Bloomberg Businessweek - USA (2019-11-18)

◼ FINANCE Bloomberg Businessweek November 18, 2019

30

Apple pitchesits newcreditcardas a model

ofsimplicity and transparency, upending every-

thing consumers think about cards. But for

Goldman Sachs, the bank that actually approves

borrowers and lends the money, it’s brought some

of the same headaches that have long plagued

the industry.

In early November, both companies got caught

up in a social media firestorm. In Twitter posts,

a tech entrepreneur named David Heinemeier

Hansson and then Apple Inc. co-founder Steve

Wozniak complained that their wives had been

given lower credit limits on their Apple Cards than

they had, despite sharing finances. Wozniak said

he and his wife report the same income and have

a joint bank account, which should mean lenders

view them as equals. The New York Department of

Financial Services soon announced it would con-

duct an investigation.

Goldman Sachs Group Inc. has said it’s done

nothing wrong, and there’s no evidence the bank

intentionally discriminated against women. But

THEBOTTOMLINE GoldmanSachssaysit doesn’tconsider—

or even know—the gender of applicants for the Apple Card. But

computer models can unintentionally entrench discrimination.

● A viral furor over the Apple Card is a reminder

of how opaque credit decisions can be

THEBOTTOMLINE Companiesthatofferadvicetoinvestorson

how to vote their shares are facing more regulation, which might

make it less likely that they’ll disagree with management.

contendshareholderproposalsarea mechanism

forholding corporations accountable. While the

SEC’s new rules won’t stop big pension funds from

getting an issue on a ballot, the worry is that they

couldstopfreshissuesandideasaboutcorporate

governancefrombubblingup.“Partofthevaluein

keepingthedollar threshold relatively low is that it

is a statement that even relatively small investors

bring something to the table,” says Jill Fisch, a law

professor at the University of Pennsylvania who stud-

ies proxy issues. A seat at that table might be getting

more exclusive. �Andrea Vittorio, with Ben Bain

thatmaybepartoftheproblem,accordingto

critics. The complex models that guide Goldman’s

lending decisions could inadvertently produce

results that disadvantage certain groups.

The increasing use of algorithms in lending

decisions has sharpened a yearslong debate about

what policymakers refer to as “disparate impact.”

Consumeradvocatesarepushingregulatorsand

companiestorethinkwhethermodelsareentrench-

ingthediscriminationthatcomputer-driven lend-

ingis meanttostampout.“Researchis startingto

revealsometroublingexamplesinwhichthereal-

ityofalgorithmicdecision-making falls short of our

expectations or is simply wrong,” Nicol Turner Lee,

a fellow at the Center for Technology Innovation at

the Brookings Institution, recently told Congress.

One reason Goldman has found itself in the hot

seat is that the Apple Card, unlike many, doesn’t

let households share accounts. That could lead

to family members getting significantly different

credit limits. Goldman says it’s considering offer-

ing the option. The bank said in a tweet it would

also reevaluate credit decisions if the borrowing

limit comes back lower than a customer expected.

“We have not and never will make decisions based

on factors like gender,” the company tweeted. “In

fact, we do not know your gender or marital status

during the Apple Card application process.”

A 2016 study by credit reporting agency

Experian found that women had higher credit

scores, less debt, and a lower rate of late mort-

gage payments than men. Still, the Federal Trade

Commission has warned that women may continue

to face difficulties getting credit.

Using complex algorithms that take into account

hundreds of variables should lead to fairer out-

comes than relying on error-prone loan officers

who may harbor biases, proponents say. “It’s hard

for humans to manually identify these characteris-

tics that would make someone more creditworthy,”

says Paul Gu, co-founder of Upstart Network Inc.,

a tech firm that uses artificial intelligence to help

banks make loans.

But with the rise of machine learning, it’s hard

to keep track of whether decisions are really fair.

“Algorithms are not only nonpublic, they are actu-

ally treated as proprietary trade secrets by many

companies,” Rohit Chopra, an FTC commissioner,

saidinOctober.“Victimsofdiscriminatoryalgo-

rithmsseldomif everknowthey havebeen

victimized.” �Shahien Nasiripour, Jenny Surane,

and Sridhar Natarajan

Goldman ,

Apple Cards,

And Algos

▲ The Apple Card