An overview of arguments for concern about automation

Epistemic sta­tus: Fairly con­fi­dent on the eco­nomics, less so on the poli­ti­cal sci­ence points. This write-up is mainly the re­sult of a class I took on au­toma­tion and the la­bor mar­ket with pro­fes­sor Nir Jaimovich as well as the book ‘The Tech­nol­ogy Trap’ by Carl Benedikt Frey. Sec­tion 3.2 on think­ing about the fu­ture is (nec­es­sar­ily) more spec­u­la­tive and un-grounded in ev­i­dence.

Qual­ifi­ca­tions: I have finished the course­work for a Master’s in Eco­nomics which gives me some abil­ity to eval­u­ate the ar­gu­ments, though I have not taken the time to eval­u­ate the method­ol­ogy of ev­ery­thing cited.

The fol­low­ing is fo­cused al­most en­tirely on the down­side risks from au­toma­tion and is not meant to rep­re­sent the en­tire space of work re­gard­ing the im­pacts of au­toma­tion. The liter­a­ture on this topic is very broad and the top­ics I am ad­dress­ing here are largely those that economists are think­ing about with some poli­ti­cal sci­ence-re­lated con­cerns thrown in as well. For ex­am­ple, I omit welfare effects and the mean­ing of work.

In sec­tion 1, I provide a nar­ra­tive of how the in­dus­trial rev­olu­tion may have shaped the so­ciety we live in to­day. In sec­tion 2, I cover how economists think au­toma­tion is chang­ing so­ciety to­day. I then touch on how these points may be con­cern­ing for our fu­ture in sec­tion 3.

1. A quick sum­mary of im­por­tant les­sons from the in­dus­trial rev­olu­tion:

In or­der to set the stage and guide think­ing about how we might ex­pect near-fu­ture tech­nolog­i­cal change to im­pact the way our so­ciety works, let’s take a look at how the in­dus­trial rev­olu­tion may have led to the cre­ation of a demo­cratic so­ciety in the West (but mostly Bri­tain). Un­less oth­er­wise in­di­cated, this sec­tion is ba­si­cally a sum­mary of Chap­ter 11 in “The Tech­nol­ogy Trap”, the rele­vant sources can be found there.

A con­vinc­ing case can be made that the rea­son we ended up with demo­cratic so­cieties in the West was due to tech­nolog­i­cal change. While this is far from be­ing widely ac­cepted among his­to­ri­ans (I spent an hour or two look­ing for his­tory books which fol­low this line of rea­son­ing and came up dry), my im­pres­sion is that it is at least tac­itly ac­cepted by many economists and poli­ti­cal sci­en­tists. I would recom­mend Dafoe’s pa­per on Tech­nolog­i­cal Deter­minism to dive deeper into the gen­eral think­ing on top­ics like these and in­ter-dis­ci­plinary differ­ences.

Tak­ing a fairly sim­plis­tic (and not-in-the-book) per­spec­tive we can think of pre-in­dus­trial work­ers (al­most all farm­ers) as hav­ing very lit­tle abil­ity to or­ga­nize or strike. They lived in fairly sparse com­mu­ni­ties, had no abil­ity to with­hold la­bor (with­out starv­ing that is, a seem­ingly un­pop­u­lar strat­egy), and were limited in their abil­ity to or­ga­nize. When the in­dus­trial rev­olu­tion hap­pened, all of the sud­den eco­nomic elites needed work­ers in their fac­to­ries. At first this was un­skil­led la­bor (of­ten chil­dren as the work was so sim­ple and chil­dren so cheap to em­ploy). At first this met with lit­tle benefit for the work­ing classes as dur­ing a pe­riod called En­gel’s Pause (roughly the first half of the 19th cen­tury) there was lit­tle to no wage growth (Allen, 2009). It’s worth not­ing that it was dur­ing this pe­riod when Marx was com­ing up with his the­o­ries about the ex­ploita­tion of the work­ing class by cap­i­tal­ists, which seems to have been largely ac­cu­rate at the time.

The la­bor de­manded was so un­skil­led that if work­ers were to strike, the fac­tory own­ers could sim­ply fire them and find some­one else, this led to lit­tle bar­gain­ing power for the work­ing class. It prob­a­bly didn’t help that many of the la­bor­ers were chil­dren un­der 13 years old. While there were ri­ots and protests at this time (no­tably the Lud­dites and Cap­tain Swing Riots) about au­toma­tion and work­ing con­di­tions more gen­er­ally, they were at a scale that the English mil­i­tary could sup­press. This took quite a lot of effort on the gov­ern­ment’s part and was the rea­son that many coun­tries ac­tively fought against the adop­tion of la­bor-sav­ing tech­nol­ogy in the past (Frey, 2019. Ch. 3). How­ever, it was not enough pres­sure that fac­tory own­ers or gov­ern­ments were forced to in­crease wages or work­ing con­di­tions.

Even­tu­ally the situ­a­tion changed in the sec­ond half of the 19th cen­tury as the skill-level needed in­creased. Work­ers were ac­tu­ally hard to re­place and could ne­go­ti­ate for wages. Even­tu­ally they were able to se­cure the abil­ity to strike legally and to unionize and be­came more or­ga­nized. There is a good case to be made here that even­tu­ally this led to them be­ing able to push for demo­cratic change. Once some classes of peo­ple were en­fran­chised there was a pos­i­tive feed­back loop among poli­ti­ci­ans to en­fran­chise more in or­der to cap­ture the new vot­ers.

Through­out the 20th cen­tury this situ­a­tion held, with most tech­nol­ogy be­ing com­ple­men­tary to la­bor rather than sub­sti­tut­ing. Typewrit­ers are a good ex­am­ple of a tech­nol­ogy that, while dis­plac­ing some pro­fes­sional writ­ers, cre­ated many new jobs that had pre­vi­ously been too ex­pen­sive to be fea­si­ble. Writ­ing a lot of cor­re­spon­dence and keep­ing ex­ten­sive files be­came eco­nom­i­cally fea­si­ble and new in­dus­tries and jobs could arise around it. Fur­ther­more, unions were pow­er­ful and were con­stantly ne­go­ti­at­ing when new tech­nol­ogy was adopted to en­sure a smooth tran­si­tion.

I think the gen­eral take­away here is a fram­ing of Western elites as be­ing dragged into the cre­ation of demo­cratic sys­tems by work­ers. It paints a pic­ture of gov­ern­ments as driven by in­cen­tives. Ac­cord­ing to this nar­ra­tive, it just so hap­pens that for the past ~150-200 years these in­cen­tives have al­igned with democ­racy fairly well. As you read fur­ther I en­courage you to think about whether or not gov­ern­men­tal and elite in­cen­tives may be shift­ing away from pro­vid­ing work­ers with power in the form of votes.

2. What seems to be hap­pen­ing due to au­toma­tion now:

It is worth not­ing that there is a lot of dis­agree­ment about what ex­actly is hap­pen­ing now, to what ex­tent it is due to au­toma­tion as op­posed to trade, and whether we would ex­pect the num­ber of jobs to even­tu­ally diminish. Be­fore pro­ceed­ing I should note that the im­pacts of trade and au­toma­tion are very similar. Both au­toma­tion and trade im­pact la­bor­ers work­ing in rou­tine jobs (jobs in­volv­ing repet­i­tive and straight­for­ward tasks, think as­sem­bly line or data en­try). In the case of trade this hap­pens due to the fact that, similar to au­toma­tion, lower skil­led fac­tory work that can be speci­fied well is eas­ier to re­place with cheaper work­ers in poorer coun­tries. So when think­ing about trends and ex­pec­ta­tions for the fu­ture, we can look at stud­ies eval­u­at­ing the effect of trade and es­sen­tially treat them as the same. We can then guess that trade is un­likely to in­crease much more while au­toma­tion likely will (Frey & Os­borne, 2017). What fol­lows is a sum­mary of the ideas cir­cu­lat­ing about the likely effects from au­toma­tion (and trade) that are oc­cur­ring now.

[Edit] The likely pace of au­toma­tion is hotly con­tested and I have not yet taken the time to as­sess the differ­ent stud­ies. To give you an idea of the range of es­ti­mates, here is an ex­cerpt from the ILO’s liter­a­ture re­view on the topic that should give you an idea of the de­bate (Bal­liester & Elsheikhi, 2018. Sec­tion 2.1.2):

…on the ba­sis of de­tailed oc­cu­pa­tions data some ob­servers es­ti­mate that 47 per cent of to­tal U.S. em­ploy­ment is at high risk of be­ing digi­tal­ised within 20 years [Da­heim & Win­ter­mann, 2017]. Globally, au­toma­tion is es­ti­mated to af­fect 1.1 billion work­ers (49 per cent of jobs) and US$12.7 billion in wages [Chui et. al, 2017]. Fur­ther­more, the World Bank (2016) es­ti­mates as much as 66.6 per cent of jobs sus­cep­ti­ble to be made re­dun­dant in the de­vel­op­ing world due to tech­nol­ogy dis­rup­tion … In con­trast, other stud­ies pro­duce much lower figures, such as Arntz et al. (2016) who find that only around 9 per cent of jobs are au­tomat­able in OECD counties

[end edit]

Of pri­mary im­por­tance (and more cer­tain) is that many rou­tine jobs are be­ing lost in the US (Jaimovich & Siu, 2012) and Europe (Fer­nan­dez-Ma­cias & Hurley, 2017). While it is likely that much of this is due to trade, there has also been some effect from au­toma­tion (Cortes et. al, 2016a). A high pro­por­tion of the peo­ple who are los­ing their jobs ei­ther leave the la­bor force or move to lower-paid non-rou­tine jobs (Cortes et. al, 2016a). It is worth not­ing that while un­em­ploy­ment is very low at the mo­ment, the mea­sures widely re­ported by the me­dia only count those who have looked for work in the past 2 weeks. Once some­one gives up look­ing for work they are cat­e­go­rized as NLF (Not in the La­bor Force). This is re­flected in a de­clin­ing la­bor force par­ti­ci­pa­tion rate in the US, though ap­par­ently not in Europe. In the US, the peo­ple leav­ing the la­bor force are of­ten liv­ing on dis­abil­ity.

Many of the lost rou­tine jobs hap­pen to be mid­dle-in­come jobs. This has led to wage po­lariza­tion (Ace­moglu & Au­tor, 2011), mean­ing that the mid­dle is drop­ping out and we are see­ing a greater con­cen­tra­tion of work­ers at lower and up­per in­come lev­els. This is a likely con­trib­u­tor to ris­ing in­equal­ity. Many jobs at the lower end of the in­come spec­trum are ei­ther not worth au­tomat­ing or are non-rou­tine and there­fore difficult to au­to­mate. For ex­am­ple, jan­i­tors are poorly paid but they do many tasks that are ex­ceed­ingly com­plex for ma­chines to ac­com­plish (f.ex, non-rou­tine tasks like clean­ing a toi­let).

It is broadly agreed among economists that in­fla­tion and pur­chas­ing power-ad­justed wages have not grown (much) in the past 30-40 years (known as wage stag­na­tion), though this is not with­out its dis­sen­ters. Au­toma­tion likely con­tributes to wage stag­na­tion but is un­likely to have been the pri­mary driv­ing force (Ace­moglu & Restrepo, 2017). Some mod­els sug­gest that wage stag­na­tion has been largely the re­sult of de­creased worker bar­gain­ing power. There is some ev­i­dence to back this up (Ben­m­elech et. al, 2019; Azar et. al, 2017).

Au­toma­tion has fur­ther been linked to the move­ment in the US to­ward poli­ti­cal po­lariza­tion (Au­tor et. al, 2016a) and in par­tic­u­lar the rise of Trump (Au­tor et. al, 2016b; Frey et. al, 2018), though it is un­clear ex­actly why this is hap­pen­ing. Given that the stud­ies are largely cor­re­la­tional and do not have con­vinc­ing causal mod­els, I would en­courage a bit of skep­ti­cism. Here are a few (of many) the­o­ries for how au­toma­tion might be driv­ing these effects:

    • Frey as­serts that an im­por­tant fac­tor was Trump’s promises to rust belt vot­ers to bring back man­u­fac­tur­ing jobs, which they liked be­cause they feared au­toma­tion was tak­ing their jobs (Frey et. al, 2018).

    • Ro­drick pro­poses that when eco­nomic con­di­tions de­te­ri­o­rate peo­ple turn to salient nar­ra­tives such as, ‘the im­mi­grants are tak­ing your jobs’ which far-right poli­ti­ci­ans are happy to provide (Ro­drik, 2018).

    • Dixit and Wie­bull (2007) ar­gue that de­te­ri­o­rat­ing con­di­tions lead peo­ple to be­come more ex­treme in what­ever be­liefs they have, which would lead to more votes for poli­ti­ci­ans on both ex­tremes.

3. Rea­sons we might worry about the con­se­quences of these trends:

The tech­nolog­i­cal de­ter­minist nar­ra­tive laid out in part 1 ar­gues that tech­nolog­i­cal de­vel­op­ments dur­ing the in­dus­trial rev­olu­tion led to democ­racy be­ing es­tab­lished. If you buy into this ar­gu­ment, it seems rea­son­able to worry about what might hap­pen should the con­di­tions that led to democ­racy no longer hold. New tech­nol­ogy (of­ten AI-en­abled) is of­ten la­bor-re­plac­ing (Frey, 2019) rather than com­ple­ment­ing, unions have lost power, there is a solid case to be made that wage stag­na­tion is due to de­creas­ing worker bar­gain­ing power, and even the demo­cratic in­fluence of vot­ers may be wan­ing (though I have not looked into the ev­i­dence on this) (Schlos­man et. al, 2012). If the chang­ing tech­nolog­i­cal con­di­tions that the in­dus­trial rev­olu­tion brought with it re­ally did shape so­ciety to the ex­tent put for­ward above, what effect might more ca­pa­ble AI sys­tems have on so­ciety? How might they shape our in­sti­tu­tions and in what di­rec­tion will they push?

3.1 Po­ten­tial harms of in­equal­ity on democ­racy & stability

Most of the more tan­gible near-term (next 10-20 years) nega­tive effects of au­toma­tion on so­ciety seem to run through its effect on in­equal­ity (though I hold this view weakly). As­sum­ing this is the case, let’s ex­plore some likely con­se­quences of in­creas­ing in­equal­ity as well as some re­lated fac­tors. It should be noted that there is likely a sig­nifi­cant amount of liter­a­ture on this topic which I am largely un­fa­mil­iar with and the points be­low are mostly drawn from Bruce Sch­neier’s blog and Frey’s, ‘The Tech­nol­ogy Trap’.

The in­creas­ing cost of cam­paigns means that poli­ti­ci­ans rely in­creas­ingly on wealthier in­di­vi­d­u­als rather than the av­er­age voter, mak­ing them more be­holden to these peo­ple (Frey—Tech trap). Inequal­ity wors­ens this trend by in­creas­ing the dis­par­ity be­tween top donors and the av­er­age ones. There is an­other pos­si­ble driver of this trend, bet­ter ad­ver­tis­ing and/​or im­proved ma­nipu­la­tion meth­ods. The more these meth­ods im­prove, the greater role money will play in win­ning an elec­tion as op­posed to policy po­si­tions and ac­countabil­ity.

Although I do not know of ev­i­dence that au­toma­tion has caused a de­cline in unions, I deem it fairly likely. If the trend of de­clin­ing unioniza­tion con­tinues, and truly rep­re­sents a weak­ened abil­ity of the lower classes to mo­bi­lize power, we should see a de­cline in the abil­ity of the lower-class to in­fluence poli­ti­cal de­ci­sions. There is some be­lief among poli­ti­cal sci­en­tists that the lower-class already has a weak­ened in­fluence on policy (Gilens & Page, 2014). Though many ar­gue that the stud­ies show­ing this are not valid. Frey uses the ex­am­ple of min­i­mum wage. Whether or not it would ac­tu­ally be good to in­crease the min­i­mum wage (a hotly de­bated is­sue among economists), the policy is heav­ily fa­vored by the pub­lic (95% of democrats, 75% of low-in­come Repub­li­cans, and 45% of Repub­li­cans earn­ing over $150,000/​yr) (Frey, 2019). In this case it would seem that or­ga­ni­za­tions like the Na­tional Res­tau­rant As­so­ci­a­tion have sig­nifi­cant efforts in place to avoid rais­ing the min­i­mum wage while work­ers seem to have no such co­or­di­nated ac­tion. We would ex­pect this dis­par­ity to worsen as in­equal­ity in­creases.

This is in part be­cause the lower class is gen­er­ally not very ac­tive in poli­tics (Kraus et. al, 2015). It there­fore helps if the mid­dle class is in­ter­ested in pass­ing the leg­is­la­tion that would help the lower class (i.e. welfare leg­is­la­tion). As dis­par­i­ties in­crease we may see less iden­ti­fi­ca­tion with other groups in so­ciety (ap­par­ently mid­dle-class and lower-class used to mix much more than they do now (Frey, 2019). Fur­ther, Frey char­ac­ter­izes the new par­ti­san di­vide to be along the lines of in­tel­lec­tu­als (Democrats) and the wealthy (Repub­li­cans) rather than along rich and poor lines. Look­ing at how few of the Demo­cratic can­di­date plat­forms take the welfare of the lower classes as a top pri­or­ity, I think this is a fairly rea­son­able as­ser­tion. The im­pli­ca­tion is that there is not a party look­ing out for the in­ter­ests of the lower-class as the Democrats used to. We would ex­pect this to be­come more of a prob­lem as in­equal­ity in­creased.

A fur­ther cause for di­vide is the per­cep­tion by whites in the US that Afri­can-Amer­i­cans are get­ting ahead (which Afri­can-Amer­i­cans do seem to feel) while they get left be­hind (Frey, 2019). This rel­a­tive power/​wealth loss is un­set­tling to white Amer­i­cans which causes some un­rest and in­creases the like­li­hood of racial back­lash. It is pos­si­ble that in­creas­ing in­equal­ity con­tinues to have this racially dis­pro­por­tionate effect which could fur­ther drive a wedge into an already di­vided so­ciety.

More gen­er­ally, if Amer­i­cans look to the fu­ture and ex­pect lit­tle im­prove­ment (which they largely seem to already), they will be un­happy about the cur­rent state of af­fairs and want to change it. There are an abun­dance of the­o­ries as to how this might hap­pen and what they might want to change. It seems clear though that an un­happy lower class could be a driver for fur­ther in­sta­bil­ity. Trump is an ob­vi­ous out­growth of this and I would wa­ger that, should cur­rent trends con­tinue, we should ex­pect worse down the road.

Many peo­ple ap­pear to be in fa­vor of heavy reg­u­la­tion for au­toma­tion tech­nol­ogy (58% think there should be re­stric­tions on how many jobs can be re­placed by ma­chines), per­haps due to per­cep­tion that au­toma­tion was the cause of their eco­nomic situ­a­tion (Frey, 2018). It may be that, given a fairly pop­ulist leader, the US could im­pose heavy reg­u­la­tion on au­toma­tion. In fact, I think there is an ar­gu­ment to be made that we should ex­pect the pub­lic per­cep­tion of AI to be closely tied to per­cep­tions of au­toma­tion. If this is the case, could we see a pur­pose­ful slow­ing down of AI re­search in the US if peo­ple per­ceive au­toma­tion to be largely harm­ful?

There is a strong case to be made, and some ev­i­dence (Car­bonero et. al, 2018), for greater in­ter­na­tional in­equal­ity due to au­toma­tion as well. Most coun­tries that have grown rich in the past sev­eral decades did so be­cause of a com­par­a­tively cheap la­bor force. As we au­to­mate more pro­cesses how­ever and the high-skill en­g­ineer run­ning an au­to­mated fac­tory is the pri­mary source of la­bor needed for pro­duc­tion, we would ex­pect in­dus­tries to move back to the de­vel­oped world (closer to mar­ket). This would cut off one of the few ways we know for poor coun­tries to catch up eco­nom­i­cally. It is un­clear whether there will be other ways for these coun­tries to grow. The con­se­quence is a more un­equal world and per­haps a ‘de­vel­op­ing’ world that is no longer de­vel­op­ing, but en­tirely de­pen­dent on wealthier na­tions. The effects of this will no doubt be far reach­ing, though it is un­clear what effect this would have on the de­vel­oped world.

Fi­nally, it is worth men­tion­ing that in­for­ma­tion war­fare will con­tinue to ad­vance (though defenses will hope­fully ad­vance in step). Bruce Sch­neier’s ex­cel­lent piece, “Democ­racy’s Dilemma” sheds some use­ful light here. He points out three ma­jor fac­tors:

  • Differ­ent sec­tions of the pop­u­la­tion have fun­da­men­tally differ­ent ideas about re­al­ity. Repub­li­cans and Democrats can­not even agree on whether base con­di­tions which a policy seeks to ad­dress are true (think an­thro­pogenic cli­mate change). There­fore the ar­gu­ment, which would ideally be about which policy will work bet­ter, is more about whether there should be a policy at all. This leaves lit­tle room for the clas­sic el­e­ments of de­bate be­tween par­ties to play out and in­cen­tivize find­ing an op­ti­mal solu­tion.

  • Trust be­tween par­ties has de­clined. As each side per­ceives the other to be act­ing in bad faith they will in­creas­ingly see it as jus­tifi­able to break demo­cratic norms in or­der to lock-in power while they have it. This could lead to a self-sus­tain­ing cy­cle of de­te­ri­o­rat­ing demo­cratic in­sti­tu­tions.

  • The pub­lic per­cep­tion of elec­tions, hear­ings, tri­als, etc are ex­tremely im­por­tant in that peo­ple need to trust them in or­der to re­spect the gov­ern­ment’s le­gi­t­i­macy. The per­cep­tion of these can be tam­pered with fairly eas­ily by out­side ac­tors in or­der to harm the le­gi­t­i­macy of demo­cratic institutions

3.2 Think­ing about the long-term fu­ture:

Above I have listed many of the more grounded fears based off of some, of­ten weak, ev­i­dence. While, taken to­gether, these are con­cern­ing, they do not cap­ture the risks we might face if we were to see true mass un­em­ploy­ment (>20%) which could fol­low more along the lines of Harare’s ‘Use­less Class’ idea (Harari, 2017). It is also worth not­ing that the im­pacts of au­toma­tion cor­re­late highly with de­vel­op­ment of AI ca­pa­bil­ities. It seems likely then that if mass un­em­ploy­ment is to oc­cur, it will hap­pen be­fore we de­velop AGI (pos­si­bly shortly be­fore), there­fore in any world where we reach AGI, we will also have had to deal with the con­se­quences of mass un­em­ploy­ment.

While I be­lieve it is en­tirely pos­si­ble that hu­man­ity could benefit greatly from high lev­els of au­toma­tion, the ex­am­ple of the in­dus­trial rev­olu­tion gives me pause in think­ing we will be able to dis­tribute the benefits of what is com­ing. If it re­ally is the case that the only rea­son we ended up with welfare-pro­mot­ing poli­cies is due to the bar­gain­ing power of work­ers, why would we ex­pect fu­ture elites to dis­tribute the benefits of in­creased pro­duc­tivity in the case that they no longer rely on the la­bor of the lower-class? While Democ­racy pro­vides a safe­guard by pro­vid­ing work­ers with vot­ing power that should the­o­ret­i­cally be able to safe­guard their rights, the threats to democ­racy de­tailed above cou­pled with the pos­si­bil­ity of an au­thor­i­tar­ian ad­van­tage from tech­nolog­i­cal growth (some­thing I will write about more in the fu­ture) should be cause for con­cern. The ex­tent to which this should be a pri­or­ity cause area when think­ing about the long-term fu­ture how­ever is still far from clear.

If you’re in­ter­ested in this topic and/​or have re­search ideas, get in touch at alex.l.lintz(at)

4. Bibliog­ra­phy:

Allen, R. (2009). “En­gels’ pause: Tech­ni­cal change, cap­i­tal ac­cu­mu­la­tion, and in­equal­ity in the british in­dus­trial rev­olu­tion”. Ex­plo­ra­tions in Eco­nomic His­tory. Vol­ume 46, Is­sue 4, Pages 418-435

Ace­moglu, D. & Au­tor, D. (2011). “Skills, Tasks and Tech­nolo­gies: Im­pli­ca­tions for Em­ploy­ment and Earn­ings”. Hand­book of La­bor Eco­nomics, Vol­ume 4b

Ace­moglu, D. & Restrepo. P (2017). “Robots and Jobs: Ev­i­dence from US Labor
Mar­kets”. NBER Work­ing Paper No. 23285

Arntz, M. et al. 2016. The Risk of Au­toma­tion for Jobs in OECD Coun­tries, OECD So­cial, Em­ploy­ment and Mi­gra­tion Work­ing Papers (Paris, OECD Pub­lish­ing).

Au­tor, D., and Dorn, D. (2013), ‘The Growth of Low-skill Ser­vice Jobs and the Po­lariza­tion of the US La­bor Mar­ket’, Amer­i­can Eco­nomic Re­view, 103(5), 1553–97.

Au­tor, D., Dorn, D., Han­son, G., and Ma­jlesi, K. (2016a), ‘Im­port­ing Poli­ti­cal Po­lariza­tion? The Elec­toral Con­se­quences of Ris­ing Trade Ex­po­sure’, tech­ni­cal re­port, Na­tional Bureau of Eco­nomic Re­search.

Au­tor, D., Dorn, D., Han­son, G., and Ma­jlesi, K. (2016b), ‘A Note on the Effect of Ris­ing Trade Ex­po­sure on the 2016 Pres­i­den­tial Elec­tion’, tech­ni­cal re­port, mimeo, MIT

Azar, J., Mari­nescu, I. and Stein­baum M., “La­bor Mar­ket Con­cen­tra­tion,” NBER Work­ing Paper 24147, 2017.

Bal­liester, T. and Elsheikhi, A. (2018). The Fu­ture of Work: A Liter­a­ture Re­view. In­ter­na­tional La­bor Organization

Ben­m­elech, E., Bergman, N and Kim, H. (2019). “Strong Em­ploy­ers and Weak Em­ploy­ees: How Does Em­ployer Con­cen­tra­tion Affect Wages?” https://​​​​ab­stract=3146679 or http://​​​​10.2139/​​ssrn.3146679

Car­bonero, F., Ernst, E., and We­ber, E. (2018). Robots wor­ld­wide: The im­pact of au­toma­tion on em­ploy­ment and trade. In­ter­na­tional Labour Organization

Cortes, G. M., Jaimovich, N., and Siu, H. E. (2016a). Dis­ap­pear­ing rou­tine jobs: Who, how, and why? Tech­ni­cal re­port, Na­tional Bureau of Eco­nomic Re­search.

Dafoe, A. (2015). On Tech­nolog­i­cal Deter­minism: A Ty­pol­ogy, Scope Con­di­tions, and a Mechanism. Science, Tech­nol­ogy, & Hu­man Values, 40(6), 1047–1076

Da­heim, C.; Win­ter­mann, O. 2017. 2050: Die Zukunft Der Ar­beit, (Guter­sloh, Ber­tels­mann Foun­da­tion).

Dixit, A. K., & Weibull, J. W. (2007). Poli­ti­cal po­lariza­tion. Pro­ceed­ings of the Na­tional Academy of Sciences, 104(18), 7351-7356.

Fer­nan­dez-Ma­cias, E, Hurley, J (2017) Rou­tine-bi­ased tech­ni­cal change and job po­lariza­tion in Europe. So­cio-Eco­nomic Re­view 15(3): 563–585.

Frey, C. & Os­borne, M. (2017) “The fu­ture of em­ploy­ment: How sus­cep­ti­ble are jobs to com­put­er­i­sa­tion?”. Tech­nolog­i­cal Fore­cast­ing and So­cial Change, Vol­ume 114

Frey, C., Berger, T., Chen, C. (2018). “Poli­ti­cal ma­chin­ery: did robots swing the 2016 US pres­i­den­tial elec­tion?”, Oxford Re­view of Eco­nomic Policy

Frey, C. (2019). “The Tech­nol­ogy Trap: Cap­i­tal, La­bor, and Power in the Age of Au­toma­tion.” Prince­ton Univer­sity Press.

Gilens, M., & Page, B. (2014). Test­ing The­o­ries of Amer­i­can Poli­tics: Elites, In­ter­est Groups, and Aver­age Ci­ti­zens. Per­spec­tives on Poli­tics, 12(3)

Harari, Noah (2017). Homo Deus: A Brief His­tory of To­mor­row. Lon­don: Vintage

Jaimovich, N. and Siu, H. (2012). Job Po­lariza­tion and Jobless Re­cov­er­ies. Na­tional Bureau of Eco­nomic Re­search. Work­ing Paper

Kraus, M., An­der­son, C., and Cal­laghan, B. (2015), The Inequal­ity of Poli­tics: So­cial Class Rank and Poli­ti­cal Par­ti­ci­pa­tion

Ro­drik, D. (2018). Pop­ulism and the Eco­nomics of Global­iza­tion. Jour­nal of In­ter­na­tional Busi­ness Policy, 1-22.

Schlos­man, K., Verba, S., & Brady, H. (2012). The Un­heav­enly Cho­rus: Unequal Poli­ti­cal Voice and the Bro­ken Promise of Amer­i­can Democ­racy. Prince­ton Univer­sity Press.

World Bank. 2016. World Devel­op­ment Re­port 2016: Digi­tal Div­i­dends, (Wash­ing­ton, D.C.).