1
00:00:00,000 --> 00:00:01,984
[SQUEAKING]

2
00:00:01,984 --> 00:00:04,464
[RUSTLING]

3
00:00:04,464 --> 00:00:10,920
[CLICKING]

4
00:00:10,920 --> 00:00:13,590
FRANK SCHILBACH: Welcome
to Lecture 18 of 14.13.

5
00:00:13,590 --> 00:00:17,160
This lecture's about gender
discrimination and identity.

6
00:00:17,160 --> 00:00:17,760
OK.

7
00:00:17,760 --> 00:00:19,050
So what's the agenda today?

8
00:00:19,050 --> 00:00:21,300
Generally, we're going to
talk about gender, identity,

9
00:00:21,300 --> 00:00:23,275
and particularly,
discrimination.

10
00:00:23,275 --> 00:00:25,650
First, I'm going to give you
a broad and a quick overview

11
00:00:25,650 --> 00:00:29,580
of why this is an
important area to study.

12
00:00:29,580 --> 00:00:33,000
Then I'm going to give you
an equally brief overview

13
00:00:33,000 --> 00:00:37,470
of the gender gap in wages,
the fact that women earn less

14
00:00:37,470 --> 00:00:43,170
money than men for equal work,
how that has evolved over time,

15
00:00:43,170 --> 00:00:45,722
and how much still is
left of this gender gap.

16
00:00:45,722 --> 00:00:47,430
Then we're going to
talk for a little bit

17
00:00:47,430 --> 00:00:50,070
about potential
technological solutions.

18
00:00:50,070 --> 00:00:54,690
Can we use technology to improve
or to reduce discrimination

19
00:00:54,690 --> 00:00:55,872
potentially?

20
00:00:55,872 --> 00:00:57,705
And the answer would
be yes, to some extent.

21
00:00:57,705 --> 00:01:00,355
But that technology will
not solve everything.

22
00:01:00,355 --> 00:01:02,730
Then we're going to talk about
two particularly important

23
00:01:02,730 --> 00:01:03,580
issues.

24
00:01:03,580 --> 00:01:05,880
One is beliefs and
people's updating

25
00:01:05,880 --> 00:01:10,957
about beliefs when it comes to
men and women and their skill.

26
00:01:10,957 --> 00:01:13,290
And then we're going to talk
about gender identity norms

27
00:01:13,290 --> 00:01:17,370
in particular, which is
the idea that people's

28
00:01:17,370 --> 00:01:21,540
deeply held beliefs about
what should be done in society

29
00:01:21,540 --> 00:01:24,460
or not might affect
their behavior.

30
00:01:24,460 --> 00:01:26,880
And finally, we're going to
talk a little bit about demand

31
00:01:26,880 --> 00:01:29,305
and supply for different tasks.

32
00:01:29,305 --> 00:01:30,930
So in particular,
I'm going to show you

33
00:01:30,930 --> 00:01:34,620
that women are more
likely to say yes

34
00:01:34,620 --> 00:01:37,290
to what's called nonpromotable
tasks, tasks that

35
00:01:37,290 --> 00:01:38,730
are good for
society, but perhaps

36
00:01:38,730 --> 00:01:41,310
not good for their career,
or tasks that are good

37
00:01:41,310 --> 00:01:43,570
for, like, a company or
a university as a whole,

38
00:01:43,570 --> 00:01:46,140
but are not good, are not
promotable in the sense

39
00:01:46,140 --> 00:01:49,420
that they help women,
in fact, get promoted.

40
00:01:49,420 --> 00:01:52,080
Women are more likely to
say yes to those things.

41
00:01:52,080 --> 00:01:55,440
In addition, on top of
that, precisely because they

42
00:01:55,440 --> 00:01:57,230
are more likely to
say yes, they're

43
00:01:57,230 --> 00:01:59,910
also more likely to be asked
in the first place, which sort

44
00:01:59,910 --> 00:02:01,540
of amplifies that difference.

45
00:02:01,540 --> 00:02:03,540
I'm going to tell you
about this in more detail.

46
00:02:03,540 --> 00:02:04,998
The first question
you might ask is

47
00:02:04,998 --> 00:02:06,630
why study gender differences?

48
00:02:06,630 --> 00:02:09,350
I have four answers for you.

49
00:02:09,350 --> 00:02:11,260
The first one is simple
equality, fairness,

50
00:02:11,260 --> 00:02:12,820
and justice issues.

51
00:02:12,820 --> 00:02:17,060
People simply should be rewarded
equally for the same output.

52
00:02:17,060 --> 00:02:20,620
So if a man and a woman
works equally many hours

53
00:02:20,620 --> 00:02:23,380
and is equally productive,
they should be paid the same.

54
00:02:23,380 --> 00:02:27,580
And if that's not the
case, that's unequal.

55
00:02:27,580 --> 00:02:30,610
That's unfair and not--

56
00:02:30,610 --> 00:02:32,710
injust, or not justified.

57
00:02:32,710 --> 00:02:37,450
Now there might be additional
equality and fairness issues

58
00:02:37,450 --> 00:02:39,110
that men and women
might not even

59
00:02:39,110 --> 00:02:41,210
be allowed to work
as much, women

60
00:02:41,210 --> 00:02:44,110
might not be allowed
to work as much as men.

61
00:02:44,110 --> 00:02:47,930
That's an additional
sort of fairness concern.

62
00:02:47,930 --> 00:02:49,960
Second, there's an
efficiency argument,

63
00:02:49,960 --> 00:02:52,450
which is overall
productivity and welfare

64
00:02:52,450 --> 00:02:55,270
falls if women and other groups
are held back by discrimination

65
00:02:55,270 --> 00:02:56,860
and other distortions.

66
00:02:56,860 --> 00:03:00,340
And indeed, a substantial
share of recent US growth

67
00:03:00,340 --> 00:03:02,800
can be explained by
improved allocation

68
00:03:02,800 --> 00:03:05,510
of talent in the economy.

69
00:03:05,510 --> 00:03:08,020
So this is just a
simple idea that, well,

70
00:03:08,020 --> 00:03:11,680
if we exclude certain
groups, including women

71
00:03:11,680 --> 00:03:16,690
from certain jobs, the most
talented people in those groups

72
00:03:16,690 --> 00:03:18,613
will not be allowed to--

73
00:03:18,613 --> 00:03:21,310
or some of the most talented
people in those groups

74
00:03:21,310 --> 00:03:23,300
might not be allowed
to do those jobs.

75
00:03:23,300 --> 00:03:25,720
And that's bad for
society overall.

76
00:03:25,720 --> 00:03:27,250
Think about, for
example, doctors.

77
00:03:27,250 --> 00:03:29,830
If women cannot be
doctors, well then,

78
00:03:29,830 --> 00:03:34,060
we will essentially just exclude
a large fraction of the most

79
00:03:34,060 --> 00:03:37,850
talented people in terms of
being doctors, the smartest,

80
00:03:37,850 --> 00:03:42,910
brightest and hardest
working women.

81
00:03:42,910 --> 00:03:46,510
And that will be worse
not just for those women,

82
00:03:46,510 --> 00:03:50,050
but it will also be bad for
medical care, innovation,

83
00:03:50,050 --> 00:03:52,540
and so on in the
medical profession

84
00:03:52,540 --> 00:03:54,940
because you just essentially
exclude some of the best

85
00:03:54,940 --> 00:03:58,572
people from that type of work.

86
00:03:58,572 --> 00:04:00,280
And if you want to
learn more about this,

87
00:04:00,280 --> 00:04:01,720
the paper by Hsieh et al.

88
00:04:01,720 --> 00:04:06,110
is an excellent
discussion of this issue.

89
00:04:06,110 --> 00:04:10,010
Third, we might learn
about the formation

90
00:04:10,010 --> 00:04:13,370
of preferences in personality
by studying gender.

91
00:04:13,370 --> 00:04:17,209
And here, the idea is that
there are, in many cases,

92
00:04:17,209 --> 00:04:19,775
differences and preferences--

93
00:04:19,775 --> 00:04:22,070
[INAUDIBLE] preferences,
social preferences--

94
00:04:22,070 --> 00:04:25,280
and taste for competition
or competitiveness

95
00:04:25,280 --> 00:04:27,920
but also attitudes
towards negotiation,

96
00:04:27,920 --> 00:04:30,860
or other things such as
identity aspirations, or over-

97
00:04:30,860 --> 00:04:33,490
or underconfidence
more generally.

98
00:04:33,490 --> 00:04:37,840
Well, so if there are these
differences at later ages,

99
00:04:37,840 --> 00:04:40,330
we can learn about the
formation of those preferences

100
00:04:40,330 --> 00:04:45,310
in personality by looking,
for instance, at people

101
00:04:45,310 --> 00:04:49,720
at different ages and trying
to understand at which ages

102
00:04:49,720 --> 00:04:51,730
those differences emerge.

103
00:04:51,730 --> 00:04:53,830
For instance, you might
look at 5-year-olds

104
00:04:53,830 --> 00:04:55,720
or 10-year-olds or
15-year-olds and look

105
00:04:55,720 --> 00:04:57,520
at their risks and
social preferences,

106
00:04:57,520 --> 00:05:01,420
and then try to understand
when those gender gaps emerge

107
00:05:01,420 --> 00:05:02,770
over time.

108
00:05:02,770 --> 00:05:06,310
And that will allow us, then, to
understand questions of nature

109
00:05:06,310 --> 00:05:08,440
versus nurture,
whether there are

110
00:05:08,440 --> 00:05:11,080
inherent differences in
preferences and personality

111
00:05:11,080 --> 00:05:14,000
overall, just men and women
are just different inherently

112
00:05:14,000 --> 00:05:15,370
for some reason.

113
00:05:15,370 --> 00:05:21,340
Or it could be that it's
really about socialization

114
00:05:21,340 --> 00:05:23,440
and education and the like.

115
00:05:23,440 --> 00:05:25,930
Girls and boys are just
treated very differently

116
00:05:25,930 --> 00:05:28,060
by their parents, by
their social environments,

117
00:05:28,060 --> 00:05:29,890
their teachers,
and so on, and that

118
00:05:29,890 --> 00:05:32,290
might create such
differences and preferences

119
00:05:32,290 --> 00:05:34,000
and other behaviors.

120
00:05:34,000 --> 00:05:36,130
So that's another
reason to study those.

121
00:05:36,130 --> 00:05:39,910
You're going to mostly focus
on number one, just simply

122
00:05:39,910 --> 00:05:42,970
the idea, the objective that
people should be rewarded

123
00:05:42,970 --> 00:05:48,710
equally for the same help.

124
00:05:48,710 --> 00:05:50,030
Now what is the gender gap?

125
00:05:50,030 --> 00:05:52,370
What do I mean when I
talk about the gender gap?

126
00:05:52,370 --> 00:05:56,390
So Claudia Goldin, who is a
eminent researcher at Harvard,

127
00:05:56,390 --> 00:05:59,270
who has done a lot of
seminal work on gender

128
00:05:59,270 --> 00:06:02,150
and the economics of gender,
has written a very nice overview

129
00:06:02,150 --> 00:06:06,350
paper in 2014 and giving an
overview of what do we know

130
00:06:06,350 --> 00:06:10,160
and how far have we come in
trying to reduce and close

131
00:06:10,160 --> 00:06:12,440
the gender gap.

132
00:06:12,440 --> 00:06:15,500
The graph that I'm showing
here summarizes this discussion

133
00:06:15,500 --> 00:06:17,443
fairly well.

134
00:06:17,443 --> 00:06:18,610
It's a bit of a messy graph.

135
00:06:18,610 --> 00:06:19,985
So let me show
you what it shows,

136
00:06:19,985 --> 00:06:21,420
or tell you what it shows.

137
00:06:21,420 --> 00:06:25,140
On the x-axis, you see
people's age from 20 to 70.

138
00:06:25,140 --> 00:06:28,860
On the y-axis, you see the log
of female to male earnings,

139
00:06:28,860 --> 00:06:31,450
or average female
to male earnings.

140
00:06:31,450 --> 00:06:33,880
Which is essentially a
measure of the gender gap,

141
00:06:33,880 --> 00:06:36,720
and so the difference
in log is essentially

142
00:06:36,720 --> 00:06:40,660
the percentage difference in
female versus male earnings.

143
00:06:40,660 --> 00:06:44,310
So minus 0.3, and we think of
that as, like, 30% difference

144
00:06:44,310 --> 00:06:48,720
in female versus male earnings.

145
00:06:48,720 --> 00:06:49,560
What do we see here?

146
00:06:49,560 --> 00:06:52,560
Well, A, we see a
pretty large gender gap

147
00:06:52,560 --> 00:06:55,290
anywhere in this graph.

148
00:06:55,290 --> 00:06:57,510
I should have also said
different lines are now

149
00:06:57,510 --> 00:06:59,280
different cohorts.

150
00:06:59,280 --> 00:07:04,260
So these are starting in 1923,
so people born in 1923, '28,

151
00:07:04,260 --> 00:07:08,640
'33, '38, '43, and so on,
at five year intervals,

152
00:07:08,640 --> 00:07:12,630
from the oldest to like
more recent cohorts.

153
00:07:12,630 --> 00:07:13,510
Now what do we see?

154
00:07:13,510 --> 00:07:16,980
We see a significant gender
gap everywhere in this graph,

155
00:07:16,980 --> 00:07:18,970
or in all lines of the graph.

156
00:07:18,970 --> 00:07:23,460
We see that the gender gap,
even for the most recent cohort

157
00:07:23,460 --> 00:07:26,040
seems to be about 20% to 30%.

158
00:07:26,040 --> 00:07:28,830
We see that the
gender gap has fallen.

159
00:07:28,830 --> 00:07:30,540
So when you look
at like the most--

160
00:07:30,540 --> 00:07:32,160
the oldest cohorts,
the gender gap

161
00:07:32,160 --> 00:07:35,550
was even larger than
the more recent cohorts.

162
00:07:35,550 --> 00:07:39,540
You also see that there's
sort of this U-shape, where

163
00:07:39,540 --> 00:07:44,430
we seem to find that the gender
gap starts emerging already

164
00:07:44,430 --> 00:07:47,610
at age 27, 28, 30.

165
00:07:47,610 --> 00:07:50,550
There's already a gender gap
of something like 10% to 15%,

166
00:07:50,550 --> 00:07:52,500
even for the most
recent cohorts.

167
00:07:52,500 --> 00:07:56,560
But then that gender gap
gets magnified towards, like,

168
00:07:56,560 --> 00:08:02,880
[INAUDIBLE] 20% 30%, 40%,
until about age 40 to 45,

169
00:08:02,880 --> 00:08:04,110
when it's at a plateau.

170
00:08:04,110 --> 00:08:08,610
So maybe it falls a bit overall.

171
00:08:08,610 --> 00:08:10,630
So what do we learn here?

172
00:08:10,630 --> 00:08:13,280
Well, there are substantial
female labor market

173
00:08:13,280 --> 00:08:14,690
gains in the last half century.

174
00:08:14,690 --> 00:08:17,670
But large gender gaps remain.

175
00:08:17,670 --> 00:08:22,410
In 2010, for instance, the
ratio of mean annual earnings

176
00:08:22,410 --> 00:08:24,840
between male and
female workers, these

177
00:08:24,840 --> 00:08:29,070
are people who worked full-time,
full year, 25 to 69 years,

178
00:08:29,070 --> 00:08:33,270
was 0.72, and that of
the median was 0.77.

179
00:08:33,270 --> 00:08:36,720
Notice that the
figure here controls

180
00:08:36,720 --> 00:08:38,742
for work time and education.

181
00:08:38,742 --> 00:08:40,659
So while of course, it
might also be the case,

182
00:08:40,659 --> 00:08:44,100
it is indeed the case that
women work fewer hours in a day

183
00:08:44,100 --> 00:08:48,180
than men do, even conditional
on work time education,

184
00:08:48,180 --> 00:08:49,980
there's a significant
gap, gender gap

185
00:08:49,980 --> 00:08:53,580
remaining of something
like 20 to 30%

186
00:08:53,580 --> 00:08:58,810
that women are paid
less for equal work.

187
00:08:58,810 --> 00:09:02,440
Now first, what are
some of the reasons why

188
00:09:02,440 --> 00:09:05,230
there have been substantial
female labor market gains

189
00:09:05,230 --> 00:09:07,150
over the last half century?

190
00:09:07,150 --> 00:09:10,540
Well, there's quite
a few technological

191
00:09:10,540 --> 00:09:12,820
and other explanatory factors.

192
00:09:12,820 --> 00:09:16,540
One big factor was a reduction
in the gender gap in education.

193
00:09:16,540 --> 00:09:23,350
Women are just more able and
can afford to receive education.

194
00:09:23,350 --> 00:09:26,150
That's true for both primary,
secondary education, but also,

195
00:09:26,150 --> 00:09:29,920
in particular, for university
or higher education,

196
00:09:29,920 --> 00:09:31,210
more generally.

197
00:09:31,210 --> 00:09:33,190
Second, there's been
technological innovations,

198
00:09:33,190 --> 00:09:35,260
such as the pill,
dishwasher, and so on,

199
00:09:35,260 --> 00:09:40,990
which allowed women to do more
work outside of the household.

200
00:09:40,990 --> 00:09:43,030
There have also been
labor demand shifts.

201
00:09:43,030 --> 00:09:45,160
In particular,
the US economy has

202
00:09:45,160 --> 00:09:49,270
been moving from
manufacturing and agriculture

203
00:09:49,270 --> 00:09:53,790
towards services.

204
00:09:53,790 --> 00:09:56,280
Women tend to have
comparative advantages

205
00:09:56,280 --> 00:10:01,110
in services compared to like,
in particular, hard manual.

206
00:10:01,110 --> 00:10:05,520
And so these labor demand shifts
then have disproportionately

207
00:10:05,520 --> 00:10:08,340
benefited women.

208
00:10:08,340 --> 00:10:10,650
Finally, there's been
also lower discrimination,

209
00:10:10,650 --> 00:10:14,280
in particular, stronger
regulatory controls

210
00:10:14,280 --> 00:10:17,160
and increased market
competitiveness.

211
00:10:17,160 --> 00:10:20,220
For the latter, it's if
there's more competition,

212
00:10:20,220 --> 00:10:24,840
firms cannot afford to
discriminate as much.

213
00:10:24,840 --> 00:10:26,340
Because it's actually
a costly thing

214
00:10:26,340 --> 00:10:29,790
to do, to hire incompetent men
compared to competent women.

215
00:10:32,810 --> 00:10:36,290
Now we're going to discuss
a paper by Goldin and Rouse,

216
00:10:36,290 --> 00:10:39,890
which is an example of a
simple technological solution,

217
00:10:39,890 --> 00:10:42,860
in particular, this question
whether blind auditions

218
00:10:42,860 --> 00:10:47,060
in orchestras increased gender
ratios in those orchestras.

219
00:10:47,060 --> 00:10:49,700
So is the fact that people--

220
00:10:49,700 --> 00:10:51,560
the blind auditions
were introduced,

221
00:10:51,560 --> 00:10:54,290
did that contribute to
the increase in the gender

222
00:10:54,290 --> 00:10:56,240
ratio in orchestras?

223
00:10:56,240 --> 00:10:58,580
What do I mean by the
increase in gender ratios

224
00:10:58,580 --> 00:11:00,380
in the orchestras?

225
00:11:00,380 --> 00:11:03,890
So even until the
1960s, '70s, and '80s,

226
00:11:03,890 --> 00:11:07,670
there were shockingly low
female ratios in top orchestras

227
00:11:07,670 --> 00:11:10,380
and orchestras more generally.

228
00:11:10,380 --> 00:11:17,070
Now these ratios were below
10% in many orchestras.

229
00:11:17,070 --> 00:11:19,320
And only recently,
in the '80s and '90s,

230
00:11:19,320 --> 00:11:21,570
in particular, there
were stark increases

231
00:11:21,570 --> 00:11:27,720
in the number of women who
play in those orchestras.

232
00:11:27,720 --> 00:11:33,470
This increase was the case
in the top five orchestras,

233
00:11:33,470 --> 00:11:36,230
but [INAUDIBLE] also has been,
have been similar improvements

234
00:11:36,230 --> 00:11:37,850
in gender ratios in
other orchestras,

235
00:11:37,850 --> 00:11:43,610
which you can read in the report
that I linked in the slides.

236
00:11:43,610 --> 00:11:47,480
Now one really
important reason here

237
00:11:47,480 --> 00:11:51,500
seems to have been blatant
sexism, even by very renowned,

238
00:11:51,500 --> 00:11:54,560
or in particular, by renowned
conductors, which you can--

239
00:11:54,560 --> 00:11:56,690
I'm not going to
repeat here, but you

240
00:11:56,690 --> 00:12:00,120
can read it in section one of
the paper by Goldin and Rouse.

241
00:12:00,120 --> 00:12:02,610
And so one question
is like, well,

242
00:12:02,610 --> 00:12:05,270
is there's this sexism
and discrimination,

243
00:12:05,270 --> 00:12:13,370
has that contributed to the
low gender ratio in orchestras?

244
00:12:13,370 --> 00:12:16,520
Now while there have been
all these improvements

245
00:12:16,520 --> 00:12:21,200
in the gender ratio [INAUDIBLE]
musicians in general?

246
00:12:21,200 --> 00:12:23,840
There still are vast
differences in conductors

247
00:12:23,840 --> 00:12:26,150
and the number of conductors
and music directors.

248
00:12:26,150 --> 00:12:29,360
They're essentially
predominantly male still.

249
00:12:29,360 --> 00:12:32,090
Moreover, while there have
been a lot of improvements

250
00:12:32,090 --> 00:12:36,650
in terms of gender, most
musicians are still white.

251
00:12:36,650 --> 00:12:40,010
So in particular,
African-Americans and Latinos

252
00:12:40,010 --> 00:12:45,690
are very much underrepresented
in orchestras.

253
00:12:45,690 --> 00:12:51,720
Now let me tell you a short
aside about women in economics.

254
00:12:51,720 --> 00:12:54,090
This is not just the
case that in orchestras

255
00:12:54,090 --> 00:12:59,160
the fraction of women is low,
that's also true in economics.

256
00:12:59,160 --> 00:13:01,140
And here's an overview of that.

257
00:13:01,140 --> 00:13:05,730
There's a [INAUDIBLE] paper by
Lundberg and Stearns in 2019,

258
00:13:05,730 --> 00:13:08,280
which essentially says that
some progress has been made,

259
00:13:08,280 --> 00:13:09,690
but lots more needs to be done.

260
00:13:09,690 --> 00:13:11,430
And more recently,
in particular,

261
00:13:11,430 --> 00:13:14,520
some of the progress
seems to have stalled.

262
00:13:14,520 --> 00:13:16,190
What do we mean by that?

263
00:13:16,190 --> 00:13:18,010
Well, here this graph
shows the fraction

264
00:13:18,010 --> 00:13:22,510
of women in different stages
of the profession ranging

265
00:13:22,510 --> 00:13:27,910
from senior mangers here at the
top to first-year PhD students

266
00:13:27,910 --> 00:13:32,140
to assistant professors,
new assistant professors,

267
00:13:32,140 --> 00:13:34,833
associate professors,
and full professors.

268
00:13:34,833 --> 00:13:36,250
And what you see
here, there seems

269
00:13:36,250 --> 00:13:38,390
to be some trend that's
upwards, in particular

270
00:13:38,390 --> 00:13:40,747
for full professors and
associate professors,

271
00:13:40,747 --> 00:13:41,830
which is, of course, good.

272
00:13:45,100 --> 00:13:48,700
there's less of a positive
trend to be seen in particular

273
00:13:48,700 --> 00:13:50,515
when it comes to
assistant professors.

274
00:13:50,515 --> 00:13:53,470
Here you essentially
see since 2000--

275
00:13:53,470 --> 00:13:57,820
about 2007 or 2006 has been
no increase, if anything,

276
00:13:57,820 --> 00:14:01,870
like a decrease in the fraction
of female assistant professors.

277
00:14:01,870 --> 00:14:08,160
But even the fraction of PhD
students is clearly below 50%.

278
00:14:08,160 --> 00:14:10,350
And so the profession
is starting

279
00:14:10,350 --> 00:14:12,630
to understand and
address the issues,

280
00:14:12,630 --> 00:14:15,880
in particular, to
try to reduce sexism,

281
00:14:15,880 --> 00:14:19,650
and other issues
holding women back.

282
00:14:19,650 --> 00:14:22,470
For example, some
particularly nice examples

283
00:14:22,470 --> 00:14:24,990
of these mentoring
programs, which

284
00:14:24,990 --> 00:14:30,990
are trying to support female and
minority assistant professors

285
00:14:30,990 --> 00:14:33,030
and in terms of mentoring
and trying to sort of

286
00:14:33,030 --> 00:14:38,220
foster their careers, which
has been shown to impact

287
00:14:38,220 --> 00:14:41,850
and significantly
improve the profession.

288
00:14:41,850 --> 00:14:45,960
For what it's worth in '14,
'13, the majority of students

289
00:14:45,960 --> 00:14:47,880
tends to be female.

290
00:14:47,880 --> 00:14:51,200
And moreover, the best students
tend to be female as well.

291
00:14:51,200 --> 00:14:52,950
So very much hope you
are still interested

292
00:14:52,950 --> 00:14:55,200
and all the female
students are very much

293
00:14:55,200 --> 00:14:57,960
still interested in economics
and would pursue a career

294
00:14:57,960 --> 00:15:02,750
in economics, perhaps academic.

295
00:15:02,750 --> 00:15:05,900
Now back to orchestra,
so many orchestras

296
00:15:05,900 --> 00:15:09,320
introduced blind auditions
in the 1970s and 1980s.

297
00:15:09,320 --> 00:15:12,590
And if you look back at the
graph that I showed you--

298
00:15:12,590 --> 00:15:16,730
this graph here-- you see in the
1970s, '80s, and '90s, that's

299
00:15:16,730 --> 00:15:20,420
also the time when there was
a sharp increase in the share

300
00:15:20,420 --> 00:15:22,970
of women in those orchestras.

301
00:15:22,970 --> 00:15:26,150
So you can see
here in the graphs,

302
00:15:26,150 --> 00:15:29,030
orchestras, if you want to
get a job in orchestras,

303
00:15:29,030 --> 00:15:31,010
you have to do auditions.

304
00:15:31,010 --> 00:15:32,930
These auditions are
very, very competitive

305
00:15:32,930 --> 00:15:34,730
and often have like
several stages.

306
00:15:34,730 --> 00:15:37,220
There are preliminary,
semifinals, and final symphony

307
00:15:37,220 --> 00:15:38,300
orchestras.

308
00:15:38,300 --> 00:15:42,180
What you see here is 11
anonymized orchestras,

309
00:15:42,180 --> 00:15:45,370
including the top
orchestras in the US,

310
00:15:45,370 --> 00:15:49,480
and the introduction of blind
auditions in those orchestras.

311
00:15:49,480 --> 00:15:51,730
What you see is
that most orchestras

312
00:15:51,730 --> 00:15:54,460
did introduce blind
auditions at least in some

313
00:15:54,460 --> 00:15:58,120
of their stages
of the auditions.

314
00:15:58,120 --> 00:16:01,970
However, many of them also only
introduce it in some stages,

315
00:16:01,970 --> 00:16:03,370
but not in others.

316
00:16:03,370 --> 00:16:07,030
Now you can think already
that that's not sufficient.

317
00:16:07,030 --> 00:16:09,760
In some cases, if you have
only like blind auditions

318
00:16:09,760 --> 00:16:12,815
in the preliminaries, but not
in the semifinals and finals,

319
00:16:12,815 --> 00:16:14,440
you might sort of
reduce discretization

320
00:16:14,440 --> 00:16:15,315
in the preliminaries.

321
00:16:15,315 --> 00:16:17,890
But then if women are
discriminated in the semifinals

322
00:16:17,890 --> 00:16:20,725
or finals, that's not going to
help them get hired, which is,

323
00:16:20,725 --> 00:16:25,770
at the end of the day,
we, of course, care about.

324
00:16:25,770 --> 00:16:29,040
Now Goldin and Rouse look
at technological solutions

325
00:16:29,040 --> 00:16:33,390
and look at the question of
whether such technological

326
00:16:33,390 --> 00:16:35,520
solutions, in fact, can help.

327
00:16:35,520 --> 00:16:39,960
Now the question to
ask now is looking

328
00:16:39,960 --> 00:16:44,010
at data from actual auditions,
often using individual fixed

329
00:16:44,010 --> 00:16:46,485
effects, do these blind
auditions, the introduction

330
00:16:46,485 --> 00:16:50,310
of blind auditions help
candidates reach the next round

331
00:16:50,310 --> 00:16:52,940
or, in particular, get hired.

332
00:16:52,940 --> 00:16:56,780
Now the data is fairly
rich and often includes

333
00:16:56,780 --> 00:17:00,320
or allows for the inclusion
of individual fixed effects.

334
00:17:00,320 --> 00:17:02,360
That is the case if you
look at the musicians,

335
00:17:02,360 --> 00:17:05,030
Goldin and Rouse are
going to ask the question,

336
00:17:05,030 --> 00:17:09,530
is a particular
musician more likely

337
00:17:09,530 --> 00:17:14,300
to make it to the next round
when auditions are blinded?

338
00:17:14,300 --> 00:17:17,119
And is that more likely
the case for female

339
00:17:17,119 --> 00:17:20,838
than for male musicians?

340
00:17:20,838 --> 00:17:22,130
Now what did they find overall?

341
00:17:22,130 --> 00:17:26,869
They find evidence that
blind audition procedures

342
00:17:26,869 --> 00:17:29,720
fostered impartiality
in hiring and increased

343
00:17:29,720 --> 00:17:33,110
the proportion of women
in symphony orchestras.

344
00:17:33,110 --> 00:17:35,330
Now there's some caveats
here that, you know,

345
00:17:35,330 --> 00:17:37,310
some of these estimates
are fairly imprecise,

346
00:17:37,310 --> 00:17:38,810
in particular, in
the later stages

347
00:17:38,810 --> 00:17:44,420
because a number of observations
was down quite a bit.

348
00:17:44,420 --> 00:17:46,580
Some of these results
are a little bit mixed.

349
00:17:46,580 --> 00:17:48,230
But overall, the
evidence suggests

350
00:17:48,230 --> 00:17:50,660
that these blind
auditions did indeed

351
00:17:50,660 --> 00:17:55,880
foster impartiality and help
women succeed in getting jobs,

352
00:17:55,880 --> 00:17:59,860
in particular in these top
orchestras in the West.

353
00:17:59,860 --> 00:18:02,720
Now why are these
results perhaps somewhat

354
00:18:02,720 --> 00:18:03,478
less clear cut?

355
00:18:03,478 --> 00:18:05,020
I already told you
like one reason is

356
00:18:05,020 --> 00:18:07,510
that in the later rounds,
the number of observations

357
00:18:07,510 --> 00:18:10,150
gets kind of small because no
longer have so many people make

358
00:18:10,150 --> 00:18:11,620
it to the later rounds.

359
00:18:11,620 --> 00:18:13,000
But there's another
reason, which

360
00:18:13,000 --> 00:18:16,690
is the auditions have several
rounds, not all of which

361
00:18:16,690 --> 00:18:18,100
were blinded.

362
00:18:18,100 --> 00:18:19,505
Now if there's one--

363
00:18:19,505 --> 00:18:21,130
if there's discrimination
in one round,

364
00:18:21,130 --> 00:18:24,940
but not in the next one, this
issue that in the next round,

365
00:18:24,940 --> 00:18:28,270
there will be like
selection, and in particular,

366
00:18:28,270 --> 00:18:30,730
the discriminated people who
make it to the next round,

367
00:18:30,730 --> 00:18:32,580
you expect them to
do better, right?

368
00:18:32,580 --> 00:18:34,420
If there's discrimination
against women

369
00:18:34,420 --> 00:18:38,080
in the first round,
but not in a second,

370
00:18:38,080 --> 00:18:40,180
you think that women
in the second round

371
00:18:40,180 --> 00:18:42,730
will do better because only
the best women will make it

372
00:18:42,730 --> 00:18:46,030
to the second round
because otherwise they

373
00:18:46,030 --> 00:18:50,260
will be screened out
in the first round

374
00:18:50,260 --> 00:18:52,782
due to discrimination.

375
00:18:52,782 --> 00:18:55,700
But if that's the case,
then making comparisons

376
00:18:55,700 --> 00:18:59,173
in later rounds is potentially
biased or problematic

377
00:18:59,173 --> 00:19:01,340
because you're comparing
essentially different types

378
00:19:01,340 --> 00:19:04,130
of people in different rounds.

379
00:19:04,130 --> 00:19:08,990
And that leads to the results
being less clear-cut overall.

380
00:19:08,990 --> 00:19:10,630
Let me give you another example.

381
00:19:10,630 --> 00:19:15,488
Suppose male and female workers
get graded on their task ,

382
00:19:15,488 --> 00:19:18,030
a coding task, online where you
essentially see their output,

383
00:19:18,030 --> 00:19:19,850
but you also see their gender.

384
00:19:19,850 --> 00:19:22,550
Now suppose women get
graded systematically

385
00:19:22,550 --> 00:19:24,520
worse due to sexism
and no other reasons.

386
00:19:24,520 --> 00:19:26,270
Suppose women and men
are exactly the same

387
00:19:26,270 --> 00:19:28,580
in the work they do,
but some men are sexist

388
00:19:28,580 --> 00:19:30,950
and they're going
to rate women worse.

389
00:19:30,950 --> 00:19:34,880
Now for any given score,
you should expect the women

390
00:19:34,880 --> 00:19:37,850
that you hire compared to any
other random man that you hire

391
00:19:37,850 --> 00:19:40,740
to do better than that man.

392
00:19:40,740 --> 00:19:43,250
And the reason being that
they have been systematically

393
00:19:43,250 --> 00:19:47,640
disadvantaged due to the sexism
in the rating to start with.

394
00:19:47,640 --> 00:19:49,110
Now you should think about this.

395
00:19:49,110 --> 00:19:51,240
Because, you know,
there's also some evidence

396
00:19:51,240 --> 00:19:55,270
of teaching evaluations
being biased against women.

397
00:19:55,270 --> 00:19:58,440
Now if you see a man and a woman
are at the exact same teaching

398
00:19:58,440 --> 00:20:01,260
rating, well, you would probably
want to take the woman's class

399
00:20:01,260 --> 00:20:02,820
because the woman has been--

400
00:20:02,820 --> 00:20:06,600
the female professor has been
discriminated to start with.

401
00:20:06,600 --> 00:20:09,480
Having that high score or
the similarly high score

402
00:20:09,480 --> 00:20:13,770
as the male professor is
more impressive and likely

403
00:20:13,770 --> 00:20:16,407
to indicate that the
class will be better.

404
00:20:16,407 --> 00:20:18,240
There's an excellent
paper by Bohren et. al.

405
00:20:18,240 --> 00:20:21,835
that considers the issue of--
the dynamics of discrimination

406
00:20:21,835 --> 00:20:24,210
in much more detail if you're
interested in learning more

407
00:20:24,210 --> 00:20:26,870
about that.

408
00:20:26,870 --> 00:20:30,700
Now despite the
technological solution

409
00:20:30,700 --> 00:20:35,890
that advances that image,
[INAUDIBLE] substantial gender

410
00:20:35,890 --> 00:20:38,170
gaps remain.

411
00:20:38,170 --> 00:20:41,080
So while, as I told
you, the gender gap

412
00:20:41,080 --> 00:20:44,230
has been reduced both in terms
of labor market or labor force

413
00:20:44,230 --> 00:20:48,100
participation, how many days and
hours, how many women overall,

414
00:20:48,100 --> 00:20:50,320
how many days and
hours those women work

415
00:20:50,320 --> 00:20:53,710
and in their wages and earnings,
how much they are being paid

416
00:20:53,710 --> 00:20:59,910
conditional on working,
there are still

417
00:20:59,910 --> 00:21:02,010
substantial gender gaps.

418
00:21:02,010 --> 00:21:05,280
In particular, women's
labor market participation

419
00:21:05,280 --> 00:21:08,850
has plateaued since
the early 1990s.

420
00:21:08,850 --> 00:21:11,160
And even now, among
entering cohorts,

421
00:21:11,160 --> 00:21:13,980
women still earn
significantly less than men.

422
00:21:13,980 --> 00:21:16,290
And that's true
even for conditional

423
00:21:16,290 --> 00:21:17,700
on work time and education.

424
00:21:17,700 --> 00:21:20,580
That's the graph that I
was showing you earlier.

425
00:21:20,580 --> 00:21:24,060
Now given that there are these
persistent remaining gender

426
00:21:24,060 --> 00:21:28,470
gaps even though a lot of
the technological and other

427
00:21:28,470 --> 00:21:31,080
barriers have been reduced--
now women are educated

428
00:21:31,080 --> 00:21:33,930
or as educated as men are and
there's less technological

429
00:21:33,930 --> 00:21:38,610
barriers that preclude
women from working,

430
00:21:38,610 --> 00:21:41,070
researchers have now started
to consider some of less

431
00:21:41,070 --> 00:21:44,170
traditional within
economic factors,

432
00:21:44,170 --> 00:21:47,550
including things such as risk
attitudes, negotiation skills,

433
00:21:47,550 --> 00:21:50,460
taste for computation,
but also--

434
00:21:50,460 --> 00:21:52,410
and this is what we're
going to focus on--

435
00:21:52,410 --> 00:21:56,268
beliefs and social
norms and identity.

436
00:21:56,268 --> 00:21:57,810
I'm going to first
talk about beliefs

437
00:21:57,810 --> 00:21:59,768
and then we're going to
talk about social norms

438
00:21:59,768 --> 00:22:02,700
and identity.

439
00:22:02,700 --> 00:22:05,430
Before we get to
beliefs about gender,

440
00:22:05,430 --> 00:22:07,170
people's-- women's
and men's skills,

441
00:22:07,170 --> 00:22:09,570
I'm going to tell you very
briefly about a seminal paper

442
00:22:09,570 --> 00:22:13,950
by Bertrand and Mullainathan
from 2004 that randomizes--

443
00:22:13,950 --> 00:22:17,230
randomized names in
job applications.

444
00:22:17,230 --> 00:22:19,650
So what they're trying to do--

445
00:22:19,650 --> 00:22:23,100
they essentially sent
out job applications

446
00:22:23,100 --> 00:22:27,180
that were otherwise identical,
male and female workers.

447
00:22:27,180 --> 00:22:29,430
These are fictitious
job applications.

448
00:22:29,430 --> 00:22:31,950
And they sent them to
employers and then looked

449
00:22:31,950 --> 00:22:35,430
at callback rates
by those employers.

450
00:22:35,430 --> 00:22:39,930
And the only thing that was like
varied in those applications

451
00:22:39,930 --> 00:22:42,660
was the name on the application.

452
00:22:42,660 --> 00:22:43,920
And what did they do?

453
00:22:43,920 --> 00:22:48,090
Well, they used white-sounding
names such as Emily and Greg,

454
00:22:48,090 --> 00:22:50,430
as well as African-American
sounding names

455
00:22:50,430 --> 00:22:53,010
such as Lakisha and Jamal.

456
00:22:53,010 --> 00:22:57,060
And then they looked at,
well, were these applications,

457
00:22:57,060 --> 00:22:58,950
when they had
white-sounding names that

458
00:22:58,950 --> 00:23:02,850
were otherwise identical, more
successful than applications

459
00:23:02,850 --> 00:23:10,860
by names with
African-American-sounding names

460
00:23:10,860 --> 00:23:14,580
and, again, that were
otherwise identical?

461
00:23:14,580 --> 00:23:18,510
And Bertrand and Mullainathan
find a striking result

462
00:23:18,510 --> 00:23:21,150
that callback rates for
white-sounding names

463
00:23:21,150 --> 00:23:24,240
were 50% higher than for
African-American-sounding

464
00:23:24,240 --> 00:23:25,090
names.

465
00:23:25,090 --> 00:23:27,240
That's a huge difference.

466
00:23:27,240 --> 00:23:29,070
And that's important
not just because it's

467
00:23:29,070 --> 00:23:31,740
a hassle because you have to
send out more applications,

468
00:23:31,740 --> 00:23:33,960
but if you get called
back more, you're

469
00:23:33,960 --> 00:23:37,630
more likely to be interviewed
or asked for an interview,

470
00:23:37,630 --> 00:23:39,640
you're more likely to
actually get the job,

471
00:23:39,640 --> 00:23:42,380
you're more likely to--

472
00:23:42,380 --> 00:23:45,980
less likely to be unemployed,
and you might also

473
00:23:45,980 --> 00:23:50,300
be less likely to stop searching
whatsoever because you might

474
00:23:50,300 --> 00:23:53,220
be disappointed or discouraged.

475
00:23:53,220 --> 00:23:57,170
So that's a huge difference
and, essentially, discrimination

476
00:23:57,170 --> 00:24:03,140
against African-Americans
was blatantly obvious

477
00:24:03,140 --> 00:24:04,580
from this research.

478
00:24:04,580 --> 00:24:08,360
Numerous other studies show
gender and racial biases

479
00:24:08,360 --> 00:24:10,400
in various ways.

480
00:24:10,400 --> 00:24:14,750
These are ranging from very
similar studies, including

481
00:24:14,750 --> 00:24:19,350
these audit studies where people
are sent resumes and the like,

482
00:24:19,350 --> 00:24:22,040
but also ranging to studies
where like people on eBay

483
00:24:22,040 --> 00:24:24,410
tried to sell things,
where they try

484
00:24:24,410 --> 00:24:27,710
to sell like an iPad
or iPhone, and they

485
00:24:27,710 --> 00:24:31,040
hold that iPhone with
hands that's either

486
00:24:31,040 --> 00:24:33,260
like white or non-white.

487
00:24:33,260 --> 00:24:35,660
And then they find
essentially that white hands

488
00:24:35,660 --> 00:24:41,510
are more likely to sell their
iPhones than non-white hands.

489
00:24:41,510 --> 00:24:44,450
Now this is a very important
and depressing result.

490
00:24:44,450 --> 00:24:47,420
One thing to flag here that it's
hard to distinguish statistical

491
00:24:47,420 --> 00:24:49,460
from case-based discrimination.

492
00:24:49,460 --> 00:24:51,500
You're going to talk more
about this recitation.

493
00:24:51,500 --> 00:24:52,980
What do I mean by that?

494
00:24:52,980 --> 00:24:55,850
Well, it might be
that there is--

495
00:24:55,850 --> 00:24:59,090
people have a preference
for white applicants.

496
00:24:59,090 --> 00:25:01,550
That is to say, controlling
for performance,

497
00:25:01,550 --> 00:25:04,370
it might just be that
people are racist

498
00:25:04,370 --> 00:25:06,290
and they'd rather
have white applicants

499
00:25:06,290 --> 00:25:10,140
than African-American
applicants.

500
00:25:10,140 --> 00:25:15,440
This is what economists would
call case-based discrimination,

501
00:25:15,440 --> 00:25:17,780
which essentially is
conditional on the performance.

502
00:25:17,780 --> 00:25:21,140
People just like certain types
of people better than others.

503
00:25:21,140 --> 00:25:22,700
And, therefore, they
call them or are

504
00:25:22,700 --> 00:25:24,670
more likely to call them back.

505
00:25:24,670 --> 00:25:28,660
A different explanation is
belief-based explanation.

506
00:25:28,660 --> 00:25:30,940
It might be that
employers think that

507
00:25:30,940 --> 00:25:34,030
African-American applicants
will perform worse,

508
00:25:34,030 --> 00:25:37,000
even controlling for all
other aspects of the resume.

509
00:25:37,000 --> 00:25:40,570
So it could be essentially
you see, people see this name

510
00:25:40,570 --> 00:25:43,180
and they think that even
if the resume is good,

511
00:25:43,180 --> 00:25:46,540
that person will not perform
work as well as somebody

512
00:25:46,540 --> 00:25:48,240
with a white-sounding name.

513
00:25:48,240 --> 00:25:51,720
Notice that these beliefs
could be correct or not,

514
00:25:51,720 --> 00:25:54,755
but in any case, we will-- and
that's-- we'll talk about this

515
00:25:54,755 --> 00:25:56,945
in recitation as well.

516
00:25:56,945 --> 00:25:58,820
Economists again, would
call this statistical

517
00:25:58,820 --> 00:25:59,690
discrimination.

518
00:25:59,690 --> 00:26:01,940
This could be, again, correct
statistical estimation

519
00:26:01,940 --> 00:26:08,150
or statistical discrimination
that's based on biased beliefs.

520
00:26:08,150 --> 00:26:11,240
But either way,
these things jointly

521
00:26:11,240 --> 00:26:16,940
led to clear discrimination of
African-American applications,

522
00:26:16,940 --> 00:26:20,330
which, in turn, leads
to striking differences

523
00:26:20,330 --> 00:26:26,740
in unemployment or job
finding rates across races.

524
00:26:26,740 --> 00:26:30,100
We're going to now focus on
beliefs and, in particular,

525
00:26:30,100 --> 00:26:31,690
about beliefs about
skills and how

526
00:26:31,690 --> 00:26:34,390
people update on those
beliefs depending

527
00:26:34,390 --> 00:26:36,780
on the gender of the
person they're updating on.

528
00:26:39,565 --> 00:26:41,190
The people we are
going to consider now

529
00:26:41,190 --> 00:26:44,250
is the paper by
Sarsons, 2019, which

530
00:26:44,250 --> 00:26:48,230
considers how people interpret
signals in the labor market.

531
00:26:48,230 --> 00:26:50,120
The question that
Sarsons asks here

532
00:26:50,120 --> 00:26:52,700
is whether somebody's
gender influences

533
00:26:52,700 --> 00:26:56,420
the way we interpret
information about this person

534
00:26:56,420 --> 00:26:58,940
and about his or her peers.

535
00:26:58,940 --> 00:27:00,240
What do I mean by that?

536
00:27:00,240 --> 00:27:04,400
Well, in many situations
when people do certain tasks

537
00:27:04,400 --> 00:27:08,790
and perform at work, good
things and bad things happen.

538
00:27:08,790 --> 00:27:11,870
Now you can infer from
these things that occur,

539
00:27:11,870 --> 00:27:14,060
from these events, you
can infer something

540
00:27:14,060 --> 00:27:17,250
about their competence
and their performance.

541
00:27:17,250 --> 00:27:19,940
While many of these situations
are ambiguous in the sense

542
00:27:19,940 --> 00:27:22,972
it could be that the person was
just really lucky or it could

543
00:27:22,972 --> 00:27:24,680
be that the person is
really, really good

544
00:27:24,680 --> 00:27:26,140
at what they're doing.

545
00:27:26,140 --> 00:27:28,890
Now what Sarsons is asking
when we see those events

546
00:27:28,890 --> 00:27:30,670
and when people
see those events,

547
00:27:30,670 --> 00:27:34,480
are people interpreting
those events differently

548
00:27:34,480 --> 00:27:37,440
for men versus women?

549
00:27:37,440 --> 00:27:41,510
So, for example, if a man-- if
something bad happens at work,

550
00:27:41,510 --> 00:27:45,260
a boss might react
to that person

551
00:27:45,260 --> 00:27:49,340
differently or
update their beliefs

552
00:27:49,340 --> 00:27:51,920
about the person's skills
differently depending

553
00:27:51,920 --> 00:27:54,080
whether it's man or woman.

554
00:27:54,080 --> 00:27:59,150
In addition, the boss might
also update differently

555
00:27:59,150 --> 00:28:00,900
about the peers.

556
00:28:00,900 --> 00:28:03,110
So if a woman makes
a mistake, the boss

557
00:28:03,110 --> 00:28:05,930
might not only
update his beliefs

558
00:28:05,930 --> 00:28:10,490
about this woman, but also
about all other women that

559
00:28:10,490 --> 00:28:11,980
work for him.

560
00:28:11,980 --> 00:28:15,666
Similarly, for a man, of course,
the boss might do the same.

561
00:28:15,666 --> 00:28:19,000
And what Heather Sarsons
now asks is the question,

562
00:28:19,000 --> 00:28:22,030
is this updating or the
interpretation of information

563
00:28:22,030 --> 00:28:26,440
differently for male
versus female workers?

564
00:28:26,440 --> 00:28:27,755
And why do we care?

565
00:28:27,755 --> 00:28:30,130
Well, we care a lot about this
because hiring, promotion,

566
00:28:30,130 --> 00:28:32,140
and wage decisions
hinge on information

567
00:28:32,140 --> 00:28:33,850
about workers' ability.

568
00:28:33,850 --> 00:28:36,160
And in particular, if
there's systematic difference

569
00:28:36,160 --> 00:28:39,130
in how information about men
and women is interpreted,

570
00:28:39,130 --> 00:28:42,070
that could lead to differences
in hiring, promotion, and wage

571
00:28:42,070 --> 00:28:45,700
decisions and, in turn, might
contribute to the gender gap.

572
00:28:49,110 --> 00:28:51,810
What does Sarsons do to
answer this question?

573
00:28:51,810 --> 00:28:54,630
She looks at how do
physicians change referrals

574
00:28:54,630 --> 00:29:00,510
to surgeons and their
peers after a patient,

575
00:29:00,510 --> 00:29:03,330
after certain patient outcomes?

576
00:29:03,330 --> 00:29:06,620
Now what I mean by
that, physicians often

577
00:29:06,620 --> 00:29:09,800
refer patients needing
surgery to a local surgeon.

578
00:29:09,800 --> 00:29:12,260
So somebody breaks
their arm or their leg,

579
00:29:12,260 --> 00:29:14,320
well, many physicians
can't do this on their own,

580
00:29:14,320 --> 00:29:18,940
so they will refer this
patient to a local surgeon.

581
00:29:18,940 --> 00:29:22,540
Now the referral choice
reflects the physician's beliefs

582
00:29:22,540 --> 00:29:25,320
about the surgeon's
ability, right?

583
00:29:25,320 --> 00:29:29,650
You want to do the
best for your patient.

584
00:29:29,650 --> 00:29:32,120
And you want to
send that patient

585
00:29:32,120 --> 00:29:34,030
to the best possible surgeon.

586
00:29:34,030 --> 00:29:35,860
So if you have a
suspicion, if you

587
00:29:35,860 --> 00:29:38,170
think a certain surgeon
is not competent,

588
00:29:38,170 --> 00:29:40,900
you better not send your
patient to that surgeon,

589
00:29:40,900 --> 00:29:43,750
in part, because you might
care a lot about the patient,

590
00:29:43,750 --> 00:29:46,780
in part, because,
you know, if you

591
00:29:46,780 --> 00:29:51,600
make bad referrals, that might
also reflect on you eventually.

592
00:29:51,600 --> 00:29:54,960
Now what does Sarsons
do specifically?

593
00:29:54,960 --> 00:29:58,380
To document whether
the reaction depends

594
00:29:58,380 --> 00:29:59,800
on the surgeon's gender.

595
00:29:59,800 --> 00:30:01,230
What does Sarsons do?

596
00:30:01,230 --> 00:30:04,260
She matches on surgeon and
patient characteristics

597
00:30:04,260 --> 00:30:05,280
on procedures.

598
00:30:05,280 --> 00:30:10,200
That is to say, she takes male
and female surgeons and patient

599
00:30:10,200 --> 00:30:14,490
characteristics and procedures--
which is like, you know,

600
00:30:14,490 --> 00:30:17,580
broken arms versus more
complicated procedures--

601
00:30:17,580 --> 00:30:20,670
so she matches all of those
characteristics and such

602
00:30:20,670 --> 00:30:23,610
that in observations set,
she has surgeons only

603
00:30:23,610 --> 00:30:24,990
differ by gender.

604
00:30:24,990 --> 00:30:28,260
And then she considers what
we call an event study, which

605
00:30:28,260 --> 00:30:30,390
is comparing how
physicians reacts

606
00:30:30,390 --> 00:30:36,240
to male and female surgeons
when good and bad things happen.

607
00:30:36,240 --> 00:30:39,470
Let me now show you more
precisely what we mean.

608
00:30:39,470 --> 00:30:42,740
So we have here on
the x-axis, quarters.

609
00:30:42,740 --> 00:30:45,500
On the y-axis, we have the
referrals from the physicians

610
00:30:45,500 --> 00:30:47,240
to the performing surgeon.

611
00:30:47,240 --> 00:30:50,540
So these are essentially
dyads or pairs of physicians

612
00:30:50,540 --> 00:30:52,250
and performing surgeons.

613
00:30:52,250 --> 00:30:56,990
And you see here minus
1 is normalized to be 0.

614
00:30:56,990 --> 00:31:00,800
And then, essentially, you see
the evolution of the referrals

615
00:31:00,800 --> 00:31:04,350
from a specific physician
to a performing surgery.

616
00:31:04,350 --> 00:31:05,523
These are men and women.

617
00:31:05,523 --> 00:31:07,190
And these are essentially
cases in which

618
00:31:07,190 --> 00:31:10,010
there is no adverse event
or the adverse events

619
00:31:10,010 --> 00:31:12,117
are like a patient's
death, for example.

620
00:31:12,117 --> 00:31:13,700
What you see are
essentially these are

621
00:31:13,700 --> 00:31:14,992
men and women that are matched.

622
00:31:14,992 --> 00:31:18,090
They're supposed to
look exactly the same.

623
00:31:18,090 --> 00:31:23,320
And what we see here is that
referrals increase over time.

624
00:31:23,320 --> 00:31:26,310
They tend to sort of increase
fairly steeply over time

625
00:31:26,310 --> 00:31:29,490
and then sort of
plateau off over time.

626
00:31:29,490 --> 00:31:32,100
The exact shape and pattern
of this relationship

627
00:31:32,100 --> 00:31:33,488
doesn't matter so much.

628
00:31:33,488 --> 00:31:35,280
However, what matters
is that men and women

629
00:31:35,280 --> 00:31:39,383
look exactly the same, right?

630
00:31:39,383 --> 00:31:40,800
So the performing
surgeon is here.

631
00:31:40,800 --> 00:31:43,590
The gender here that
varies is the gender

632
00:31:43,590 --> 00:31:46,130
of the performing surgery.

633
00:31:46,130 --> 00:31:48,170
Now then she looks at--

634
00:31:48,170 --> 00:31:51,410
Sarsons looks at cases in which
there is an adverse event.

635
00:31:51,410 --> 00:31:56,160
Again, the adverse event
happens around here.

636
00:31:56,160 --> 00:31:59,630
So what you see here is then
when there is an adverse event,

637
00:31:59,630 --> 00:32:03,380
there's a clear
punishment in the sense

638
00:32:03,380 --> 00:32:05,270
of like the male
surgeon now gets

639
00:32:05,270 --> 00:32:11,690
fewer referrals from
the physician right

640
00:32:11,690 --> 00:32:13,700
after the adverse event happens.

641
00:32:13,700 --> 00:32:16,100
So this is manifested
by the blue bars

642
00:32:16,100 --> 00:32:19,880
being lower than the gray bars.

643
00:32:19,880 --> 00:32:21,830
Notice that the number
doesn't go down.

644
00:32:21,830 --> 00:32:25,110
It's still above-- everything
here is above the red line.

645
00:32:25,110 --> 00:32:29,960
So it's just the
physician is not

646
00:32:29,960 --> 00:32:32,975
increasing the referrals
over time anymore

647
00:32:32,975 --> 00:32:36,762
or that increase gets
dampened by the adverse event.

648
00:32:36,762 --> 00:32:37,970
This is for the male surgeon.

649
00:32:37,970 --> 00:32:40,520
Let's look at now what
happens to the female surgeon.

650
00:32:40,520 --> 00:32:43,910
What we see here is a clearer
difference that opens up

651
00:32:43,910 --> 00:32:46,310
after the adverse event.

652
00:32:46,310 --> 00:32:48,230
Notice that there's
no pre-trends here.

653
00:32:48,230 --> 00:32:49,940
Before, it looks--
men and women,

654
00:32:49,940 --> 00:32:52,430
or female and male
surgeons, including

655
00:32:52,430 --> 00:32:55,380
the red and the blue bars
look exactly the same.

656
00:32:55,380 --> 00:32:57,800
But the bar, the
difference only opens up

657
00:32:57,800 --> 00:32:59,320
after the adverse event.

658
00:32:59,320 --> 00:33:01,430
And there's a pretty
large gap, you can see,

659
00:33:01,430 --> 00:33:03,770
that shows up in the referrals.

660
00:33:03,770 --> 00:33:07,130
And this gap is way
larger if you compare

661
00:33:07,130 --> 00:33:09,410
the red versus the gray lines--

662
00:33:09,410 --> 00:33:11,150
it's way larger
compared to the gap

663
00:33:11,150 --> 00:33:14,450
with the divergence between
the blue and the gray lines.

664
00:33:14,450 --> 00:33:18,070
So what we see here is
that both men and women

665
00:33:18,070 --> 00:33:21,400
are punished in the sense
of receiving fewer referrals

666
00:33:21,400 --> 00:33:22,900
after an adverse event.

667
00:33:22,900 --> 00:33:25,690
That punishment is way
more severe for female

668
00:33:25,690 --> 00:33:28,620
than for male surgeons.

669
00:33:28,620 --> 00:33:32,183
Now let's do the same for
the absence of a good event

670
00:33:32,183 --> 00:33:33,850
aren't actually good
events that happen.

671
00:33:33,850 --> 00:33:34,683
What's a good event?

672
00:33:34,683 --> 00:33:36,992
Well, it's essentially a
surprisingly good thing

673
00:33:36,992 --> 00:33:38,700
that happened where
there's a complicated

674
00:33:38,700 --> 00:33:41,810
surgery and the person
does really well.

675
00:33:41,810 --> 00:33:45,750
What we see here is, again,
sort of the men and women,

676
00:33:45,750 --> 00:33:48,000
they look pretty much
exactly the same.

677
00:33:48,000 --> 00:33:52,710
You can see here maybe men are
doing slightly better overall.

678
00:33:52,710 --> 00:33:55,880
But overall, these gray lines
look very similar over time.

679
00:33:55,880 --> 00:33:57,630
Now, if there's a good
event that happens,

680
00:33:57,630 --> 00:34:00,630
you see, essentially, men
get very much rewarded.

681
00:34:00,630 --> 00:34:06,280
If the male surgeon
does well, he

682
00:34:06,280 --> 00:34:11,139
gets rewarded in the sense of
receiving a lot more referrals

683
00:34:11,139 --> 00:34:13,929
from the PCP, from the doctor.

684
00:34:13,929 --> 00:34:16,870
Now if you do the same
for female surgeon.

685
00:34:16,870 --> 00:34:20,420
Again, female surgeons
benefit from that.

686
00:34:20,420 --> 00:34:26,409
But the increase in referrals
is a lot lower than for men.

687
00:34:26,409 --> 00:34:28,600
So, again, both men
and women, there

688
00:34:28,600 --> 00:34:30,550
is a reaction to
that information,

689
00:34:30,550 --> 00:34:32,560
to a good or a bad event.

690
00:34:32,560 --> 00:34:35,710
But the reaction tends to
be much stronger, much more

691
00:34:35,710 --> 00:34:37,659
favorable for men.

692
00:34:37,659 --> 00:34:40,690
In terms of the good events,
men get rewarded more

693
00:34:40,690 --> 00:34:41,860
for the good events.

694
00:34:41,860 --> 00:34:46,239
And women get punished more
for bad events, [INAUDIBLE]

695
00:34:46,239 --> 00:34:48,330
bad events that happen.

696
00:34:48,330 --> 00:34:50,219
Let me summarize
the main results.

697
00:34:50,219 --> 00:34:53,550
After a bad outcome, after, for
example like a patient death,

698
00:34:53,550 --> 00:34:57,100
there is a 34 decrease in
referrals to female surgeon--

699
00:34:57,100 --> 00:34:57,600
surgeons.

700
00:34:57,600 --> 00:35:00,000
And there's a stagnation in
referrals to male surgeons.

701
00:35:00,000 --> 00:35:02,850
So, essentially, women tend
to be punished quite a lot.

702
00:35:02,850 --> 00:35:05,200
Men, a little bit in
the sense that referrals

703
00:35:05,200 --> 00:35:08,190
don't go up anymore, but
the punishment is much less.

704
00:35:08,190 --> 00:35:10,530
Now, strikingly,
physicians are also

705
00:35:10,530 --> 00:35:14,880
less likely to refer to
other female surgeons.

706
00:35:14,880 --> 00:35:16,560
That is to say, if
a female surgeon,

707
00:35:16,560 --> 00:35:19,290
if a patient dies
from a female surgeon,

708
00:35:19,290 --> 00:35:22,290
the physician is also
less likely to refer

709
00:35:22,290 --> 00:35:25,230
patients, future patients
to other female surgeons.

710
00:35:25,230 --> 00:35:28,420
These other female surgeons,
nobody has died there.

711
00:35:28,420 --> 00:35:30,000
So, essentially,
what seems to be

712
00:35:30,000 --> 00:35:35,620
the case is that the physicians
also are updating about women

713
00:35:35,620 --> 00:35:40,360
or female surgeons in general if
there is a negative event that

714
00:35:40,360 --> 00:35:43,520
happens, which is
sort of very striking.

715
00:35:43,520 --> 00:35:48,020
After a good outcome, such
as unanticipated survival,

716
00:35:48,020 --> 00:35:50,450
there's a doubling of
referrals to male surgeons

717
00:35:50,450 --> 00:35:53,900
but only a 70% increase in
referrals to female surgeons.

718
00:35:53,900 --> 00:35:55,980
And there are here--
there's no spillovers

719
00:35:55,980 --> 00:35:59,560
to other female surgeons.

720
00:35:59,560 --> 00:36:01,720
So, again, what do we find here?

721
00:36:01,720 --> 00:36:04,930
Well, we find that
women are punished more

722
00:36:04,930 --> 00:36:07,300
for unexpected bad events.

723
00:36:07,300 --> 00:36:13,127
And they're rewarded less
for unexpected good events.

724
00:36:13,127 --> 00:36:14,710
That is to say, it's
not just the case

725
00:36:14,710 --> 00:36:17,860
that there's more updating
going on, in the sense

726
00:36:17,860 --> 00:36:19,840
maybe you learn more
from good and bad events

727
00:36:19,840 --> 00:36:22,840
because maybe you have like less
information about men and women

728
00:36:22,840 --> 00:36:23,890
to start with.

729
00:36:23,890 --> 00:36:25,390
That's not the case here.

730
00:36:25,390 --> 00:36:29,030
What is here the case
is that there's more

731
00:36:29,030 --> 00:36:34,010
updating or more punishment
for mistakes for women and less

732
00:36:34,010 --> 00:36:36,620
reward for good
things to happen,

733
00:36:36,620 --> 00:36:41,010
for good events
for women as well.

734
00:36:41,010 --> 00:36:45,070
Now, why do these
asymmetries matter?

735
00:36:45,070 --> 00:36:49,540
Well, one important issue here
is that women have more chances

736
00:36:49,540 --> 00:36:51,320
to make mistakes.

737
00:36:51,320 --> 00:36:54,430
And, in particular, if
there is like dropping out,

738
00:36:54,430 --> 00:36:57,960
if women don't get promoted
or if they even drop out,

739
00:36:57,960 --> 00:37:00,170
that leads to lower
skill accumulation

740
00:37:00,170 --> 00:37:02,480
and, you know, women will
be just less qualified

741
00:37:02,480 --> 00:37:04,970
overall in the end because
they have less chances

742
00:37:04,970 --> 00:37:09,653
to get promoted and
get more responsibility

743
00:37:09,653 --> 00:37:12,070
and so on and so forth and
then they even drop out overall

744
00:37:12,070 --> 00:37:15,280
and we just have fewer
women in the profession.

745
00:37:15,280 --> 00:37:19,150
Now, moreover,
usually wage gaps are

746
00:37:19,150 --> 00:37:22,330
measured conditional skills,
industry, and position.

747
00:37:22,330 --> 00:37:25,320
But if there is
biased evaluations

748
00:37:25,320 --> 00:37:27,880
that then lead to differences
in skills and positions

749
00:37:27,880 --> 00:37:29,950
in the workplace, not
only is it the case

750
00:37:29,950 --> 00:37:35,210
that women are receiving wage
gaps conditional on the skills,

751
00:37:35,210 --> 00:37:37,570
but they also have different
skills and positions

752
00:37:37,570 --> 00:37:39,910
in the workplace because
of biased evaluations.

753
00:37:39,910 --> 00:37:42,700
So that's an additional
scope for discrimination

754
00:37:42,700 --> 00:37:45,930
and differences in
earnings and so on.

755
00:37:45,930 --> 00:37:47,640
And then, importantly,
this is also

756
00:37:47,640 --> 00:37:50,340
not a mistake that can
get corrected over time.

757
00:37:50,340 --> 00:37:53,640
Because if women drop
out in particular, then--

758
00:37:53,640 --> 00:37:56,100
or if there's no referrals
that happen anymore,

759
00:37:56,100 --> 00:37:58,410
then the doctor or
any other person

760
00:37:58,410 --> 00:38:01,770
might not receive any
signals anymore from work.

761
00:38:01,770 --> 00:38:04,350
If you sort of update
on a particular person

762
00:38:04,350 --> 00:38:07,970
that that patient of mine
died because I sent them

763
00:38:07,970 --> 00:38:09,600
to that person,
if you then think,

764
00:38:09,600 --> 00:38:11,460
you know, this
person is terrible

765
00:38:11,460 --> 00:38:14,430
and don't send any
further patients anymore,

766
00:38:14,430 --> 00:38:16,800
you will never learn that
this woman was, in fact, very

767
00:38:16,800 --> 00:38:19,020
much highly qualified
and that was just

768
00:38:19,020 --> 00:38:21,840
an accident that could
have happened to anyone.

769
00:38:21,840 --> 00:38:25,010
So that's a really
important issue that--

770
00:38:25,010 --> 00:38:26,660
so when women are
underrepresented

771
00:38:26,660 --> 00:38:29,730
employers see fewer
outcomes overall,

772
00:38:29,730 --> 00:38:33,116
and if there is updates
more of the bad outcomes,

773
00:38:33,116 --> 00:38:37,130
or, in particular, if there's
also updates about all women,

774
00:38:37,130 --> 00:38:42,560
then there will be fewer chances
to learn and then women are let

775
00:38:42,560 --> 00:38:45,890
go and so on or less
likely to be hired,

776
00:38:45,890 --> 00:38:48,620
then there will be no chance
of actually correcting

777
00:38:48,620 --> 00:38:50,360
those [INAUDIBLE].

778
00:38:50,360 --> 00:38:54,140
Strikingly, as I told you,
what's really important

779
00:38:54,140 --> 00:38:57,170
as a result is that seems
to be like updating,

780
00:38:57,170 --> 00:39:00,030
not just about
one specific woman

781
00:39:00,030 --> 00:39:02,120
overall after the
negative events,

782
00:39:02,120 --> 00:39:08,330
but also about all other
women that get punished

783
00:39:08,330 --> 00:39:11,210
for the bad event that happens.

784
00:39:11,210 --> 00:39:13,130
And if that's the
case, if there's

785
00:39:13,130 --> 00:39:15,710
several women in the
profession, then all women

786
00:39:15,710 --> 00:39:17,990
always have to pay for
any bad events that

787
00:39:17,990 --> 00:39:18,950
happen to any woman.

788
00:39:18,950 --> 00:39:21,140
And sometimes, you know,
mistakes, in fact, happen.

789
00:39:21,140 --> 00:39:25,160
That's, of course, terrible
because that sort of amplifies

790
00:39:25,160 --> 00:39:26,210
a lot of the updating.

791
00:39:26,210 --> 00:39:30,470
Because even if somebody
makes no mistakes whatsoever,

792
00:39:30,470 --> 00:39:33,240
negative updates happen
on their performance,

793
00:39:33,240 --> 00:39:37,130
other women perform
negatively in some ways.

794
00:39:42,190 --> 00:39:46,470
Now one question you
might ask is, well,

795
00:39:46,470 --> 00:39:49,820
couldn't we just use algorithms,
machine learning, et cetera,

796
00:39:49,820 --> 00:39:52,250
to overcome such biases?

797
00:39:52,250 --> 00:39:54,170
In particular, instead
of hiring and so on,

798
00:39:54,170 --> 00:39:57,080
couldn't you use
algorithms and machine

799
00:39:57,080 --> 00:40:01,280
learning to train those
algorithms and instead

800
00:40:01,280 --> 00:40:04,310
of use computers instead of
humans and then, you know,

801
00:40:04,310 --> 00:40:06,320
we might get rid
of those biases?

802
00:40:06,320 --> 00:40:08,240
Well, one key
problem with that is

803
00:40:08,240 --> 00:40:10,940
that algorithms
themselves can be biased.

804
00:40:10,940 --> 00:40:11,810
Now why is that?

805
00:40:11,810 --> 00:40:16,470
Well, we train algorithms
based on human decisions.

806
00:40:16,470 --> 00:40:19,900
As an example, if you
train an algorithm--

807
00:40:19,900 --> 00:40:22,610
this is, in fact, what
some companies did,

808
00:40:22,610 --> 00:40:24,505
based on actual
decisions, you get

809
00:40:24,505 --> 00:40:26,630
asked the question, this
[INAUDIBLE],, if you train

810
00:40:26,630 --> 00:40:29,210
that as an actual hiring
decisions or interview

811
00:40:29,210 --> 00:40:33,590
decisions or you say, well,
would this person be invited

812
00:40:33,590 --> 00:40:37,610
or is this person, is this
resume likely to be invited

813
00:40:37,610 --> 00:40:39,050
for an interview?

814
00:40:39,050 --> 00:40:41,840
Well, if those interviewing
decisions to start with

815
00:40:41,840 --> 00:40:44,550
were biased based
on sexism, racism,

816
00:40:44,550 --> 00:40:47,690
and so on, if you train
your algorithm on it, then

817
00:40:47,690 --> 00:40:51,770
very much you make that
algorithm biased as well.

818
00:40:51,770 --> 00:40:56,680
Now, crucially, that's,
of course, problematic

819
00:40:56,680 --> 00:40:58,810
and I very much would
like to avoid that.

820
00:40:58,810 --> 00:41:00,430
And we want to be very careful.

821
00:41:00,430 --> 00:41:02,680
But, crucially,
some of these biases

822
00:41:02,680 --> 00:41:05,470
can be perhaps easier
or fixed and perhaps

823
00:41:05,470 --> 00:41:09,660
fixed more easily than biases
in human decision-making, right?

824
00:41:09,660 --> 00:41:11,230
If somebody is sexist
or racist, it's

825
00:41:11,230 --> 00:41:15,260
very hard to fix this, at
least in the short run.

826
00:41:15,260 --> 00:41:18,220
However, an algorithm,
essentially,

827
00:41:18,220 --> 00:41:21,010
is just doing whatever
you train them to do.

828
00:41:21,010 --> 00:41:24,910
So if you build equity concerns
into the objective function

829
00:41:24,910 --> 00:41:27,170
of an algorithm, then
we can essentially

830
00:41:27,170 --> 00:41:32,680
reduce or even eliminate
these biases explicitly

831
00:41:32,680 --> 00:41:35,968
because the algorithm is
explicitly asked to do so.

832
00:41:35,968 --> 00:41:37,510
And so there's a
fascinating-- if you

833
00:41:37,510 --> 00:41:38,890
want to learn more
about this overall,

834
00:41:38,890 --> 00:41:40,723
there's a fascinating
talk on discrimination

835
00:41:40,723 --> 00:41:43,090
by algorithms and people
by Sendhil Mullainathan

836
00:41:43,090 --> 00:41:45,760
that I've linked
here in the slides

837
00:41:45,760 --> 00:41:48,860
that you can watch it
and learn more about.

838
00:41:48,860 --> 00:41:51,940
But overall, so just
to say, to summarize,

839
00:41:51,940 --> 00:41:54,635
algorithms can be biased,
but they can also be fixed.

840
00:41:54,635 --> 00:41:56,260
And that's an important
research agenda

841
00:41:56,260 --> 00:41:59,160
that people have been
working on more recently,

842
00:41:59,160 --> 00:42:02,380
the design of fair algorithms.

843
00:42:02,380 --> 00:42:04,390
Next, we're going to talk
about gender identity

844
00:42:04,390 --> 00:42:10,070
norms and a very nice paper
by Bertrand et. al. from 2015.

845
00:42:10,070 --> 00:42:13,610
Identity considerations were
imported from social psychology

846
00:42:13,610 --> 00:42:17,843
to economics in multiple
papers by Akerlof and Kranton

847
00:42:17,843 --> 00:42:19,880
in particular in 2000.

848
00:42:19,880 --> 00:42:22,220
There's a very nice book
about identity economics

849
00:42:22,220 --> 00:42:25,310
if you're interested in
learning more about that.

850
00:42:25,310 --> 00:42:28,990
Gender identity norms
are an important example

851
00:42:28,990 --> 00:42:33,380
or application of such norms
or identity considerations

852
00:42:33,380 --> 00:42:34,880
more generally.

853
00:42:34,880 --> 00:42:38,870
In particular, Akerlof and
Kranton, or we here as well,

854
00:42:38,870 --> 00:42:43,040
focus on two social
categories, men and women.

855
00:42:43,040 --> 00:42:46,220
For simplicity, we can also just
talk about husbands and wives,

856
00:42:46,220 --> 00:42:49,050
understanding things are
more complicated in reality.

857
00:42:49,050 --> 00:42:51,850
But for simplicity, we're
going to focus on that for now.

858
00:42:54,420 --> 00:42:55,890
What do we mean by
gender identity?

859
00:42:55,890 --> 00:42:57,307
Well, gender
identity is something

860
00:42:57,307 --> 00:43:00,090
that changes payoffs from
different action as dictated

861
00:43:00,090 --> 00:43:02,190
by some prescriptive norms.

862
00:43:02,190 --> 00:43:03,360
What might those norms be?

863
00:43:03,360 --> 00:43:05,460
Well, one norm would
be men should not

864
00:43:05,460 --> 00:43:07,980
do women's work in the
house, in the home,

865
00:43:07,980 --> 00:43:11,820
like cooking,
cleaning, and so on.

866
00:43:11,820 --> 00:43:14,520
Some men might consider
that as women's work.

867
00:43:14,520 --> 00:43:17,490
And they might feel
particularly high disutility

868
00:43:17,490 --> 00:43:21,390
from doing so if they
violate that law.

869
00:43:21,390 --> 00:43:26,370
Another norm would be men should
earn more than their wives.

870
00:43:26,370 --> 00:43:30,690
Now husbands now lose identity,
experience lower utility

871
00:43:30,690 --> 00:43:33,330
in circumstances when these
prescriptions are violated,

872
00:43:33,330 --> 00:43:37,890
when he does housework, or when
his wife earns more than half

873
00:43:37,890 --> 00:43:40,030
of the household income.

874
00:43:40,030 --> 00:43:44,030
Notice that wives might
also lose identity.

875
00:43:44,030 --> 00:43:45,780
For simplicity, we're
going to talk more--

876
00:43:45,780 --> 00:43:49,150
only about husbands,
I'm doing so for now.

877
00:43:49,150 --> 00:43:51,360
Now Bertrand et. al.

878
00:43:51,360 --> 00:43:54,070
particularly focus
on the prescription

879
00:43:54,070 --> 00:43:56,940
that men should earn
more than their wives.

880
00:43:56,940 --> 00:44:00,120
And it's important to notice
that this particular gender

881
00:44:00,120 --> 00:44:04,650
identity norm would only matter
in a world where women could--

882
00:44:04,650 --> 00:44:06,060
would not matter
in a world where

883
00:44:06,060 --> 00:44:09,360
women could never earn more
than her actual or potential

884
00:44:09,360 --> 00:44:10,260
husband.

885
00:44:10,260 --> 00:44:12,840
But if women are held back
by technological factors,

886
00:44:12,840 --> 00:44:17,100
by education, et cetera,
and never even earn

887
00:44:17,100 --> 00:44:20,430
close to as much as their
husbands or potential husbands,

888
00:44:20,430 --> 00:44:24,440
well, then the gender identity
norm is kind of irrelevant.

889
00:44:24,440 --> 00:44:27,920
However, as women make
gains in the labor market,

890
00:44:27,920 --> 00:44:30,650
these slow moving
gender identity norms

891
00:44:30,650 --> 00:44:34,550
can become increasingly relevant
and increasingly relevant

892
00:44:34,550 --> 00:44:35,700
constraints.

893
00:44:35,700 --> 00:44:38,690
So now as women have more
and more earnings potential

894
00:44:38,690 --> 00:44:40,760
and are much--

895
00:44:40,760 --> 00:44:43,370
be quite likely to earn
more than their husbands,

896
00:44:43,370 --> 00:44:47,480
these norms really bite
and might hold women back.

897
00:44:47,480 --> 00:44:50,990
Now how would you study
this question empirically?

898
00:44:50,990 --> 00:44:53,120
The idea here in Bertrand et.

899
00:44:53,120 --> 00:44:58,920
al. is that, well,
if husbands loose

900
00:44:58,920 --> 00:45:02,880
identity when their wives earn
more than half the household

901
00:45:02,880 --> 00:45:06,960
income, well, then
they will try to avoid

902
00:45:06,960 --> 00:45:09,180
such situations, either--

903
00:45:09,180 --> 00:45:11,310
in various ways,
and particularly

904
00:45:11,310 --> 00:45:14,430
by avoiding such marriages
in the first place.

905
00:45:14,430 --> 00:45:16,260
But also if those
marriages happen,

906
00:45:16,260 --> 00:45:18,330
that might lead to
divorce and so on

907
00:45:18,330 --> 00:45:20,190
and those marriages
might break up.

908
00:45:20,190 --> 00:45:22,590
So the idea of a
paper then is to look

909
00:45:22,590 --> 00:45:27,060
for a missing mass in couples
where, particularly in

910
00:45:27,060 --> 00:45:29,820
places where the husband
earns just a little bit more

911
00:45:29,820 --> 00:45:31,030
than their wife.

912
00:45:31,030 --> 00:45:32,200
Now how do they do that?

913
00:45:32,200 --> 00:45:35,190
Well, they have to look at the
distribution of relative income

914
00:45:35,190 --> 00:45:37,733
using US admin data.

915
00:45:37,733 --> 00:45:39,150
What do I mean by
relative income?

916
00:45:39,150 --> 00:45:41,400
Relative income is the
income between the husband

917
00:45:41,400 --> 00:45:42,750
and the wife.

918
00:45:42,750 --> 00:45:46,080
They have essentially,
there's a very nice survey,

919
00:45:46,080 --> 00:45:49,170
data of the survey income
and program participation.

920
00:45:49,170 --> 00:45:53,670
These are a series of
representative national panels

921
00:45:53,670 --> 00:45:59,130
that about 70,000 couple
observations from 1992 to 2004.

922
00:45:59,130 --> 00:46:02,460
This includes couples only
in which both the husbands

923
00:46:02,460 --> 00:46:05,810
and wives are earning
positive income.

924
00:46:05,810 --> 00:46:08,730
So notice that this was
not about husband and wife

925
00:46:08,730 --> 00:46:09,630
working at all.

926
00:46:09,630 --> 00:46:11,880
It's not about women being
at home versus working.

927
00:46:11,880 --> 00:46:17,370
But it's about cases where
both couples or both spouses

928
00:46:17,370 --> 00:46:18,540
have positive income.

929
00:46:21,180 --> 00:46:24,330
Their income measure is
annual total labor income

930
00:46:24,330 --> 00:46:27,060
plus self-employed
income as well.

931
00:46:27,060 --> 00:46:29,070
Now using that data,
Bertrand et. al.

932
00:46:29,070 --> 00:46:31,140
can now compute the
shares of couples

933
00:46:31,140 --> 00:46:34,665
earning different
fractions of total income.

934
00:46:34,665 --> 00:46:36,540
So what I'm going to
show you is the fraction

935
00:46:36,540 --> 00:46:40,380
of the wife as a share of
total income in the household.

936
00:46:40,380 --> 00:46:43,170
And that's graphed
using 20 bits.

937
00:46:43,170 --> 00:46:44,530
What does this graph look like?

938
00:46:44,530 --> 00:46:48,990
You can see here the
share earned by the wife

939
00:46:48,990 --> 00:46:50,490
is just a fraction of couples.

940
00:46:50,490 --> 00:46:51,920
This overall.

941
00:46:51,920 --> 00:46:53,290
So this is all the couples.

942
00:46:53,290 --> 00:46:54,180
This is [INAUDIBLE].

943
00:46:54,180 --> 00:46:57,450
This all adds up
to 100% overall.

944
00:46:57,450 --> 00:47:01,370
And what you see is, perhaps not
surprisingly, that the share--

945
00:47:01,370 --> 00:47:05,700
the fraction of couple for which
the share of earned by the wife

946
00:47:05,700 --> 00:47:09,950
is higher than 50%
is not super high.

947
00:47:09,950 --> 00:47:12,530
That's not surprising
given that men on average

948
00:47:12,530 --> 00:47:14,540
are earning more than women.

949
00:47:14,540 --> 00:47:15,970
So that's not the question here.

950
00:47:15,970 --> 00:47:21,110
The question here is, is there
a missing mass just above 50%?

951
00:47:21,110 --> 00:47:26,840
The idea is, suppose there are
some couples where women earn

952
00:47:26,840 --> 00:47:31,610
45% or 46% or even 49%,
there the norm does not

953
00:47:31,610 --> 00:47:34,370
cause an issue or there's
no identity problems

954
00:47:34,370 --> 00:47:36,200
because the husband
earns somewhat more

955
00:47:36,200 --> 00:47:37,770
and feels happy about it.

956
00:47:37,770 --> 00:47:41,553
But as soon as you cross
the threshold to like 51%,

957
00:47:41,553 --> 00:47:42,470
now there's a problem.

958
00:47:42,470 --> 00:47:43,640
Now the woman earns more.

959
00:47:43,640 --> 00:47:47,010
The husband feels uncomfortable,
unhappy and so on.

960
00:47:47,010 --> 00:47:52,260
And that might lead to
those marriages disappearing

961
00:47:52,260 --> 00:47:53,760
in various ways.

962
00:47:53,760 --> 00:47:56,457
A, they might not form
in the first place.

963
00:47:56,457 --> 00:47:58,040
Potential husband
and wife might never

964
00:47:58,040 --> 00:48:00,890
even marry because husband
gets uncomfortable or does not

965
00:48:00,890 --> 00:48:03,410
want to marry somebody
who earns more.

966
00:48:03,410 --> 00:48:04,825
Second, there might be divorces.

967
00:48:04,825 --> 00:48:06,200
These marriages
might essentially

968
00:48:06,200 --> 00:48:09,680
dissolve because there's lots
of conflict in the household.

969
00:48:09,680 --> 00:48:11,333
Now, how does that
manifest here?

970
00:48:11,333 --> 00:48:12,750
You can look at
like if you come--

971
00:48:12,750 --> 00:48:16,760
if you look at the graph and
look at this line here coming

972
00:48:16,760 --> 00:48:19,700
from the left was 50%, you
can kind of look at like,

973
00:48:19,700 --> 00:48:23,360
try to predict-- and it's kind
of the idea of the empirical

974
00:48:23,360 --> 00:48:24,170
strategy--

975
00:48:24,170 --> 00:48:27,080
you can try to predict what
fraction would you expect

976
00:48:27,080 --> 00:48:29,020
there to be just above 50%?

977
00:48:29,020 --> 00:48:30,770
And when you do that,
you think, you know,

978
00:48:30,770 --> 00:48:33,710
the fraction should be
about something like 7%,

979
00:48:33,710 --> 00:48:35,060
should be about here.

980
00:48:35,060 --> 00:48:38,630
But, instead, there's this
jump down to about 6%.

981
00:48:38,630 --> 00:48:41,630
So there's about 1% of
couples are missing here.

982
00:48:41,630 --> 00:48:44,250
There's a missing
mass going on here,

983
00:48:44,250 --> 00:48:47,600
which suggests that there
are some marriages that

984
00:48:47,600 --> 00:48:52,520
are do not happen or dissolve
because of those identity

985
00:48:52,520 --> 00:48:53,035
concerns.

986
00:48:53,035 --> 00:48:54,410
This is what I
already said here.

987
00:48:54,410 --> 00:48:56,540
There's a cliff to
the right of 0.5

988
00:48:56,540 --> 00:48:59,990
in the distribution of relative
earnings in most couples.

989
00:48:59,990 --> 00:49:02,000
And that's an implication
of the prescription

990
00:49:02,000 --> 00:49:04,700
that men should earn
more than their wives.

991
00:49:04,700 --> 00:49:08,420
Now what are the mechanisms
behind this missing mass?

992
00:49:08,420 --> 00:49:11,210
Bertrand et. al. had a wealth
of evidence, additional evidence

993
00:49:11,210 --> 00:49:12,800
showing these mechanisms.

994
00:49:12,800 --> 00:49:15,840
In particular, they
show three mechanisms,

995
00:49:15,840 --> 00:49:18,830
which is, first,
missing couples did not

996
00:49:18,830 --> 00:49:20,280
form in the first place.

997
00:49:20,280 --> 00:49:21,320
When you look at--

998
00:49:21,320 --> 00:49:25,330
when you just randomly match
men and women, some of which

999
00:49:25,330 --> 00:49:28,170
were men have higher earnings
potential compared to women

1000
00:49:28,170 --> 00:49:30,080
and the other way
around, the couples

1001
00:49:30,080 --> 00:49:33,020
where women are earning just
somewhat less than women

1002
00:49:33,020 --> 00:49:34,970
are just much less
likely or less

1003
00:49:34,970 --> 00:49:38,150
likely to marry
compared to couples

1004
00:49:38,150 --> 00:49:41,440
where women are just earning
somewhat less than men.

1005
00:49:41,440 --> 00:49:44,440
Second, such couples are less
happy and stable and more

1006
00:49:44,440 --> 00:49:45,502
likely to end in divorce.

1007
00:49:45,502 --> 00:49:46,960
So conditional on
marrying, they're

1008
00:49:46,960 --> 00:49:50,977
more likely to get divorced as
marriages just end unhappily.

1009
00:49:50,977 --> 00:49:52,810
And that's, of course,
then also contributes

1010
00:49:52,810 --> 00:49:54,720
to this missing mass.

1011
00:49:54,720 --> 00:49:58,860
And then third, in cases where
wives have higher earnings

1012
00:49:58,860 --> 00:50:01,200
potential than
their husbands, they

1013
00:50:01,200 --> 00:50:05,700
decide now to work less outside
of the household or, you know,

1014
00:50:05,700 --> 00:50:08,700
put in less [INAUDIBLE] work so
they get less promoted, which

1015
00:50:08,700 --> 00:50:12,300
then sort of pushes
down their earnings

1016
00:50:12,300 --> 00:50:15,720
or increased husband's
earnings in relative terms,

1017
00:50:15,720 --> 00:50:19,590
such as done in such cases
the woman earns less than 50%.

1018
00:50:19,590 --> 00:50:21,720
And that also leads
to that missing mass.

1019
00:50:21,720 --> 00:50:23,730
In fact, what you see
here is this increase

1020
00:50:23,730 --> 00:50:26,680
in mass going from
like 20% to 40%,

1021
00:50:26,680 --> 00:50:28,467
which perhaps is
also like the reason

1022
00:50:28,467 --> 00:50:30,300
because some of the
mass that's missing here

1023
00:50:30,300 --> 00:50:33,530
is pushed here to the left.

1024
00:50:33,530 --> 00:50:39,160
Now, in addition to working
less outside of the household

1025
00:50:39,160 --> 00:50:43,870
and just working fewer hours and
working less overtime, it's on,

1026
00:50:43,870 --> 00:50:47,140
women then are also held
back by non-marketing

1027
00:50:47,140 --> 00:50:47,980
in childcare work.

1028
00:50:47,980 --> 00:50:49,840
So women, in
particular, women who

1029
00:50:49,840 --> 00:50:54,880
are earning just a little
bit more than their husbands

1030
00:50:54,880 --> 00:50:58,390
are doing more nonmarket and
childcare work, which is often

1031
00:50:58,390 --> 00:51:00,020
called the second shift.

1032
00:51:00,020 --> 00:51:01,690
So the idea here
is that husbands

1033
00:51:01,690 --> 00:51:05,470
even when his identity gets
threatened by the woman earning

1034
00:51:05,470 --> 00:51:07,660
more than he does,
and then he can

1035
00:51:07,660 --> 00:51:10,510
particularly not to the
childcare or nonmarket work.

1036
00:51:10,510 --> 00:51:13,302
So the woman ends up not
only working more outside

1037
00:51:13,302 --> 00:51:15,760
of the household and earning
more and being more productive

1038
00:51:15,760 --> 00:51:19,210
in that work, but also
doing more work at home

1039
00:51:19,210 --> 00:51:21,910
and in addition to that.

1040
00:51:21,910 --> 00:51:26,820
That's essentially what's called
the second or double shift.

1041
00:51:26,820 --> 00:51:34,550
Now an additional issue that is
a really important contributor

1042
00:51:34,550 --> 00:51:39,080
to the gender gap is
the arrival of children.

1043
00:51:39,080 --> 00:51:41,260
And this is a very nice
paper by Kleven et. al. that

1044
00:51:41,260 --> 00:51:44,960
looks at people's
earning, in particular,

1045
00:51:44,960 --> 00:51:48,440
the earning, how earnings
devolve after the birth

1046
00:51:48,440 --> 00:51:50,910
of a couple's first child.

1047
00:51:50,910 --> 00:51:54,020
This is an event study
that looks at years

1048
00:51:54,020 --> 00:51:57,260
over time for men and women.

1049
00:51:57,260 --> 00:52:00,290
And what you see essentially
for men, these are earnings

1050
00:52:00,290 --> 00:52:02,480
and on the right here
you have hours worked.

1051
00:52:02,480 --> 00:52:04,850
For men, essentially,
there's barely

1052
00:52:04,850 --> 00:52:09,690
any difference, if any
difference, of earnings

1053
00:52:09,690 --> 00:52:12,260
once the first child is born.

1054
00:52:12,260 --> 00:52:16,580
Essentially, men's earnings
tend to be pretty flat.

1055
00:52:16,580 --> 00:52:18,280
In contrast, of
course, that could be

1056
00:52:18,280 --> 00:52:19,530
for the growth that's avoided.

1057
00:52:19,530 --> 00:52:21,200
But, you know, at least
there's no reduction

1058
00:52:21,200 --> 00:52:22,190
in men's [INAUDIBLE].

1059
00:52:22,190 --> 00:52:25,100
This is, by the
way, Danish data.

1060
00:52:25,100 --> 00:52:29,870
Now for women, however,
there's a clearer reduction

1061
00:52:29,870 --> 00:52:32,600
in earnings.

1062
00:52:32,600 --> 00:52:40,410
And the long-run child
penalty is about almost 20%.

1063
00:52:40,410 --> 00:52:41,810
So, essentially, women after--

1064
00:52:41,810 --> 00:52:43,280
if you look at
men and women that

1065
00:52:43,280 --> 00:52:47,990
earn the same to start
with, women earn about 20%

1066
00:52:47,990 --> 00:52:51,200
less in the long run.

1067
00:52:51,200 --> 00:52:54,170
That is due to the
first child or after--

1068
00:52:54,170 --> 00:52:57,530
this gap arises after
the first child was born.

1069
00:52:57,530 --> 00:53:00,320
This is true for earnings.

1070
00:53:00,320 --> 00:53:02,060
It's also true for hours worked.

1071
00:53:02,060 --> 00:53:04,340
And there the
penalty is about 10%.

1072
00:53:04,340 --> 00:53:07,160
It's also true for
wages, as in how much

1073
00:53:07,160 --> 00:53:10,760
are people paid
conditional on work.

1074
00:53:10,760 --> 00:53:13,160
So part of that is
women working less.

1075
00:53:13,160 --> 00:53:16,970
But part of it is also women
earning less, a la being

1076
00:53:16,970 --> 00:53:20,480
promoted less, being paid less,
getting fewer raises and so on

1077
00:53:20,480 --> 00:53:23,120
once the child is born.

1078
00:53:23,120 --> 00:53:25,610
One very interesting
fact in this paper

1079
00:53:25,610 --> 00:53:29,540
is that child penalties are
transmitted through generations

1080
00:53:29,540 --> 00:53:32,540
from parents to
daughters, suggesting

1081
00:53:32,540 --> 00:53:36,270
an influence of child
environment on gender identity.

1082
00:53:36,270 --> 00:53:38,600
There seems to be something
about the gender identity

1083
00:53:38,600 --> 00:53:43,250
about the woman is supposed
to take care of the child

1084
00:53:43,250 --> 00:53:45,080
and that runs in the family.

1085
00:53:45,080 --> 00:53:46,970
And that leads to
particularly large

1086
00:53:46,970 --> 00:53:54,740
what they call child penalties
for those kinds of women

1087
00:53:54,740 --> 00:53:56,890
compared to men.

1088
00:53:56,890 --> 00:53:59,060
Now one thing you
might say is, well,

1089
00:53:59,060 --> 00:54:01,810
couldn't we have some gender
neutral family policies

1090
00:54:01,810 --> 00:54:03,400
that alleviate that?

1091
00:54:03,400 --> 00:54:08,590
And, in particular, in
academia, for instance, one idea

1092
00:54:08,590 --> 00:54:13,660
is one could use gender neutral
tenure clock stopping policies

1093
00:54:13,660 --> 00:54:16,630
to alleviate such issues.

1094
00:54:16,630 --> 00:54:18,400
And the idea is
that, in particular,

1095
00:54:18,400 --> 00:54:20,030
sort of early
years of the career

1096
00:54:20,030 --> 00:54:21,765
are particularly valuable.

1097
00:54:21,765 --> 00:54:23,140
So the very least
we can do is we

1098
00:54:23,140 --> 00:54:27,430
can stop the tenure clocks
to help women or couples

1099
00:54:27,430 --> 00:54:29,830
to deal with childcare.

1100
00:54:29,830 --> 00:54:33,550
And many, many universities and
research intensive universities

1101
00:54:33,550 --> 00:54:35,440
in the US have
adopted such what's

1102
00:54:35,440 --> 00:54:39,910
called gender neutral tenure
clock policies in the US.

1103
00:54:39,910 --> 00:54:43,010
Now what do I mean
by gender neutral?

1104
00:54:43,010 --> 00:54:50,620
Essentially, if a couple has a
child, not only does the woman,

1105
00:54:50,620 --> 00:54:53,110
if she has given
birth to the child,

1106
00:54:53,110 --> 00:54:56,750
does she get an extension
in the tenure clock,

1107
00:54:56,750 --> 00:55:00,720
but also the spouse will get
such an extension as well.

1108
00:55:00,720 --> 00:55:03,370
So it's gender
neutral in that sense.

1109
00:55:03,370 --> 00:55:05,770
Now, what would such
policies then do?

1110
00:55:05,770 --> 00:55:08,800
Well, they're intended to
involve men more in child care

1111
00:55:08,800 --> 00:55:12,490
and foster that, the idea
being like if you only

1112
00:55:12,490 --> 00:55:16,120
give that extension to
women, it sort of pushes

1113
00:55:16,120 --> 00:55:21,700
women to take care
of their children

1114
00:55:21,700 --> 00:55:23,860
and sort of not to work.

1115
00:55:23,860 --> 00:55:27,880
And perhaps if we involve
both spouses or parents

1116
00:55:27,880 --> 00:55:32,050
in the childcare, providing
those gender neutral

1117
00:55:32,050 --> 00:55:34,757
tenure clock stopping policies
might contribute to that.

1118
00:55:34,757 --> 00:55:36,340
But, of course,
there's no enforcement

1119
00:55:36,340 --> 00:55:38,990
and thus, potentially,
such policies

1120
00:55:38,990 --> 00:55:40,900
in that potentially
these policies

1121
00:55:40,900 --> 00:55:43,702
might be even enhancing
gender inequality.

1122
00:55:43,702 --> 00:55:45,160
And, particularly,
if it's the case

1123
00:55:45,160 --> 00:55:49,330
that if men take advantage
of this tenure clock

1124
00:55:49,330 --> 00:55:52,000
stopping policy by essentially
just having another year to do

1125
00:55:52,000 --> 00:55:56,050
research, but still don't
do very much in supporting

1126
00:55:56,050 --> 00:55:59,120
the child, then men might
actually benefit from it,

1127
00:55:59,120 --> 00:56:02,343
while women might not
benefit at all in part

1128
00:56:02,343 --> 00:56:03,760
because, relatively,
they're going

1129
00:56:03,760 --> 00:56:05,620
to do worse compared to men.

1130
00:56:05,620 --> 00:56:11,920
And so Antecol look, in fact,
at the impact of these adoptions

1131
00:56:11,920 --> 00:56:18,300
of these supposedly gender
neutral tenure clock

1132
00:56:18,300 --> 00:56:19,630
stopping policies.

1133
00:56:19,630 --> 00:56:23,100
And they find that
the introduction

1134
00:56:23,100 --> 00:56:26,250
of this policy at [INAUDIBLE]
economics department

1135
00:56:26,250 --> 00:56:30,060
substantially reduced female
tenure rates while increasing

1136
00:56:30,060 --> 00:56:31,200
male tenure rates.

1137
00:56:31,200 --> 00:56:34,762
So essentially what looks like
a neutral policy, in fact,

1138
00:56:34,762 --> 00:56:35,970
is not gender neutral at all.

1139
00:56:35,970 --> 00:56:39,870
It makes things worse for
women and better for men.

1140
00:56:39,870 --> 00:56:43,800
Let me tell you about
one final paper on gender

1141
00:56:43,800 --> 00:56:46,260
identity and the labor market.

1142
00:56:46,260 --> 00:56:48,480
This paper is called "'Acting
wife'-- marriage market

1143
00:56:48,480 --> 00:56:50,730
incentives and labor
market investments."

1144
00:56:50,730 --> 00:56:52,960
It's by Bursztyn et. al.
and it asks the question

1145
00:56:52,960 --> 00:56:56,190
whether women avoid
career-enhancing actions

1146
00:56:56,190 --> 00:56:59,160
because these actions signal
undesirable or seemingly

1147
00:56:59,160 --> 00:57:02,220
undesirable traits such as
ambition to the marriage

1148
00:57:02,220 --> 00:57:03,770
market.

1149
00:57:03,770 --> 00:57:06,660
There's two parts in the paper.

1150
00:57:06,660 --> 00:57:09,390
One part is just
observational data

1151
00:57:09,390 --> 00:57:12,720
that finds that while married
and unmarried female MBA

1152
00:57:12,720 --> 00:57:15,420
students perform similarly
when their performance is

1153
00:57:15,420 --> 00:57:18,990
unobserved by classmates, such
on some exams or problem sets,

1154
00:57:18,990 --> 00:57:22,830
unmarried women have lower
participation grades which are

1155
00:57:22,830 --> 00:57:24,880
observed by their classmates.

1156
00:57:24,880 --> 00:57:30,060
And so the idea here is that
only unmarried female MBA

1157
00:57:30,060 --> 00:57:34,960
students have incentives
on the marriage market.

1158
00:57:34,960 --> 00:57:38,775
And, therefore, when their
participation is observed,

1159
00:57:38,775 --> 00:57:41,130
when their performance
is observed,

1160
00:57:41,130 --> 00:57:44,220
they might want to scale
back their ambition

1161
00:57:44,220 --> 00:57:46,890
because they look too ambitious
on the marriage market.

1162
00:57:46,890 --> 00:57:52,140
And that might be adversely
interpreted by potential dates.

1163
00:57:52,140 --> 00:57:55,150
And, of course, married women,
since they are already married,

1164
00:57:55,150 --> 00:57:57,090
don't have that incentive.

1165
00:57:57,090 --> 00:57:59,520
So they don't engage
in such behavior.

1166
00:57:59,520 --> 00:58:01,900
As you might know, there's
lots of dating going on.

1167
00:58:01,900 --> 00:58:06,180
So lots of MBA students are
actively looking for a spouse

1168
00:58:06,180 --> 00:58:09,590
during their experience,
both men and women.

1169
00:58:09,590 --> 00:58:12,620
Now in addition, they have
field experiments with MBAs

1170
00:58:12,620 --> 00:58:17,180
in which they vary the
expectation whether responses

1171
00:58:17,180 --> 00:58:20,390
in a real stake
placement questionnaire

1172
00:58:20,390 --> 00:58:23,540
is going to be made public, in
particular, whether it's going

1173
00:58:23,540 --> 00:58:26,690
to be observed by their peers.

1174
00:58:26,690 --> 00:58:30,620
And now female,
single female students

1175
00:58:30,620 --> 00:58:33,890
reported or desired
salary and willingness

1176
00:58:33,890 --> 00:58:38,420
to travel and work long hours
on those questionnaires when

1177
00:58:38,420 --> 00:58:41,750
they expected their classmates
to see their preferences.

1178
00:58:41,750 --> 00:58:45,230
That is to say, when these
questions are public,

1179
00:58:45,230 --> 00:58:52,370
single female students reported
lower ambition as manifested

1180
00:58:52,370 --> 00:58:54,830
by desired salaries and
willingness to travel and work

1181
00:58:54,830 --> 00:58:56,960
hours, presumably
because they want

1182
00:58:56,960 --> 00:59:00,300
to look more favorable
on the marriage market.

1183
00:59:00,300 --> 00:59:03,780
Notably, other groups'
responses were affected

1184
00:59:03,780 --> 00:59:06,658
by such peer observability.

1185
00:59:06,658 --> 00:59:08,700
So in other groups, it
didn't matter whether it's

1186
00:59:08,700 --> 00:59:10,840
in public and in private.

1187
00:59:10,840 --> 00:59:12,580
In addition, there's
a second experiment

1188
00:59:12,580 --> 00:59:16,780
that indicates the effects were
driven by the observability

1189
00:59:16,780 --> 00:59:18,610
by a single male peers.

1190
00:59:18,610 --> 00:59:22,780
That is to say, if another
woman saw the information, that

1191
00:59:22,780 --> 00:59:24,080
didn't matter so much.

1192
00:59:24,080 --> 00:59:26,770
But if this other
single male peer

1193
00:59:26,770 --> 00:59:30,160
was likely or possibly going
to see that information,

1194
00:59:30,160 --> 00:59:33,470
then these effects
would show up.

1195
00:59:33,470 --> 00:59:37,840
So that's overall showing
that, essentially,

1196
00:59:37,840 --> 00:59:41,900
what's perceived to
be undesirable traits,

1197
00:59:41,900 --> 00:59:44,440
such as ambition,
if that is perceived

1198
00:59:44,440 --> 00:59:49,000
to be damaging on the marriage
market, that can then, in turn,

1199
00:59:49,000 --> 00:59:51,010
reduce ambition and
then contributes

1200
00:59:51,010 --> 00:59:54,760
to the gender gap overall.

1201
00:59:59,990 --> 01:00:02,330
OK, let me tell you
now about a final paper

1202
01:00:02,330 --> 01:00:07,470
on saying no and on demand and
supply for different tasks.

1203
01:00:07,470 --> 01:00:09,470
This is a very nice paper
by Vesterlund et. al.

1204
01:00:09,470 --> 01:00:13,050
And it asks the question whether
women say no often enough.

1205
01:00:13,050 --> 01:00:15,440
And so that's motivated
by the observation

1206
01:00:15,440 --> 01:00:20,030
that female faculty member
spend fewer hours in research

1207
01:00:20,030 --> 01:00:23,450
and more hours on
university survey committees

1208
01:00:23,450 --> 01:00:25,020
than male faculty.

1209
01:00:25,020 --> 01:00:27,770
They are also more
likely to have positions

1210
01:00:27,770 --> 01:00:29,660
on university-wide committees.

1211
01:00:29,660 --> 01:00:31,460
They advise more
undergrad students

1212
01:00:31,460 --> 01:00:33,770
and participate more in
departments and college level

1213
01:00:33,770 --> 01:00:37,100
committees than male faculty.

1214
01:00:37,100 --> 01:00:39,500
More generally,
in mid-level jobs,

1215
01:00:39,500 --> 01:00:41,930
men more than women
evaluate the individual task

1216
01:00:41,930 --> 01:00:43,790
assignments as challenging.

1217
01:00:43,790 --> 01:00:46,940
And this is partially
attributed to differential task

1218
01:00:46,940 --> 01:00:49,190
assignments by supervisors.

1219
01:00:49,190 --> 01:00:55,720
So if you sort of see
these differences,

1220
01:00:55,720 --> 01:01:01,180
you might ask the
question, why do women

1221
01:01:01,180 --> 01:01:04,810
decide to spend or spends
their work time differently?

1222
01:01:04,810 --> 01:01:07,100
And there's two types
of dimensions here.

1223
01:01:07,100 --> 01:01:10,100
One is demand and one is supply.

1224
01:01:10,100 --> 01:01:11,950
So demand is
essentially the question

1225
01:01:11,950 --> 01:01:14,630
whether sex differences
in the types of tasks

1226
01:01:14,630 --> 01:01:18,940
that women and men are asked to
do at work is that [INAUDIBLE]

1227
01:01:18,940 --> 01:01:21,573
essentially what
they're being asked.

1228
01:01:21,573 --> 01:01:22,990
And are women more
likely than men

1229
01:01:22,990 --> 01:01:25,720
to be asked to do what's called
non-promotable tasks, tasks

1230
01:01:25,720 --> 01:01:27,970
that are not really
helpful for their career,

1231
01:01:27,970 --> 01:01:29,320
for getting promotions.

1232
01:01:29,320 --> 01:01:31,790
And then you might ask
why that's the case.

1233
01:01:31,790 --> 01:01:33,610
Supply is are there
sex differences

1234
01:01:33,610 --> 01:01:36,440
in the willingness to agree to
perform non-promotable tasks

1235
01:01:36,440 --> 01:01:36,940
when asked?

1236
01:01:36,940 --> 01:01:38,565
[INAUDIBLE] conditional
on being asked,

1237
01:01:38,565 --> 01:01:41,350
were women more
likely to say yes.

1238
01:01:41,350 --> 01:01:46,240
And, again, are women more
likely to demand to say yes

1239
01:01:46,240 --> 01:01:48,430
to non-promotable tasks.

1240
01:01:48,430 --> 01:01:50,350
Now why do we care?

1241
01:01:50,350 --> 01:01:53,180
Well, there's an individual
decision-making perspective,

1242
01:01:53,180 --> 01:01:55,810
which is that people might
make suboptimal decisions how

1243
01:01:55,810 --> 01:01:57,430
to allocate their work--

1244
01:01:57,430 --> 01:01:58,360
time at work.

1245
01:01:58,360 --> 01:02:00,210
And understanding these
underlying reasons

1246
01:02:00,210 --> 01:02:04,060
and [INAUDIBLE] understand
these underlying reasons,

1247
01:02:04,060 --> 01:02:06,310
which will lead them potentially
to some interventions

1248
01:02:06,310 --> 01:02:08,685
to improve decision-making
for certain individuals, which

1249
01:02:08,685 --> 01:02:10,780
have a lot of what
behavioral economics is.

1250
01:02:10,780 --> 01:02:13,900
But there's also managerial
and social planner perspectives

1251
01:02:13,900 --> 01:02:17,230
that organizations may not
be using their resources most

1252
01:02:17,230 --> 01:02:18,070
effectively.

1253
01:02:18,070 --> 01:02:22,390
And if you sort of
reallocated some of the tasks,

1254
01:02:22,390 --> 01:02:24,988
that would increase
output overall.

1255
01:02:24,988 --> 01:02:27,280
And then finally, there is
a public policy perspective.

1256
01:02:27,280 --> 01:02:30,760
Well, if these sex differences
in the allocation of time

1257
01:02:30,760 --> 01:02:34,570
explain vertical sex
segregation, essentially women

1258
01:02:34,570 --> 01:02:36,250
not being promoted
enough, then it

1259
01:02:36,250 --> 01:02:38,920
would help us sort of
improve gender equity overall

1260
01:02:38,920 --> 01:02:43,220
and perhaps also try to
reduce the gender gap.

1261
01:02:43,220 --> 01:02:45,920
Now what are these promotable
and non-promotable tasks

1262
01:02:45,920 --> 01:02:47,540
in the [INAUDIBLE] academics?

1263
01:02:47,540 --> 01:02:49,900
A promotable task is a
task that's doing research.

1264
01:02:49,900 --> 01:02:53,240
That's essentially what people
are evaluated on eventually,

1265
01:02:53,240 --> 01:02:55,400
but also any other task,
essentially, that's

1266
01:02:55,400 --> 01:02:58,190
seen where other
people will reward you

1267
01:02:58,190 --> 01:02:59,630
for it when you do it.

1268
01:02:59,630 --> 01:03:01,800
And non-promotable tasks
are essentially tasks

1269
01:03:01,800 --> 01:03:05,150
that often [INAUDIBLE],, many
people could do the task

1270
01:03:05,150 --> 01:03:06,930
and everybody wants
the task to be done,

1271
01:03:06,930 --> 01:03:09,320
yet everybody prefers
somebody else to do it.

1272
01:03:09,320 --> 01:03:12,440
An example would be an ethics
committee of the university.

1273
01:03:12,440 --> 01:03:14,000
Surely, we should have one.

1274
01:03:14,000 --> 01:03:15,500
But surely, sitting
on the committee

1275
01:03:15,500 --> 01:03:17,150
will not help
people get promoted

1276
01:03:17,150 --> 01:03:20,020
or get tenure when
it comes to research.

1277
01:03:20,020 --> 01:03:23,950
Now the authors had to do
a field study of faculty

1278
01:03:23,950 --> 01:03:25,330
at a large public university.

1279
01:03:25,330 --> 01:03:26,747
And they sent,
essentially, emails

1280
01:03:26,747 --> 01:03:31,300
from the chair of the faculty
senate asking them to volunteer

1281
01:03:31,300 --> 01:03:34,000
to join one of several
university-wide faculty senate

1282
01:03:34,000 --> 01:03:34,990
committees.

1283
01:03:34,990 --> 01:03:37,360
And the clear
answer in this study

1284
01:03:37,360 --> 01:03:42,500
is that women are much more
likely to volunteer when asked.

1285
01:03:42,500 --> 01:03:44,490
Now, that is an
interesting fact.

1286
01:03:44,490 --> 01:03:47,090
But it doesn't necessarily
let us disentangle the demand

1287
01:03:47,090 --> 01:03:50,610
and supply explanation that
I showed you previously.

1288
01:03:50,610 --> 01:03:52,550
So in addition
then, the authors do

1289
01:03:52,550 --> 01:03:54,110
a lot of experiments
in which they

1290
01:03:54,110 --> 01:03:58,190
do what's called the
threshold public good scheme.

1291
01:03:58,190 --> 01:04:02,390
And this is sort of very much
trying to mirror or resemble

1292
01:04:02,390 --> 01:04:03,890
the reality.

1293
01:04:03,890 --> 01:04:05,810
In this game, a small
group that needs

1294
01:04:05,810 --> 01:04:08,180
to find a volunteer for a task.

1295
01:04:08,180 --> 01:04:12,320
And participants are anonymously
matched into groups of three.

1296
01:04:12,320 --> 01:04:16,580
They're randomly every
match for each of 10 rounds.

1297
01:04:16,580 --> 01:04:19,010
The game is set up
such that everyone

1298
01:04:19,010 --> 01:04:21,050
prefers that the task
be undertaken by someone

1299
01:04:21,050 --> 01:04:22,970
other than themselves.

1300
01:04:22,970 --> 01:04:25,500
People get two minutes to
decide whether to invest.

1301
01:04:25,500 --> 01:04:27,920
And so the task is essentially
you have to click a button.

1302
01:04:27,920 --> 01:04:29,780
And then you invest the money.

1303
01:04:29,780 --> 01:04:31,270
Only one person can invest.

1304
01:04:31,270 --> 01:04:33,290
And the round ends
when somebody invests.

1305
01:04:33,290 --> 01:04:36,560
If no one invests, group
members all earn $1.

1306
01:04:36,560 --> 01:04:39,800
If a person invests,
that person earns

1307
01:04:39,800 --> 01:04:43,970
$1.25 and the remaining
group members earn $2.

1308
01:04:43,970 --> 01:04:46,040
The clock ticks down
until one person invests

1309
01:04:46,040 --> 01:04:48,235
or no investment is
made in two minutes.

1310
01:04:48,235 --> 01:04:49,610
So the game is
essentially set up

1311
01:04:49,610 --> 01:04:53,135
in a way such that everybody
wants that somebody--

1312
01:04:53,135 --> 01:04:54,950
and it's clear that
somebody should invest.

1313
01:04:54,950 --> 01:04:57,158
But nobody wants to invest
because if you invest, you

1314
01:04:57,158 --> 01:04:58,017
only $1.25.

1315
01:04:58,017 --> 01:05:00,350
But if you don't invest or
somebody else invests and you

1316
01:05:00,350 --> 01:05:03,290
don't invest, you get $2.

1317
01:05:03,290 --> 01:05:05,780
Now in this game, women
are significantly more

1318
01:05:05,780 --> 01:05:06,717
likely to invest.

1319
01:05:06,717 --> 01:05:09,050
We can essentially see the
probability of investing here

1320
01:05:09,050 --> 01:05:10,730
on the left side by round.

1321
01:05:10,730 --> 01:05:13,310
And the red line, which is
the upper line, the woman

1322
01:05:13,310 --> 01:05:15,507
line, the female line
is way, way higher

1323
01:05:15,507 --> 01:05:16,590
compared to the male line.

1324
01:05:16,590 --> 01:05:18,140
So there's much
higher probability

1325
01:05:18,140 --> 01:05:19,930
of investing in this game.

1326
01:05:19,930 --> 01:05:22,610
Now why are women
more likely to invest?

1327
01:05:22,610 --> 01:05:24,050
Is it the case
that women believe

1328
01:05:24,050 --> 01:05:26,660
that their cooperation is
necessary for an optimal group

1329
01:05:26,660 --> 01:05:29,300
decision, but men believe
that their cooperation isn't

1330
01:05:29,300 --> 01:05:29,820
required.

1331
01:05:29,820 --> 01:05:33,020
So it could be that women
think that, well, somebody--

1332
01:05:33,020 --> 01:05:34,937
if I don't do it,
nobody else does it.

1333
01:05:34,937 --> 01:05:37,520
And then men essentially think,
well, somebody else can do it,

1334
01:05:37,520 --> 01:05:40,560
in particular, women could do
it if they're in the group.

1335
01:05:40,560 --> 01:05:44,970
So then the authors do the
same experiment again using

1336
01:05:44,970 --> 01:05:51,300
single sex sessions where
essentially all three

1337
01:05:51,300 --> 01:05:54,300
people in the session are female
or all people in the session

1338
01:05:54,300 --> 01:05:55,495
are male.

1339
01:05:55,495 --> 01:05:57,120
And then what they
find is essentially,

1340
01:05:57,120 --> 01:06:01,050
interestingly, these differences
in gender now go away entirely.

1341
01:06:01,050 --> 01:06:05,340
There's no gender differences
in same sex sessions anymore.

1342
01:06:05,340 --> 01:06:09,300
So just to summarize,
so experiment one,

1343
01:06:09,300 --> 01:06:12,147
when there's a
mixed gender, women

1344
01:06:12,147 --> 01:06:13,230
are more likely to invest.

1345
01:06:13,230 --> 01:06:15,150
And experiment
two, women and men

1346
01:06:15,150 --> 01:06:17,280
are equally likely to invest.

1347
01:06:17,280 --> 01:06:19,980
Now, one question you
might ask is, well,

1348
01:06:19,980 --> 01:06:22,410
is it the case that
really, isn't this really

1349
01:06:22,410 --> 01:06:23,170
about beliefs?

1350
01:06:23,170 --> 01:06:26,290
So it could be that
there's no, in fact,

1351
01:06:26,290 --> 01:06:28,290
no differences in the
preferences for investing.

1352
01:06:28,290 --> 01:06:31,050
So it's not like women are
necessarily nicer than men.

1353
01:06:31,050 --> 01:06:33,330
But, rather, there are
differences in beliefs

1354
01:06:33,330 --> 01:06:35,242
that the women will invest.

1355
01:06:35,242 --> 01:06:37,200
in particular, if there's
a woman in the group,

1356
01:06:37,200 --> 01:06:40,080
everybody else will think, oh,
the woman is going to invest.

1357
01:06:40,080 --> 01:06:43,530
And the woman might think like,
well, if others don't invest

1358
01:06:43,530 --> 01:06:45,648
or others will not
invest, and she

1359
01:06:45,648 --> 01:06:47,190
might believe that,
she might have to

1360
01:06:47,190 --> 01:06:50,460
or think that she needs to
do that herself, otherwise

1361
01:06:50,460 --> 01:06:53,890
everybody will end up
with $1 and less money.

1362
01:06:53,890 --> 01:06:56,970
So now there's a
third experiment here

1363
01:06:56,970 --> 01:07:01,500
where people are ask
to essentially pick

1364
01:07:01,500 --> 01:07:04,080
whom they would like to ask.

1365
01:07:04,080 --> 01:07:06,550
And so there's four
people per group.

1366
01:07:06,550 --> 01:07:07,740
And three people can invest.

1367
01:07:07,740 --> 01:07:09,240
And the incentives
are [INAUDIBLE]..

1368
01:07:09,240 --> 01:07:11,323
These are like-- they're
called the green players.

1369
01:07:11,323 --> 01:07:14,100
And one person is
unable to invest,

1370
01:07:14,100 --> 01:07:17,400
but asks one of the
three to invest.

1371
01:07:17,400 --> 01:07:20,010
Importantly, the requests
are not even binding.

1372
01:07:20,010 --> 01:07:20,970
That's the red player.

1373
01:07:20,970 --> 01:07:23,400
The red player, essentially,
once the [INAUDIBLE] happens,

1374
01:07:23,400 --> 01:07:25,890
you incentivize
that that happens.

1375
01:07:25,890 --> 01:07:27,760
So this is kind of
what this looks like.

1376
01:07:27,760 --> 01:07:29,760
You would get like three
different players here.

1377
01:07:29,760 --> 01:07:32,977
And you can essentially decide
who you would like to trust.

1378
01:07:32,977 --> 01:07:35,310
And it's set up, of course,
that you can see the gender.

1379
01:07:35,310 --> 01:07:38,470
Some are male and
some are female.

1380
01:07:38,470 --> 01:07:41,520
Now women are way more
likely to be asked.

1381
01:07:41,520 --> 01:07:44,220
This is the total
times asked to invest

1382
01:07:44,220 --> 01:07:45,688
for male and female players.

1383
01:07:45,688 --> 01:07:46,980
This is the relative frequency.

1384
01:07:46,980 --> 01:07:49,200
And essentially, you can
see their distribution here

1385
01:07:49,200 --> 01:07:54,090
is way shifted to the right
for women compared to men.

1386
01:07:54,090 --> 01:07:57,030
Now is it better to ask a woman?

1387
01:07:57,030 --> 01:07:59,790
Well, absent a request, the
investment rate actually

1388
01:07:59,790 --> 01:08:01,680
does not differ by gender.

1389
01:08:01,680 --> 01:08:05,170
But when asked to invest, women
are more likely to comply.

1390
01:08:05,170 --> 01:08:08,040
So when the woman is asked
to invest, 76% of women

1391
01:08:08,040 --> 01:08:10,440
invest compared to
14% when not asked.

1392
01:08:10,440 --> 01:08:13,350
In contrast, men are also more
likely to invest when asked,

1393
01:08:13,350 --> 01:08:16,930
but it's only 51$ compared
to 14% when not asked.

1394
01:08:16,930 --> 01:08:20,700
So the marginal increase of
being asked is higher for women

1395
01:08:20,700 --> 01:08:21,731
than for men.

1396
01:08:21,731 --> 01:08:23,939
So if you're thinking about,
like whom should we ask?

1397
01:08:23,939 --> 01:08:25,890
Well, you kind of
want to ask a woman

1398
01:08:25,890 --> 01:08:30,300
because a woman is more likely
to comply with your request.

1399
01:08:30,300 --> 01:08:32,430
And so now if
women-- since women

1400
01:08:32,430 --> 01:08:37,723
are more likely to be expected
to say yes, they also ask more.

1401
01:08:37,723 --> 01:08:40,140
So in that sort of sense, you
know, the gender differences

1402
01:08:40,140 --> 01:08:43,979
get amplified by increased
demand for women to contribute.

1403
01:08:43,979 --> 01:08:46,359
Well, particularly, if there's
a man and a woman where

1404
01:08:46,359 --> 01:08:48,750
both can think about like
one of them could do it,

1405
01:08:48,750 --> 01:08:51,090
the man will think, well,
the woman will do it anyway.

1406
01:08:51,090 --> 01:08:53,507
The woman will think, well,
the man is not going to do it.

1407
01:08:53,507 --> 01:08:55,720
And he's going to think
that I will do it anyway.

1408
01:08:55,720 --> 01:08:58,740
And so then the woman
ends up doing it more.

1409
01:08:58,740 --> 01:09:02,609
In addition, when there's an
opportunity to ask somebody,

1410
01:09:02,609 --> 01:09:04,500
but asking is kind of
costly and you just

1411
01:09:04,500 --> 01:09:06,390
want to find somebody
who will do it.

1412
01:09:06,390 --> 01:09:07,450
But who will you ask?

1413
01:09:07,450 --> 01:09:08,310
You ask the woman.

1414
01:09:08,310 --> 01:09:13,850
Because she's more likely to
comply, more likely to say yes.

1415
01:09:13,850 --> 01:09:17,340
Let me sort of summarize
what we discussed.

1416
01:09:17,340 --> 01:09:21,020
So first is large gender,
wage, and earnings gaps.

1417
01:09:21,020 --> 01:09:24,564
They have been reduced due
to technological advances

1418
01:09:24,564 --> 01:09:25,939
and other improvements
over time.

1419
01:09:25,939 --> 01:09:30,950
But there are still persistent
gender differences in the US

1420
01:09:30,950 --> 01:09:32,689
and many other countries.

1421
01:09:32,689 --> 01:09:35,120
Bias beliefs and
identity concerns

1422
01:09:35,120 --> 01:09:40,729
play a major role in
explaining these differences.

1423
01:09:40,729 --> 01:09:42,740
In addition, there's
some feedback mechanisms

1424
01:09:42,740 --> 01:09:45,740
between the demand and supply
of non-promotable tasks

1425
01:09:45,740 --> 01:09:48,413
that could be, potentially
quite important.

1426
01:09:48,413 --> 01:09:50,330
Of course, I showed you
only a lab experiment.

1427
01:09:50,330 --> 01:09:51,788
but these could be
really important

1428
01:09:51,788 --> 01:09:53,550
in real-world situations.

1429
01:09:53,550 --> 01:09:55,460
So not only is it
that what matters

1430
01:09:55,460 --> 01:09:58,040
is what women decide to
do, but also what they're

1431
01:09:58,040 --> 01:10:00,650
asked to do by others,
in part as a response

1432
01:10:00,650 --> 01:10:05,060
to their propensity to say
yes or no to certain tasks.

1433
01:10:05,060 --> 01:10:06,620
Better understanding
these issues

1434
01:10:06,620 --> 01:10:08,622
can help us mitigate
the gender gap.

1435
01:10:08,622 --> 01:10:10,330
And we can sort of
then, by understanding

1436
01:10:10,330 --> 01:10:12,122
these issues better
and particular beliefs,

1437
01:10:12,122 --> 01:10:16,970
you might be able to improve
or correct biased beliefs

1438
01:10:16,970 --> 01:10:21,600
and, therefore, close
the gender gap potential.

1439
01:10:21,600 --> 01:10:24,240
What's next in the next
few lectures, lecture 19,

1440
01:10:24,240 --> 01:10:26,220
we'll talk about frames,
defaults, nudges,

1441
01:10:26,220 --> 01:10:27,360
and mental accounting.

1442
01:10:27,360 --> 01:10:29,870
Please read Madrian
and Shea 2001.

1443
01:10:29,870 --> 01:10:31,680
And lecture 20, we'll
look about-- talk

1444
01:10:31,680 --> 01:10:36,060
about malleability and
inaccessability of preferences.

1445
01:10:36,060 --> 01:10:39,930
Please read Ariely
2003 for that.

1446
01:10:39,930 --> 01:10:41,470
That was all I have to say.

1447
01:10:41,470 --> 01:10:43,520
Thank you so much.