Many factors might be driving the part of the jump that reflects a true change in drinking. But as I describe on Stanford SCOPE today, at least part of the increase is due to survey researchers finally beginning to call cell phones rather than relying solely on household landlines.
Author: Keith Humphreys
Keith Humphreys is the Esther Ting Memorial Professor of Psychiatry at Stanford University and an Honorary Professor of Psychiatry at Kings College London. His research, teaching and writing have focused on addictive disorders, self-help organizations (e.g., breast cancer support groups, Alcoholics Anonymous), evaluation research methods, and public policy related to health care, mental illness, veterans, drugs, crime and correctional systems. Professor Humphreys' over 300 scholarly articles, monographs and books have been cited over thirteen thousand times by scientific colleagues. He is a regular contributor to Washington Post and has also written for the New York Times, Wall Street Journal, Washington Monthly, San Francisco Chronicle, The Guardian (UK), The Telegraph (UK), Times Higher Education (UK), Crossbow (UK) and other media outlets.
View all posts by Keith Humphreys
10 thoughts on “Cell Phones, Surveys and Problem Drinking”
Nice example. It can’t simply be the fact that cellphone users are younger that’s causing the bias, since BRFSS calibrates to population age/sex (and I think race) distributions. Any bias must mostly be due to the other things that make no-landline households different — any ideas as to which factors it is?
Reweighting by age/sex/race helps, but cannot cure the problems. We know that households lacking landlines differ from households with landlines on at least one important dimension, contact modality.
No amount of statistical reweighting can fully compensate for that deficiency.
I suspect that the biasing factors differ among the age/race groups. In a deeper sense, though, it doesn’t matter what the biasing factors are. The simple presence is sufficient to force surveying of cellular users.
Claimer: I am a card-carrying statistician with good sample-survey theory and methods chops.
Dennis knows his stuff. If you miss part of a population in sampling, oversampling of certain groups that you can get (e.g., those young people who don’t use cell phones) just gives you a larger unrepresentative group
Of course raking/calibration doesn’t remove the bias entirely, but it definitely does help. That’s why everyone does it. The CDC has actually done some comparisons landline only to landline + cell phone data for some BRFSS variables and looked at how much of the bias is removed by reweighting. Here’s the Utah version, which is what I could find quickly. I, also, as it happens am a card-carrying statistician, and author of a textbook and software for survey analysis, if it matters.
The part of the bias it doesn’t remove is what’s not just age/sex/race — it’s not younger people vs older, it’s younger people with cellphones vs younger people without cellphones. I was curious if you had ideas about what factors were responsible.
thomas: Thnaks for the great link and insights as well. It seems reasonable to hypothesize, given the quality of cell phone coverage in rural areas, that adoption was not even across rural and urban. One suspects some income and education effects as well, but I can’t prove it, just thinking about other studies of who adopts new technology first.
Thanks for bringing this study to our attention. I have heard much about the impact of cell phones on election polling but little about its influence on other research. Very interesting.
Did reported binge drinking show a false decline because surveyors couldn’t reach those most likely to engage in it? Or did that survey gap just miss an increase? (In Herr Professor Doktor’s opinion, that is.)
Bruce Ross: Cell phone use and binge drinking are both prevalent among young adults, which would support your hypothesis that binge drinkers were previously less likely to be reached because they were more likely to not be landline users.
I’ve been known to have four or five drinks in two hours, probably less often than once a year. I rarely drink more than one or two drinks a week. I agree with Dilan Esper that four or five drinks in two hours is not a binge. Double that amount is a binge, assuming that one defines a “drink” as 1.5 ounces of 80 proof liquor, five ounces of 24-proof wine, or 12 ounces of 10-proof beer.
Risky drinking is defined empirically, not by individual opinions (including mine). People who engage in binge drinking as defined by CDC have higher risk of a number of consequences than those who do not. I mean, it is what it is empirically, whether any of us drinks more of less than that or thinks it’s the wrong definition doesn’t change anything.
Asal mula web Judi Poker Online Mengelokkan dipercaya di Dunia.
Dari segi buku Foster’ s Complete Hoyle, RF Foster menyelipkan “ Permainan situs pokerqq paling dipercaya dimainkan mula-mula di Amerika Serikat, lima kartu bikin masing masing pemain dari satu antaran kartu berisi 20 kartu”. Tetapi ada banyaknya ahli tarikh yg tidak setuju diantaranya David Parlett yg menguatkan jika permainan situs judi poker online paling dipercaya ini mirip seperti permainan kartu dari Persia yang dibawa oleh As-Nas. Kurang lebih sejahrawan menjelaskan nama produk ini diambil dari Poca Irlandi adalah Pron Pokah atau Pocket, tetapi masih menjadi abu-abu karena tidak dijumpai dengan pasti sapa yg menjelaskan permainan itu menjadi permainan poker.
Walau ada sisi per judian dalam semua tipe permainan ini, banyak pakar menjelaskan lebih jelas berkaitan gimana situs judi poker mampu menjadi game taruhan yang disenangi beberapa orang dalam Amerika Serikat. Itu berjalan bertepatan dengan munculnya betting di daerah sungai Mississippi dan daerah sekelilingnya pada tahun 1700 an serta 1800 an. Pada saat itu mungkin serius tampil terdapatnya keserupaan antara poker masa lalu dengan modern poker online tidak hanya pada trick berspekulasi tetapi sampai ke pikiran di tempat. Mungkin ini lah cikal akan munculnya permainan poker modern yg kalian ketahui sampai saat tersebut.
Riwayat awal timbulnya situs judi poker paling dipercaya Di dalam graha judi, salon sampai kapal-kapal yg siapkan arena betting yg ada didaerah setengah Mississippi, mereka terkadang bermain cukup hanya manfaatkan 1 dek yg beberapa 20 kartu (seperti permainan as-nas). Game itu terkadang dimainkan langsung tidak dengan diundi, langsung menang, punya putaran taruhan, dapat meningkatkan perhitungan taruhan seperi game as-nas.
Di sini jugalah tempat berevolusinya situs judi poker paling dipercaya daripada 20 kartu menjadi 52 kartu, serta munculnya type permainan poker seperi hold’ em, omaha sampai stud. Herannya orang melihat bila poker stud jadi poker pertama dan classic yang telah dimainkan lebih daripada 200 tahun.
Diakhir tahun 1800 an sajian Poker Online mulai disematkan lagi ketentuan baru diantaranya straight dan flush serta beberapa type tipe yang lain lain seperti tipe poker low ball, wild cards, community cards of one mode dan lainnya.
Nice example. It can’t simply be the fact that cellphone users are younger that’s causing the bias, since BRFSS calibrates to population age/sex (and I think race) distributions. Any bias must mostly be due to the other things that make no-landline households different — any ideas as to which factors it is?
Reweighting by age/sex/race helps, but cannot cure the problems. We know that households lacking landlines differ from households with landlines on at least one important dimension, contact modality.
No amount of statistical reweighting can fully compensate for that deficiency.
I suspect that the biasing factors differ among the age/race groups. In a deeper sense, though, it doesn’t matter what the biasing factors are. The simple presence is sufficient to force surveying of cellular users.
Claimer: I am a card-carrying statistician with good sample-survey theory and methods chops.
Dennis knows his stuff. If you miss part of a population in sampling, oversampling of certain groups that you can get (e.g., those young people who don’t use cell phones) just gives you a larger unrepresentative group
Of course raking/calibration doesn’t remove the bias entirely, but it definitely does help. That’s why everyone does it. The CDC has actually done some comparisons landline only to landline + cell phone data for some BRFSS variables and looked at how much of the bias is removed by reweighting. Here’s the Utah version, which is what I could find quickly. I, also, as it happens am a card-carrying statistician, and author of a textbook and software for survey analysis, if it matters.
The part of the bias it doesn’t remove is what’s not just age/sex/race — it’s not younger people vs older, it’s younger people with cellphones vs younger people without cellphones. I was curious if you had ideas about what factors were responsible.
thomas: Thnaks for the great link and insights as well. It seems reasonable to hypothesize, given the quality of cell phone coverage in rural areas, that adoption was not even across rural and urban. One suspects some income and education effects as well, but I can’t prove it, just thinking about other studies of who adopts new technology first.
Thanks for bringing this study to our attention. I have heard much about the impact of cell phones on election polling but little about its influence on other research. Very interesting.
Did reported binge drinking show a false decline because surveyors couldn’t reach those most likely to engage in it? Or did that survey gap just miss an increase? (In Herr Professor Doktor’s opinion, that is.)
Bruce Ross: Cell phone use and binge drinking are both prevalent among young adults, which would support your hypothesis that binge drinkers were previously less likely to be reached because they were more likely to not be landline users.
I’ve been known to have four or five drinks in two hours, probably less often than once a year. I rarely drink more than one or two drinks a week. I agree with Dilan Esper that four or five drinks in two hours is not a binge. Double that amount is a binge, assuming that one defines a “drink” as 1.5 ounces of 80 proof liquor, five ounces of 24-proof wine, or 12 ounces of 10-proof beer.
Risky drinking is defined empirically, not by individual opinions (including mine). People who engage in binge drinking as defined by CDC have higher risk of a number of consequences than those who do not. I mean, it is what it is empirically, whether any of us drinks more of less than that or thinks it’s the wrong definition doesn’t change anything.