EA Hotel Fundraiser 3: Estimating the relative Expected Value of the EA Hotel (Part 1)

Ex­ec­u­tive summary

Here in Part 1, we will define a for­mula for the ex­pected value (EV) of the EA Ho­tel, rel­a­tive to donat­ing to other EA or­gani­sa­tions, called . We will then fac­tor each of the el­e­ments of as much as we can, and pre­sent our best es­ti­mate for each of these fac­tors, us­ing data gath­ered from cur­rent res­i­dents. In Part 2, we will use these es­ti­mates as an an­chor for defin­ing a plau­si­ble range of val­ues for each fac­tor, and do a Monte Carlo simu­la­tion on them. We aim to show 1) that un­der a wide range of as­sump­tions, the ho­tel is higher EV than the al­ter­na­tive and 2) a few show­cased val­ues for that will match differ­ent po­ten­tial donors.

1. What is the rel­a­tive EV of the EA Ho­tel, ?

Since the ho­tel is a meta-char­ity that hosts peo­ple who are work­ing on pro­jects that range as broadly as EA it­self, we will not try to calcu­late ac­tual EV, as in QALYs or some other stan­dard­ised mea­sure. We will in­stead at­tempt to calcu­late rel­a­tive EV, which is EV di­vided by coun­ter­fac­tual EV.

Let us first es­tab­lish the coun­ter­fac­tual. We will as­sume that most read­ers have a dona­tion bud­get, which they already in­tend to spend on EA. We will also as­sume that the EA or­gani­sa­tion they would oth­er­wise fund is fund­ing-con­strained, and would spend these ad­di­tional funds on hiring. We imag­ine a hy­po­thet­i­cal philan­thropist P, who has $X to spend. This amount would al­low ei­ther hires to be made ( be­ing the cost of a marginal hire at the coun­ter­fac­tual EA org), or it would al­low res­i­dents to live at the ho­tel (note that is low—it costs us <£6000/​year all in to host some­one at the EA Ho­tel).

Coun­ter­fac­tual EV would be the added value of these marginal hires, which is the im­pact of their work sub­tracted by the im­pact they would be mak­ing when not em­ployed by an EA or­gani­sa­tion. We call it .

Ho­tel EV would be the added value of these ad­di­tional res­i­dents, which we will call . We will as­sume that the amount of res­i­dents we can host scales lin­early with the amount of fund­ing we re­ceive. Since economies of scale ap­ply, and since our fund­ing gap ar­guably wouldn’t be closed un­til well into the hun­dreds of thou­sands, this is a con­ser­va­tive as­sump­tion.

Rel­a­tive EV would be di­vid­ing the lat­ter by the former. Sim­plify­ing, we get .

In the rest of this post, we will be fur­ther elab­o­rat­ing and es­ti­mat­ing all of these vari­ables.

1.1 Mak­ing more intuitive

In many of our es­ti­mates, we will defer to our in­tu­ition to es­ti­mate an an­swer. This is a nec­es­sary evil, for we are calcu­lat­ing the EV of a group of pro­jects that is suffi­ciently het­ero­ge­neous to be be­yond prop­erly for­mal­is­ing.

We found that the terms in our equa­tion are eas­ier to in­tuit if we define each one of them as rel­a­tive to the coun­ter­fac­tual. For each of our vari­ables, we define a rel­a­tive ver­sion:

2 Defin­ing

Re­mem­ber that stands for the value of the work of one ho­tel res­i­dent, di­vided by the value of the work of one coun­ter­fac­tual EA hire.

The value of the ho­tel comes from:

  • Allow­ing more peo­ple to do char­ity work

  • Allow­ing those peo­ple to do char­ity work more re­source­fully, by giv­ing them:

    1. Time (be­cause the ho­tel does their house­keep­ing, and be­cause they don’t have to travel much for so­cial ac­tivi­ties)

    2. Fo­cus (by en­g­ineer­ing an en­vi­ron­ment that boosts men­tal health and pro­duc­tivity)

    3. A net­work (by liv­ing with ~20 res­i­dents and an ar­ray of vis­i­tors pass­ing through)

The ques­tion of EV of the ho­tel, per res­i­dent, can thus be fac­tored as fol­lows:

1) How re­source­ful does the ho­tel en­vi­ron­ment al­low a res­i­dent to be, com­pared to a coun­ter­fac­tual EA hire?

2) If they were given the same re­sources as a coun­ter­fac­tual EA hire, how much value would the work of a ho­tel res­i­dent cre­ate?

We spec­ify as fol­lows:

  • be­ing the re­sources that a ho­tel res­i­dent could in­vest into their work, rel­a­tive to a coun­ter­fac­tual EA hire
  • be­ing the value gen­er­ated by their work, af­ter con­trol­ling for re­sources.

2.1 What is , the rel­a­tive value of a pro­ject en­abled by the EA Ho­tel?

This vari­able has a very pre­cise mean­ing, so al­low us to de­scribe it in de­tail.

should cap­ture:

  • The qual­ity of the pro­ject that a ho­tel res­i­dent wants to carry out

  • The qual­ity of the judg­ment of the res­i­dent in particular

It should not cap­ture the amount of re­sources that this par­tic­u­lar ho­tel res­i­dent can spend on their work. Any differ­ence in that doesn’t stem from their pro­ject, should stem from their gen­eral abil­ity to carry out pro­jects.

To have a good in­tu­ition for , imag­ine that we would have a ho­tel res­i­dent, and put them in a “nor­mal­ised” work­ing en­vi­ron­ment. That is: they have their own house to keep, ac­cess to an office at a nor­mal dis­tance, a so­cial life that re­quires travel, a pro­fes­sional net­work of an av­er­age amount of loose ties, etc. Then the im­pact of their par­tic­u­lar pro­ject, di­vided by the im­pact of a marginal EA hire, would be . This is per­haps the crux of the whole EV calcu­la­tion.

We have found that for­mally es­ti­mat­ing is nearly im­pos­si­ble to do prop­erly, be­cause it in­volves fac­tor­ing it into all the com­plex­ity of the var­i­ous pro­jects that EA Ho­tel guests are do­ing. Th­ese pro­jects change over time, are of­ten highly ex­plo­ra­tive and/​or en­trepreneurial and cover a broad range of cause ar­eas in­clud­ing those for which no es­tab­lished met­rics ex­ist yet.

Yet, we have an in­tu­itive sense that most of these pro­jects clearly meet EA stan­dards, and aren’t nec­es­sar­ily less im­pact­ful than a marginal EA hire.

Now EA prides it­self for do­ing the mea­sure­ments to de­ter­mine whether a strat­egy for im­pact is ac­tu­ally worth­while, but we feel that this risks ne­glect­ing the black box of in­tu­ition that of­ten gives bet­ter an­swers.

If you are con­sid­er­ing the ho­tel as a re­cip­i­ent of your dona­tion, we ask you to make your own in­tu­itive es­ti­mate for . Th­ese are the steps:

  1. Iden­tify your coun­ter­fac­tual. Which char­ity would you donate to, if not the EA Ho­tel?

  2. Imag­ine that your dona­tion would be ex­actly enough to make the char­ity hire one ad­di­tional em­ployee. In­tuit the amount of im­pact this em­ployee would make.

  3. Imag­ine that your dona­tion would also al­low ex­actly one ad­di­tional guest to stay at the ho­tel. Look at our last post for refer­ence. In­tuit the amount of im­pact the work of this guest would make, if it were car­ried out by this guest with a nor­mal level of mo­ti­va­tion, in a reg­u­lar office, with a net­work of nor­mal size.

  4. Now di­vide the former to the lat­ter. This is your per­sonal in­tu­itive value for .

You may be more com­fortable us­ing bounds in­stead of num­bers. Bar nega­tive out­comes, it is ob­vi­ous that is more than 0.01. It is also ob­vi­ous that is less than 100. Where would you put the bounds of , so that you are still 90% cer­tain that is be­tween them?

You may be think­ing: what about ho­tel guests who are study­ing, or skil­ling up in a new area? In these cases we could need to fac­tor out a dis­count rate, , and a lead time in years, , to work­ing on ob­ject level work of ex­pected rel­a­tive qual­ity :



Dis­count rates should be set at a level you think is rea­son­able given the Haste Con­sid­er­a­tion (note that the Haste Con­sid­er­a­tion also favours in­vest­ing in more peo­ple want­ing to tran­si­tion to full time di­rect EA work, as the ho­tel is do­ing). Per­haps = 0.9, mean­ing a 10% yearly dis­count, is rea­son­able. Or maybe even =0.8 (20%) if you think the ur­gency of di­rect EA work is par­tic­u­larly high. might typ­i­cally be be­tween 0.5 and 2 years.

2.2 What is , the fac­tor by which the pro­duc­tivity of EA Ho­tel res­i­dents is mul­ti­plied?

Imag­ine a marginal EA hire would spend 30 fo­cused hours on work. If a ho­tel res­i­dent finds the means to work 40 fo­cused hours, due to less time spent on house­keep­ing and travel, and due to a mo­ti­vat­ing en­vi­ron­ment, that is an amount of added value equal to 33% of a marginal EA hire work­ing full-time. With this ex­am­ple, would be 1.33. Since the ho­tel en­vi­ron­ment of­ten causes a con­sid­er­able in­crease in pro­duc­tivity, we want to add this el­e­ment to our calcu­la­tions.

While res­i­dents re­port that they are able to do 2.2 times as much work at the ho­tel on av­er­age, this num­ber does not map nicely to our anal­y­sis. For some res­i­dents, it com­pares their over­all im­pact with the case in which they would self-fund. For oth­ers it takes into ac­count that they’re do­ing differ­ent work. We are mod­el­ling as a fac­tor on pro­duc­tivity that comes from the ho­tel as com­pared to a reg­u­lar job, in which the em­ployee has to or­ganise their own life to a much larger ex­tent. Some fac­tors that go into are:

  • Freed up time be­cause house­work is largely done by the ho­tel, and be­cause less time is spent on travel

  • In­creased well-be­ing, from an en­vi­ron­ment in­ten­tion­ally de­signed for it, and free ac­cess to counselling

  • An im­proved pro­fes­sional and so­cial net­work, from liv­ing with 20 EA’s and fre­quent visitors

We will gen­er­al­ise the first two into a vari­able we would like to call “per­sonal pro­duc­tivity”, which is the amount of con­cen­trated effort one can put into one’s day to day work. The lat­ter will have an anal­y­sis on its own. We ex­pect these fac­tors to be roughly in­de­pen­dent, mean­ing that they both feed into :

be­ing the amount of pro­duc­tive hours worked rel­a­tive to an EA org hire, and be­ing the value of one’s net­work rel­a­tive to the coun­ter­fac­tual.

2.2.1 Es­ti­mat­ing rel­a­tive pro­duc­tivity,

Note that . The nu­mer­a­tor be­ing the amount of pro­duc­tive hours a ho­tel res­i­dent can work, and the de­nom­i­na­tor be­ing the amount of pro­duc­tive hours a marginal EA hire can work.

Our hope is that a healthy and time-sav­ing ho­tel en­vi­ron­ment causes an in­crease in the amount of fo­cused hours worked.

We no­ticed that our cul­ture strongly in­clines us to­wards pro­duc­tivity. Sim­ply sit­ting in the liv­ing room com­pels one to work. We ex­pect that ac­cess to an en­vi­ron­ment like this leads to an in­crease in hours worked.

We asked res­i­dents to es­ti­mate their amounts of pro­duc­tive hours per week, and found an av­er­age of 40:

We also asked how many pro­duc­tive hours they would have worked oth­er­wise, and found an av­er­age of 29.

While there is some rea­son to be­lieve that the former will go up as we fur­ther de­velop the ho­tel, we will now use it as a con­ser­va­tive es­ti­mate for . While it may be tempt­ing to use the lat­ter as an es­ti­mate for , we don’t think this would be fair: 29 is the hours that our res­i­dents would oth­er­wise work from home. Com­pared to a good office, work­ing from home pre­sum­ably low­ers their pro­duc­tivity, so might be higher. It’s also worth con­sid­er­ing that EA or­gani­sa­tion em­ploy­ees are se­lected for con­scien­tious­ness, hard work and ded­i­ca­tion at a high rate.

So we con­clude that , leav­ing the value of to our simu­la­tion in Part 2.

2.2.2 es­ti­mat­ing net­work value,

Hav­ing more like-minded peo­ple around gives you more op­por­tu­ni­ties. Ac­cord­ing to the so­cial ver­sion of Met­calfe’s law, the value of a net­work of nodes scales as , or per node. We take this to mean that, for each per­son in a net­work each in­cre­ment in size will in­crease their num­ber of op­por­tu­ni­ties by one, but since their band­width is limited and in­ter­ests don’t always al­ign, they will only be able to cap­i­tal­ise on the equiv­a­lent value of op­por­tu­ni­ties.

Ob­serve that, in some ab­stract sense, all the value of the EA com­mu­nity can be stated as an ac­cu­mu­la­tion of pos­i­tive sum trades be­tween peo­ple. Our law states that, if one’s net­work size changes from to , the value of their work is mul­ti­plied by .

So we re­define as fol­lows:

be­ing the pro­fes­sional net­work size of a ho­tel res­i­dent, and be­ing the pro­fes­sional net­work size of a coun­ter­fac­tual EA hire.

While for some, “pro­fes­sional net­work size” has an in­tu­itive mean­ing, we found that this isn’t the case for ev­ery­one, so al­low us to define that more pre­cisely. What we mean is the set of peo­ple that you could eas­ily in­ter­act with in a pro­fes­sional ca­pac­ity, if in­cen­tives lined up that way. To ap­prox­i­mate this num­ber, take each per­son you know. Do the two of you have com­mon knowl­edge about the rough out­line of both of each other’s work? Then they’re in your pro­fes­sional net­work. Note that this means that peo­ple in a wildly differ­ent cause area, or out­side of EA, can be in your pro­fes­sional net­work.

We can’t es­ti­mate the value of , be­cause that de­pends on the coun­ter­fac­tual of the donor, but we have asked our res­i­dents to es­ti­mate their pro­fes­sional net­work size:

Note the log­a­r­ith­mic scale.

2.3 Con­clu­sion for

Put­ting it all to­gether, we get .

For each guest, we have gath­ered their an­swers for and and es­ti­mated their based on our in­tu­ition, given what we know about them.

The data sug­gests that and are mostly un­cor­re­lated (r = 0.15), and so are and (r = 0.04), but and are strongly cor­re­lated (r = 0.6). This may be owed to our pos­si­bly bi­ased as­sump­tion that the most value is in build­ing in­fras­truc­ture, which nat­u­rally re­quires a larger net­work.

We will as­sume that the same pat­tern of in­de­pen­den­cies will hold for hires. Sim­plify­ing fur­ther:

We will calcu­late un­der a range of as­sump­tions in Part 2.

3. What are and , the rel­a­tive val­ues of work done by peo­ple if they’re not housed or hired?

Both of these can be given a similar struc­ture to :

, , … be­ing the val­ues of these vari­ables in the coun­ter­fac­tual situ­a­tion where the res­i­dent or hire is do­ing some­thing else with their ca­reer.

Prop­erly es­ti­mat­ing these vari­ables will re­quire gath­er­ing data from EA’s. This wasn’t an op­tion for us given short run­ways, so we will have to make do with rough es­ti­mates in­stead.

3.1 Rough es­ti­mates for

3.1.1

15 of the ho­tel res­i­dents are do­ing work that they would oth­er­wise self-fund by work­ing part time. For these, we can define .

Of those that are do­ing some­thing else at the ho­tel, three would be work­ing a nor­mal job out­side EA, one would con­tinue do­ing promis­ing AI ca­pa­bil­ities re­search, and one would con­tinue to scout for op­por­tu­ni­ties in the EA land­scape. We es­ti­mate that their val­ues for would be , and re­spec­tively.

Us­ing the av­er­age of our cur­rent guests as an es­ti­mate, we will guess that is around

3.1.2

Since most of the ho­tel res­i­dents would self-fund, they would spend con­sid­er­ably less time work­ing on their pro­jects, with the rest of their time spent on work­ing part-time in­stead. They re­port an av­er­age speedup of 2.2, which we will use to guess that .

3.1.3

We would ex­pect that ev­ery­one meets ap­prox­i­mately the same amount of peo­ple when they move to the ho­tel, so we will define to be a con­stant in­crease over . Our sur­vey shows that this in­crease is an av­er­age of 20.6, which we will use to say .

3.1.4

Put­ting the above to­gether, we get:

3.2 Rough es­ti­mates for

3.2.1

We don’t have any data what­so­ever about likely val­ues for this, so we will leave it to the simu­la­tion in Part 2.

3.2.2

Given the strong com­pe­ti­tion, we will as­sume that a marginal EA hire is highly con­scien­tious. They might not have a pro­ject available to them that is as good as work­ing at an EA org, but what­ever pro­ject they will be work­ing, they will work on it pro­duc­tively. Our best guess for is that it equals .

3.2.3

Again, we’re as­sum­ing that this num­ber is some con­stant differ­ence from , the size of which de­pends on the par­tic­u­lar or­gani­sa­tion that the coun­ter­fac­tual EA hire joins. If they’re do­ing re­mote work, it may be just a few, or only 1. If they get to work at an office of a ma­jor EA org, it may be as high as 50. Even more if they re­lo­cate to a large hub like the Bay or Lon­don.

3.2.4

Put­ting the above to­gether, we get:

3.3 Con­clu­sions for and

We will calcu­late and un­der a range of as­sump­tions in Part 2.

4. Put­ting it all together

We get:


In Part 2 of this post we will give some ex­am­ple calcu­la­tions un­der a range of as­sump­tions and sce­nar­ios, and at­tempt to draw some con­clu­sions. We will also link to an on­line calcu­la­tor for read­ers to calcu­late their own es­ti­mates.

Thanks to Toon Alfrink for de­vel­op­ing the method­ol­ogy and draft­ing the post, Greg Colbourn for edit­ing, and Sasha Cooper and Florent Ber­thet for com­ments.


Our Ask

Do you like this ini­ti­a­tive, and want to see it con­tinue for longer, on a more sta­ble foot­ing? Do you want to cheaply buy time spent work­ing full time on work re­lat­ing to EA, whilst si­mul­ta­neously fa­cil­i­tat­ing a thriv­ing EA com­mu­nity hub? Is your rel­a­tive EV es­ti­mate for the pro­ject high? Then we would like to ask for your sup­port.

We are get­ting crit­i­cally low on run­way. Our cur­rent short­fall is ~£5k/​month from May on­ward. We will have to start giv­ing cur­rent guests no­tice in 1 month.

To donate, please see our GoFundMe or get in touch if you’d like to make a di­rect bank trans­fer.

If you’d like to give reg­u­lar sup­port, we also have a Pa­treon.

Pre­vi­ous posts in this se­ries: