This looks to be a correlational study. As an exercise, let’s try thinking of confounding effects that would point in both directions.
If I had ED, then I imagine (a) I would try looking within porn to see if there was some way to sustain arousal, and (b) the problems with ED might mean I’d have less sex with a partner (or even break up, in marginal relationships; or lack the confidence to form new ones), and have to satisfy my sex drive by myself more often. Thus, ED causes more porn use, and we’d expect them to be correlated. Therefore, a null result means there must be a counter effect: porn use must reduce ED!
(a) If I had ED, and I believed the “common wisdom” that porn causes ED, then I would avoid porn; in other words, ED causes less porn use. (b) I would guess that a lower sex drive can cause both ED and reduced porn use. Both of these effects imply anticorrelation, and therefore the study’s result of a null correlation means porn use must cause ED.
Some of these might be investigated and/or controlled for. Let’s imagine controlling for 1(b), by looking at single men. Now let’s try to imagine what could screw up the analysis. Consider these two worlds:
For those with a high sex drive, ED is no handicap to relationships, because one compensates with oral sex and other measures. But having low sex drive and ED will lead to a breakup. Therefore, by restricting our sample to single men, we’re creating an extra correlation between ED and low sex drive; and low sex drive causes less porn use, so we expect this to yield an anticorrelation between ED and porn use (and therefore null result means porn use causes ED).
Having a high sex drive has no effect on whether ED causes you and your partner to break up. Therefore, null result means null causation.
In general, if there’s a piece of the causal chain you don’t know about and could go either way, then whatever analysis you do can’t yield the correct answer in both worlds. If you have enough data, accurate measurements of all relevant variables, then you might be able to account for all confounding effects and end up isolating the causation you want.
Checking out the actual study… they controlled for exactly two things: age and education. The sample was also restricted to “sexually active” men, which it doesn’t seem to define. (Does it mean they’re currently in a relationship? Have had sex in the last 12 months? If a bad case of ED means a man hasn’t had sex in years despite wanting to, does this exclude him from the study? Surely such men are the most important ones for the study’s goals?) They did ask about sex drive… but in study 1, they lumped in “lack of sexual desire” with “sexual difficulties” that include ED, and in study 2, they asked specifically about a reduction in sex drive in the last 12 months, but seemingly nothing about overall sex drive.
Well, I was going to say: (a) if the “high sex drive protects relationships from ED” hypothesis is true (which I just made up; I suspect it’s a real but weak effect), then this would leave us with a sample where ED is extra-correlated with high sex drive, which could be relevant; more importantly, (b) within a relationship, I expect things like “the man becomes less attracted to his partner” or “emotional conflicts or other relationship problems interfere with the attraction” (which have many possible causes, and I think are not rare) to cause both “instances of ED when the man tries to have sex with his partner” and “the man to use porn more often”. I expect (b) is a significant effect.
And even if there were a study that controlled for all the above, I can come up with more effects, at least some of which would be plausibly significant. Not to mention, controlling for something requires measuring it, and I’m not sure things like “the man feeling emotionally distant in a way that may interfere with attraction” could be accurately quantified in survey questions. This is why I want studies with a randomized intervention.
Technically, the title of the study asks about an “association”—that is, a correlation—and it delivered on that. But I don’t think anyone seriously cares about the association except insofar as it sheds light on causation. (If they truly didn’t care about causation, then why do any controls?) Thus, despite the study’s size, in terms of causality it looks pretty impotent.
Discussion and Conclusions
[...] The only significant relationship was observed in the 2011 Croatian sample (Study 1) between pornography use and ED. The direction of this association is unclear, as pornography use may also be a way to cope with sexual difficulties or decreased sexual satisfaction.
Ya think?! (And are they using the word “association” to mean “causation” in that sentence? I’m not sure what a “directed association” would be otherwise. Is this a motte and bailey on the word “association”? The paper comes from a Croatian university—maybe a translation issue? Google says half of Croatians speak English fluently.) Surely this was foreseeable. I guess the charitable explanation is that they would have done followups to tease out the details if it looked like there was a big effect.
The study shows there can’t be a huge effect that outweighs all confounders, but I think some potential confounders are at least medium-sized… and I do worry that the “sexually active” criterion might have excluded a lot of central examples of the phenomenon they’re supposed to be investigating. In study 1, only 52% of men who took the survey met all their criteria (including filling out the whole survey), and in study 2 it’s only 26%. Also, lumping in low sex drive with ED, as study 1 did, would probably reduce the correlation of ED with porn use (given that low sex drive probably reduces porn use).
Of course, nobody listens to science.
I’m afraid I agree with those who don’t listen to whatever this study is an exemplar of. I want to believe their conclusion, and I think it seems likely based on my “armchair theorizing and amateur observation”, but their contribution has not really updated my beliefs or confidence. Actually, I feel a little more worried than I did before.
This looks to be a correlational study. As an exercise, let’s try thinking of confounding effects that would point in both directions.
If I had ED, then I imagine (a) I would try looking within porn to see if there was some way to sustain arousal, and (b) the problems with ED might mean I’d have less sex with a partner (or even break up, in marginal relationships; or lack the confidence to form new ones), and have to satisfy my sex drive by myself more often. Thus, ED causes more porn use, and we’d expect them to be correlated. Therefore, a null result means there must be a counter effect: porn use must reduce ED!
(a) If I had ED, and I believed the “common wisdom” that porn causes ED, then I would avoid porn; in other words, ED causes less porn use. (b) I would guess that a lower sex drive can cause both ED and reduced porn use. Both of these effects imply anticorrelation, and therefore the study’s result of a null correlation means porn use must cause ED.
Some of these might be investigated and/or controlled for. Let’s imagine controlling for 1(b), by looking at single men. Now let’s try to imagine what could screw up the analysis. Consider these two worlds:
For those with a high sex drive, ED is no handicap to relationships, because one compensates with oral sex and other measures. But having low sex drive and ED will lead to a breakup. Therefore, by restricting our sample to single men, we’re creating an extra correlation between ED and low sex drive; and low sex drive causes less porn use, so we expect this to yield an anticorrelation between ED and porn use (and therefore null result means porn use causes ED).
Having a high sex drive has no effect on whether ED causes you and your partner to break up. Therefore, null result means null causation.
In general, if there’s a piece of the causal chain you don’t know about and could go either way, then whatever analysis you do can’t yield the correct answer in both worlds. If you have enough data, accurate measurements of all relevant variables, then you might be able to account for all confounding effects and end up isolating the causation you want.
Checking out the actual study… they controlled for exactly two things: age and education. The sample was also restricted to “sexually active” men, which it doesn’t seem to define. (Does it mean they’re currently in a relationship? Have had sex in the last 12 months? If a bad case of ED means a man hasn’t had sex in years despite wanting to, does this exclude him from the study? Surely such men are the most important ones for the study’s goals?) They did ask about sex drive… but in study 1, they lumped in “lack of sexual desire” with “sexual difficulties” that include ED, and in study 2, they asked specifically about a reduction in sex drive in the last 12 months, but seemingly nothing about overall sex drive.
Well, I was going to say: (a) if the “high sex drive protects relationships from ED” hypothesis is true (which I just made up; I suspect it’s a real but weak effect), then this would leave us with a sample where ED is extra-correlated with high sex drive, which could be relevant; more importantly, (b) within a relationship, I expect things like “the man becomes less attracted to his partner” or “emotional conflicts or other relationship problems interfere with the attraction” (which have many possible causes, and I think are not rare) to cause both “instances of ED when the man tries to have sex with his partner” and “the man to use porn more often”. I expect (b) is a significant effect.
And even if there were a study that controlled for all the above, I can come up with more effects, at least some of which would be plausibly significant. Not to mention, controlling for something requires measuring it, and I’m not sure things like “the man feeling emotionally distant in a way that may interfere with attraction” could be accurately quantified in survey questions. This is why I want studies with a randomized intervention.
Technically, the title of the study asks about an “association”—that is, a correlation—and it delivered on that. But I don’t think anyone seriously cares about the association except insofar as it sheds light on causation. (If they truly didn’t care about causation, then why do any controls?) Thus, despite the study’s size, in terms of causality it looks pretty impotent.
Ya think?! (And are they using the word “association” to mean “causation” in that sentence? I’m not sure what a “directed association” would be otherwise. Is this a motte and bailey on the word “association”? The paper comes from a Croatian university—maybe a translation issue? Google says half of Croatians speak English fluently.) Surely this was foreseeable. I guess the charitable explanation is that they would have done followups to tease out the details if it looked like there was a big effect.
The study shows there can’t be a huge effect that outweighs all confounders, but I think some potential confounders are at least medium-sized… and I do worry that the “sexually active” criterion might have excluded a lot of central examples of the phenomenon they’re supposed to be investigating. In study 1, only 52% of men who took the survey met all their criteria (including filling out the whole survey), and in study 2 it’s only 26%. Also, lumping in low sex drive with ED, as study 1 did, would probably reduce the correlation of ED with porn use (given that low sex drive probably reduces porn use).
I’m afraid I agree with those who don’t listen to whatever this study is an exemplar of. I want to believe their conclusion, and I think it seems likely based on my “armchair theorizing and amateur observation”, but their contribution has not really updated my beliefs or confidence. Actually, I feel a little more worried than I did before.