Bagels and Steroids

I find myself in an awkward position in discussions on steroids. I’ve written quite a bit on the subject, mostly critical of those who claim steroids are ruining the game. The reason for this is that I don’t see anything in the stats that shows players are getting much help from PEDs. However, this is a different issue from whether or not players are using them.

I understand that the incentives are such that if steroids do improve performance that we ought to expect players to use them. If you think you can get a boost over your competitors, many of whom you suspect of juicing, why not? This leads me to wonder: how many players use steroids? Ken Caminiti suggested a at least 50%. Tony Gwynn estimated it’s 20%. At a minimum, MLB’s tests from the 2003 season show 5-7% of players “failed” steroid tests, though “failure” included those who refused to be tested.

Quantifying dishonest behavior is difficult. Even when you promise amnesty for revealing bad behavior, people just don’t like to cop to it. So, all we have to go on in baseball is information from three sources. First, we have the testimony of players with every incentive to bias their estimates upwards (i.e., Caminiti and Conseco)—”hey, everybody is doing it, I’m not such bad guy.” Then we have the guessers like Gwynn, which reminds me of smelling marijuana in the in high school locker room. You smell a funny smell so you know it’s going on. You have a general idea of who’s responsible, but you don’t want any problems so keep your head down and don’t ask any questions. Names and numbers are based on hunches. Then we have the “anonymous” tests—tests that everyone knows are coming with logistics that make instant testing impossible. And let’s not forget that these tests can be beaten.

I think between 10-15% of major league baseball players use some type of illegal PED. Where do I get this number? From looking at cheating where we can identify it: the bagel man from Freakonomics. If you haven’t read the book or don’t recall, Paul Feldman is the economist who quit his job to be a full-time bagel supplier on a quasi-honor system in Washington, DC. And like an economist should, he payed close attention to the data he was generating. For the specifics of his findings read the book or the linked article, but the in the end he found that about 13% of his customers didn’t pay for his bagels. It seems to me that this is a good place to start in looking for the percentage of players willing to juice in MLB. If 87% of the population is honest enough not to steal bagels, that’s probably close to the percentage of players who are unwilling to use steroids to get ahead.

Certainly, the rewards are higher for juicers than for bagel thieves, but I believe that the ethical constraints that prevent cheating are similar despite the reward. Though I think there is some moral wiggle room, either you are honest or you are not. And I don’t think ballplayers are any more or less honest than the general population.

16 Responses “Bagels and Steroids”

  1. Mike F. says:

    The bagel example from Freakonomics is a good starting point, however, the 13% of customers who do not pay for the bagels create no incentive for “honest” people to become cheaters. In baseball it is possible that 13% of the population cheated with no remorse or hesitation, yet it also forced other players to consider using them in order to keep up with competition. If Jim Leyritz is taking steroids than my best response is to take them as well (assuming no negative consequences)or else I will perceive that Jim Leyritz will put up better numbers than me…and that’s embarrasing.

  2. NKW says:

    This might also be a self selecting group. If we grant that PED’s do, in fact, enhance performance, then it would be safe to assume that the 13% who are dishonest would rank among the best players in the world.

    In simpler terms, if 13 of 100 people take steroids, it is safe to assume that they would be among the better of the 100. Maybe not the top 13, maybe someone’s natural ability vaulted them to the top spot, but that the 13 are, in some way more likely to fall in the top. Or, put differently, the top is more likely than 13% to be dishonest because the entire population of dishonest people is in “the top”.

    Now, If we select major leaguers by rank with respect to the population, it would be a safe assumption to say we select the cream of the crop. But we’ve established that the cream of this crop is more likely to be tainted.

    I hope what I’m saying comes through and I don’t come off as a complete crackpot. But basically, while I accept the 13% argument, I think it is either incorrect to apply it here, or at least slightly misleading.

  3. jcombs says:

    I don’t think you can really compare the two situations. The payoff difference is just so large, that it leads me to believe that Gwynn’s number of 15-20% (and maybe a little higher) is accurate.

    When one cheats for one season in baseball, his payoff from this one chemically altered season can be disproportionate to the investment (risk of getting caught, suspension, lost pay, and fines). The theives in the Freakonomics example are only receiving one bagel for one theft. Cheating baseball players in contract years can be paid (to extend the analogy) four or five bagels for one bagel of investment.

  4. Andrew says:

    I like NKW’s argument and agree with it. Maybe 13% of all people who played baseball, say starting in high school, chose to use PEDs. Assuming PEDs do what they do, then essentially the “Quad-A” player becomes a big leaguer.

  5. John says:

    Is there a material probability of being caught stealing the bagels? If not, then the cost-benefit analysis disintegrates into the value of the bagel weighed against moral qualms of stealing. The 13% bagel figure is an artifact of those specialized conditions.

    In PED usage, the benefit is the projected performance increase. And the cost is the probability of being caught times the penalty associated with being caught.

  6. J. Cross says:

    I think the 13% might be a good pre-testing estimate for the reasons explained in the post but, like John says in #5, once there’s a real chance of getting caught I don’t think this scenario is too close to the bagel scenario any longer.

  7. NKW says:

    J. Cross: I too agree with John, but I’m not entirely sure which way you lean. It seems you lean towards larger risks than rewards and less than 13% usage. But really, how big a chance there is to get caught? HGH is untestable, and samples aren’t saved, so what is the probability of being caught? I’d say not very high unless you’re friends with Jason Grimsley.

    Also, you have to figure in that the rewards are significantly higher too. I’m sure steroid users increase their income by more than a couple bagels.

    So while I do agree with John (except I would argue the benefit is the salary bump, not the performance increase… which correlate but aren’t the same), I don’t think its safe to say it clearly is higher or clearly is lower based on that argument. I personally believe the argument leads to it being higher in John’s case as well as in my original argument. Either way, steroid use is probably higher than the 13% figure given.

  8. Jim says:

    Aren’t there some statistical techniques that pollsters often use to reduce the incentive to lie? One I seem to recall is where they ask the responder to flip a coin privately, and answer truthfully only if it comes up heads. I thought they used this sort of polls to determine rates of drug use, sexual activity, etc. among teens. It would be interesting to see this type of a poll done with MLB players and PEDs.

  9. J. Cross says:

    NKW, I agree that there might be other reasons that this doesn’t entirely match the bagel situation but I think the key component of the bagel guy story is it tells how many people are willing to get cheat when there’s no chance of getting caught. I think it also shows that cheating isn’t completely a risk/reward thing. Most people won’t cheat even when there is no risk of getting caught at all. I also think that people *might* be more willing to cheat when it comes to things that are considered small potatoes than things that are higher reward (and correspondingly higher level cheating) so I wouldn’t necessarily assume that the higher stakes of steroid use implies more cheating.

    My WAG is that the 13% is a reasonable estimate for pre-testing cheating levels* but the cheating has been approximately cut in half by testing. Maybe I’m overly optimistic about the chance (or perceived chance) of catching cheaters though.

    *I’d actually guess that pre-testing cheating levels we’re slightly higher than 13% for reasons stated above 1) cheaters are slightly more likely to be in the majors as opposed to AAA and 2) cheating goes up with the perception that other players are gaining an advantage from cheating. Still, these might be small factors and 13% could be close.

    Jim, is the responder told that they MUST lie if the coin flip comes up the tails? If so, wouldn’t all yes/no questions get 50% results since half the responders are giving the opposite answer?

  10. NKW says:

    J. Cross:

    Firstly, the coin flip test. I believe that if its outcome A (say, heads), they are told to say they do, in fact, take drugs/steroids/whatever. If its tails, then they are told to tell the truth, whatever it may be. This way, each responder could claim that his/her admission of guilt was actually due to outcome A (thus never admitting guilt), and the tester can determine with fairly good accuracy the amount of drug usage/whatever (ie, over half the responses are “guilty”. If we remove enough guilties to represent 50% of the whole sample, the remaining guilty/innocents provide a relatively good ratio).

    It’s interesting, I hadn’t considered that people would be less likely to cheat on “less important” items. While I would personally disagree with that, its something to consider. Personally, I’m of the belief that if I’m going to sacrifice my morals by cheating, I’m only going to do it for a damn high risk.

    As for the liklihood of being caught for steroids, I take the pessimistic view. Yes, 99% of steroids will be tested for and cheaters caught. However, HGH (and other undetectable steroids) can (and without any evidence, I believe they do) represent a disproportionately high ratio of steroid users. I mean, if given the choice between a ‘roid that can get me caught and one that won’t, its a no brainer. So yeah, the testing works, but it serves only (in my pessimistic view) to introduce would-be users to HGH and Balco-like companies.

    Of course, the question to refute my argument would be the case study of Mr. Palmeiro. Why would a man with so much reputation to lose, with so much money at his disposal to buy any concoction he so desires, buy an “inferior” steroid? Does his positive test represent the entirety of the “stars” who use? Was he just stupid? I don’t know and it is a question I’ve considered quite a bit.

    Also, one last thing J. Cross- I don’t mean to attack or anything, I hope it doesn’t appear that way. It just seems we’ve got polar opposite opinions.

  11. NKW says:

    Error in my previous comment – paragraph 2 – “damn high reward”… I really ought to proofread.

  12. Jim says:

    As I recall the way it works is if the coin flip comes up tails, the respondants are to answer “yes” regardless of what the truth is. So if 76% respond yes, we can estimate that 13% answered yes truthfully. That randomized response would preserve the anonymity of anyone who is guilty.

  13. Jim says:

    Actually, I think I got the math wrong. If 76% respond yes you would estimate 26/50 = 52% guilty. To get 13% guilty you’d expect 56.5% yes responses.

  14. NKW says:

    Jim, I believe the second version of the math is right.

  15. J. Cross says:

    NKW, thanks for the coin flip explanation.

    I guess my thought that some people might be more willing to cheat on small things comes from personal experience. If I sell something for $20 on eBay I don’t bother reporting it on my taxes. If I made thousands of dollars on eBay I’d feel morally compelled to report it. To give another example, the other day the balls popped out of a bar pool table before I inserted the $1.50 in quarters. I didn’t bother to bring $1.50 up to the bar. If, on the other hand, I found $100 that belonged to the bar I would have brought it to the bartender.

    Maybe I’m just a lazy person. I would certainly pay for my donuts, however.

  16. rob says:

    i’m thinking the same as a few other folks who posted here. there is plenty of incentive for an athlete to cheat (i.e., take peds) where as the incentive is much less for one of us to cheat and steal a bagel. given that, 13% is probably a starting point.

    we know that professional sports is all about being better than the next guy and that means you must keep up (common sense tells me that if player a is using and getting better, player b might do the same. for every cal ripken there is probably a wally joyner who will at least try it). we also know baseball’s ped-testing program is pourous at best. there is not a huge risk of getting caught (as i understand it as a layman) if you’re smart and taking the right *stuff*.