Robots vs. the Underclass

Pundits say automation is decimating the middle class. The truth is even crueler.

This illustration can only be used with the John B. Judis piece that originally ran in the 2/28/2015 issue of National Journal magazine.
National Journal
Add to Briefcase
John B. Judis
Feb. 27, 2015, 12:01 a.m.

Ever since Gen­er­al Elec­tric in­stalled the first in­dus­tri­al ro­bot in 1961, Amer­ic­ans have been wor­ry­ing that auto­ma­tion could des­troy the coun­try’s labor force. Dur­ing the Great Re­ces­sion and its af­ter­math, these voices grew even louder. “We’re not go­ing to have a job­less re­cov­ery,” busi­ness writer Jeff Jar­vis pre­dicted in 2011. “We’re go­ing to have a job­less fu­ture.” “Smart ma­chines won’t kill us all, but they’ll def­in­itely take our jobs and soon­er than you think,” Moth­er Jones warned in 2013.

But which jobs, ex­actly, are go­ing to dis­ap­pear? To hear many pun­dits tell it, the ad­vance of tech­no­logy is spe­cific­ally threat­en­ing the middle ranks of the work­force. Auto­ma­tion, warned The Eco­nom­ist last Oc­to­ber, will lead to “the fur­ther erosion of the middle class.” “Ro­bots won’t des­troy jobs, but they may des­troy the middle class,” a Vox story was titled. The As­so­ci­ated Press pro­duced a series of art­icles head­lined, “What’s des­troy­ing the middle class? Not taxes. Not China. Think tech­no­logy.”

Of the two polit­ic­al parties, Demo­crats have most loudly echoed the fears about auto­ma­tion and the middle class. Three ma­jor Demo­crat­ic-aligned policy groups — the Third Way, the Cen­ter for Amer­ic­an Pro­gress, and the Hamilton Pro­ject — sponsored re­search in re­cent years ar­guing that the middle class was be­ing harmed by auto­ma­tion.

But in the past year or two, some of the eco­nom­ists who study this is­sue have be­gun to veer in a dif­fer­ent dir­ec­tion: They have pro­duced re­search sug­gest­ing that auto­ma­tion is go­ing to be re­l­at­ively good to the middle class — cre­at­ing new op­por­tun­it­ies not only for pro­fes­sion­als and man­agers but also for what MIT eco­nom­ist Dav­id Autor de­scribes as “re­l­at­ively well-re­mu­ner­ated, middle-skill” work­ers. These work­ers have at­ten­ded some col­lege or have bach­el­or’s de­grees, and they make at or above the me­di­an wage. They are the heart of what Amer­ic­ans con­sider the middle class — and it now ap­pears they aren’t dis­ap­pear­ing at all.

This doesn’t mean auto­ma­tion is cost-free, however. Ac­cord­ing to this new line of think­ing, the main vic­tims of auto­ma­tion will not be middle-in­come work­ers but rather those with lower levels of in­come and edu­ca­tion — that is, Amer­ic­ans who make less than $35,000 a year and have only a high school de­gree.

On the sur­face, the ques­tion of which demo­graph­ic group is be­ing most harmed by auto­ma­tion may sound like just a highly tech­nic­al ar­gu­ment among eco­nom­ists. But, in fact, it sheds light on the ut­ter in­ad­equacy of U.S. polit­ics and policy. While Demo­crats worry that the middle class is be­ing threatened by ro­bots, and Re­pub­lic­ans woo the middle class with tax cuts, neither party is do­ing nearly enough to ad­dress the tech­no­logy-driv­en dis­aster fa­cing lower-in­come, less-edu­cated Amer­ic­ans.

BROADLY SPEAK­ING, auto­ma­tion ar­rived in the U.S. eco­nomy in three over­lap­ping waves. The first wave began in the 1960s with the in­tro­duc­tion of the main­frame com­puter, the tran­sist­or, and the sil­ic­on chip. Firms star­ted in­stalling com­puter-coded ro­bots and ma­chine tools that took over much of the work of the fact­ory floor — from the pro­duc­tion of a phys­ic­al good to its pack­aging for sale. The ob­ject­ive even­tu­ally be­came the cre­ation of a “lights-out” fact­ory, in which pro­duc­tion would take place 24 hours a day without hu­man su­per­vi­sion. While no Amer­ic­an firm has yet achieved this for its en­tire op­er­a­tion, many now boast un­lit, un­heated, foot­ball-field-size areas where coded ma­chines have taken over the tasks that hu­mans used to per­form.

At Amazon's warehouses, orange robots the size of ottomans transport goods to stations where they are packed and sent off. (David Paul Morris/Bloomberg via Getty Images) Bloomberg via Getty Images

The ef­fects of this first wave on the eco­nomy were pro­found. From 1979 to 2013, U.S. man­u­fac­tur­ing em­ploy­ment fell 37 per­cent, from 19.3 mil­lion to 12.1 mil­lion. Some of that was due to out­sourcing — one study es­tim­ates that Chinese-im­port com­pet­i­tion ac­coun­ted for a quarter of the loss of man­u­fac­tur­ing em­ploy­ment between 1990 and 2007 — but a great deal of the rest was due to the auto­ma­tion of ba­sic, re­pet­it­ive fact­ory labor, the kind you would find on an auto­mobile as­sembly line. In­deed, largely be­cause of auto­ma­tion, Amer­ic­an man­u­fac­tur­ers today pro­duce far more goods with far less labor than they did just a few dec­ades ago. (Pro­ductiv­ity in man­u­fac­tur­ing in­creased 75 per­cent­age points between 1987 and 2007.)

The second wave of the auto­ma­tion re­volu­tion came in the 1980s. This time, the key change was the in­tro­duc­tion of the per­son­al com­puter, which gradu­ally began to re­place lower-level of­fice staff: sec­ret­ar­ies, tax-pre­parers, type­set­ters, and file clerks.

You can see evid­ence of both waves at com­pan­ies like Con­vey­ers & Auto­ma­tion — a Towson, Mary­land, firm that uses ro­bots to cre­ate lights-out pack­aging fa­cil­it­ies for Coca-Cola, Pep­si-Cola, Corn­ing Glass, and oth­er big man­u­fac­tur­ers. When I vis­ited re­cently, what was most strik­ing was not only the way the large yel­low ro­bots picked up and put down cans and bottles, but also the fact that at the firm’s headquar­ters, I couldn’t spot a single cler­ic­al or blue-col­lar work­er.

The third wave ar­rived in the early 1990s with the rise of the In­ter­net. It trans­formed the dis­tri­bu­tion as well as the pro­duc­tion of goods and ser­vices, cre­at­ing what eco­nom­ist W. Bri­an Ar­thur calls a “sec­ond­ary eco­nomy.” “Busi­ness pro­cesses that once took place among hu­man be­ings are now be­ing ex­ecuted elec­tron­ic­ally,” Ar­thur writes. “They are tak­ing place in an un­seen do­main that is strictly di­git­al.” The sec­ond­ary eco­nomy can make fin­an­cial de­cisions, do in­vent­ory, dia­gnose ill­nesses, wage war, reg­u­late elec­tri­city use, and sell everything from books to auto­mo­biles to ma­chine tools.

Amazon began as the ar­chetyp­ic­al third-wave firm. It sold books and later oth­er goods through the In­ter­net, threat­en­ing the ex­ist­ence of book­stores and shop­ping malls. But it has also be­come a pi­on­eer in ful­filling the prom­ise of auto­ma­tion’s first wave. The com­pany bought Kiva Sys­tems, which pro­duces or­ange ro­bots the size of ot­to­mans that roam un­lit, un­heated sec­tions of Amazon’s ware­houses all day and night, trans­port­ing shelves of goods to sta­tions where they are packed and sent off. Amazon won’t dis­cuss how much labor these devices have saved, but a man­ager of Amazon’s sub­si­di­ary Zap­pos has es­tim­ated that they cut labor in half. Moreover, the ro­bots are sup­posed to speed up pack­aging by 400 per­cent.

In all these re­spects, auto­ma­tion has cut a wide swath through the eco­nomy. Like elec­tri­city, it’s a gen­er­al-pur­pose tech­no­logy; its ef­fects are per­vas­ive and not con­fined to a single in­dustry. But while it has elim­in­ated the jobs of cler­ic­al staff at firms like Con­vey­ors & Auto­ma­tion, as well as the jobs of work­ers who used to scurry around Amazon ware­houses grabbing pack­ages off shelves, auto­ma­tion has not elim­in­ated mid-skill, me­di­an-wage, middle-class po­s­i­tions.

That, at least, is the con­clu­sion now drawn by MIT’s Autor and by two re­search­ers at Ox­ford, Carl Be­ne­dikt Frey and Mi­chael A. Os­borne. In a present­a­tion last sum­mer to the Fed­er­al Re­serve Bank of Kan­sas City, Autor, draw­ing on cat­egor­ies he had de­veloped earli­er with Har­vard’s Frank Levy and Richard Murnane, set out the ways that auto­ma­tion has elim­in­ated, ad­ded, mod­i­fied, or left un­touched vari­ous jobs. His cat­egor­ies (which I’ll use with minor vari­ations) provide a use­ful way to see how auto­ma­tion has changed rather than des­troyed the middle class.

First, there are the parts of the eco­nomy where auto­ma­tion has sup­ple­men­ted the hu­man role — but not made it ob­sol­ete. These oc­cu­pa­tions fall in­to roughly three cat­egor­ies: com­plex ab­stract tasks (sur­geons, dent­ists, law­yers, en­gin­eers, sci­ent­ists, ed­it­ors, ar­chi­tects, stock brokers, loan of­ficers, ther­ap­ists, school teach­ers, sales rep­res­ent­at­ives, and a bevy of dif­fer­ent kinds of tech­ni­cians, es­pe­cially in health care); jobs in­volving non­routine per­son­al in­ter­ac­tions (spe­cial­ized store clerks and tech­nic­al-sup­port per­son­nel, home health aides, per­son­al train­ers, po­lice, para­med­ics, and fire­fight­ers); plus those skilled crafts that can­not eas­ily be re­duced to routine in­struc­tions and now of­ten re­quire com­puter train­ing (com­puter, util­ity, or tele­com­mu­nic­a­tions re­pair per­son­nel, truck drivers, pi­lots, elec­tri­cians, mech­an­ics, and ma­chin­ists).

Of course, some of these oc­cu­pa­tions may even­tu­ally fall vic­tim to auto­ma­tion. Google, for in­stance, has de­veloped a driver­less car that could even­tu­ally be used to trans­port goods. But for now, none of these jobs are likely to be elim­in­ated by tech­no­logy.

So what jobs are in danger? Autor defines them as “routine tasks “¦ that fol­low an ex­haust­ive set of rules and hence are read­ily amen­able to com­pu­ter­iz­a­tion.” These in­clude low-level cler­ic­al and sec­ret­ari­al work, rule-driv­en in­ter­per­son­al en­coun­ters at banks, stores, and any­where tick­ets are sold, and much blue-col­lar work in factor­ies and ware­houses.

In oth­er words, the oc­cu­pa­tions that are safe from auto­ma­tion are all over the map in terms of in­come and edu­ca­tion level — ran­ging from home health aides to tele­com re­pair per­son­nel to sur­geons. But the jobs most threatened by auto­ma­tion are dis­pro­por­tion­ately un­skilled and low-wage, re­quir­ing only a high school de­gree. In fact, nine of the 10 jobs that Autor and an­oth­er MIT eco­nom­ist, Dav­id Dorn, cite as the most sus­cept­ible to com­pu­ter­iz­a­tion are low-skill and low-wage.

Production of cars has become automated in the past 50 years. They can now function without humans, too: Google has developed a driverless car. (Popperfoto/Getty Images) Popperfoto/Getty Images

As re­cently as 2013, Autor had main­tained that auto­ma­tion was po­lar­iz­ing the job mar­ket — cre­at­ing a situ­ation in which there was little middle zone between low-skilled ser­vice work and pro­fes­sion­al or ma­na­geri­al work. But last sum­mer, in his present­a­tion to the Fed­er­al Re­serve Bank, he took a dif­fer­ent tack. He now ar­gues that “em­ploy­ment po­lar­iz­a­tion will not con­tin­ue in­def­in­itely. While many middle-skill tasks are sus­cept­ible to auto­ma­tion, many middle-skill jobs” in­clude tasks that are not. Many of these jobs, he ex­plains, “will com­bine routine tech­nic­al tasks with the set of non­routine tasks in which work­ers hold com­par­at­ive ad­vant­age — in­ter­per­son­al in­ter­ac­tion, flex­ib­il­ity, ad­apt­ab­il­ity and prob­lem-solv­ing.”

Frey and Os­borne reached a sim­il­ar con­clu­sion. In a 2013 pa­per in which they ranked oc­cu­pa­tions on the basis of wheth­er they were com­pu­ter­iz­able, they con­cluded that the jobs “least sus­cept­ible” to com­pu­ter­iz­a­tion are high-skill pro­fes­sion­al and ma­na­geri­al jobs, but they also pre­dicted that a host of middle-in­come oc­cu­pa­tions would sur­vive. “Rather than re­du­cing the de­mand for middle-in­come oc­cu­pa­tions,” they wrote, “our mod­el pre­dicts that com­pu­ter­iz­a­tion will mainly sub­sti­tute for low-skill and low-wage jobs in the near fu­ture.”

Autor still main­tains that, in the past, auto­ma­tion was des­troy­ing the middle class. But eco­nom­ist Steph­en Rose ar­gues that the middle class was nev­er be­ing des­troyed, only gradu­ally trans­formed. The dif­fer­ence lies partly in how they count the loss of many blue-col­lar pro­duc­tion jobs. Autor defines these jobs as middle-in­come and middle-skilled — and sees their loss as evid­ence of the de­struc­tion of the middle class. Rose, on the oth­er hand, sees these jobs as middle-in­come (be­cause of fleet­ing cor­por­ate ac­qui­es­cence to uni­on wages) but lower-skilled. That changes the whole cal­cu­la­tion of what has happened over the past few dec­ades.

There is little dis­agree­ment among eco­nom­ists about how to help the vic­tims of auto­ma­tion.

Rose has cre­ated an ex­tens­ive data­base of oc­cu­pa­tions based not on Labor De­part­ment clas­si­fic­a­tions, but on a job’s spe­cif­ic func­tion, the skill re­quired to per­form it, and its place in the chain of out­put. Ac­cord­ing to his fig­ures, from 1967 to 2007 — roughly the peri­od dur­ing which the auto­ma­tion re­volu­tion took hold — the share of ma­na­geri­al and pro­fes­sion­al jobs in the labor force grew from 22 to 35 per­cent; the share of mid-skill jobs de­clined only slightly from 39 to 36 per­cent; and the share of low-skill jobs fell from 39 to 29 per­cent.

As Rose de­scribes it, the trend has been gradu­ally mov­ing, with some stops and starts — for in­stance, dur­ing the last re­ces­sion — to­ward an up­graded rather than a po­lar­ized work­force. It’s hard for me to say con­clus­ively who is right, but Rose’s pic­ture fits more with what I’ve seen as I’ve traveled around the United States over the years cov­er­ing cam­paigns. The very wealth­i­est have be­come even wealth­i­er; but be­low them, many Amer­ic­ans have gradu­ally been mov­ing up­ward. The middle class has not dis­ap­peared but edged ahead in edu­ca­tion, in­come, and re­spons­ib­il­ity on the job. Amer­ic­ans’ stand­ard of liv­ing has stead­ily ris­en — and that shows up in how they spend their money: Ac­cord­ing to Rose, Amer­ic­ans spent 46 per­cent of their in­come in 1947 on food, drink, and cloth­ing. By 2007, they were only spend­ing 18 per­cent. Much more of their money is now de­voted to what John Maynard Keynes de­scribed as “re­l­at­ive” rather than “ab­so­lute” needs.

But amidst this im­prove­ment in Amer­ic­ans’ stand­ard of liv­ing, there are also pock­ets of high un­em­ploy­ment and poverty throughout the coun­try — in cit­ies like De­troit or East St. Louis, in much of West Vir­gin­ia or east­ern Ken­tucky, and in rur­al parts of the Deep South. The people in these places have not be­nefited from the prom­ise of auto­ma­tion, and if the eco­nom­ists are right about the rap­id dis­ap­pear­ance of low-skill jobs, things are only go­ing to get much worse. The ques­tion, then, is: What should be done to help these people? And it’s a ques­tion that neither Demo­crats nor Re­pub­lic­ans — both of whom would rather talk about the middle class than the poor — seem es­pe­cially eager to face.

Transportation Secretary Anthony Foxx (right) and Google Chairman Eric Schmidt (left) get out of a Google self-driving car at the Google headquarters on Feb. 2 in Mountain View, Calif. (Justin Sullivan/Getty Images) Getty Images

IN TRUTH, THERE is little dis­agree­ment among eco­nom­ists about how to help the vic­tims of auto­ma­tion. “Our find­ings “¦ im­ply that as tech­no­logy races ahead, low-skill work­ers will real­loc­ate to tasks that are non-sus­cept­ible to com­pu­ter­iz­a­tion, that is, tasks re­quir­ing cre­at­ive and so­cial in­tel­li­gence,” Frey and Os­borne write. “For work­ers to win the race, however, they will have to ac­quire cre­at­ive and so­cial skills.” And the first place they must do that is in school.

As the auto­ma­tion re­volu­tion has de­veloped, it has stead­ily raised work­ers’ edu­ca­tion­al re­quire­ments — and work­ers have re­spon­ded by go­ing to school more than they had. Ac­cord­ing to the census, in 1967, only about half of Amer­ic­ans over 25 had gradu­ated from high school; by 2009, that num­ber was 87 per­cent. In 1967, less than 10 per­cent over 25 had gradu­ated from col­lege; by 2009, it was 30 per­cent.

But that still leaves a lot of work­ers un­pre­pared for today’s job mar­ket. Cur­rently, 34 per­cent of the labor force over the age of 25 has not made it bey­ond high school. Many of these work­ers are fall­ing through the cracks: The labor par­ti­cip­a­tion rate of high school gradu­ates is 17 per­cent­age points lower than the par­ti­cip­a­tion rate of four-year-col­lege gradu­ates.

The prob­lem is, in part, the wide dis­par­ity in the qual­ity of pub­lic schools. And the chal­lenge is starkest in the states where high school and col­lege gradu­ation rates are the low­est — states such as Mis­sis­sippi, Alabama, Louisi­ana, Ken­tucky, Arkan­sas, West Vir­gin­ia, South Car­o­lina, Ten­ness­ee, and Nevada. In these loc­ales, many chil­dren don’t have a chance to move up the lad­der of the new work­force that auto­ma­tion is cre­at­ing.

What’s needed is ob­vi­ous: an enorm­ous, sus­tained, and prob­ably ex­pens­ive ef­fort to in­crease the qual­ity of the coun­try’s worst schools. That would in­clude money to at­tract out­stand­ing teach­ers and to make it pos­sible for lower-in­come chil­dren to go to col­lege. (It would also in­volve re­vital­iz­ing the com­munit­ies around the schools.) Yet so far, Re­pub­lic­ans and Demo­crats have mustered only half-meas­ures — and even those half-meas­ures have proved astound­ingly dif­fi­cult to im­ple­ment.

In George W. Bush’s first year in of­fice, for in­stance, Demo­crats and Re­pub­lic­ans passed the No Child Left Be­hind Act, which im­posed min­im­um stand­ards on all states. Each state would ad­min­is­ter stand­ard­ized tests, and stu­dents would have to show pro­gress for a school to re­ceive fed­er­al fund­ing. But the con­tent of the tests was left up to the states, and some of the states that most needed to im­prove their school sys­tems balked at do­ing so.

Then, in 2009, with the sup­port of the Obama ad­min­is­tra­tion, 44 states and Wash­ing­ton, D.C., ad­op­ted com­mon stand­ards for math and read­ing; known as the Com­mon Core, the bench­marks were de­signed to pre­pare stu­dents for col­lege and the new eco­nomy. “These parsi­mo­ni­ous stand­ards give great­er weight to today’s found­a­tion­al skills than the weight giv­en by most state stand­ards,” Levy and Murnane wrote re­cently.

But the pro­pos­al has be­come a ma­jor source of con­tro­versy on the right, gen­er­at­ing enorm­ous op­pos­i­tion from Re­pub­lic­an gov­ernors and from tea-party groups. Nine­teen states — all with GOP gov­ernors at the time — have either re­jec­ted out­right or down­graded the Com­mon Core stand­ards. These in­clude many of the states — such as Mis­sis­sippi, Louisi­ana, and South Car­o­lina — for which the pro­pos­al was de­signed. While former Flor­ida Gov. Jeb Bush re­mains sup­port­ive, two oth­er Re­pub­lic­an pres­id­en­tial as­pir­ants who had pre­vi­ously backed Com­mon Core — Louisi­ana Gov. Bobby Jin­dal and New Jer­sey Gov. Chris Christie — now either op­pose or ex­press “grave con­cerns” about the pro­gram.

Pres­id­ent Obama and many Demo­crat­ic poli­cy­makers con­tin­ue to back the Com­mon Core, but some Demo­crat­ic politi­cians, moved in part by teach­ers’ war­i­ness of stand­ard­ized tests, have be­gun to dis­tance them­selves. Last fall, New York Gov. An­drew Cuomo claimed he had “noth­ing to do with Com­mon Core” and prom­ised to “dis­reg­ard Com­mon Core scores for at least five years.” In Wash­ing­ton state, the Demo­crat­ic Party passed a res­ol­u­tion re­ject­ing Com­mon Core. Hil­lary Clin­ton has yet to take a clear po­s­i­tion on the is­sue.

The Com­mon Core stand­ards cer­tainly don’t rep­res­ent a fi­nal an­swer to the ques­tion of what to do about the 34 per­cent of Amer­ic­ans who are go­ing to suf­fer most from auto­ma­tion. These stand­ards are merely a small start­ing point, but the fact that they are be­ing re­jec­ted where they are most needed — and that some Demo­crats are join­ing the many Re­pub­lic­ans who have backed away from them — says much about our polit­ic­al sys­tem’s in­ab­il­ity to ad­dress the so­cial chal­lenge of auto­ma­tion.

One ele­ment of the prob­lem may be Amer­ic­ans’ tra­di­tion­al aver­sion to gov­ern­ment, but part of it may also be struc­tur­al. In a par­lia­ment­ary sys­tem with pro­por­tion­al rep­res­ent­a­tion, the 34 per­cent might have a party that would sum­mon them to the polls and rep­res­ent their dis­tinct in­terests. By con­trast, in Amer­ica’s two-party, win­ner-take-all sys­tem, these voters are in­ev­it­ably ig­nored in fa­vor of lar­ger, wealth­i­er groups that are more likely to vote.

And so, Demo­crats be­moan the de­struc­tion of a middle class that ac­tu­ally re­mains quite vi­brant, and Re­pub­lic­ans cater to middle-class com­plaints about taxes and spend­ing, while try­ing to lure poor whites with rhet­or­ic about abor­tion and gun con­trol. In the mean­time, the re­volu­tion in auto­ma­tion will con­tin­ue — and the 34 per­cent of Amer­ic­ans who are not pre­pared to join it will be lost in the shuffle.


Welcome to National Journal!

You are currently accessing National Journal from IP access. Please login to access this feature. If you have any questions, please contact your Dedicated Advisor.