Empire Theory, Part I: Competitive Landscape

[Cross­posted from my Medium blog]

Em­pire The­ory is a frame­work for un­der­stand­ing and prac­tic­ing com­pet­i­tive strat­egy. Com­pet­i­tive strat­egy is the art of defeat­ing op­po­nents. once you have cho­sen a do­main of com­pe­ti­tion, good com­pet­i­tive strat­egy en­ables you to win.

Com­pet­i­tive strat­egy re­quires un­der­stand­ing how ac­tors be­have based on their po­si­tion in a strate­gic land­scape. This knowl­edge serves two clear pur­poses. First, by rec­og­niz­ing the pat­terns of these strate­gic play­ers, it’s pos­si­ble to in­fer a vast amount about the strate­gic land­scape on the ba­sis of rel­a­tively lit­tle ev­i­dence. Se­cond, a deeper un­der­stand­ing of strate­gic moves and op­po­nents’ in­cen­tives al­lows us to bet­ter craft our own com­pet­i­tive strat­egy, through pre­dict­ing, plan­ning for, and re­spond­ing to be­hav­ior.


Here we use em­pire to mean a group of co­or­di­nated ac­tors that op­er­ate around some cen­tral power. Co­or­di­nated ac­tors are those peo­ple us­ing dis­cernible mechanisms for al­ign­ing their ac­tions to achieve par­tic­u­lar goals. A cen­tral power is an ac­tor or set of ac­tors caus­ing oth­ers in a given re­gion to co­or­di­nate. The ac­tual cen­tral power may not be the os­ten­si­ble cen­tral power; for ex­am­ple, a startup might be de facto run by its CTO rather than its CEO. An em­pire then, be­ing a group of co­or­di­nated ac­tors, will among those ac­tors always have some kind of cen­tral power that is main­tain­ing co­or­di­na­tion. Let’s list some ex­am­ple em­pires to illus­trate:

A com­pany:

Co­or­di­nated ac­tors: Em­ploy­ees, busi­ness part­ners, customers

Cen­tral power: The CEO /​ executives

A gov­ern­ment:

Co­or­di­nated ac­tors: The civil ser­vice, the mil­i­tary, cor­po­ra­tions, citizens

Cen­tral power: the king /​ the pres­i­dent /​ the legislature

The Muski­verse:

Co­or­di­nated ac­tors: Peo­ple at SpaceX, Tesla, So­larcity, and the Bor­ing Com­pany, per­haps others

Cen­tral power: Elon Musk

The Frac­tal Na­ture of Empires

Em­pires are frac­tal. There will be sub-em­pires within any given em­pire. In the Catholic church, for ex­am­ple, we could con­sider the co­or­di­nated ac­tors to be the global Catholic clergy plus lay peo­ple and the cen­tral power to be the lead­ers at the Vat­i­can. How­ever, it also makes sense to con­sider a sin­gle parish as an em­pire where the co­or­di­nated ac­tors are the mem­bers of the parish and the cen­tral power is the priest. Like­wise, a so­cial move­ment like Effec­tive Altru­ism could be con­sid­ered an em­pire where the co­or­di­nated ac­tors are the mem­bers of the move­ment and the cen­tral power is the cluster of peo­ple and or­ga­ni­za­tions guid­ing the ide­ol­ogy and strate­gies of the rest. That said, an in­di­vi­d­ual or­ga­ni­za­tion within the move­ment could also be con­sid­ered an em­pire.

The frac­tal na­ture of em­pires fol­lows from the frac­tal na­ture of co­or­di­na­tion mechanisms. An em­pire can be iden­ti­fied ei­ther by notic­ing a group co­or­di­nat­ing, or by iden­ti­fy­ing a co­or­di­na­tion mechanism and then iden­ti­fy­ing the ac­tors co­or­di­nated by that mechanism. As there will be differ­ent co­or­di­na­tion mechanisms pre­sent at var­i­ous places within an em­pire, and thus sub-clusters of tighter co­or­di­na­tion, em­pires will be frac­tal.

The Con­tents of Empires

Em­pires are com­posed of play­ers, re­sources, and other em­pires. Play­ers are the in­di­vi­d­u­als with enough power to be rele­vant to the over­all func­tion­ing of the em­pire. Re­sources are as­sets that can be drawn upon for the em­pire to func­tion. This cat­e­gory fea­tures many things be­sides phys­i­cal re­sources, in­clud­ing money, in­for­ma­tion, and per­sonal re­la­tion­ships. Co­or­di­na­tion mechanisms (both nat­u­ral and con­structed) and peo­ple that are not suffi­ciently pow­er­ful to be rele­vant for the over­all func­tion­ing of the em­pire are also con­sid­ered re­sources. Fi­nally, be­cause em­pires are frac­tal, em­pires con­tain other em­pires.

The Prob­lem of Lo­cal Focus

In a given em­pire, the dy­nam­ics of the most cen­tral sub-em­pire have a large effect on the rest of the em­pire, and con­trol of the cen­tral sub-em­pire is im­por­tant to top strate­gic play­ers as it yields con­trol of the rest of the em­pire. As a re­sult, the top play­ers in an em­pire tend to pri­ori­tize con­trol­ling the cen­tral sub-em­pire. This phe­nomenon re­peats in a frac­tal man­ner. To illus­trate, con­sider the United States an em­pire, and the pres­i­dent of the United States a player seek­ing to con­trol the em­pire. Within the United States, let’s say the cen­tral sub-em­pire is the ex­ec­u­tive branch. Within the ex­ec­u­tive branch, let’s say the cen­tral sub-em­pire is the cab­i­net. If the pres­i­dent can­not con­trol the cab­i­net, then it will be much more difficult for him to con­trol the ex­ec­u­tive branch. If he can­not con­trol the ex­ec­u­tive branch, then it will be much more difficult to con­trol the United States gov­ern­ment.

A great deal of re­sources then tends to be spent on con­trol of the cen­tral sub-em­pire. This al­lo­ca­tion of re­sources de­tracts from the proper func­tion­ing of the rest of the em­pire and hurts the em­pire’s ex­pan­sion, as more re­sources spent on cen­tral in­fight­ing means fewer re­sources spent on other things es­sen­tial to the em­pire’s func­tion­ing. Un­for­tu­nately, this out­sized ex­pen­di­ture is not the re­sult of cor­rup­tion and whimsy, but poli­ti­cal ne­ces­sity (a lot of what we usu­ally call “cor­rup­tion” stems from poli­ti­cal ne­ces­sity). This prob­lem of lo­cal fo­cus is one of strongest limit­ing fac­tors on the sizes of em­pires, be­cause the prob­lem tends to get worse as an em­pire gets larger. The prob­lem of lo­cal fo­cuses in­creases in larger em­pires be­cause the more power an em­pire has,the more skil­led play­ers are at­tracted to it. The more skil­led play­ers are at­tracted to a given em­pire, the more difficult it is to con­trol the cen­tral sub-em­pire. The more difficult it is to con­trol the cen­tral sub-em­pire, the more difficult it is to pre­serve and ex­pand the em­pire. As a re­sult, the prob­lem of lo­cal fo­cus hugely limits the ex­pan­sion of em­pires.

Power Classes

The co­or­di­nated ac­tors in an em­pire will have differ­ing amounts of power. For ex­am­ple, con­sider a tech startup as an em­pire. The founder can hire and fire peo­ple, will usu­ally play the lead role in de­ter­min­ing the startup’s strat­egy, and can con­tribute di­rectly to the cre­ation of the com­pany’s product. In con­trast, a newly hired pro­gram­mer may only be able to con­tribute to the product. As such, the founder has more power in the em­pire than the newly hired em­ployee. Power classes are a ty­pol­ogy of the co­or­di­nated ac­tors in an em­pire on the ba­sis of their rel­a­tive power lev­els.

High is the cen­tral power that defines an em­pire’s zone of co­or­di­na­tion. Without high, the em­pire would not ex­ist and the other ac­tors would not be co­or­di­nated. High also plays the largest role in de­ter­min­ing the dis­tri­bu­tion of re­sources within the em­pire. High can be an in­di­vi­d­ual (e.g. a force­ful CEO) or a group (e.g. the board of di­rec­tors of a foun­da­tion). It will of­ten make sense to model high as an em­pire in it­self,be­cause there are nat­u­rally oc­cur­ring co­or­di­na­tion mechanisms that cause high to be its own cluster of co­or­di­na­tion within an em­pire, and there are usu­ally a small num­ber of in­di­vi­d­u­als in high that co­or­di­nate the other high play­ers (a high within high). Th­ese nat­u­ral co­or­di­na­tion mechanisms in­clude that high play­ers are mu­tu­ally threat­ened by mid­dle play­ers and by ag­gres­sive out­side em­pires.

Mid is the col­lec­tion of in­di­vi­d­u­als or groups that have suffi­cient power to challenge high’s con­trol. Mid play­ers will of­ten have smaller em­pires of their own. Mid plays an im­por­tant role in con­strain­ing the ac­tion of high. In our tech startup ex­am­ple, mid play­ers might be the man­agers of the en­g­ineer­ing and sales teams.

It does not usu­ally make sense to model mid as a sin­gle em­pire. They are very sel­domly co­or­di­nated as such.

Be­cause mid play­ers con­trol fewer re­sources than high play­ers, any mid player will have to ex­pend a greater por­tion of their re­sources to se­cure the co­or­di­na­tion of a fel­low mid player. An in­vest­ment of $1M is a no­table and risky ven­ture when your net worth is $20M. It might be an af­terthought if your net worth is $2B.

Each in­di­vi­d­ual mid player con­trols no­tably fewer re­sources than high, you have to co­or­di­nate more of them to reach the same ca­pa­bil­ities a sin­gle high player can provide. Co­or­di­na­tion costs are su­per­lin­ear, so pool­ing any­thing ex­cept the sim­plest re­sources in this way is un­eco­nom­i­cal. Co­or­di­nat­ing 30 differ­ent strate­gic play­ers rather than 3, is like­lier to in­crease costs by a fac­tor of 100 rather than 10.

For any given mid player, high is usu­ally a prefer­able ally to other mid play­ers. Given these known prob­lems and the ex­ist­ing un­cer­tainty in mu­tual eval­u­a­tion, a mid player must then not only match, but out­bid the offer high mid give. This is in­fre­quent.

Low is the col­lec­tion of play­ers that can challenge mid but can­not challenge high. Low has the largest pop­u­la­tion and the least power. In our tech startup ex­am­ple, the low play­ers would be in­di­vi­d­ual pro­gram­mers or sales peo­ple. The pro­gram­mers on an en­g­ineer­ing team could plau­si­bly challenge their man­ager, but they could not plau­si­bly challenge the founder. Like mid, it does not make sense to model low as an em­pire.

Out­side is any ac­tor that is not co­or­di­nated by the high power. In our ex­am­ple, this could be an em­ployee at an­other com­pany or the mayor of a town in France. Out­side play­ers may still seek to af­fect an em­pire, in­clud­ing by med­dling in its in­ter­nal af­fairs.

As men­tioned ear­lier, cer­tain ac­tors are best mod­eled as re­sources. Any ac­tor that can­not in­de­pen­dently challenge mid is best un­der­stood as a re­source, be­cause these ac­tors will not be rele­vant for un­der­stand­ing the em­pire. They can be un­der­stood as re­sources, be­cause they will be used by low, mid, and high play­ers to ac­com­plish their ob­jec­tives. For ex­am­ple, they might provide la­bor or be weaponized by play­ers against each other.

Ex­am­ples of clas­sify­ing by Power Class

In the United States to­day, high is best un­der­stood as be­ing com­posed of key fed­eral agen­cies. Heads of ma­jor in­sti­tu­tions such as large com­pa­nies, banks, uni­ver­si­ties or gov­er­nors of in­di­vi­d­ual states can be un­der­stood as mid. State offi­cials, heads of lo­cal groups and smaller or­ga­ni­za­tions can be un­der­stood as low. Every­one else is best mod­eled as a re­source. Rele­vant out­side pow­ers con­sist of key for­eign gov­ern­ments such as China or Rus­sia.

At Har­vard Univer­sity, per­haps high is oc­cu­pied by the pres­i­dent, provost, deans, vice pres­i­dents, or trustees. Mid might be key pro­fes­sors, long-time staff, heads of de­part­ments and ma­jor donors. Low might be stu­dent or­ga­niz­ers or less im­por­tant pro­fes­sors. Other stu­dents, as­sis­tant pro­fes­sors, re­place­able staff, and smaller donors. Rele­vant out­side play­ers might be com­pa­nies that re­cruit from the uni­ver­sity or the lo­cal city gov­ern­ment.

Power Classes are Fractal

Like em­pires, power classes are frac­tal. The same ac­tor can be clas­sified as low, mid, or high de­pend­ing on the frame of refer­ence. For ex­am­ple, a parish priest in New York might be low if con­sid­er­ing the en­tire Catholic church, mid if con­sid­er­ing the Arch­dio­cese of New York, and high if con­sid­er­ing the priest’s parish it­self.

Cau­tions in Classification

The offi­cial story of who is and is not pow­er­ful does not always match the ac­tual story. For ex­am­ple, it might be that the pres­i­dent of Har­vard has only mod­er­ate in­ter­nal in­fluence and that one of the deans has by far the most in­ter­nal in­fluence. In this case, the pres­i­dent might be bet­ter clas­sified as a mid-player. When as­sign­ing in­di­vi­d­u­als and groups to power classes in an em­pire, be skep­ti­cal of your as­sess­ments, as it is easy to as­sume power dis­tri­bu­tions based on the offi­cial story.

Strate­gic Landscapes

A strate­gic land­scape is a do­main of com­pe­ti­tion among play­ers. A do­main of com­pe­ti­tion is a re­gion in which play­ers com­pete for scarce re­sources.

Try­ing to an­a­lyze a strate­gic land­scape with­out spec­i­fy­ing a do­main of com­pe­ti­tion will yield con­fu­sion and er­ror. If the do­main of com­pe­ti­tion isn’t speci­fied, ends and means can­not be dis­t­in­guished. Most ac­tions are am­bigu­ous, so un­less they are in­ter­preted through a definite hy­poth­e­sis, in­ves­ti­ga­tion has no clear di­rec­tion and un­cer­tainty can­not be re­solved.

This ap­proach dis­t­in­guishes the mere ac­cu­mu­la­tion of facts from anal­y­sis. The cru­cial task is de­ter­min­ing which facts are rele­vant and pri­ori­tiz­ing them. While you might imag­ine a lo­gis­ti­cal anal­y­sis that doesn’t spec­ify a do­main of com­pe­ti­tion, it will fail to pre­dict the range of in­ter­ac­tions be­tween play­ers.

You might cor­rectly note the in­dus­trial ca­pac­i­ties in a par­tic­u­lar re­gion, but if you are not keep­ing track of whether the fac­to­ries are al­igned ei­ther through an owner, a state or an oli­gopoly or whether they are mis­al­igned, you will fail to pre­dict which prod­ucts can be built or which pro­jects will be car­ried out.

Since play­ers can mod­ify any mere lo­gis­ti­cal fact, the ac­cu­mu­la­tion of facts with­out knowl­edge of the do­main of com­pe­ti­tion will per­haps cor­rectly show the func­tion­ing of some sys­tems but will fail to pre­dict changes in the sys­tem.

For ex­am­ple, an­a­lyz­ing the strate­gic land­scape that in­cludes the oil in­dus­try and the so­cial jus­tice move­ment with­out spec­i­fy­ing a re­source they are com­pet­ing over will re­sult in some­thing like a list of re­ports of me­dia events and gen­eral be­liefs. To un­der­stand their dy­nam­ics or even cor­rectly eval­u­ate the facts on the ground you have to iden­tify ei­ther a definite con­flict point, or their over­all strate­gic aims and po­si­tion.

You might be­gin to an­a­lyze them as a part of the poli­ti­cal strate­gic land­scape in which the re­source com­peted over is the alle­giance of a par­tic­u­lar con­gress­man.

The oil in­dus­try might have the abil­ity to offer pos­i­tive re­sources in the form of fi­nan­cial or le­gal sup­port for the con­gress­man. Their pur­pose for com­pet­ing in the poli­ti­cal land­scape be­ing fa­vor­able leg­is­la­tion for their in­dus­try. The so­cial jus­tice move­ment might be able to mount a cam­paign against the con­gress­man, at­tack­ing their char­ac­ter. Their pur­pose be­ing so­cial re­form, per­haps through leg­is­la­tion.

In the com­pet­i­tive sce­nario laid out it would only makes sense for an un­pop­u­lar or weak con­gress­man to go with so­cial jus­tice, and only tem­porar­ily, since all they can offer is to stay their hand, while the oil in­dus­try can provide use­ful re­sources, that im­proves the con­gress­man’s long term po­si­tion.

Em­pires are do­mains of com­pe­ti­tion, and do­mains of com­pe­ti­tion tend to be em­pires; em­pires are always do­mains of com­pe­ti­tion in which play­ers are com­pet­ing for power, and do­mains of com­pe­ti­tion al­most always have co­or­di­nat­ing mechanisms bind­ing the com­peti­tors to­gether (for ex­am­ple, com­peti­tors in the oil in­dus­try co­or­di­nat­ing to defeat clean-air leg­is­la­tion).

The term “land­scape” pro­vides a use­ful metaphor for think­ing about these do­mains of com­pe­ti­tion. You can think of the ter­rain of a strate­gic land­scape as be­ing de­ter­mined by the com­peti­tors and their rel­a­tive power. Imag­ine your­self stand­ing on a precipice over­look­ing a strate­gic land­scape of a uni­ver­sity. You see rol­ling hills off to the left, some of which are larger than oth­ers, rep­re­sent­ing the heads of the var­i­ous hu­man­i­ties’ de­part­ments. In the mid­dle is a tower moun­tain rep­re­sent­ing the cen­tral ad­minis­tra­tion, upon which there is high rocky out­crop rep­re­sent­ing the pres­i­dent of the uni­ver­sity. The land­scape is not static, but dy­namic, with the ter­rain shift­ing as play­ers make moves and gain or lose power. If you want to com­pete in this strate­gic land­scape, you will have to nav­i­gate it, tak­ing into ac­count the pow­ers of the other play­ers in de­ter­min­ing your path, your com­pet­i­tive strat­egy. The same goes for the other com­peti­tors.


Ear­lier I claimed that ac­tors ex­hibit com­mon pat­terns of be­hav­ior de­pend­ing upon their rel­a­tive po­si­tion in a strate­gic land­scape. Now we can parse this: in a do­main of com­pe­ti­tion, as­pects of the be­hav­ior of high, mid, and low play­ers will be con­sis­tent and rec­og­niz­able. This means, for in­stance, that there are pat­terns of in­ter­ac­tion be­tween high and mid, and that, if we iden­tify high and mid in a par­tic­u­lar do­main, we will im­me­di­ately learn much about how those play­ers will be­have. The com­mon be­hav­iors of play­ers are a con­se­quence of what works and does not work for play­ers given their po­si­tion on the land­scape. Un­der­stand­ing such pat­terns thus sub­stan­tially broad­ens one’s range of available strate­gic op­tions. We will ex­plore these dy­nam­ics in de­tail in part two.

Con­tinues with Part 2: Power Dynamics