2011 Survey Results

A big thank you to the 1090 peo­ple who took the sec­ond Less Wrong Cen­sus/​Sur­vey.

Does this mean there are 1090 peo­ple who post on Less Wrong? Not nec­es­sar­ily. 165 peo­ple said they had zero karma, and 406 peo­ple skipped the karma ques­tion—I as­sume a good num­ber of the skip­pers were peo­ple with zero karma or with­out ac­counts. So we can only prove that 519 peo­ple post on Less Wrong. Which is still a lot of peo­ple.

I apol­o­gize for failing to ask who had or did not have an LW ac­count. Be­cause there are a num­ber of these failures, I’m putting them all in a com­ment to this post so they don’t clut­ter the sur­vey re­sults. Please talk about changes you want for next year’s sur­vey there.

Of our 1090 re­spon­dents, 972 (89%) were male, 92 (8.4%) fe­male, 7 (.6%) tran­sex­ual, and 19 gave var­i­ous other an­swers or ob­jected to the ques­tion. As abysmally male-dom­i­nated as these re­sults are, the per­cent of women has tripled since the last sur­vey in mid-2009.

We’re also a lit­tle more di­verse than we were in 2009; our per­cent non-whites has risen from 6% to just be­low 10%. Along with 944 whites (86%) we in­clude 38 His­pan­ics (3.5%), 31 East Asi­ans (2.8%), 26 In­dian Asi­ans (2.4%) and 4 blacks (.4%).

Age ranged from a sup­posed min­i­mum of 1 (they start mak­ing ra­tio­nal­ists early these days?) to a more plau­si­ble min­i­mum of 14, to a max­i­mum of 77. The mean age was 27.18 years. Quar­tiles (25%, 50%, 75%) were 21, 25, and 30. 90% of us are un­der 38, 95% of us are un­der 45, but there are still eleven Less Wrongers over the age of 60. The av­er­age Less Wronger has aged about one week since spring 2009 - so clearly all those anti-agath­ics we’re tak­ing are work­ing!

In or­der of fre­quency, we in­clude 366 com­puter sci­en­tists (32.6%), 174 peo­ple in the hard sci­ences (16%) 80 peo­ple in fi­nance (7.3%), 63 peo­ple in the so­cial sci­ences (5.8%), 43 peo­ple in­volved in AI (3.9%), 39 philoso­phers (3.6%), 15 math­e­mat­i­ci­ans (1.5%), 14 statis­ti­ci­ans (1.3%), 15 peo­ple in­volved in law (1.5%) and 5 peo­ple in medicine (.5%).

48 of us (4.4%) teach in academia, 470 (43.1%) are stu­dents, 417 (38.3%) do for-profit work, 34 (3.1%) do non-profit work, 41 (3.8%) work for the gov­ern­ment, and 72 (6.6%) are un­em­ployed.

418 peo­ple (38.3%) have yet to re­ceive any de­grees, 400 (36.7%) have a Bach­e­lor’s or equiv­a­lent, 175 (16.1%) have a Master’s or equiv­a­lent, 65 peo­ple (6%) have a Ph.D, and 19 peo­ple (1.7%) have a pro­fes­sional de­gree such as an MD or JD.

345 peo­ple (31.7%) are sin­gle and look­ing, 250 (22.9%) are sin­gle but not look­ing, 286 (26.2%) are in a re­la­tion­ship, and 201 (18.4%) are mar­ried. There are strik­ing differ­ences across men and women: women are more likely to be in a re­la­tion­ship and less likely to be sin­gle and look­ing (33% men vs. 19% women). All of these num­bers look a lot like the ones from 2009.

27 peo­ple (2.5%) are asex­ual, 119 (10.9%) are bi­sex­ual, 24 (2.2%) are ho­mo­sex­ual, and 902 (82.8%) are het­ero­sex­ual.

625 peo­ple (57.3%) de­scribed them­selves as monog­a­mous, 145 (13.3%) as polyamorous, and 298 (27.3%) didn’t re­ally know. Th­ese num­bers were similar be­tween men and women.

The most pop­u­lar poli­ti­cal view, at least ac­cord­ing to the much-ma­ligned cat­e­gories on the sur­vey, was liber­al­ism, with 376 ad­her­ents and 34.5% of the vote. Liber­tar­i­anism fol­lowed at 352 (32.3%), then so­cial­ism at 290 (26.6%), con­ser­va­tivism at 30 (2.8%) and com­mu­nism at 5 (.5%).

680 peo­ple (62.4%) were con­se­quen­tial­ist, 152 (13.9%) virtue ethi­cist, 49 (4.5%) de­on­tol­o­gist, and 145 (13.3%) did not be­lieve in moral­ity.

801 peo­ple (73.5%) were athe­ist and not spiritual, 108 (9.9%) were athe­ist and spiritual, 97 (8.9%) were ag­nos­tic, 30 (2.8%) were deist or pan­the­ist or some­thing along those lines, and 39 peo­ple (3.5%) de­scribed them­selves as the­ists (20 com­mit­ted plus 19 luke­warm)

425 peo­ple (38.1%) grew up in some fla­vor of non­the­ist fam­ily, com­pared to 297 (27.2%) in com­mit­ted the­ist fam­i­lies and 356 in luke­warm the­ist fam­i­lies (32.7%). Com­mon fam­ily re­li­gious back­grounds in­cluded Protes­tantism with 451 peo­ple (41.4%), Catholi­cism with 289 (26.5%) Jews with 102 (9.4%), Hin­dus with 20 (1.8%), Mor­mons with 17 (1.6%) and tra­di­tional Chi­nese re­li­gion with 13 (1.2%)

There was much de­ri­sion on the last sur­vey over the av­er­age IQ sup­pos­edly be­ing 146. Clearly Less Wrong has been dumbed down since then, since the av­er­age IQ has fallen all the way down to 140. Num­bers ranged from 110 all the way up to 204 (for refer­ence, Mar­ilyn vos Sa­vant, who holds the Guin­ness World Record for high­est adult IQ ever recorded, has an IQ of 185).

89 peo­ple (8.2%) have never looked at the Se­quences; a fur­ther 234 (32.5%) have only given them a quick glance. 170 peo­ple have read about 25% of the se­quences, 169 (15.5%) about 50%, 167 (15.3%) about 75%, and 253 peo­ple (23.2%) said they’ve read al­most all of them. This last num­ber is ac­tu­ally lower than the 302 peo­ple who have been here since the Over­com­ing Bias days when the Se­quences were still be­ing writ­ten (27.7% of us).

The other 72.3% of peo­ple who had to find Less Wrong the hard way. 121 peo­ple (11.1%) were referred by a friend, 259 peo­ple (23.8%) were referred by blogs, 196 peo­ple (18%) were referred by Harry Pot­ter and the Meth­ods of Ra­tion­al­ity, 96 peo­ple (8.8%) were referred by a search en­g­ine, and only one per­son (.1%) was referred by a class in school.

Of the 259 peo­ple referred by blogs, 134 told me which blog referred them. There was a very long tail here, with most blogs only refer­ring one or two peo­ple, but the over­whelming win­ner was Com­mon Sense Athe­ism, which is re­spon­si­ble for 18 cur­rent Less Wrong read­ers. Other im­por­tant blogs and sites in­clude Hacker News (11 peo­ple), Marginal Revolu­tion (6 peo­ple), TV Tropes (5 peo­ple), and a three way tie for fifth be­tween Red­dit, Se­bas­ti­anMar­shall.com, and You Are Not So Smart (3 peo­ple).

Of those peo­ple who chose to list their karma, the mean value was 658 and the me­dian was 40 (these num­bers are pretty mean­ingless, be­cause some peo­ple with zero karma put that down and other peo­ple did not).

Of those peo­ple will­ing to ad­mit the time they spent on Less Wrong, af­ter elimi­nat­ing one out­lier (sorry, but you don’t spend 40579 min­utes daily on LW; even I don’t spend that long) the mean was 21 min­utes and the me­dian was 15 min­utes. There were at least a dozen peo­ple in the two to three hour range, and the win­ner (well, ex­cept the 40579 guy) was some­one who says he spends five hours a day.

I’m go­ing to give all the prob­a­bil­ities in the form [mean, (25%-quar­tile, 50%-quar­tile/​me­dian, 75%-quar­tile)]. There may have been some prob­lems here re­volv­ing around peo­ple who gave num­bers like .01: I didn’t know whether they meant 1% or .01%. Ex­cel helpfully rounded all num­bers down to two dec­i­mal places for me, and af­ter a while I de­cided not to make it stop: un­less I wanted to do ge­o­met­ric means, I can’t do jus­tice to re­ally small grades in prob­a­bil­ity.

The Many Wor­lds hy­poth­e­sis is true: 56.5, (30, 65, 80)
There is in­tel­li­gent life el­se­where in the Uni­verse: 69.4, (50, 90, 99)
There is in­tel­li­gent life el­se­where in our galaxy: 41.2, (1, 30, 80)
The su­per­nat­u­ral (on­tolog­i­cally ba­sic men­tal en­tities) ex­ists: 5.38, (0, 0, 1)
God (a su­per­nat­u­ral cre­ator of the uni­verse) ex­ists: 5.64, (0, 0, 1)
Some re­vealed re­li­gion is true: 3.40, (0, 0, .15)
Aver­age per­son cry­on­i­cally frozen to­day will be suc­cess­fully re­vived: 21.1, (1, 10, 30)
Some­one now liv­ing will reach age 1000: 23.6, (1, 10, 30)
We are liv­ing in a simu­la­tion: 19, (.23, 5, 33)
Sig­nifi­cant an­thro­pogenic global warm­ing is oc­cur­ring: 70.7, (55, 85, 95)
Hu­man­ity will make it to 2100 with­out a catas­tro­phe kil­ling >90% of us: 67.6, (50, 80, 90)

There were a few sig­nifi­cant de­mo­graph­ics differ­ences here. Women tended to be more skep­ti­cal of the ex­treme tran­shu­man­ist claims like cry­on­ics and an­ti­a­gath­ics (for ex­am­ple, men thought the cur­rent gen­er­a­tion had a 24.7% chance of see­ing some­one live to 1000 years; women thought there was only a 9.2% chance). Older peo­ple were less likely to be­lieve in tran­shu­man­ist claims, a lit­tle less likely to be­lieve in an­thro­pogenic global warm­ing, and more likely to be­lieve in aliens liv­ing in our galaxy. Com­mu­nity vet­er­ans were more likely to be­lieve in Many Wor­lds, less likely to be­lieve in God, and—sur­pris­ingly—less likely to be­lieve in cry­on­ics (sig­nifi­cant at 5% level; could be a fluke). Peo­ple who be­lieved in high ex­is­ten­tial risk were more likely to be­lieve in global warm­ing, more likely to be­lieve they had a higher IQ than av­er­age, and more likely to be­lieve in aliens (I found that same re­sult last time, and it puz­zled me then too.)

In­trigu­ingly, even though the sam­ple size in­creased by more than 6 times, most of these re­sults are within one to two per­cent of the num­bers on the 2009 sur­vey, so this sup­ports tak­ing them as a di­rect line to pre­vailing ra­tio­nal­ist opinion rather than the con­tin­gent opinions of one ran­dom group.

Of pos­si­ble ex­is­ten­tial risks, the most feared was a bio­eng­ineered pan­demic, which got 194 votes (17.8%) - a nat­u­ral pan­demic got 89 (8.2%), mak­ing pan­demics the over­whelming leader. Un­friendly AI fol­lowed with 180 votes (16.5%), then nu­clear war with 151 (13.9%), ecolog­i­cal col­lapse with 145 votes (12.3%), eco­nomic/​poli­ti­cal col­lapse with 134 votes (12.3%), and as­ter­oids and nan­otech bring­ing up the rear with 46 votes each (4.2%).

The mean for the Sin­gu­lar­ity ques­tion is use­less be­cause of the very high num­bers some peo­ple put in, but the me­dian was 2080 (quar­tiles 2050, 2080, 2150). The Sin­gu­lar­ity has got­ten later since 2009: the me­dian guess then was 2067. There was some dis­cus­sion about whether peo­ple might have been an­chored by the pre­vi­ous men­tion of 2100 in the x-risk ques­tion. I changed the or­der af­ter 104 re­sponses to pre­vent this; a t-test found no sig­nifi­cant differ­ence be­tween the re­sponses be­fore and af­ter the change (in fact, the trend was in the wrong di­rec­tion).

Only 49 peo­ple (4.5%) have never con­sid­ered cry­on­ics or don’t know what it is. 388 (35.6%) of the re­main­der re­ject it, 583 (53.5%) are con­sid­er­ing it, and 47 (4.3%) are already signed up for it. That’s more than dou­ble the per­cent signed up in 2009.

231 peo­ple (23.4% of re­spon­dents) have at­tended a Less Wrong meetup.

The av­er­age per­son was 37.6% sure their IQ would be above av­er­age—un­der­con­fi­dent! Imag­ine that! (quar­tiles were 10, 40, 60). The mean was 54.5% for peo­ple whose IQs re­ally were above av­er­age, and 29.7% for peo­ple whose IQs re­ally were be­low av­er­age. There was a cor­re­la­tion of .479 (sig­nifi­cant at less than 1% level) be­tween IQ and con­fi­dence in high IQ.

Isaac New­ton pub­lished his Prin­cipia Math­e­mat­ica in 1687. Although peo­ple guessed dates as early as 1250 and as late as 1960, the mean was...1687 (quar­tiles were 1650, 1680, 1720). This marks the sec­ond con­sec­u­tive year that the av­er­age an­swer to these difficult his­tor­i­cal ques­tions has been ex­actly right (to be fair, last time it was the me­dian that was ex­actly right and the mean was all of eight months off). Let no one ever say that the wis­dom of crowds is not a pow­er­ful tool.

The av­er­age per­son was 34.3% con­fi­dent in their an­swer, but 41.9% of peo­ple got the ques­tion right (again with the un­der­con­fi­dence!). There was a highly sig­nifi­cant cor­re­la­tion of r = -.24 be­tween con­fi­dence and num­ber of years er­ror.

This graph may take some work to read. The x-axis is con­fi­dence. The y-axis is what per­cent of peo­ple were cor­rect at that con­fi­dence level. The red line you rec­og­nize as perfect cal­ibra­tion. The thick green line is your re­sults from the New­ton prob­lem. The black line is re­sults from the gen­eral pop­u­la­tion I got from a differ­ent cal­ibra­tion ex­per­i­ment tested on 50 ran­dom trivia ques­tions; take the in­ter­com­pa­ra­bil­ity of the two with a grain of salt.

As you can see, Less Wrong does sig­nifi­cantly bet­ter than the gen­eral pop­u­la­tion. How­ever, there are a few ar­eas of failure. First is that, as usual, peo­ple who put zero and one hun­dred per­cent had nonzero chances of get­ting the ques­tion right or wrong: 16.7% of peo­ple who put “0” were right, and 28.6% of peo­ple who put “100″ were wrong (in­ter­est­ingly, peo­ple who put 100 did worse than the av­er­age of ev­ery­one else in the 90-99 bracket, of whom only 12.2% erred). Se­cond of all, the line is pretty hori­zon­tal from zero to fifty or so. Peo­ple who thought they had a >50% chance of be­ing right had ex­cel­lent cal­ibra­tion, but peo­ple who gave them­selves a low chance of be­ing right were poorly cal­ibrated. In par­tic­u­lar, I was sur­prised to see so many peo­ple put num­bers like “0”. If you’re pretty sure New­ton lived af­ter the birth of Christ, but be­fore the pre­sent day, that alone gives you a 1% chance of ran­domly pick­ing the cor­rect 20-year in­ter­val.

160 peo­ple wanted their re­sponses kept pri­vate. They have been re­moved. The rest have been sorted by age to re­move any in­for­ma­tion about the time they took the sur­vey. I’ve con­verted what’s left to a .xls file, and you can down­load it here.