top of page

Can astrology predict life outcomes? We tested it.

Updated: Feb 13



Many people claim astrology is nonsense, whereas others swear by it. But is there actually a simple way to test if it works? It turns out, it’s not too hard to put the simple sun sign (a.k.a., zodiac sign) version of astrology to the test. 


So, we ran a study to do so.


If you want to see what we found out, or if you want to learn how these results speak to broader issues about belief systems and how to be a careful and critical thinker, then keep reading! If all you care about are the results, simply scroll down now to the section titled: The verdict: can astrology predict life outcomes?



Which version of astrology did we test?


Astrology has many traditions that vary in their methods, but one that is very popular (and relatively simple) is sun sign astrology. According to this belief system, the position of the sun at the time you were born determines which one of twelve different zodiac signs you are (capricorn, aquarius, pisces, etc), and people of those different signs are alleged to have distinct personalities and dispositions. YouGov polling reveals that almost a third of US adults "somewhat" or "strongly" agree  that sun sign astrology is accurate, with an additional 16% saying they "don't know":




Numerous popular websites and newspapers explicitly promote sun sign astrology, making claims like:


People think Geminis are two-faced, but this is perhaps more driven by their super-powered perceptiveness. They can simply see all sides to things, and even what might lie ahead, so they hedge their bets till the cards are on the table, for sure. Gemini is a shrewd player.


Bold, pioneering and courageous, Aries is the leader of the zodiac. They are daring and adventurous and are unafraid to strike out into unknown territory where others would be unable to go.


Numerous websites like these claim that there are very significant and fundamental differences between the personalities of people born at different times of the year. If this was the case, then we would expect to see differences in the life outcomes of people with different sun signs. 


That’s what we put to the test. We didn't test other astrological systems, some of which are much more complicated than sun sign astrology. 



How we ran this study


We started by measuring 37 life outcomes for a total of 308 people, along with their astrological sun signs. The outcomes we measured included things like: 


  • Levels of satisfaction with one’s (i) social, (ii) romantic, (iii) work, and (iv) general life circumstances.

  • Political alignment (on a progressive to conservative spectrum)

  • Religiosity

  • Amount that one hosts or plans events for friends, family, or coworkers

  • Whether one has received public recognition or awards for achievement

  • Time spent volunteering in one’s community in the last year

  • Number of distinct crimes for which one has received conviction

  • The presence of suicidal ideation in the last 10 years


And more. For a full list of outcomes measured, see the appendix at the end of this blog post.


Then we attempted to determine whether participants’ astrological sun signs could be used to predict their life outcomes. For instance, according to some sources Libras are the most social of the zodiac signs and Capricorn the least social. If this is really true, then we'd expect zodiac signs to be predictive of life outcomes such as how satisfied people are with their friendships, how often people host events, and how many close friends they report having.    


To perform these tests, we employed a statistical method known as linear regression (in particular, we used a type of linear regression called "L2" or "Ridge Regression" with cross-validation to avoid overfitting). Linear regression is used to try to predict a variable (in this case, any one of the life outcomes) using a set of other variables (in this case, whether or not a person was born into each of the 12 zodiac signs). Regression analysis also tells us how much each of these variables contributes to the prediction (which in this case indicates how associated each of the 12 zodiac signs is with that life outcome).  We applied this process to each of the 37 life outcomes one by one to see how well zodiac signs can predict them each.   


We quantified how well sun signs predicted outcomes using a statistical measure called the correlation coefficient, denoted as ‘r’. This coefficient measures the strength and direction of the relationship between two variables — here, the prediction of each life outcome using astrological sun signs and each life outcome itself. An r value close to 1 or -1 indicates a strong positive or negative correlation, respectively, while a value close to 0 suggests no correlation. 


Sometimes linear regression can "overfit" data, meaning it learns to fit not just the signal in the data but the noise as well. To help ensure our analysis didn't have this problem, we used Ridge Regression instead of ordinary least squares regression. But Ridge Regression has a parameter to be set (called "alpha", which determines the level of overfitting vs. underfitting). We varied this value over a very wide range to help ensure our conclusions were robust.  



The verdict: can astrology predict life outcomes?


In our data Astrological sun signs had literally zero predictive ability across all 37 outcomes. The correlation coefficients, or ‘r values’, were all 0, meaning that the zodiac signs had no ability whatsoever to predict any of the 37 outcomes! Take a look at the chart below:  



Average Accuracy of Sun Signs in Predicting Life Outcomes of 308 U.S. Participants




What the zeros indicate is that when testing the predictions on new people (i.e., people whose data was not used to fit the linear regression model), the accuracy was not better than just predicting the average value for each person (e.g., predicting that each person is of the average age, each person has the average BMI, etc.) In other words, taking into account astrological sun signs added nothing to the accuracy of the predictions.


How reliable is this method? Well, to test it out, we did an additional analysis, where instead of using people's real zodiac signs, we tried assigning each person a zodiac sign at random and then applied the exact same method to see what happened. When we did so, we got results almost exactly like the table shown above (for real zodiac signs). The only difference is that when we used random (fake) zodiac signs, there was one false positive (i.e., 1 of the 37 life outcomes appeared to be predictable, which we know it can't have been because the data was set up to be random).


Could it just be that these life outcomes are inherently unpredictable, so our study results tell us just about the unpredictability of life itself, rather than about the effectiveness of astrology? No, because we also collected the Big 5 personality test scores of our study participants, and as the chart below shows, using personality scores, we were able to predict life outcomes with a decent level of reliability. So the issue is not that these life outcomes are wholly unpredictable - it's that zodiac signs can't be used to predict them.



If we apply the same analysis (trying to predict one life outcome at a time) to the Big 5 personality traits that we did to zodiac signs, we see a very different pattern of results. In fact, we found that 22 of the 37 life outcomes had some level of predictability using people's Big 5 personality scores, including satisfaction with life, the amount they meditate, trauma, winning awards and work, and work promotions. Contrast this with our finding that none of the life outcomes were predictable using zodiac signs.


Another way to look at the data we collected is using classical statistics. In that paradigm, you can use what's known as an ANOVA test to see if each outcome varies more than you'd expect by chance based on astrological sun sign. When we do that, we get a p-value for each outcome reflecting the probability that you'd get as much variation in that outcome (when dividing people up by sun sign) as we found in our data, if in fact astrological sun signs had no predictive power at all. A lower p-value is associated with a greater probability of a real effect. By chance alone, even if astrological sun signs were completely useless for predicting these life outcomes, we'd expect to find roughly 2 outcomes with p<0.05. As you can see in the table below, we find only one p-value with p<0.05, which means we failed to find significant evidence of sun signs being predictive at all:





Our study provides evidence that sun sign astrology does not make accurate predictions about any of a wide variety of aspects of people’s lives and, therefore, cannot be relied on as a method for developing accurate beliefs about those aspects. However, please note that it cannot rule out tiny effects, even though it does provide evidence against moderate and large-sized effects. So if, for instance, being a Pisces made people 3% more likely to have deep emotional connections with people than other star signs, our method would not be able to detect that effect. On the other hand, if the effects were small that would make them far less useful for any practical purposes.


This inability of our method to identify small effects could also explain why our study didn't pick up on effects that previous studies have documented about the date of birth influencing outcomes like athletic success (due to being either one of the youngest on the team or one of the oldest on the team depending on whether you are born just before or just after an arbitrary cutoff). Such effects may well exist and may be especially important for special populations (such as elite athletes), but if, when averaged across the adult population, these effects are small, our method would not have been likely to identify them.


Of course, this doesn't completely exclude the possibility that sun sign astrology might predict something beyond the outcomes we looked at. But what could that be? If sun sign astrology were to predict a certain factor, let's call it ‘x’, and if x were related to any of the outcomes we examined, then by extension, sun sign astrology should have shown some predictive power for that outcome as well. So while we can't rule out that sun sign astrology predicts something, this study provides evidence that if it does predict something, what it predicts is wholly unrelated to any of the 37 outcomes we tested.  Considering the diverse range of these outcomes, it's challenging to imagine what unexplored territory could potentially be influenced by sun sign astrology.


We're not the only ones to fail to find any effects for sun sign astrology. For example, the online dating OkCupid once put it to the test by using their match algorithm (which they had previously shown had some significant ability to predict something reasonably construed as dating 'compatibility') to predict the compatibility between all pairs of zodiac signs, and here is what they found:




Zodiac signs had absolutely no influence on compatibility.


A major limitation of both our work here, and OkCupid's data analysis is that it is only testing sun signs. Many people who believe in astrology use much more complicated systems. For some of these more complicated systems (which still partially rely on sun signs), the complete failure to find any associations whatsoever between zodiac signs and life outcomes may still be evidence against the theory, whereas other astrological systems may be sufficiently unrelated to sun signs that this evidence doesn't say anything about them. Such systems are, unfortunately, often much more difficult to put to the test, in part because of their complexity, but also because each practitioner may apply them in somewhat different ways.



What does this mean for you?


Maybe the results of our study are not news to you. Astrology is, after all, a favorite punching bag of rationalists, scientists, and philosophers (perhaps because astrology is both very popular and appears to rely on no mechanism accepted by our current best scientific theories).


Our study helps to further show that criticism of sun sign astrology may be justified, but If you really care about thinking critically and rationally, you should reflect on how you use this information in your own life (beyond simply not believing in astrology).


It’s likely that, if you already believed that astrology is pseudoscience, then people interested in astrology are part of an ‘outgroup’ for you (a group of people you do not identify with or belong to), so you will probably find it easier to criticize them and their views than your own systems of belief. If you want to improve your critical thinking, you shouldn’t merely look for unreliable systems of belief in outgroups: you should look for unreliable systems of belief in yourself and your ingroups too. (Here is a good essay that we like on this subject.)


For example, maybe you like to apply principles from economics to your daily life. Some such applications may well be fruitful and have good predictive power, but economists have long pointed out that the predictive power of many economic principles is extremely limited without in-depth study, typically at an advanced university level, and many of the foundational assumptions of various economic models don’t apply in lots of situations (a much-discussed example of this is the limitations of the rational actor assumption). If you’re someone who likes using economic concepts in your daily life, it might be worth considering whether there are cases where you might be too quick to assume the applicability of those concepts. Do they make predictions you can check?


So, if you already rejected astrology, rather than letting this study lead you to self-congratulation for not believing in astrology, we suggest letting it prompt you to look for similar systems of belief in your own life. What theories about the world do you rely on regularly? Is there good evidence that they make reliable predictions?


For help thinking about this sort of thing broadly, you can try our Question Your Identity tool, here free:



Data analysis for this post was conducted by Emmanuel Nnaemeka.


If you'd like to analyze our astrological sun sign data for yourself, you can click here to download our data.



Appendix: List of Outcomes Tested


In order to measure outcomes for this study, we asked participants the following questions. Some are phrased as statements rather than questions and, in those cases, participants were asked to select how strongly they agree or disagree (on a 7-point Likert scale from “Totally agree” to “Totally disagree”).


  1. On average, how many hours do you sleep every night?

  2. What is the total number of income in U.S. dollars made by your entire household last year before taxes?

  3. What is your body mass index (BMI)?

  4. In politics, where do your views generally fall on the scale from "left" (progressive) to "right" (conservative)?

  5. Does the Low Carb diet align most closely with your regular eating habits?

  6. What is your current employment status (not including work through Amazon Mechanical Turk)?

  7. Do you currently own a home or have owned a home in the past?

  8. How much time per week do you dedicate to exercise of at least moderate intensity?

  9. On average, how much time do you spend meditating each day?

  10. Were there any crimes (e.g., arson, theft, fraud, etc.) in the last ten years where you were a victim?

  11. In the last 10 years, how many distinct crimes (e.g., theft, drug dealing, fraud, homicide, etc., but not traffic violations) were you convicted of?

  12. In the past year, how many different types of services (e.g., credit card bills, phone bills, utilities, car payments, insurance premiums, mortgage installments) have you failed to make a payment for?

  13. I am satisfied with my life right now.

  14. I am satisfied with my physical health.

  15. I am satisfied with my romantic life.

  16. I am satisfied with the quality and quantity of my social interactions and friendships.

  17. I am satisfied with my work life.

  18. In the last 5 years, have you used physical force against someone during an argument or disagreement?

  19. Have you had any traumatic experiences, and if so, to what extent do they continue to influence your daily life?

  20. How old are you?

  21. I am religious.

  22. In terms of my biological sex, I am a female.

  23. What is your highest level of education?

  24. In the past 10 years, to what extent have you considered committing suicide?

  25. In the past year, have you received public recognition or awards for your achievements?

  26. How many people do you share a deep emotional connection with?

  27. In the past year, how many times have you hosted or planned an event (e.g., birthday party, holiday, picnic) for your friends, family, or coworkers?

  28. In the last year, how many times have you actively sought out and received feedback from others to improve your skills?

  29. In the last 10 years, how long have you spent in jail or prison?

  30. On average, how many hours of paid work do you do per week (including all kinds of work you do that you get paid for)?

  31. How many close friends or family members do you have that would be willing to immediately help you in an emergency?

  32. In the last 5 years, how many times have you received a promotion at work?

  33. In the last 5 years, how many times were you fired (*not* including layoffs or company closures)?

  34. In the last 5 years, how many times did you have to stay at the hospital for medical care?

  35. In the past year, have you enrolled in a training or personal development course, including MOOCs (e.g., on Coursera), but excluding university courses?

  36. In the past year, how many times have you volunteered in your community?

  37. What is your current romantic relationship status?



8 Comments


Diego Oliveira
Diego Oliveira
Jan 15

I'm surprised by the full wall of zeroes, too. Not only would I expect some fluctuations by chance (not so sure about this. Even though I have studied statistics, I still have much more to learn about it), but I would also expect that people would be influenced by their zodiac sun sign via a social mechanism - perhaps people of a certain sign would feel inclined to act in a certain way, thereby producing some non-zero correlations despite the possibility of no real effect between date of birth and life outcomes or personality.

Like
Spencer Greenberg
Spencer Greenberg
Jan 16
Replying to

Thanks for your comment. We updated the post to include more information about the method, including showing what happened when we ran the exact same method on random data (where we assign people's zodiac sign at random instead of using their real zodiac sign), as well as on the big 5 personality data. Our method cannot rule out very small relationships, but it does provide evidence against moderate and large sized relationships.a

Like

Sylwia Pro
Sylwia Pro
Jan 12

Astrologers will tell you that someone's sun sign (i.e. what is normally referred to as the zodiac sign) is just one parameter of someone astrological profile and even not the most important one. "Life outcomes" are predicted by something else in astrology - planets in houses responsible for different areas of life and so on.

Like
Spencer Greenberg
Spencer Greenberg
Jan 16
Replying to

Yes, this approach does not test all forms of astrology, though it does test a common belief that many people have, in particular: "YouGov polling finds that almost a third of US adults "somewhat" or "strongly" agree  that sun sign astrology is accurate, with an additional 16% saying they "don't know""

Like

Jerry Feist
Jerry Feist
Jan 12

When you report correlations of 0, what is your rounding point? If something has a correlation of .14 in your chart, is that reported as 0?

Like
Spencer Greenberg
Spencer Greenberg
Jan 16
Replying to

Thanks for your comment. We use a 0 whenever the R (similar to a multivariate version of the correlatoin, r) is negative or zero. This occurs when the model's predictions on new people are worse than just assigning the mean value as the prediction for each person. We updated the post to include more information about the method, including showing what happened when we ran the exact same method on random data (where we assign people's zodiac sign at random instead of using their real zodiac sign), as well as on the big 5 personality data. You can see in the new image we link to that applying the method to totally random data produces a result that is very…

Like

randomcrap01
Jan 10

It seems pretty surprising that there would be zero correlation across the board with a sample size of 323 people -- even OK Cupid had a BIT of variation with a sample size of 500k. With 323 people and 37 possible outcomes to correlate to, I would have expected some of them to be different from 0 just by chance. Is this an artifact of using a Ridge regression rather than a more typical regression? Or is my intuition on that mistaken?


(PS -- pardon the username, it was derived automatically from my email & is not a comment on the quality of the site.)

Like
Spencer Greenberg
Spencer Greenberg
Jan 16
Replying to

Hi, good question. We've updated the post with more about the method we use, and show what happens when it is applied to random data (fake zodiac signs when we give each person a random sign rather than their real sign). As you can see, it produces a result that looks very much like the zodiac sign result. In the big table of zeros, we use a 0 whenever the R (similar to a multivariate version of the correlation, r) is negative or zero. This occurs when the model's predictions on new people are worse than just assigning the mean value as the prediction for each person.

Like
bottom of page