2016 LessWrong Diaspora Survey Analysis: Part One (Meta and Demographics)

2016 LessWrong Di­as­pora Sur­vey Analysis

Overview


Sur­vey Meta

Introduction

Hello ev­ery­body, this is part one in a se­ries of posts an­a­lyz­ing the 2016 LessWrong Di­as­pora Sur­vey. The sur­vey ran from March 24th to May 1st and had 3083 re­spon­dents.

Al­most two thou­sand eight hun­dred and fifty hours were spent sur­vey­ing this year and you’ve all waited nearly two months from the first sur­vey re­sponse to the re­sults writeup. While the re­sults have been available for over a week, they haven’t seen wide­spread dis­sem­i­na­tion in large part be­cause they lacked a suc­cinct sum­mary of their con­tents.

When we started the sur­vey in march I posted this graph show­ing the dropoff in ques­tion re­sponses over time:

So it seems only rea­son­able to post the same graph with this years sur­vey data:

(I should note that this anal­y­sis counts cer­tain things as ques­tions that the other chart does not, so it says there are many more ques­tions than the pre­vi­ous sur­vey when in re­al­ity where are about as many as last year.)

2016 Di­as­pora Sur­vey Stats

Sur­vey hours spent in to­tal: 2849.818888888889

Aver­age num­ber of min­utes spent on sur­vey: 102.14404619673437

Me­dian num­ber of min­utes spent on sur­vey: 39.775

Mode min­utes spent on sur­vey: 20.266666666666666

The take­away here seems to be that some peo­ple take a long time with the sur­vey, rais­ing the av­er­age. How­ever, most peo­ple’s sur­vey time is some­where be­low the forty five minute mark. LessWrong does a very long sur­vey, and I wanted to make sure that in­vest­ment was re­warded with a deep de­tailed anal­y­sis. Weigh­ing in at over four thou­sand lines of python code, I hope the anal­y­sis I’ve put to­gether is worth the wait.

Credits

I’d like to thank peo­ple who con­tributed to the anal­y­sis effort:

Bar­tosz Wroblewski

Ku­udes on #lesswrong

Obor­mot on #lesswrong

Two anony­mous contributors

And any­body else who I may have for­got­ten. Thanks again to Scott Alexan­der, who wrote the ma­jor­ity of the sur­vey and ran it in 2014, and who has also been gen­er­ous enough to li­cense his part of the sur­vey un­der a cre­ative com­mons li­cense along with mine.


Demographics

Age

The 2014 sur­vey gave these num­bers for age:

Age: 27.67 + 8.679 (22, 26, 31) [1490]

In 2016 the num­bers were:

Mean: 28.108772669759592
Me­dian: 26.0
Mode: 23.0

Most LWers are in their early to mid twen­ties, with some older LWers bring­ing up the av­er­age. The av­er­age is close enough to the former figure that we can prob­a­bly say the LW de­mo­graphic is in their 20′s or 30′s as a gen­eral rule.

Sex and Gender

In 2014 our gen­der ra­tio looked like this:

Fe­male: 179, 11.9%
Male: 1311, 87.2%

In 2016 the pro­por­tion of women in the com­mu­nity went up by over four per­cent:

Male: 2021 83.5%
Fe­male: 393 16.2%

One hy­poth­e­sis on why this hap­pened is that the 2016 sur­vey fo­cused on the di­as­pora rather than just LW. Di­as­pora com­mu­ni­ties plau­si­bly have marginally higher rates of fe­male mem­ber­ship. If I had more time I would write an anal­y­sis in­ves­ti­gat­ing the de­mo­graph­ics of each di­as­pora com­mu­nity, but to an­swer this par­tic­u­lar ques­tion I think a cou­ple of SQL queries are illus­tra­tive:

(Note: Ac­tiveMem­ber­ships one and two are ‘LessWrong’ and ‘LessWrong Mee­tups’ re­spec­tively.)
sqlite> se­lect count(birth­sex) from data where (Ac­tiveMem­ber­ships_1 = “Yes” OR Ac­tiveMem­ber­ships_2 = “Yes”) AND birth­sex=”Male”;
425
sqlite> se­lect count(birth­sex) from data where (Ac­tiveMem­ber­ships_1 = “Yes” OR Ac­tiveMem­ber­ships_2 = “Yes”) AND birth­sex=”Fe­male”;
66
>>> 66 /​ (425 + 66)\
0.13441955193482688\

Well, maybe. Of course, be­fore we wring our hands too much on this ques­tion it pays to re­mem­ber that as­signed sex at birth isn’t the whole story. The gen­der ques­tion in 2014 had these re­sults:

F (cis­gen­der): 150, 10.0%
F (trans­gen­der MtF): 24, 1.6%
M (cis­gen­der): 1245, 82.8%
M (trans­gen­der FtM): 5, 0.3%
Other: 64, 4.3%

In 2016:

F (cis­gen­der): 321 13.3%
F (trans­gen­der MtF): 65 2.7%
M (cis­gen­der): 1829 76%
M (trans­gen­der FtM): 23 1%
Other: 156 6.48%

Some things to note here. 16.2% of re­spon­dents were as­signed fe­male at birth but only 13.3% still iden­tify as women. 1% are trans­men, but where did the other 1.9% go? Pre­sum­ably into the ‘Other’ field. Let’s find out.

sqlite> se­lect count(birth­sex) from data where birth­sex = “Fe­male” AND gen­der = “Other”;
57
sqlite> se­lect count(*) from data;
3083
>>> 57 /​ 3083
0.018488485241647746

Seems to be the case. In gen­eral the pro­por­tion of men is down 6.1% from 2014. We also gained 1.1% transwomen and .7% trans­men in 2016. Mov­ing away from bi­nary gen­ders, this sur­veys non­bi­nary gen­der count gained in pro­por­tion by nearly 2.2%. This means that over one in twenty LWers iden­ti­fied as a non­bi­nary gen­der, mak­ing it a larger de­mo­graphic than bi­nary trans­gen­der LWers! As ex­cit­ing as that may sound to some ears the num­bers tell one story and the write ins tell quite an­other.

It pays to keep in mind that non­bi­nary gen­ders are a com­mon troll op­tion for peo­ple who want to write in crit­i­cism of the ques­tion. A quick look at the write ins ac­com­pa­ny­ing the other op­tion in­di­cates that this is what many peo­ple used it for, but by no means all. At 156 re­sponses, that’s small enough to be worth do­ing a quick man­ual tally.

<table bor­der=”0″> <cap­tion>”Other” Gen­ders, Sam­ple Size: 156 </​cap­tion> <thead> <tr> <th>Clas­sifi­ca­tion</​th><th>Count</​th> </​tr> </​thead> <tbody> <tr> <td>Agen­der</​td> <td>35</​td> </​tr> <tr> <td>Eso­teric</​td> <td>6</​td> </​tr> <tr> <td>Fe­male</​td> <td>6</​td> </​tr> <tr> <td>Male</​td> <td>21</​td> </​tr> <tr> <td>Male-To-Fe­male</​td> <td>1</​td> </​tr> <tr> <td>Non­bi­nary</​td> <td>55</​td> </​tr> <tr> <td>Ob­jec­tion on Ba­sis Gen­der Doesn’t Ex­ist</​td> <td>6</​td> </​tr> <tr> <td>Ob­jec­tion on Ba­sis Gen­der Is Bi­nary</​td> <td>7</​td> </​tr> <tr> <td>in Pro­cess of Tran­si­tion­ing</​td> <td>2</​td> </​tr> <tr> <td>Re­fusal</​td> <td>7</​td> </​tr> <tr> <td>Un­de­cided</​td> <td>10</​td> </​tr> </​tbody> </​table> So de­pend­ing on your com­fort zone as to what con­sti­tutes a countable gen­der, there are 90 to 96 valid ‘other’ an­swers in the sur­vey dataset. (La­beled dataset)

>>> 90 /​ 3083
0.029192345118391177

With some cleanup the num­ber trails be­hind the bi­nary trans­gen­der one by the greater part of a per­centage point, but only by. I bet that if you went through and did the same sort of tally on the 2014 sur­vey re­sults you’d find that the pro­por­tion of valid non­bi­nary gen­der write ins has gone up be­tween then and now.

Some in­ter­est­ing ‘es­o­teric’ an­swers: At­tack He­lo­copter, Black­star, Elizer, spi­der­man, Agenderfluid

For the rest of this sec­tion I’m go­ing to just fo­cus on differ­ences be­tween the 2016 and 2014 sur­veys.

2014 De­mo­graph­ics Ver­sus 2016 Demographics

Country

United States: −1.000% 1298 53.700%
United King­dom: −0.100% 183 7.600%
Canada: +0.100% 144 6.000%
Aus­tralia: +0.300% 141 5.800%
Ger­many: −0.600% 85 3.500%
Rus­sia: +0.700% 57 2.400%
Fin­land: −0.300% 25 1.000%
New Zealand: −0.200% 26 1.100%
In­dia: −0.100% 24 1.000%
Brazil: −0.300% 16 0.700%
France: +0.400% 34 1.400%
Is­rael: +0.200% 29 1.200%
Other: 354 14.646%

[Sum­ming these all up to one shows that nearly 1% of change is un­ac­counted for. My hy­poth­e­sis is that this 1% went into the other coun­tries not in the list, this can’t be eas­ily con­firmed be­cause the 2014 anal­y­sis does not list the other coun­try per­centage.]

Race

Asian (East Asian): −0.600% 80 3.300%
Asian (In­dian sub­con­ti­nent): +0.300% 60 2.500%
Mid­dle Eastern: 0.000% 14 0.600%
Black: −0.300% 12 0.500%
White (non-His­panic): −0.300% 2059 85.800%
His­panic: +0.300% 57 2.400%
Other: +1.200% 108 4.500%

Sex­ual Orientation

Hetero­sex­ual: −5.000% 1640 70.400%
Ho­mo­sex­ual: +1.300% 103 4.400%
Bi­sex­ual: +4.000% 428 18.400%
Other: +3.880% 144 6.180%

[LessWrong got 5.3% more gay, 9.1% if you’re more loose with the defi­ni­tion. Be­fore we start any wild spec­u­la­tion, the 2014 ques­tion in­cluded asex­u­al­ity as an op­tion and it got 3.9% of the re­sponses, we spun this off into a sep­a­rate ques­tion on the 2016 sur­vey which should ex­plain a sig­nifi­cant por­tion of the change.]

Are you asex­ual?

Yes: 171 0.074
No: 2129 0.926

[Scott said in 2014 that he’d prob­a­bly ‘vastly un­der­counted’ our asex­ual read­ers, a near dou­bling in our count would seem to sup­port this.]

Re­la­tion­ship Style

Pre­fer mono­go­mous: −0.900% 1190 50.900%
Pre­fer polyamorous: +3.100% 426 18.200%
Uncer­tain/​no prefer­ence: −2.100% 673 28.800%
Other: +0.426% 45 1.926%

[Polyamorous gained three points, pre­sum­ably the drop in un­cer­tain peo­ple went into that bin.]

Num­ber of Partners

0: −2.300% 1094 46.800%
1: −0.400% 1039 44.400%
2: +1.200% 107 4.600%
3: +0.900% 46 2.000%
4: +0.100% 15 0.600%
5: +0.200% 8 0.300%
Lots and lots: +1.000% 29 1.200%

Re­la­tion­ship Goals

...and seek­ing more re­la­tion­ship part­ners: +0.200% 577 24.800%
…and pos­si­bly open to more re­la­tion­ship part­ners: −0.300% 716 30.800%
…and cur­rently not look­ing for more re­la­tion­ship part­ners: +1.300% 1034 44.400%

Are you mar­ried?

Yes: 443 0.19
No: 1885 0.81

[This ques­tion ap­peared in a differ­ent form on the pre­vi­ous sur­vey. Mar­riage went up by .8% from last year.]

Who do you cur­rently live with most of the time?

Alone: −2.200% 487 20.800%
With par­ents and/​or guardians: +0.100% 476 20.300%
With part­ner and/​or chil­dren: +2.100% 687 29.400%
With room­mates: −2.000% 619 26.500%

[This would seem to line up with the re­sult that sin­gle LWers went down by 2.3%]

How many chil­dren do you have?

Sum: 598 or greater
0: +5.400% 2042 87.000%
1: +0.500% 115 4.900%
2: +0.100% 124 5.300%
3: +0.900% 48 2.000%
4: −0.100% 7 0.300%
5: +0.100% 6 0.300%
6: 0.000% 2 0.100%
Lots and lots: 0.000% 3 0.100%

[In­ter­est­ingly enough, child­less LWers went up by 5.4%. This would seem in­con­gru­ous with the pre­vi­ous re­sults. Not sure how to in­ves­ti­gate though.]

Are you plan­ning on hav­ing more chil­dren?

Yes: −5.400% 720 30.700%
Uncer­tain: +3.900% 755 32.200%
No: +2.800% 869 37.100%

[This is an in­ter­est­ing re­sult, ei­ther nearly 4% of LWers are sud­denly less en­thu­si­as­tic about hav­ing kids, or new en­trants to the sur­vey are less likely and less sure if they want to. Pos­si­bly both.]

Work Status

Stu­dent: −5.402% 968 31.398%
Aca­demics: +0.949% 205 6.649%
Self-em­ployed: +4.223% 309 10.023%
In­de­pen­dently wealthy: +0.762% 42 1.362%
Non-profit work: +1.030% 152 4.930%
For-profit work: −1.756% 954 30.944%
Govern­ment work: +0.479% 135 4.379%
Homemaker: +1.024% 47 1.524%
Unem­ployed: +0.495% 228 7.395%

[Most in­ter­est­ing re­sult here is that 5.4% of LWers are no longer stu­dents or new sur­vey en­trants aren’t.]

Profession

Art: +0.800% 51 2.300%
Biol­ogy: +0.300% 49 2.200%
Busi­ness: −0.800% 72 3.200%
Com­put­ers (AI): +0.700% 79 3.500%
Com­put­ers (other aca­demic, com­puter sci­ence): −0.100% 156 7.000%
Com­put­ers (prac­ti­cal): −1.200% 681 30.500%
Eng­ineer­ing: +0.600% 150 6.700%
Fi­nance /​ Eco­nomics: +0.500% 116 5.200%
Law: −0.300% 50 2.200%
Math­e­mat­ics: −1.500% 147 6.600%
Medicine: +0.100% 49 2.200%
Neu­ro­science: +0.100% 28 1.300%
Philos­o­phy: 0.000% 54 2.400%
Physics: −0.200% 91 4.100%
Psy­chol­ogy: 0.000% 48 2.100%
Other: +2.199% 277 12.399%
Other “hard sci­ence”: −0.500% 26 1.200%
Other “so­cial sci­ence”: −0.200% 48 2.100%

[The largest pro­fes­sion growth for LWers in 2016 was art, that or this is a con­se­quence of new sur­vey en­trants.]

What is your high­est ed­u­ca­tion cre­den­tial earned?

None: −0.700% 96 4.200%
High School: +3.600% 617 26.700%
2 year de­gree: +0.200% 105 4.500%
Bach­e­lor’s: −1.600% 815 35.300%
Master’s: −0.500% 415 18.000%
JD/​MD/​other pro­fes­sional de­gree: 0.000% 66 2.900%
PhD: −0.700% 145 6.300%
Other: +0.288% 39 1.688%

[Hm, the aca­demic cre­den­tials of LWers seems to have gone down some since the last sur­vey. As usual this may also be the re­sult of new sur­vey en­trants.]


Footnotes

  1. The 2850 hour es­ti­mate of sur­vey hours is very naive. It mea­sures the time be­tween start­ing and turn­ing in the sur­vey, a per­son didn’t nec­es­sar­ily sit there dur­ing all that time. For ex­am­ple this could eas­ily be in­clud­ing peo­ple who spent mul­ti­ple days do­ing other things be­fore fi­nally finish­ing their sur­vey.

  2. The apache he­li­copter image is li­censed un­der the Open Govern­ment Li­cense, which re­quires at­tri­bu­tion. That par­tic­u­lar edit was done by Wub­bles on the LW Slack.

  3. The first pub­lished draft of this post made a ba­sic stats er­ror calcu­lat­ing the pro­por­tion of women in ac­tive mem­ber­ships one and two, di­vid­ing the num­ber of women by the num­ber of men rather than the num­ber of women by the num­ber of men and women.