across the classifications, the paper found no consistent lower-risk pattern. That does not mean all marriages last equally long. It means these sign labels did not reliably separate divorce risk in this sample. The registers contained no information about whether participants knew their signs or consulted compatibility advice. The analysis tests whether date-based combinations aligned with recorded marriage choices and divorce risk, not whether people consciously chose partners by astrology. The authors argue awareness should not be necessary for the influence they tested; that is an assumption in their design, not a finding about how a reader responds to a chart interpretation. The Swedish study strengthens the population comparison by testing several published rules and following divorce risk over time. Yet it leaves other chart layers untouched and does not measure affection, satisfaction, or the meaning partners give a bond. Its conclusion is specific: among these Nordic-born Swedish couples, tested Sun-sign classifications did not consistently predict which combinations appeared in marriage records or which observed marriages had lower divorce risk. That is evidence about those rules, not a reading of complete synastry.

Sources: The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data; The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data; The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data

Why marriage and divorce need separate denominators

A marriage table asks which sign pairs appear among people who are already married. A divorce model asks whether a marriage ends during the time researchers observe it. Those are different outcomes, with different denominators. Neither can stand in for the other. In “Ten million marriages: A test of astrological ‘love signs’,” David Voas analyzed the 2001 England and Wales census records for more than 10.3 million married couples living together. The denominator is the married, co-resident couples present in that census, not everyone who might have met, dated, or married. The table can show whether certain sign pairs were overrepresented among those observed couples, compared with expected counts based on the signs’ marginal frequencies. It cannot by itself estimate how often people of each pairing marry, because it does not follow an unmarried population through courtship and marriage. Nor does it track how long each marriage lasts. A large count describes the couples included; it does not turn a snapshot into a marriage-entry rate. Someone still married on census day appears in the table, while people whose marriages had ended and people who never married do not. The census describes representation among current co-resident marriages, not the chance that a particular pairing will marry or remain together. Helgertz and Scott’s “The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data” separates those questions. Their marriage analysis compares sign combinations among 66,063 observed unions with expected combinations based on the sign distribution of the index individuals, used as an approximation of the marriage market. It asks whether favorable categories appear more often than that comparison predicts. This observational comparison of unions is closer to a marriage-choice question than Voas’s snapshot, but it does not follow every possible meeting into marriage. The Swedish divorce analysis starts with a different risk set. It includes about 46,000 unions observed from the beginning of the marriage, with roughly 15,000 ending in legal divorce during follow-up. The researchers measure elapsed time from marriage and follow unions until divorce or censoring. They censor when either spouse dies or emigrates, or when the index person turns 70. Requiring observation from the marriage’s start lets the analysis measure time at risk of dissolution; it also means this result applies to that restricted set of unions, not to every marriage in the register. | Study question | Who is counted? | What the result can address | | --- | --- | --- | | Voas: sign-pair representation | Married couples living together at the 2001 census | Whether some date-derived sign pairs appear more often in that snapshot than expected | | Sweden: marriage pairing | 66,063 observed unions, compared with an estimated marriage-market distribution | Whether categories called favorable are overrepresented among recorded unions | | Sweden: divorce risk | About 46,000 unions observed from marriage start, followed until divorce or censoring | Whether sign categories are associated with legal divorce risk over observed time | Identify the denominator before repeating a compatibility finding. An overrepresented pairing among married couples does not automatically mean those partners had a higher chance of marrying; that requires comparison with people who could have become partners. Fewer divorces, if found, would describe dissolution among marriages at risk, not a greater chance of meeting or marrying. A divorce study begins after legal marriage, so dating relationships and couples who never marry are outside its risk set. Requiring observation from the marriage’s start also excludes unions that began before the observation window. Follow-up ends for specified reasons, including death, emigration, or the age limit. These boundaries define the estimate: it concerns the unions and time actually observed. Read “common pairing” as a statement about a specified marriage sample, and “lower divorce risk” as a statement about unions observed from their start. Neither is a universal verdict about compatibility. Voas’s cross-section and the Swedish study’s separate marriage and divorce analyses answer related, but non-interchangeable, questions. When a report says a sign pair is common or durable, ask whether it measured choosing, entering marriage, or remaining married. That one check prevents a population count from being mistaken for a forecast about one couple.

A thought to begin with

Accessible population studies show no consistent marriage or divorce advantage for conventionally compatible Western Sun-sign pairs. The England and Wales census analysis found its initial same-sign excess weakened after birthday-recording patterns were addressed. A Swedish register study found no consistent favorable pattern across separate marriage and divorce analyses. A reported Swiss result is a meaningful but contested exception. These studies use birth-date-derived Sun signs, not complete natal-chart synastry, and their aggregate findings don’t interpret an individual couple’s charts.

Sources: Ten million marriages: A test of astrological ‘love signs’; The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data

Why compatibility rules and many comparisons matter

A sign-pair claim becomes testable only when its rule is clear: which combinations count as favorable, and what result would count as support? Without that definition, a researcher can label the interesting pairs after seeing the numbers. That may produce a memorable pattern, but it doesn't show that the pattern was predicted in advance. Twelve Sun signs can form 144 directional pairings, since Aries with Taurus and Taurus with Aries are counted separately in the register study. A table with that many cells invites attention to individual highs and lows. Even when the overall distribution follows its expected pattern, some cells may sit above or below expectation by chance. The existence of one unusual cell is therefore a different claim from evidence that a defined group of compatible pairs consistently appears more often. The study *The Validity of Astrological Predictions on Marriage and Divorce* tackles the definition problem by testing six compatibility classifications drawn from free astrology sources. The authors selected sources that clearly assigned compatibility levels to sign combinations, so each scheme could be translated into a rule for analysis without the researchers deciding which pairs seemed harmonious. The schemes did not all divide the pairings in the same way. Some marked a small group as favorable; others used several grades or labeled a much larger share of combinations as compatible. Two schemes were asymmetric, rating a pair differently when its order was reversed; the authors kept the original rules and checked symmetry separately. If one guide calls a pairing favorable and another does not, “compatible signs” is not yet one shared population category. A null finding across several explicit schemes tells us that those particular rules did not produce a consistent marriage or divorce advantage in the Swedish records. It cannot settle every possible sign-based method. But selecting whichever rule looks strongest after inspecting results would answer a weaker question than testing a rule stated in advance. The Swiss findings make the multiple-comparison issue concrete. In *The Astrology File*, as summarized by *The Validity of Astrological Predictions on Marriage and Divorce*, Sachs reported 25 marriage pairings and 13 divorce pairings as individually significant among 144 cells for each outcome. He also noted that 12 of the 13 marriage pairings with higher-than-expected probabilities were considered compatible. The accessible critique *Compatibility Sun Signs Fail World's Biggest Test on 27 Million Couples* disputes the threshold used to flag those cells. Its author argues that Sachs's published results imply a cutoff near p = 0.13 rather than the stated 0.05. At that threshold, the critique calculates that about 18.7 flagged pairings would be expected by chance among 144 for each outcome, or roughly 37 across both. Sachs reported 38. The critique therefore argues that the overall count of flagged cells offers little support for a sign effect. This is a secondary methodological challenge, not a new analysis of Sachs's original data, and the accessible material does not let us independently settle the dispute. The Swiss report remains a counterexample, but its significance is contested. What would make a population claim stronger? Researchers could publish the favorable-pair list before examining outcomes, state how all 144 comparisons are handled, report the size and uncertainty of any pattern, and test the same rule in an independent population. Those steps would distinguish a forecast from a pattern selected after the fact. So when a headline points to one unusually common pairing, ask for the full table and the rule behind it. A cell can be interesting without proving predictive value. Replication under a fixed definition would carry more weight than a striking exception on its own.

Sources: The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data; Compatibility Sun Signs Fail World's Biggest Test on 27 Million Couples

How the older Michigan and Swiss results fit

The older studies make the record more complicated, but they do not establish a dependable Sun-sign rule. In its review of earlier relationship research, “The Validity of Astrological Predictions on Marriage and Divorce” describes Silverman’s 1971 Michigan analysis as finding no indication that two astrologers’ predictions matched marriage and divorce records. The same review reports a striking Swiss exception: Gunter Sachs found that 12 of the 13 Sun-sign pairs with higher-than-expected marriage probabilities were classified as compatible. That result deserves a fair hearing. The question is whether it shows a repeatable pattern once we consider the full set of comparisons and how the study flagged them. The Michigan study appeared as “Studies of Astrology” in the Journal of Psychology in 1971. The publisher’s record confirms its bibliographic details, but the full text was not accessible here. The substantive finding therefore comes from Helgertz and Scott’s review, not direct inspection of Silverman’s paper. They say Silverman examined Michigan marriage and divorce records from 1967 and 1968 against predictions from two astrologers. They call it a rudimentary compatibility test and report no indication that the predictions corresponded to the observed records. This is context, not a detailed assessment of Silverman’s methods. Sachs’s Swiss analysis is the strongest affirmative case in this record. As summarized by “The Validity of Astrological Predictions on Marriage and Divorce,” Sachs examined zodiac combinations in marriage, divorce, and other outcomes, and reported that 12 of 13 pairings with higher-than-expected marriage probabilities were astrologically compatible. A later methodological critique, “Compatibility Sun Signs Fail World’s Biggest Test on 27 Million Couples,” describes the marriage and divorce samples as 358,763 and 109,030 couples. It reports that 25 of 144 pairings were individually significant for marriage and 13 for divorce. The Swiss result came from a large analysis, and its reported pattern aligned with a compatibility classification. The difficulty is the number of comparisons. Twelve signs for each partner yield 144 pairings. When each cell is tested separately, some may cross a significance threshold by chance, even if the overall pattern is not stable. The later critique argues that Sachs’s effective cutoff was about p = 0.13 rather than the stated p = 0.05. Using that reconstructed cutoff, it estimates about 18.7 flagged cells by chance for each outcome, or 37.4 across marriage and divorce. It compares that expectation with 38 reported significant cells across both outcomes. This is the critique’s calculation, not a new analysis of Sachs’s data here; the original Sachs analysis is not available in the accessible source set. The argument raises a serious question about the cell counts, but does not independently show what another method would find. The critique also says husband and wife sign-pair preferences showed little agreement in Sachs’s results. That challenges a broad compatibility interpretation: a stable pairing pattern would be more persuasive if both partners’ preferences aligned. This too comes through a secondary critique, so it is a reason for caution, not a final verdict. Taken together, the reports do not converge on a reproducible finding. Silverman’s result, as summarized by Helgertz and Scott, was null. Sachs reported a favorable subset, but the later critique disputes whether the significant cells exceed what many comparisons could produce. The Swedish authors present Sachs as a challenge to earlier negative findings; their own register analysis, however, finds no consistent evidence across the compatibility schemes it tests. The reports are mixed, not a settled population rule. What would strengthen the Swiss claim? Researchers would need to specify sign-pair classifications and tests before examining outcomes, account for all 144 cells and both outcome families, then reproduce the same result in an independent population. Access to Sachs’s data and analysis would also let others check how expected counts and significance thresholds were handled. Until then, the fair conclusion is narrow: Sachs reported a meaningful counterexample, while a later critique identified a multiple-comparisons concern that weakens its force. On the accessible evidence, neither establishes a dependable marriage advantage for compatible Sun signs.

Sources: The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data; Studies of Astrology; Compatibility Sun Signs Fail World's Biggest Test on 27 Million Couples

What a demographic result says about an individual chart

A population study can test whether date-derived Sun-sign pairs appear more often in marriage records or have a different divorce pattern. It cannot tell you what two complete charts say about one another. The studies here assign each person a Sun sign from a birth date, then compare groups. They do not interpret two people's Moon, Venus, or Mars placements, or the aspects between their charts. In “Ten million marriages: A test of astrological ‘love signs’,” David Voas describes the Sun sign as one element of a natal chart and distinguishes popular sign-pair claims from synastry, which compares horoscopes. His census analysis asks whether date-based sign combinations are represented differently among married couples. After examining recording errors and imputed dates, he reports that adjusted sign-pair differences were essentially random. That result bears on whether conventional Sun-sign pairings show an aggregate pattern in those marriage records. It does not read any individual chart. Helgertz and Scott’s “The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data” also assigns signs from birth dates and tests published compatibility classifications against marriage and divorce outcomes. The authors report no consistent evidence that favorably classified pairs were more common among marriages or less likely to divorce. Their study adds a second population and a separate look at divorce risk. Its chart input remains the Sun sign. Neither study calculates birth-time-dependent placements, compares Moon or Venus contacts, or weighs the full set of factors an astrologer may consider in synastry. That boundary cuts both ways. A null result for date-derived Sun-sign categories does not establish what a couple’s broader chart comparison would mean. But a full-chart interpretation also cannot be presented as though these studies tested it. Each claim belongs with the layer and outcome examined. Marriage representation, divorce risk, and the meaning of a particular pair of charts are different questions. Relationship satisfaction is another distinct outcome. Helgertz and Scott mention a separate study of married people that reported some support for an association between astrological compatibility and marital satisfaction while accounting for belief in astrology. They cite it in a literature review, not as a new analysis in the Swedish register study. It does not overturn the marriage and divorce results or show that a particular sign combination causes satisfaction or predicts a couple’s experience. For an individual comparison, start with the placements actually known. A Sun-sign pairing is one layer. Moon placements can be considered for emotional habits; Venus and Mars for affection, attraction, and ways of pursuing desire; inter-chart aspects show how placements meet. These are interpretive questions, not outcomes answered by a population table. When degrees are available, they can help identify selected major aspects under a stated orb. Missing birth information sets a limit: a comparison based on supplied placements cannot fill in what was never provided. Population evidence limits claims about aggregate marriage and divorce patterns under the tested Sun-sign rules. It neither promises a couple’s future nor dismisses reading their known placements. The Synastry Comparison Report compares supplied Moon, Venus, and Mars signs, with optional degrees for selected major aspects; it does not calculate natal charts from birth details. Keep the population result in its frame, then use the chart information you have to ask a separate interpretive question.

Sources: Ten million marriages: A test of astrological ‘love signs’; The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data

The useful conclusion for a couple comparing signs

Population studies have not established a dependable marriage or divorce advantage for conventionally compatible Sun-sign pairs. The 2001 England and Wales census analysis, “Ten million marriages: A test of astrological ‘love signs’,” counted couples already married and living together; after addressing birth-date recording patterns, remaining pair differences were essentially random. “The validity of astrological predictions on marriage and divorce” found no consistent compatibility advantage across Swedish analyses of observed marriages and divorce risk. The Swiss result is a qualified exception, not a settled counter-verdict. When you encounter a pairing claim, ask what it measures: who enters marriage, or which existing marriages end in divorce? Those are different questions. Then check the comparison. Does the study explain its expected number of couples and which sign combinations count as favorable? Finally, did the same rule produce a similar result in another population? A striking cell in one table matters less than a pattern that survives a clear rule and independent replication. This evidence has a defined reach. The studies assigned Sun signs from birth dates; they did not compare complete natal charts or interpret two people's Moon, Venus, Mars, and inter-chart aspects. Their results narrow claims about aggregate marriage and divorce records. They cannot tell you what a particular connection means. For a personal reading, start with the placements you know. Compare supplied Moon, Venus, and Mars signs, with optional degrees for selected major aspects. The Synastry Comparison Report uses those placements; it doesn't calculate natal charts from birth details. Read population evidence as context, then let the chart comparison address its own question.

Sources: The validity of astrological predictions on marriage and divorce: a longitudinal analysis of Swedish register data; Ten million marriages: A test of astrological ‘love signs’

Sources & further reading

Synastry Comparison Report

Enter known placements for two people. Get an input-by-input comparison, selected aspect calculations and a clear record of what remains uncertain.

Compare our placements

Keep exploring

Understanding synastry: explore the guide

Aries Love Horoscope for October 2026: Choose Clarity

Aries love in October 2026 favors honest pacing over instant certainty. Use Venus retrograde, the Libra New Moon, and Taurus’s Full Moon to test attraction against consistency.

Parallel and Contraparallel Contacts in Synastry

Learn how declination parallels and contraparallels add a second layer to zodiac-degree synastry, with orbs, examples, and a clear ranking method.

What Uranus, Neptune and Pluto Change in Synastry

Learn how Uranus, Neptune and Pluto reshape Moon, Venus, Mars, Sun and Mercury contacts in synastry, with degrees, orbs and practical questions.

How to Read a Retrograde Planet in a Synastry Contact

A retrograde planet in synastry isn't a relationship warning. Learn how natal motion, planet meaning, signs, degrees, orbs, and the wider contact change the reading.

Check a known aspect angle · How this publication reviews its articles