Long Bets as Charitable Giving Opportunity

Evaluating Long Bets as a prediction market shows it is dysfunctional and poorly-structured; despite the irrationality of many users, it is not good even as a way to raise money for charity.
statistics, predictions, charity, R
2017-02-242018-02-24 in progress certainty: likely importance: 5


Long Bets is a 15-year-old real-­money pre­dic­tion mar­ket run by the Long Now Foun­da­tion for incen­tiviz­ing forecasts/bets about long-term events of social impor­tance such as tech­nol­ogy or the envi­ron­ment. I eval­u­ate use of Long Bets as a char­i­ta­ble giv­ing oppor­tu­nity by win­ning bets and direct­ing the earn­ings to a good char­ity by mak­ing fore­casts for all avail­able bet oppor­tu­ni­ties and rank­ing them by expected value after adjust­ing for oppor­tu­nity cost (de­fined by expected return of stock mar­ket index­ing) and tem­po­rally dis­count­ing. I find that while there are ~41 open bets which I expect have pos­i­tive expected value if coun­ter-­bets were accept­ed, few or none of my coun­ter-­bets were accept­ed. In gen­er­al, LB has had almost zero activ­ity for the past decade, and has not incen­tivized much fore­cast­ing. This fail­ure is likely caused by its extreme restric­tion to even-odds bets, no return on bet funds (re­sult­ing in enor­mous oppor­tu­nity cost­s), and lack of main­te­nance or pub­lic­i­ty. All of these issues are highly likely to con­tinue bar­ring exten­sive changes to Long Bets, and I sug­gest that Long Bets should be wound down.

2016 annual let­ter includes a long sec­tion on his $1m long bet with Pro­tege Part­ner­s’s Ted Sei­des that the S&P 500 index will out­per­form hedge funds (as does his 2017 let­ter); unsur­pris­ing­ly, he was about to win, and in May 2017, Sei­des con­ceded the bet, blam­ing the S&P 500 for turn­ing in its “unex­pected” but his­tor­i­cally aver­age per­for­mance.

This reminded me of in gen­eral (ab­bre­vi­ated LB; about; FAQ; rules/pro­ce­dure), a /registry launched in April 2002 but which never really caught on (in part because their web­site was effec­tively unus­able for sev­eral years), but is still run­ning and still resolv­ing bets and still has any num­ber of in-progress bets and open chal­lenges. It is one of sev­eral projects of the Long Now Foun­da­tion, but less high pro­file than oth­ers like the being built for 1

For a Long Bet, to cre­ate a ‘pre­dic­tion’ one pays a $50 fee; then a chal­lenger and you put up the same bet (min­i­mum of $200) each—even odds is manda­to­ry—and des­ig­nate a pre­ferred char­i­ty, for a bet with dura­tion >=2 years, the bets are invested in a ultra­-­con­ser­v­a­tive invest­ment fund (un­spec­i­fied return, I’d guessti­mate 3% opti­misti­cal­ly), and the win­nings+in­vest­ment go to the win­ner’s char­i­ty.

Exploiting irrational bets

You can’t make money off them directly because of the char­ity restric­tion nec­es­sary to keep LB legal in the USA (reg­u­la­tion is why we can’t have nice things like pre­dic­tion mar­ket­s), but you could try to replace your cur­rent char­i­ta­ble giv­ing with canny long bet­s—as it is not a pre­dic­tion mar­ket, after all, the bets can be quite irra­tional. Indeed, some of the exist­ing pre­dic­tions strike me as quite dumb and easy money (eg bet #312, “Turkey will join the Euro­pean Union and become a model for a Demo­c­ra­tic Islamic State and lead the way for advanc­ing rela­tions between the Islamic world and ‘West­ern World’”, 2007–2018, which was unlikely in 2007 and has ~0% prob­a­bil­ity now). You don’t have to pay the $50 fee or worry about no one tak­ing up your chal­lenge, and you dou­ble your mon­ey, so if you wanted to donate $1k to , you could instead try to find a bet like #312 and offer a bet of $500 and, almost cer­tainly win­ning, send ~$1000 to AMF, thereby both pro­vid­ing account­abil­ity & increas­ing fund­ing for effec­tive char­i­ties.

One could also try to argue that the bet­ting will have addi­tional ben­e­fits like estab­lish­ing a track record as a good pre­dic­tor or get­ting pub­lic­ity for one’s views, a few bets might be inter­est­ing as a learn­ing expe­ri­ence, and taking/making bets could be use­ful as a com­pan­ion to any research/essays/books one might write as a form of account­abil­i­ty.

Challenges

The down­sides would be:

  • LB invests funds super-­con­ser­v­a­tive­ly, so the oppor­tu­nity cost is non­triv­ial—they def­i­nitely don’t get the stan­dard 7% stock mar­ket returns, so one is for­feit­ing some­thing like 4–7% annual returns. (Pre­sum­ably if one thinks that spend­ing one’s money in some other way such as donat­ing to a char­ity right now so the char­i­ta­ble activ­i­ties can be done right away is bet­ter than invest­ing & donat­ing lat­er, one thinks the oppor­tu­nity cost is >7% and so the LB oppor­tu­nity cost is even greater than just 4%!)

    For exam­ple, Buf­fett reports for his bet that as of Feb­ru­ary 2017, if they had sim­ply put their [$1]($2008)m into the S&P 500 index, with the cumu­la­tive return of 85.4% it would be worth $1.854m; in other words, Buf­fett will win for his favored char­ity lit­tle more than he would have earned if he had not used LB at all & sim­ply held the mon­ey, and sub­stan­tially less than if he had kept the money in Berk­shire Hath­away which has seen a cumu­la­tive return of 221% (BRK.A has appre­ci­ated from $145,929 2007-12-31 to $244,121 2017-02-27). Or to put it another way, by not doing a hand­shake bet, Buffett/Seides gave LB a float of $1m over an long time period with a large oppor­tu­nity cost2 and thus would have paid LB ~$800,000 for its ser­vice; it is not clear that this would have been money well spent, and indeed, it turns out that Buf­fett did not allow LB to hold the money but instead invested the orig­i­nal wager of $827,752 into a US Trea­sury bond (worth $1m by 2012) & then rein­vested into Berk­shire Hath­away.3

    • the ‘long’ aspect exac­er­bates this, as the longer a bet goes on, the less likely it is that you wind up donat­ing more to your char­ity than just sav­ing & giv­ing later

    • 3% is opti­mistic, as the rules state that half of the ‘con­ser­v­a­tive’ return goes to the Long Now Foun­da­tion in addi­tion to the $50 pre­dic­tion fee:

      The entire amount goes into a long-term invest­ment port­fo­lio called the Far­sight Fund—its assets are in “Endow­ments”, a mutual fund man­aged by Cap­i­tal Research and Man­age­ment Com­pa­ny. Half of the growth of that fund is drawn off to The Long Now Foun­da­tion, which main­tains the LB ser­vice; the other half accrues so that the even­tual pay­ment to the win­ner’s pre­ferred char­ity may be sig­nif­i­cantly larger than the orig­i­nal bet stakes.

    I asked LB for more infor­ma­tion and they replied that they intend to pub­licly review Far­sight Fund per­for­mance at the end of 2017.

  • there are not many pre­dic­tions in gen­er­al, and only a few are added each year, of which many are clearly unse­ri­ous despite the reg­is­tra­tion fee & LB’s mod­er­a­tion, lim­it­ing pos­si­ble invest­ments

  • all bets are 1:1, sharply upper bound­ing how many are worth­while (eg if some­one thinks that a thing is only 60% likely and you think it’s 100% likely and they reg­is­ter a pre­dic­tion for it, you can’t do any­thing about it) and how much you can mon­ey-pump peo­ple for (only 2x)

  • many of the pre­dic­tions seem to be invalid or out­dat­ed; bet #312 has­n’t fin­ished yet, but I think it would tech­ni­cally be invalid to make that bet with him now because a bet expir­ing in 2018 can’t be at least 2 years long

    • of the valid ones which would be prof­itable to take, I would expect most of them to back out or refuse to bet or quib­ble about tech­ni­cal­i­ties or delay as long as pos­si­ble; LB requires a $50 fee but they don’t require the $200 bet to be pre­paid and you have to nego­ti­ate with each pre­dic­tor before any bet goes into effect. So you can’t bet with most of them. (A major down­side of per­sonal bet­ting as com­pared to pre­dic­tion mar­kets is the huge fric­tion & trans­ac­tion cost­s.)
    • LB could still col­lapse or embez­zle it all or decide fla­grantly wrong (that said, the Long Now Foun­da­tion has been oper­at­ing since 1996, for 21 years, and the lat­est Form 990 for 20144 indi­cates their finan­cial health looks good, and at least there have been no com­plaints about mis­judg­ments for the hand­ful of expired bets, ~9, thus far)
  • you might be the irra­tional one and see your bets go to worth­less char­i­ties (eg Buf­fett des­ig­nated Girls Inc as his win­ner; but that $1m is going to do far less good than if it was going to the Gates Foun­da­tion with the rest of Buf­fet­t’s mon­ey!)

  • the pub­lic­ity ben­e­fits are min­i­mal: the Buf­fett bet has been widely reported on, but because of the amount and it being War­ren Buf­fett, not because LB mod­er­ates it (if any­thing, LB has ben­e­fited far more, both finan­cially and pub­lic­i­ty-­wise); other bets have not received much pub­lic­ity

Minimum required returns

The oppor­tu­nity cost is the big one. Since the min­i­mum is 2 years, and guessti­mat­ing 3% return for the fund and using the usual stock mar­ket index return of ~7%, you are for­feit­ing ~4% return for >=2 years, so the min­i­mum oppor­tu­nity cost is or ~8.4% of your bet. That is quite a lot.

The ques­tion becomes, even if you think you’ve found a bet you have >50% chance of win­ning (re­quired to make it +EV given the manda­tory 1:1 bet­ting), how much of an edge do you need to just out­pace the oppor­tu­nity cost?

You need to solve for p in a value expres­sion like . b drops out as a con­stant, so for t = 2, you need… p = 0.54. In gen­er­al, . So even if you assume you are cer­tain to win, p = 1, you can’t bet any fur­ther out than ~18.19 years (so in 2017, any bet past 2035 is out of the ques­tion).

Or to put it another way, the lost return dou­bles in 18 years (), so since you can never earn more than twice your bet on LB, past 18 years it becomes impos­si­ble to earn a return (be­fore, it is merely very dif­fi­cult).

data.frame(Year=2:19, Minimum.Win.Prob=sapply(2:19, function(t) {
    optimize(function(p) { abs(1.07^t - 2*p*1.03^t) }, seq(0,1,by=0.001))$minimum }))
# Year Minimum.Win.Prob
#    2 0.5395891434
#    3 0.5605221301
#    4 0.5822973072
#    5 0.6049417583
#    6 0.6284340139
#    7 0.6527947343
#    8 0.6781593805
#    9 0.7045338182
#   10 0.7318640396
#   11 0.7602945746
#   12 0.7898129251
#   13 0.8204982106
#   14 0.8523813174
#   15 0.8854592825
#   16 0.9198648951
#   17 0.9555596565
#   18 0.9926832986
#   19 0.9999338930
1.07^10.5
# [1] 2.034837186
data.frame(Year=2:11, Minimum.Win.Prob=sapply(2:11, function(t) {
    optimize(function(p) { abs(1.07^t - 2*p*1.00^t) }, seq(0,1,by=0.001))$minimum }))
# Year Minimum.Win.Prob
#    2     0.5724575464
#    3     0.6124911939
#    4     0.6553833666
#    5     0.7012821290
#    6     0.7503808673
#    7     0.8028891537
#    8     0.8591051279
#    9     0.9192578559
#   10     0.9836032026
#   11     0.9999338930

In the worst-­case of 0% return from the LB invest­ment fund after fees & Long Now tak­ing its half, the upper bound is 11 years (2028).

Opportunities

Already-made Bets

As of 2017-02-25, LB has 29 reg­is­tered bets, of which 9 have fin­ished & 20 are out­stand­ing with a total bal­ance (ex­clud­ing any invest­ment returns) of $1,041,800 (or $41,800 exclud­ing the Buf­fett bet). The bets matur­ing soon:

  1. 2017: the Buf­fett index vs hedge fund bet ($1000k)

  2. 2018:

  3. 2019: none

  4. 2020:

  5. 2021: none

  6. 2022: “In 2022 the US econ­omy will no longer be held to be the world leader in global envi­ron­men­tal degra­da­tion” ($2k)

  7. 2023: none

  8. 2024: “Dri­ver­less cars will be com­mer­cially avail­able in Las Veg­as, NV by May 27 2024. Trips may be point to point out­side of the city cen­ter with no require­ment for any pas­sen­ger to take over man­ual con­trol of the vehi­cle.” ($0.5k)

  9. 2025:

  10. 2026: none

  11. 2027: none

  12. 2028: none

  13. 2029: “By 2029 no com­put­er—or ‘machine intel­li­gence’—will have passed the Tur­ing Test” ($20k)

  14. 2030: “By 2030, com­mer­cial pas­sen­gers will rou­tinely fly in pilot­less planes” ($2k)

  15. 2031: none

  16. 2032: none

  17. 2033: none

  18. 2034: none

  19. 2035: none

Given the implied 50% prob­a­bil­i­ty, I dis­agree strongly on many of the­se, but unfor­tu­nately there is no addi­tional bet­ting; bets must be done on pre­dic­tions only.

Expired or Invalid Bets

There are ~207 out­stand­ing pre­dic­tions but a decent frac­tion are ill-de­fined, expired, or expir­ing too soon to reg­is­ter a bet on. Review­ing the ones with due dates before 2035:

Expired, right:

Expired, wrong:

Expired:

About to expire: Can­not be bet on and so should be removed, or the pre­dic­tion needs to be edit­ed, because it matures in <=2 years and the LB rules man­date bets must be >=2 years long:

Should be removed for being ill-defined/unjudgeable/unfalsifiable, vague, emo­tiz­ing, moron­ic, spam or just plain awful:

I have emailed this list to LB so they can clean up their pre­dic­tions page; they said they will keep it in mind should they ever update the site.

Amus­ing­ly, even though pre­dic­tors pre­sum­ably are at least 50% con­fi­dent when they make binary pre­dic­tions at 1:1 odds, by my count, over 3x as many expired pre­dic­tions were false rather than true, even judg­ing gen­er­ously and skip­ping the hard­est. So, the pre­dic­tors as a whole are not just badly but sys­tem­at­i­cally over­con­fi­dent, and one could’ve made a con­sid­er­able profit just bet­ting against every­one blind­ly.

Valid and Available Predictions

I have of all kinds over thou­sands of pre­dic­tions (with 2,263 judged pre­dic­tions on Pre­dic­tion­Book alone as of 2017-05-16) and am well-­cal­i­brat­ed, so for each avail­able pre­dic­tion, I will make a quick pre­dic­tion of their prob­a­bil­ity to help form my expect­ed-­value esti­mates.

This gives (ex­clud­ing all the expired ones, but includ­ing the well-de­fined-but-­moronic ones, which are winnable should the pre­dic­tor be suf­fi­ciently fool­ish as to accept bets) as viable can­di­dates, going by year (the main deter­mi­nant of prof­itabil­i­ty) / pre­dic­tion / ini­tial prob­a­bil­ity esti­mate:

Optimal selection of bets & amounts

Opti­mal choice of which bets to make and how much to offer is sim­pli­fied by the upper bound of 18 years, the high prob­a­bil­ity that most bet offers will be declined, the inabil­ity to rein­vest gains (since all win­nings go imme­di­ately to char­i­ty), and risk neu­tral­ity since it’s for a char­i­ty. As far as I can tell, there’s no need for advanced appa­ra­tus like or the ; one sim­ply takes one’s avail­able total invest­ment, esti­mates the prob­a­bil­ity of win­ning to get the expected val­ue, divides by years to get annual return, and offers to bet every­thing on what­ever has the high­est implied annual return >0%; then with what­ever is left over, one moves onto the next bet, and so on until one runs out of money or bets. (So one would start with the near­est ~0% pre­dic­tion­s…)

lb <- read.csv(stdin(), header=TRUE, colClasses=c("integer", "character", "factor", "numeric"))
Year,Prediction,ID,P
2020,"By 2020 a completely propellant-less (no material particles) ...",92,0.00
2020,"By 2020, 75% of all incremental new generation will come fr...",141,0.50
2020,"By 2020, GE crops will be grown on 10 fold more acreage tha...",522,0.50
2020,"By 2020, across at least 25% of the continental US, the cos...",655,0.50
2020,"By 2020, historians will agree that the start of the 21st c...",295,0.00
2020,"By 2020, in real or artificial life, sustainable evolutiona...",44,0.00
2020,"By December 31, 2020, synthetic solar energy (fusion) will ...",605,0.50
2020,"By the year 2020 solar electricity will be as cheap or chea...",76,0.50
2020,"By the year 2020 the technology will exist that will allow ...",143,0.00
2020,"By the year 2020, the tickets to space travel - at the leas...",27,0.00
2020,"China's current regime will be overthrown or reformed (anot...",588,0.02
2020,"My prediction is that artificial life; i.e., life from non-...",195,0.50
2020,"Self-replicating manufacturing machines will be commonplace...",345,0.00
2020,"The first ice-free Arctic day (as defined by NSIDC) will oc...",653,0.40
2020,"By 2020, Urban and vertical farms will replace 10% of city ...",661,0.01
2020,"Turkish president Recep Tayyip Erdogan will be found guilty...",685,0.05
2020,"The CFR (Delivered) China price of Iron Ore (62% Fe) as pub...",718,0.15
2021,"A commercial airliner will be destroyed by terrorists hacki...",615,0.05
2021,"The rate of fatalities for seafarers will be ten times that...",724,0.50
2022,"By 02022, Aether Physics (or its equivalent) will take the ...",627,0.00
2022,"In ten years, people will be watching more live video than ...",623,0.05
2023,"Patents will be abolished by at least two leading economies...",343,0.00
2023,"Roe v. Wade will not celebrate a 50th anniversary.",347,0.33
2023,"advances in personal aviation will cause the cost of land i...",433,0.00
2024,"That by 2024 'artificial' life emerging somewhere out of th...",42,0.5
2024,"US accounting and banking regulations will not require that...",149,0.50
2024,"Apple will release an electric car within the next decade.",678,0.15
2025,"A third party candidate will be elected President of the Un...",583,0.05
2025,"By 2025 at least 50% of all U.S. citizens residing within t...",106,0.01
2025,"By 2025, Albert Einstein's theory of relativity will be con...",427,0.00
2025,"By 2025, Quantum field theory will no longer be part of The...",429,0.00
2025,"By 2025, Stellar nucleosynthesis will no longer be consider...",431,0.00
2025,"By 2025, The Aether Physics Model will not have replaced qu...",432,0.00
2025,"By 2025, The giant impact hypothesis will no longer be the ...",430,0.00
2025,"By 2025, at least 25 of the 30 stocks currently in the Dow ...",112,0.50
2025,"By 2025, more than 50% of desktop computers in the world wi...",270,0.05
2025,"By 2025, new astronomical observations and theories will re...",425,0.01
2025,"By 2025, the very first human being will be cloned and this...",284,0.10
2025,"By the end of 2024 there will be a Single Global Currency m...",226,0.00
2025,"Major online internet usage research firms will record tha...",88,0.00
2025,"Neither the Google car nor any other fully autonomous car w...",656,0.20
2025,"The concept of time as a linear dimension will be replaced ...",173,0.00
2025,"The nation with the highest (nominal) GDP in the world in 2...",491,0.50
2025,"By 2025, the underground economy in the USA will be worth m...",662,0.40
2025,"By 2025, oil and gas companies will find that their carbon ...",674,0.00
2025,"Due to a favorable political & social conditions, India's e...",684,0.03
2025,"The June, 2025 Labor Force Participation Rate and unemploym...",687,0.50
2029,"By December 31 02029 one of the world's top ten car manufac...",710,0.30
2030,"..effective computerized individual citizen profiling, with...",289,0.05
2030,"A nuclear explosive weapon will be used in war or terrorism...",521,0.25
2030,"By 2030 all surgical anesthesia will be administered and mo...",43,0.3
2030,"China will break apart by 2030",297,0.05
2030,"...China will lay an ownership claim to the entire moon.",624,0.01
2030,"By 2030, 30% of libraries existing today will not have wall...",657,0.10
2030,"By 2030, the apostrophe will have functionally disappeared ...",658,0.00
2030,"That the average level of innovation between 2025 and 2035 ...",321,0.35
2030,"Iranian women will become the dominant force in Iranian pol...",659,0.02
2031,"The average annual Palmer Drought Severity Index (PDSI) com...",652,0.60
2032,"By 02032, in response to studies correlating interactive me...",629,0.10
2032,"By 2032 you will be able to purchase a mass produced car th...",622,0.70
2035,"By the year 2035 non-invasive devices will allow us to inte...",285,0.30
2035,"One or more space agencies will send a manned mission to Ma...",351,0.40
2035,"There will be a Chernobyl National Park by 2035.",511,0.33
2035,"The United States will NOT adopt a single-payer health care...",682,0.75
2035,"The amount of geologically-derived crude oil consumed by th...",726,0.50
lb$Return <- with(lb, (2*(1-P)*1.03^(Year-2017) - 1.07^(Year-2017)) / (Year-2017))
lb[order(lb$Return, decreasing=TRUE),]
#    Year                                                     Prediction  ID    P          Return
# 1  2020 By 2020 a completely propellant-less (no material particles)... 92 0.00  0.320137000000
# 5  2020 By 2020, historians will agree that the start of the 21st c... 295 0.00  0.320137000000
# 6  2020 By 2020, in real or artificial life, sustainable evolutiona...  44 0.00  0.320137000000
# 9  2020 By the year 2020 the technology will exist that will allow ... 143 0.00  0.320137000000
# 10 2020 By the year 2020, the tickets to space travel - at the leas...  27 0.00  0.320137000000
# 13 2020 Self-replicating manufacturing machines will be commonplace... 345 0.00  0.320137000000
# 15 2020 By 2020, Urban and vertical farms will replace 10% of city ... 661 0.01  0.312852153333
# 11 2020 China's current regime will be overthrown or reformed...       588 0.02  0.305567306667
# 16 2020 Turkish president Recep Tayyip Erdogan will be found guilty... 685 0.05  0.283712766667
# 17 2020 The CFR (Delivered) China price of Iron Ore (62% Fe) as pub... 718 0.15  0.210864300000
# 18 2021 A commercial airliner will be destroyed by terrorists hacki... 615 0.05  0.206917682250
# 20 2022 By 02022, Aether Physics (or its equivalent) will take the ... 627 0.00  0.183199283580
# 21 2022 In ten years, people will be watching more live video than ... 623 0.05  0.160013802094
# 22 2023 Patents will be abolished by at least two leading economies... 343 0.00  0.147895706868
# 24 2023 advances in personal aviation will cause the cost of land i... 433 0.00  0.147895706868
# 30 2025 By 2025, Albert Einstein's theory of relativity will be con... 427 0.00  0.101919247868
# 31 2025 By 2025, Quantum field theory will no longer be part of The... 429 0.00  0.101919247868
# 32 2025 By 2025, Stellar nucleosynthesis will no longer be consider... 431 0.00  0.101919247868
# 33 2025 By 2025, The Aether Physics Model will not have replaced qu... 432 0.00  0.101919247868
# 34 2025 By 2025, The giant impact hypothesis will no longer be the ... 430 0.00  0.101919247868
# 39 2025 By the end of 2024 there will be a Single Global Currency m... 226 0.00  0.101919247868
# 40 2025 Major online internet usage research firms will record tha...   88 0.00  0.101919247868
# 42 2025 The concept of time as a linear dimension will be replaced ... 173 0.00  0.101919247868
# 45 2025 By 2025, oil and gas companies will find that their carbon ... 674 0.00  0.101919247868
# 29 2025 By 2025 at least 50% of all U.S. citizens residing within t... 106 0.01  0.098752322664
# 37 2025 By 2025, new astronomical observations and theories will re... 425 0.01  0.098752322664
# 46 2025 Due to a favorable political & social conditions, India's e... 684 0.03  0.092418472258
# 28 2025 A third party candidate will be elected President of the Un... 583 0.05  0.086084621851
# 36 2025 By 2025, more than 50% of desktop computers in the world wi... 270 0.05  0.086084621851
# 38 2025 By 2025, the very first human being will be cloned and this... 284 0.10  0.070249995833
# 27 2024     Apple will release an electric car within the next decade. 678 0.15  0.069286299249
# 55 2030 By 2030, the apostrophe will have functionally disappeared ... 658 0.00  0.040555571286
# 41 2025 Neither the Google car nor any other fully autonomous car w... 656 0.20  0.038580743799
# 53 2030       ...China will lay an ownership claim to the entire moon. 624 0.01  0.038296288650
# 57 2030 Iranian women will become the dominant force in Iranian pol... 659 0.02  0.036037006014
# 49 2030 ..effective computerized individual citizen profiling, with... 289 0.05  0.029259158105
# 52 2030                                 China will break apart by 2030 297 0.05  0.029259158105
# 14 2020 The first ice-free Arctic day (as defined by NSIDC) will oc... 653 0.40  0.028743133333
# 54 2030 By 2030, 30% of libraries existing today will not have wall... 657 0.10  0.017962744925
# 23 2023             Roe v. Wade will not celebrate a 50th anniversary. 347 0.33  0.016549954250
# 59 2032 By 02032, in response to studies correlating interactive me... 629 0.10  0.003020653944
# 50 2030 A nuclear explosive weapon will be used in war or terrorism... 521 0.25 -0.015926494616
# 48 2029 By December 31 02029 one of the world's top ten car manufac... 710 0.30 -0.021343862281
# 44 2025 By 2025, the underground economy in the USA will be worth m... 662 0.40 -0.024757760271
# 51 2030 By 2030 all surgical anesthesia will be administered and mo...  43 0.30 -0.027222907797
# 56 2030 That the average level of innovation between 2025 and 2035 ... 321 0.35 -0.038519320977
# 2  2020 By 2020, 75% of all incremental new generation will come fr... 141 0.50 -0.044105333333
# 3  2020 By 2020, GE crops will be grown on 10 fold more acreage tha... 522 0.50 -0.044105333333
# 4  2020 By 2020, across at least 25% of the continental US, the cos... 655 0.50 -0.044105333333
# 7  2020 By December 31, 2020, synthetic solar energy (fusion) will ... 605 0.50 -0.044105333333
# 8  2020 By the year 2020 solar electricity will be as cheap or chea...  76 0.50 -0.044105333333
# 12 2020 My prediction is that artificial life; i.e., life from non-... 195 0.50 -0.044105333333
# 19 2021 The rate of fatalities for seafarers will be ten times that... 724 0.50 -0.046321800000
# 25 2024 That by 2024 'artificial' life emerging somewhere out of th...  42 0.50 -0.053701087293
# 26 2024 US accounting and banking regulations will not require that... 149 0.50 -0.053701087293
# 61 2035 By the year 2035 non-invasive devices will allow us to inte... 285 0.30 -0.055362555000
# 35 2025 By 2025, at least 25 of the 30 stocks currently in the Dow ... 112 0.50 -0.056427012306
# 43 2025 The nation with the highest (nominal) GDP in the world in 2... 491 0.50 -0.056427012306
# 47 2025 The June, 2025 Labor Force Participation Rate and unemploym... 687 0.50 -0.056427012306
# 63 2035               There will be a Chernobyl National Park by 2035. 511 0.33 -0.061037331871
# 62 2035 One or more space agencies will send a manned mission to Ma... 351 0.40 -0.074278477902
# 65 2035 The amount of geologically-derived crude oil consumed by th... 726 0.50 -0.093194400805
# 58 2031 The average annual Palmer Drought Severity Index (PDSI) com... 652 0.60 -0.097747312166
# 60 2032 By 2032 you will be able to purchase a mass produced car th... 622 0.70 -0.121616739384
# 64 2035 The United States will NOT adopt a single-payer health care... 682 0.75 -0.140484208062

I began offer­ing bets with a cap­i­tal of $1200 with AMF as the ben­e­fi­cia­ry, roughly what I intended to donate in 2017. Offered bets:

  1. 2017-04-24: #143, “By the year 2020 the tech­nol­ogy will exist that will allow for the ‘fax­ing’ (teleportation—sending/receiving) of actual inan­i­mate objects, such as text books, cloth­ing, jew­elry and the like.”, Rob Schnitzer

  2. 2017-05-02: #345, “Self­-repli­cat­ing man­u­fac­tur­ing machines will be com­mon­place by 2020”, Adrian Bowyer

  3. 2017-05-11: #661, “By 2020, Urban and ver­ti­cal farms will replace 10% of city pro­duce in Chica­go.”, Ed Hub­bard

  4. 2017-05-16: #685, “Turk­ish pres­i­dent Recep Tayyip Erdo­gan will be found guilty at a court of law or will flee from the coun­try by 2020.”, Ibrahim Okuyucu

    Ibrahim expressed inter­est in accept­ing on 2017-05-19, and we began nego­ti­at­ing updated terms.

Long Bets future

Status and accomplishment

Exam­in­ing the full set of pre­dic­tions, the total activ­ity on LB can be described as rang­ing from slim to none: the ser­vice is essen­tially sta­tic as it receives no new pre­dic­tions, no new bets, and not even com­ments on exist­ing bets.

Indeed, with the expi­ra­tion of the Buf­fett bet, LB will have a grand total of ~20 bets active (9 expired), and, inas­much as the Buf­fett bet rep­re­sents ~94% of all out­stand­ing bet funds (al­beit not under the con­trol of LB, see pre­vi­ous about Buf­fett con­trol­ling & invest­ing his bet’s stake), col­lapse to a total out­stand­ing sum of ~$60k. One has to won­der if this is enough to even oper­ate the “Far­sight Fund” with, as even with ~3% returns, the annual return would be some­where <$2k, of which half goes to LB.

This implies that LB has seen, over its 15 year lifes­pan, 1.9 bets per year; and, since the major­ity date to 2002–2005 after its launch, few bets post-2005 (~10), or 0.8 bets per year.

Nor have the bets, aside from the Buf­fett bet, prompted any sub­stan­tial media inter­est or changed dis­cus­sions. (Aside from the lack of com­ments or gen­eral activ­ity on the site itself, we can also look at Wikipedia search queries—al­most immea­sur­ably small and sev­eral of which were prob­a­bly me—and at Google Trends, world­wide search trends never exceed 100 searches a day and aver­age ~25 all the way back to 2004; on Google News, I could not find any men­tions of “Long Bets” which weren’t nor­mal expres­sions; and even on Hacker News, where the orig­i­nal coder of the LB site appar­ently hangs out, there are few men­tions of it which aren’t pro­voked by ongo­ing inter­est in the Buf­fett bet.)

Com­pared to other pre­dic­tion mar­kets such as Intrade, Good Judg­ment Pro­ject, or Pre­dic­tIt, LB has had orders of mag­ni­tude fewer par­tic­i­pants, con­tracts, or wagered amounts. (Even indi­vid­ual Intrade con­tracts, such as Obama win­ning re-­elec­tion in 2012, often had an order of mag­ni­tude more vol­ume than LB’s life­time total, and just the 2012 pres­i­den­tial elec­tion on Intrade alone rep­re­sented $$291 mil­lion, or >100x more.)

So over the past 15 years, despite a splashy launch with cov­er­age in and well-­con­nected par­tic­i­pants, LB has sank like a stone, is almost totally inac­tive, and with the expi­ra­tion of the Buf­fett bet, even less atten­tion will be paid to it.

Should Long Bets continue to exist?

This raises a seri­ous ques­tion: should LB con­tinue to exist at all?

Has LB changed any con­ver­sa­tions or had any mate­r­ial impact on the world worth the con­sid­er­able invest­ment by Long Now Foun­da­tion & prin­ci­pals to cre­ate, launch, and main­tain it? LB is of course in a bet­ter posi­tion to say this than I am, but as far as I can tell (speak­ing as some­one closely inter­ested in LB since its 2002 launch and who has made pre­dic­tions on most of the LB top­ics on Pre­dic­tion­Book.­com as part of ), I would have to say no.

It is true that by the nature of their long-term goals, Long Now Foun­da­tion projects can­not and should not be eval­u­ated on annual met­rics or killed off-hand­ed­ly, but LB is fron­t-loaded in the sense that bets must be made before the long future hap­pens in order to have any effects on dis­cus­sion or fos­ter account­abil­i­ty. From this per­spec­tive, from the lack of dis­cus­sion now and the lack of bet­ting activ­i­ty, no one is being held account­able for any­thing in the near or mid-term future; so it is crys­tal-­clear that as of April 2017, LB has failed. Even if LB were to expe­ri­ence a sud­den burst of activ­ity in 2017 (not that I can imag­ine why or who it would be by), it would still be many years before any new bets matured and account­abil­ity was pos­si­ble This is a trend that has been clearly vis­i­ble since at least 2005, and over a decade lat­er, noth­ing has changed and with the end of the Buf­fett bet, pub­lic inter­est in LB will decline even fur­ther.

No orga­ni­za­tion can afford to run failed ser­vices indef­i­nitely at the expense of oth­er, much more suc­cess­ful projects such as the Clock of the Long Now. Par­tic­u­larly not non-prof­its. LB is par­a­sitic on the Long Now Foun­da­tion’s resources & activ­i­ties; with the dis­ap­pear­ance of the float from the Buf­fett bet, the fees likely will not even cover a part-­time employee work­ing 2 weeks a year on LB-re­lated mat­ters such as main­te­nance or judg­ing.

So the Long Now Foun­da­tion has 3 options:

  1. main­tain the LB sta­tus quo as a zom­bie site limp­ing along, unmod­er­at­ed, with near-zero activ­i­ty, get­ting per­haps 1 or 2 bets in a good year, steadily drain­ing Foun­da­tion labor and increas­ing over­head, for no par­tic­u­lar ben­e­fit other than stub­born­ness.

  2. shut down LB

    As LB has so few bets, this would be easy. LB can sim­ply close bets which are clearly going to be won by one side, and can ‘push’ all bets where it’s unclear who will win by trans­fer­ring the funds to the pre-des­ig­nated char­i­ties. This is per­mit­ted by the legal agree­ment signed by all bet­tors, and they can hardly com­plain if their favored char­i­ties get the funds ear­lier than expected (and the char­i­ties them­selves will be very pleased to have the money ear­ly, because of the oppor­tu­nity cost, before too much time has been wasted or infla­tion evis­cer­ated the dona­tion­s). Then the site can be con­verted to a sta­tic HTML site for his­tor­i­cal archive pur­pos­es.5 Long Now Foun­da­tion can then mon­i­tor ongo­ing attempts to fill the pre­dic­tion mar­ket niche such as , or decen­tral­ized cryp­tocur­ren­cy-based projects like Hive­Mind/ & Gno­sis, and use them, either by back­ing par­tic­u­lar con­tracts and serv­ing as an oracle/judge (in cen­tral­ized mod­els like Pre­dic­tIt or Gno­sis) or by endors­ing & sub­si­diz­ing con­tracts (HiveMind/Augur).

  3. invest in LB fur­ther and attempt to revi­tal­ize or reboot it

Fixes

What could the Long Now Foun­da­tion do to improve?

  • remove the arbi­trary and unnec­es­sary even-odds restric­tion. There is no legal require­ment for this that I am aware of, and it bizarrely lim­its what bets can be made and also elim­i­nates any mean­ing­ful expres­sion of con­fi­dence. Two peo­ple who think an event might hap­pen with 40% and 60% can make bets, but two peo­ple who think 50% and 100% can’t!

  • a non-­con­ser­v­a­tive invest­ment fund, and specif­i­cal­ly, a global Van­guard stock mar­ket index: the ‘low risk’ invest­ment vehi­cle is coun­ter­pro­duc­tive, unnec­es­sary, and risky:

    1. coun­ter­pro­duc­tive: the low return makes it irre­spon­si­ble and uneth­i­cal to reg­is­ter pre­dic­tions, par­tic­u­larly for long-­dated pre­dic­tions, dis­cour­ag­ing any­one seri­ous about char­ity & the future from using LB

    2. unnec­es­sary:

      • legal­ly: as LB makes clear, it has no legal oblig­a­tion to the ‘bet­tors’, who are actu­ally dona­tors to the Long Now Foun­da­tion, to pay any­thing to the des­ig­nated char­i­ties, much less at a par­tic­u­lar return
      • moral­ly: dona­tors should be risk-neu­tral and seek to max­i­mize their even­tual impact, so trad­ing off enor­mous long-term growth for some reduc­tion in volatil­ity reduces the expected value of donating/betting
      • finan­cial­ly: for short­-term bets, most bets are <=$1k, and the pos­si­ble loss triv­ially small either as a frac­tion of either the Long Now Foun­da­tion’s assets or annual rev­enue so it can sim­ply self­-in­sure (or ask one of its wealthy donors to guar­an­tee it); for medi­um-term bets, fluc­tu­a­tions down­ward can be com­pen­sated for by bor­row­ing from long-term bets; and for long-term bets (which may be cen­turies away from matur­ing), that very aspect implies that the final invest­ment will far exceed the nom­i­nal bet and con­verge on the long-term aver­ages of 7% annu­ally (as indeed LB’s largest and most promi­nent bet demon­strates on a much shorter time-s­cale!)
    3. and risky: con­ser­v­a­tive ‘safe’ invest­ments mak­ing bet­ting highly unat­trac­tive caus­ing lack of use is itself a highly risky invest­ment (where do they go if the Long Now Foun­da­tion shuts down LB or itself dis­solves?); low growth rates in ‘safe’ invest­ments com­mit the fal­lacy inher­ent in the “pre­cau­tion­ary prin­ci­ple” by being intrin­si­cally risky as they build up no reserves against ran­dom shocks and extra­or­di­nary expens­es, entire classes of invest­ments being zeroed out by dis­as­ters or hyper­in­fla­tion.

      It is worth not­ing that many classes of bond hold­ers through­out invest­ing his­tory have failed to out­pace infla­tion & tax­a­tion, or been sim­ply zeroed out (Credit Suisse 2013), and in the long run, “safe” assets returns have been extremely low and more volatile than so-­called “risky assets” (Jordà et al 2017). As War­ren Buf­fett remarks in his 2011 annual let­ter and again in his 2017 annual let­ter, invest­ments in pro­duc­tive equi­ties & busi­ness rather than bonds or com­modi­ties are not just lucra­tive in the long run but they “will be by far the safest.”67 This is not sur­pris­ing, con­sid­er­ing the long-term aver­age real return rates cal­cu­lated by Jordà et al 2017 of 1.94% vs 4.64% over 16 coun­tries 1870–2014 (surely a long and event­ful time peri­od), imply­ing that an ‘risky’ equity invest­ment would yield a total return 44x than an equity one8, which clearly would be much safer in any mean­ing­ful sense of the term.

  • real mod­er­a­tion: many pre­dic­tions should never have been approved, and even the decent ones often need to be tight­ened up and made more objec­tive & fal­si­fi­able; it’s absurd to charge $50 to sub­mit a pre­dic­tion and then let through crack­pot­tery and out­right balder­dash like “The con­cept of time as a lin­ear dimen­sion will be replaced by one of time as a polar­ity between con­tent and con­text”. (How could any­one take such a site seri­ous­ly?)

    • pos­si­bly require min­i­mum bets to be paid up front, to avoid peo­ple delib­er­ately weasel­ing out or being incen­tivized to kill nego­ti­a­tions with prospec­tive bet­tors; since in the­o­ry, peo­ple mak­ing pre­dic­tions are will­ing to bet money on the pre­dic­tion they are mak­ing, they should have no grounds to object. If no bet­tors appear, the money can be kept until expi­ra­tion and then donated to the pre­dic­tor’s pre-spec­i­fied char­i­ty. (This mech­a­nism would be far more appeal­ing with the invest­ment fund fixed.)
  • more active pro­mo­tion by Long Now Foun­da­tion mem­bers, direc­tors, and promi­nent donors: look­ing over the bet dates, many date from a nar­row time period in the early 2000s, after which LB appears to’ve fallen into obscu­rity and gone dor­mant; but many of the peo­ple on the Long Now board or who have been involved in Long Now are well-off, donate to char­ity reg­u­lar­ly, and remain heav­ily involved in writ­ing and activism on issues rang­ing from tech­no­log­i­cal unem­ploy­ment to nuclear power to the Russ­ian Inter­net—thus, they could eas­ily route their char­i­ta­ble giv­ing through long bets on top­ics of cur­rent inter­est, as well as elim­i­nat­ing the worse pre­dic­tions & enforc­ing a basic level of san­ity and seri­ous­ness

  • com­mu­nica­tive more clearly to the pub­lic what Long Now Foun­da­tion projects are actu­ally active, and what con­crete progress is being made; the Long Now web­site cur­rently fails at this, as it appar­ently is devoted to telling peo­ple about talks they can­not attend (and which are never made avail­able online) and improve­ments at a bar they can­not drink at, while peo­ple are more inter­ested in things like progress on the Rosetta stones or the Clock


  1. Although it’s not always clear how active any of the projects are, except their bar and talk­s—they are very proud of their bar & talks and the web­site is fre­quently updated about them.↩︎

  2. Buf­fett repeat­edly empha­sizes in his let­ters that one of the keys to Berk­shire Hath­away’s busi­ness suc­cess is its insur­ers, who in turn are enabled by the ‘float’ of insur­ance pre­mi­ums which are paid long before any claims are made, and this ‘free’ money can be invested as Buf­fett wish­es; so the cost of giv­ing LB $1m for such an extended time surely was not lost on him. Sim­i­lar­ly, he notes in his 2013 annual let­ter that his estate will not be put into a long-term fund for invest­ment but will be dis­trib­uted for phil­an­thropic invest­ment as quickly as pos­si­ble: “all of my Berk­shire shares will be fully dis­trib­uted to cer­tain phil­an­thropic orga­ni­za­tions over the ten years fol­low­ing the clos­ing of my estate”.↩︎

  3. At this point, one has to won­der what func­tion LB plays in the bet at all, if it does not hold the funds, does not invest them in its Fore­sight Fund, and Buf­fett (not LB) is the one with access to the con­fi­den­tial audited results from the hedge fund side of the bet in order to judge results?↩︎

  4. The Long Now Foun­da­tion does­n’t pro­vide its Form 990s online, so I have copied them from Cit­i­zenAu­dit.org: 2001 / 2002 / 2003 / 2004 / 2005 / 2006 / 2007 / 2008 / 2009 / 2010 / 2011 / 2012 / 2013 / 2014.↩︎

  5. Or sim­ply taken down; I have already .↩︎

  6. pg18–19, Buf­fett 2011 (em­pha­sis in orig­i­nal):

    Invest­ments that are denom­i­nated in a given cur­rency include mon­ey-­mar­ket funds, bonds, mort­gages, bank deposits, and other instru­ments. Most of these cur­ren­cy-based invest­ments are thought of as “safe.” In truth they are among the most dan­ger­ous of assets. Their beta may be zero, but their risk is huge. Over the past cen­tury these instru­ments have destroyed the pur­chas­ing power of investors in many coun­tries, even as the hold­ers con­tin­ued to receive timely pay­ments of inter­est and prin­ci­pal. This ugly result, more­over, will for­ever recur…Even in the U.S., where the wish for a sta­ble cur­rency is strong, the dol­lar has fallen a stag­ger­ing 86% in value since 1965, when I took over man­age­ment of Berk­shire. It takes no less than $$9 today to buy what $$6 did at that time. Con­se­quent­ly, a tax-free insti­tu­tion would have needed 4.3% inter­est annu­ally from bond invest­ments over that period to sim­ply main­tain its pur­chas­ing pow­er. Its man­agers would have been kid­ding them­selves if they thought of any por­tion of that inter­est as “income.”… Cur­rent rates, how­ev­er, do not come close to off­set­ting the pur­chas­ing-power risk that investors assume. Right now bonds should come with a warn­ing label… Today, a wry com­ment that Wall Streeter Shelby Cul­lom Davis made long ago seems apt: “Bonds pro­moted as offer­ing risk-free returns are now priced to deliver return-free risk.”

    …My own pref­er­ence—and you knew this was com­ing—is our third cat­e­go­ry: invest­ment in pro­duc­tive assets, whether busi­ness­es, farms, or real estate. Ide­al­ly, these assets should have the abil­ity in infla­tion­ary times to deliver out­put that will retain its pur­chas­ing-power value while requir­ing a min­i­mum of new cap­i­tal invest­ment. Farms, real estate, and many busi­nesses such as Coca-­Co­la, IBM and our own See’s Candy meet that dou­ble-bar­reled test. Cer­tain other com­pa­nies—­think of our reg­u­lated util­i­ties, for exam­ple—­fail it because infla­tion places heavy cap­i­tal require­ments on them. To earn more, their own­ers must invest more. Even so, these invest­ments will remain supe­rior to non­pro­duc­tive or cur­ren­cy-based assets. Whether the cur­rency a cen­tury from now is based on gold, seashells, shark teeth, or a piece of paper (as today), peo­ple will be will­ing to exchange a cou­ple of min­utes of their daily labor for a Coca-­Cola or some See’s peanut brit­tle. In the future the U.S. pop­u­la­tion will move more goods, con­sume more food, and require more liv­ing space than it does now. Peo­ple will for­ever exchange what they pro­duce for what oth­ers pro­duce.

    Our coun­try’s busi­nesses will con­tinue to effi­ciently deliver goods and ser­vices wanted by our cit­i­zens. Metaphor­i­cal­ly, these com­mer­cial “cows” will live for cen­turies and give ever greater quan­ti­ties of “milk” to boot. Their value will be deter­mined not by the medium of exchange but rather by their capac­ity to deliver milk. Pro­ceeds from the sale of the milk will com­pound for the own­ers of the cows, just as they did dur­ing the 20th cen­tury when the Dow increased from 66 to 11,497 (and paid loads of div­i­dends as well). Berk­shire’s goal will be to increase its own­er­ship of first-­class busi­ness­es. Our first choice will be to own them in their entire­ty—but we will also be own­ers by way of hold­ing siz­able amounts of mar­ketable stocks. I believe that over any extended period of time this cat­e­gory of invest­ing will prove to be the run­away win­ner among the three we’ve exam­ined. More impor­tant, it will be by far the safest.

    ↩︎
  7. pg 12–14, dis­cussing the rea­son Buf­fett & his bet­tor agreed to sell the orig­i­nal US Trea­sury back­ing the bet for BRK shares (all empha­sis in orig­i­nal):

    Orig­i­nal­ly, Pro­tégé and I each funded our por­tion of the ulti­mate $1 mil­lion prize by pur­chas­ing $660,908 face amount of zero-­coupon U.S. Trea­sury bonds (some­times called “strips”). These bonds cost each of us $420,668—a bit less than 64¢ on the dol­lar—with the $500,000 payable in ten years.

    As the name implies, the bonds we acquired paid no inter­est, but (be­cause of the dis­count at which they were pur­chased) deliv­ered a 4.56% annual return if held to matu­ri­ty. Pro­tégé and I orig­i­nally intended to do no more than tally the annual returns and dis­trib­ute $1 mil­lion to the win­ning char­ity when the bonds matured late in 2017.

    After our pur­chase, how­ev­er, some very strange things took place in the bond mar­ket. By Novem­ber 2012, our bond­s—now with about five years to go before they matured—were sell­ing for 95.7% of their face val­ue. At that price, their annual yield to matu­rity was less than 1%. Or, to be pre­cise, 0.88%.

    Given that pathetic return, our bonds had become a dum­b—a really dum­b­—in­vest­ment com­pared to Amer­i­can equi­ties. Over time, the S&P 500—which mir­rors a huge cross-­sec­tion of Amer­i­can busi­ness, appro­pri­ately weighted by mar­ket val­ue—has earned far more than 10% annu­ally on share­hold­ers’ equity (net worth).

    In Novem­ber 2012, as we were con­sid­er­ing all this, the cash return from div­i­dends on the S&P 500 was 2.5% annu­al­ly, about triple the yield on our U.S. Trea­sury bond. These div­i­dend pay­ments were almost cer­tain to grow. Beyond that, huge sums were being retained by the com­pa­nies com­pris­ing the 500. These busi­nesses would use their retained earn­ings to expand their oper­a­tions and, fre­quent­ly, to repur­chase their shares as well. Either course would, over time, sub­stan­tially increase earn­ings-per-share. And—as has been the case since 1776—what­ever its prob­lems of the min­ute, the Amer­i­can econ­omy was going to move for­ward.

    Pre­sented late in 2012 with the extra­or­di­nary val­u­a­tion mis­match between bonds and equi­ties, Pro­tégé and I agreed to sell the bonds we had bought five years ear­lier and use the pro­ceeds to buy 11,200 Berk­shire “B” shares. The result: Girls Inc. of Omaha found itself receiv­ing [$2,222,279]($2017) last month rather than the $1 mil­lion it had orig­i­nally hoped for.

    Berk­shire, it should be empha­sized, has not per­formed bril­liantly since the 2012 sub­sti­tu­tion. But bril­liance was­n’t need­ed: After all, Berk­shire’s gain only had to beat that annual 0.88% bond bogey—hardly a Her­culean achieve­men­t…After that kinder­garten-­like analy­sis, Pro­tégé and I made the switch and relaxed, con­fi­dent that, over time, 8% was cer­tain to beat 0.88%. By a lot.

    …In­vest­ing is an activ­ity in which con­sump­tion today is fore­gone in an attempt to allow greater con­sump­tion at a later date. “Risk” is the pos­si­bil­ity that this objec­tive won’t be attained.

    By that stan­dard, pur­port­edly “risk-free” long-term bonds in 2012 were a far riskier invest­ment than a long-term invest­ment in com­mon stocks. At that time, even a 1% annual rate of infla­tion between 2012 and 2017 would have decreased the pur­chas­ing-power of the gov­ern­ment bond that Pro­tégé and I sold.

    I want to quickly acknowl­edge that in any upcom­ing day, week or even year, stocks will be riskier—­far riskier—than short­-term U.S. bonds. As an investor’s invest­ment hori­zon length­ens, how­ev­er, a diver­si­fied port­fo­lio of U.S. equi­ties becomes pro­gres­sively less risky than bonds, assum­ing that the stocks are pur­chased at a sen­si­ble mul­ti­ple of earn­ings rel­a­tive to then-pre­vail­ing inter­est rates.

    It is a ter­ri­ble mis­take for investors with long-term hori­zon­s—a­mong them, pen­sion funds, col­lege endow­ments and sav­ings-­minded indi­vid­u­al­s—to mea­sure their invest­ment “risk” by their port­fo­lio’s ratio of bonds to stocks. Often, high­-­grade bonds in an invest­ment port­fo­lio increase its risk.

    ↩︎
  8. Over the 145 year peri­od, the mean returns from Table 3, pg13 (ge­o­met­ric mean of real returns in full sam­ple), imply a com­pounded return of for bonds, for equi­ties, and .↩︎