00:00:05
and so for the group of us uh at a very
00:00:07
high level we're interested in
00:00:08
understanding the relationship between
00:00:10
the trade-offs people are willing to
00:00:11
choose to make and the trade-offs that
00:00:13
determine their happiness or a little
00:00:15
more practically we're interested in
00:00:16
thinking about situations where you have
00:00:18
data on what makes people happy but you
00:00:20
don't have data on what people choose
00:00:22
and you want to try to use this
00:00:23
happiness data to forecast things about
00:00:25
their their decision- Mak process uh so
00:00:27
to give you a little bit of background
00:00:28
uh there's an increasing practice in
00:00:30
economics of trying to use happiness
00:00:32
data to augment uh typical choice-based
00:00:34
analysis uh so to give you a you
00:00:37
particular example let's say we were
00:00:38
trying to think about how we should uh
00:00:40
value a public policy and figure out how
00:00:42
much a particular person would be
00:00:43
willing to pay to have that public
00:00:45
policy in place so one way of
00:00:47
approximating this is by taking uh
00:00:49
existing happiness data estimating how
00:00:52
policies like this affect happiness in
00:00:54
the past estimating how money plays into
00:00:56
determining happiness and then figuring
00:00:57
out the trade-off between these two
00:00:59
things and we could use that to
00:01:01
basically determine uh the amount of
00:01:03
money a person would be willing to pay
00:01:04
to put that policy in place under the
00:01:06
assumption that choices and happiness
00:01:08
are aligned right so that people are
00:01:10
choosing whatever makes them most happy
00:01:12
so that's the practice that's currently
00:01:13
being followed uh by by some economists
00:01:16
and we're basically interested in
00:01:17
assessing that trying to figure out if
00:01:18
these two trade-offs really do reflect
00:01:20
one another uh so to get at that we ran
00:01:22
a large scale study of medical students
00:01:24
as they were participating in the
00:01:26
medical residency match uh so in case
00:01:28
you're unfamiliar with the training of
00:01:30
doctors in the United States uh after uh
00:01:33
young doctors uh graduate from medical
00:01:35
school they go through a period of
00:01:37
several years uh where they get really
00:01:39
intensive Hands-On training in their
00:01:40
specialty called a medical residency
00:01:42
they have to go through an elaborate
00:01:43
process to match from their medical
00:01:45
school to a residency and so at this
00:01:47
point you're probably asking well why
00:01:48
are we talking about medical students
00:01:50
now and we were just talking about
00:01:51
happiness and happiness data and the
00:01:53
reason why we're looking at this
00:01:55
particular setting is because it has
00:01:56
these really nice properties that allow
00:01:58
us to get really high quality Cho data
00:02:00
side by side with really high quality
00:02:02
happiness data so in order to go through
00:02:04
this matching process I just described
00:02:07
medical students have to basically uh go
00:02:08
through a period of interviewing with
00:02:10
medical schools and consider really
00:02:11
carefully the trade-offs they're willing
00:02:13
to make about say The Prestige of a
00:02:15
school versus a city location or things
00:02:17
like this and then determine their
00:02:19
choice ordering over those schools and
00:02:21
say this is my first choice school this
00:02:22
is my second choice school and so on and
00:02:25
actually make a list like that and
00:02:26
submit it to a centralized matching
00:02:28
agency and that list is to determine the
00:02:30
final assignment of where everyone will
00:02:32
go uh and that mechanism was designed
00:02:34
very carefully to make sure that
00:02:35
students uh it's in their own best
00:02:37
interest to report their preferences
00:02:39
truthfully so basically we'll be
00:02:41
piggybacking off of that existing field
00:02:43
mechanism we'll be seeing their choice
00:02:44
ordering uh and pairing that up with
00:02:46
excellent survey data about how happy
00:02:49
they think they'll be at these different
00:02:50
options and the trade-offs they're
00:02:51
making as they're making this decision
00:02:53
so to generate that happiness data I
00:02:55
just described we ran a large large
00:02:57
scale survey of medical students as they
00:02:59
participated in the 2012 residency match
00:03:03
uh so in the leadup to that residency
00:03:05
match I contacted basically all the
00:03:06
medical schools in the United States uh
00:03:09
and uh you know due to my uh my my uh
00:03:12
requests I got 23 schools to agree to
00:03:14
participate and at those schools
00:03:16
students going through the matching
00:03:17
process uh were given the opportunity to
00:03:19
participate in a web survey where they
00:03:21
could uh basically report their top four
00:03:23
choices we get to see their choice
00:03:25
ordering over them we also get to ask
00:03:27
them how happy do you think you would be
00:03:28
if you went to this res how happy do you
00:03:30
think you would be if you went to this
00:03:31
other residency and also for these
00:03:33
residencies we collect all the features
00:03:34
that they're trading off when they're
00:03:35
making this decision things like
00:03:37
Prestige or stady quality or how much
00:03:39
their spouse cares about being in that
00:03:40
location things like that so with this
00:03:43
data we now have everything we need to
00:03:45
basic uh to basically put happiness data
00:03:47
and choice data side by side and uh kind
00:03:50
of compare the analysis that you could
00:03:51
do with with each of these these types
00:03:53
of data so our main results from this uh
00:03:55
are basically twofold so the first main
00:03:57
one is that uh Happy data is actually
00:04:00
reasonably useful for forecasting
00:04:02
choices in this setting uh so if you
00:04:04
were to say know that some resident or
00:04:06
some future medical Resident was
00:04:08
considering between two options we don't
00:04:10
know which one he chooses but we just
00:04:11
know that one of them uh he thinks will
00:04:13
make him happier 70 to 80% of the time
00:04:16
that's the one he will choose um now on
00:04:19
one hand you know we know that people
00:04:20
like to be happy right that's not that
00:04:21
shouldn't be surprising but on the other
00:04:23
hand this is a particular setting where
00:04:24
we don't really think decisions are are
00:04:27
necessarily made just to be happy right
00:04:29
no one's going to to a medical residency
00:04:30
to have fun uh this is a big investment
00:04:33
and important decision in their lives
00:04:34
and even in this situation uh happiness
00:04:36
data is very useful for forecasting so
00:04:38
that's you know a positive uh spin on
00:04:41
how to use happiness data in economic
00:04:42
settings there's a bit of a negative
00:04:44
side uh in our data too that's the
00:04:46
second main result and that's if you try
00:04:47
to use happiness data to infer the
00:04:49
tradeoffs people are willing to make so
00:04:51
again like how how they tradeoff
00:04:53
Prestige against City quality for
00:04:55
example we find that the trade-offs you
00:04:57
would estimate from happiness data are
00:04:59
pretty dramatic rically different than
00:05:00
the trade-offs you would estimate from
00:05:01
Choice data uh these different factors
00:05:04
wait very different uh very differently
00:05:06
in determining these two things uh and
00:05:08
that's a bit of a problem for a lot of
00:05:09
economic analysis because for uh you
00:05:12
know many questions economists are
00:05:13
asking understanding tradeoffs is really
00:05:15
the key thing we're trying to understand
00:05:17
how we how we price various things you
00:05:18
know how we trade off one attribute of a
00:05:20
of a good versus another things like
00:05:22
that and for these types of questions
00:05:24
happiness data is not getting us the
00:05:26
kind of answers we need um so kind of
00:05:28
overall our main goal in this entire
00:05:30
project was trying to get a better sense
00:05:32
of how to how to use happiness data in
00:05:34
economic applications and how far you
00:05:36
can go with that and we found some
00:05:38
positive results that it is actually
00:05:39
useful for just raw forecasting of
00:05:41
choices uh but some negative results
00:05:43
that it doesn't do a great job in
00:05:45
answering nuanced questions about how
00:05:46
you trade off different attributes of a
00:05:48
of an option you
00:05:52
face so one way or one dimension where
00:05:55
you can see differences in the way
00:05:56
trade-offs are made across happiness
00:05:58
data and across Choice data is looking
00:06:00
at the importance of say considerations
00:06:02
about your family's wellbeing or your
00:06:03
spouse's well-being uh so if we compare
00:06:06
how important uh you know say your
00:06:08
spouse's happiness is in determining
00:06:10
your choices it's actually dramatically
00:06:12
more important uh in determining your
00:06:15
choices than it is in determining your
00:06:16
predictions about how happy you'll be so
00:06:18
if we're thinking about trading off say
00:06:20
going to a more prestigious residency in
00:06:22
a location that your spouse likes less
00:06:24
you're more likely to choose that and
00:06:25
you're more likely to wait your spouse's
00:06:27
uh your spouse's opinion more heavily in
00:06:29
your choice then you would in your
00:06:30
forecast about how happy you'll be in
00:06:32
the future so in some sense this could
00:06:33
be evidence that people are willing to
00:06:35
uh to sacrifice their own happiness uh
00:06:37
you know to to benefit their spouse in
00:06:39
these kind of
00:06:43
decisions I came at this problem from
00:06:46
the point of view of an economist but it
00:06:47
does have a lot of implications for how
00:06:49
people uh conduct you know General
00:06:50
exercises and marketing so uh reasonably
00:06:53
commonly uh you know both economists and
00:06:55
marketers find themselves in situations
00:06:57
where we have lots of data about say
00:06:59
customer satisfaction consumer
00:07:00
satisfaction happiness things like that
00:07:02
and we're trying to infer from that how
00:07:04
people you know value different
00:07:05
attributes of a product we're trying to
00:07:06
sell uh how people make trade-offs in
00:07:09
various economic environments and the
00:07:11
results I was discussing for economists
00:07:12
translate immediately to the kind of
00:07:14
same kind of decisions in marketing
00:07:16
environments so basically if you're
00:07:17
trying to forecast what your consumers
00:07:19
will choose or what they will like our
00:07:21
results suggest that happiness data can
00:07:23
help you make those kind of forecasting
00:07:25
uh forecasts accurately however uh if
00:07:27
you're trying to infer more nuanced
00:07:29
question questions about how say
00:07:31
customers are trading off different
00:07:32
attributes of a product you're trying to
00:07:33
sell our results suggest that happiness
00:07:35
data doesn't really get you all the way
00:07:37
there to estimating those kind of uh
00:07:42
tradeoffs so I think this study speaks
00:07:45
to uh to to countering misperceptions on
00:07:47
two ends of a spectrum right on on one
00:07:48
end of the spectrum uh some people
00:07:50
believe that happiness is basically
00:07:52
everything that maximizing happiness is
00:07:55
the ultimate goal of all of our actions
00:07:57
and our research suggests that at least
00:07:58
when we're thinking about how happiness
00:08:00
is currently measured in surveys and
00:08:01
things like that that's not the case
00:08:03
that happiness is something that's very
00:08:04
important to people and it's an
00:08:05
important goal they're pursuing uh but
00:08:07
people will explicitly trade off the
00:08:09
pursuit of happiness against uh you know
00:08:11
to to pursue other goals uh now on the
00:08:13
other end of the spectrum I think some
00:08:15
people believe that happiness is not
00:08:17
particularly informative for economic
00:08:18
analysis uh this is a view held by by
00:08:20
many economists this is sort of a
00:08:22
frivolous psychological variable that
00:08:24
isn't really fundamentally related to
00:08:25
how we make choices and our results
00:08:28
actually suggest that's not quite right
00:08:29
either um so even in this setting that
00:08:32
is unambiguously you know a very serious
00:08:34
decision a very important decision not
00:08:35
made for fun by any by any stretch of
00:08:37
the imagination uh understanding or
00:08:39
having access to happiness data really
00:08:41
helps us understand how the decision
00:08:42
process is made and helps us forecast
00:08:44
the choices people will
00:08:48
make so there are a few things that set
00:08:50
my research apart from uh other analysis
00:08:53
done looking at the alignment between
00:08:54
choice and happiness and I think the
00:08:56
main thing is we're coming at it from a
00:08:57
little bit of a different attitude uh so
00:08:59
there's a great deal of research out
00:09:01
there demonstrating that people don't
00:09:02
always choose what'll make them happy
00:09:04
but the way that's typically discussed
00:09:06
and the way that's typically presented
00:09:07
is as evidence that people are making
00:09:09
some sort of mistake or or they're bad
00:09:11
at forecasting so the idea is that they
00:09:13
would they're trying to to maximize
00:09:15
their happiness and the only thing
00:09:16
that's stopping them is either they they
00:09:18
messed up the decision somehow or they
00:09:21
they guessed wrong about what would make
00:09:22
them most happy and so on and thus we
00:09:24
can attribute the entire gap between
00:09:26
Choice data and happiness data just to
00:09:27
these mispredictions
00:09:29
we were coming at it from a little bit
00:09:30
of a different a different kind of Bas
00:09:32
background we were thinking well okay
00:09:34
you know maybe mispredictions are
00:09:36
important but also it could be that
00:09:37
people aren't even necessarily trying to
00:09:39
maximize happiness maybe they could just
00:09:41
be thinking of it as one of many goals
00:09:42
they're pursuing and explicitly trade
00:09:44
off that goal uh against other goals and
00:09:47
so this led us to look at a situation
00:09:49
where people are making uh a decision
00:09:51
that's very considered very deliberated
00:09:53
very high stakes where we're not
00:09:55
particularly worried about mistakes
00:09:56
driving any deviations we see
00:09:59
uh and basically trying to see if we
00:10:01
still see a wedge in that setting uh and
00:10:03
of course we did and by by looking in
00:10:05
this particular type of environment we
00:10:07
were better able to get at uh how much
00:10:09
of this wedge between choice and
00:10:11
happiness is driven by people's initial
00:10:13
intentions versus just their uh their
00:10:15
mispredictions about what actually makes
00:10:17
them
00:10:21
happy in terms of what's next we're
00:10:23
thinking about continuing this line of
00:10:24
research uh by continuing to investigate
00:10:26
basically the relationship between
00:10:28
happiness data
00:10:29
uh and choice data which is the more
00:10:31
typical object in economic analysis we
00:10:33
want to continue to think about uh how
00:10:35
much you can infer about how choices are
00:10:36
made based only on happiness data uh a
00:10:40
particular Dimension that I'm interested
00:10:41
in is in trying to build build happiness
00:10:43
data into a more standard price
00:10:45
theoretic economic analysis so rather
00:10:47
than thinking about happiness as just a
00:10:49
way of approximating the thing people
00:10:50
are trying to maximize think of it as
00:10:52
this kind of abstract good that people
00:10:54
are willing to uh you know in some sense
00:10:56
buy and also trade uh and explicitly
00:10:58
trade off against other goals in their
00:11:00
life and uh we're starting to work on a
00:11:02
theoretical approach to uh to modeling
00:11:04
that and thinking about how we can uh
00:11:06
better use that type of framework to
00:11:08
import uh psychological data and
00:11:10
happiness data into economic analysis
00:11:16
[Music]
