00:00:01
i'm here at wharton professor Joe
00:00:03
Harrington to talk about his paper what
00:00:06
can a duration of discovered cartels
00:00:08
tell us about the duration of all
00:00:10
cartels welcome professor I'm glad to be
00:00:14
here so tell us about your paper well it
00:00:17
starts with the issue of collusion and
00:00:19
the probably was best stated by recently
00:00:23
deceased Justice Antonin Scalia who
00:00:25
referred to collusion as the supreme
00:00:27
evil of antitrust so collusion is all
00:00:30
about the fact that firms which we count
00:00:32
upon to compete for the business of
00:00:33
customers and in doing so result in
00:00:36
lower prices better products and the
00:00:37
like in some industries they decide that
00:00:40
that leads to profits that are too low
00:00:42
and so what they do is they engage an
00:00:44
unlawful coordination of their behavior
00:00:46
they decide to set of driving prices
00:00:49
down in order to get customers business
00:00:52
a sigh well well how about we all set
00:00:54
artificially high prices share the
00:00:56
market and we'll all earn high profits
00:00:58
and so collusion is a real challenge has
00:01:01
become particularly so in the last
00:01:02
several decades I would say 30 years ago
00:01:05
if you asked an official the Antitrust
00:01:07
Division the Department of Justice they
00:01:08
whether there are any global cartels
00:01:10
other than the type like OPEC they would
00:01:13
have said no but the fact is we observed
00:01:15
many global cartels as well as many
00:01:17
domestic cartels in in the last couple
00:01:20
decades so a challenge in terms of
00:01:23
understanding cartels you know asking
00:01:26
well how many unlawful cartels are there
00:01:28
out there you know how bad are they how
00:01:30
much do they raise prices how long do
00:01:32
they last we face a challenge which is
00:01:36
common to criminal behavior which is
00:01:38
these criminals want to hide themselves
00:01:40
so what we observe are just those
00:01:44
cartels that are unfortunate enough to
00:01:46
be discovered and convicted so a
00:01:49
challenge gets to be which is well we
00:01:51
know how many discover cartels are out
00:01:53
there we know how long they last we know
00:01:55
how high a price they said but is that
00:01:58
representative of the actual latent
00:02:00
population of all cartels so where this
00:02:03
again kind of specifically for example
00:02:04
manifest self in terms of a policy
00:02:06
challenge is suppose the Antitrust
00:02:09
Division puts in place a new program and
00:02:11
they want to know is it kind of is it
00:02:14
proving to be beneficial
00:02:15
is it actually helping to reduce the
00:02:17
presence of collusion well all we can
00:02:19
observe is what's happening to the kind
00:02:23
of population that of discover cartels
00:02:25
so if for example we observe more
00:02:28
discover cartels is that because the
00:02:30
program maybe is working to make
00:02:32
discovery more likely and that's good or
00:02:34
maybe is counterproductive maybe what's
00:02:37
doing is actually creating more cartels
00:02:38
and that's why there's more discover
00:02:40
cartels so that's kind of the starting
00:02:42
point to this research which is trying
00:02:45
to determine what is it what it is we
00:02:47
can learn about the underlying latent
00:02:50
you know Universal cartels from those
00:02:52
that we actually discover and
00:02:54
specifically what the research is
00:02:56
concerned with is the issue of the
00:02:58
duration of cartels so we can measure
00:03:01
and there have been many studies that
00:03:02
have measured how long discovered
00:03:04
cartels last these studies if the
00:03:07
average duration ranges from five to
00:03:09
eight years depending on the study and
00:03:11
the issue as well as that kind of a good
00:03:13
proxy for the duration of all cartels
00:03:16
and so what this research is designed to
00:03:18
do is to try to provide a kind of a
00:03:22
method to kind of getting at you know
00:03:24
how much bias there might be in looking
00:03:27
at discover cartels with regards to the
00:03:30
universe of cartels so what are some key
00:03:33
takeaways of the paper well the the
00:03:36
paper kind of to answer that question
00:03:38
they say that the paper kind of big stew
00:03:40
contributions first what it does is it
00:03:42
puts forth a theoretical framework
00:03:44
that's kind of think through these
00:03:45
issues and understand well when will
00:03:48
bias occur in what direction and what in
00:03:50
the way it does this it constructs a
00:03:52
model of the birth death and discovery
00:03:55
of cartels and and one of the things
00:03:57
that it shows and this is in the context
00:04:00
of if you read many papers and the
00:04:02
literature some economists will say that
00:04:04
well the measured duration of discover
00:04:08
cartels is an over estimate of the
00:04:10
actual true duration and some will say
00:04:12
it's an underestimate and so what we
00:04:14
want to do is put forth a theoretical
00:04:16
framework where we can say precisely you
00:04:17
know when is it an over under estimate
00:04:19
what drives that and so what we find is
00:04:21
that if cartels differ a lot in terms of
00:04:26
their the likelihood that they'll
00:04:28
collapse
00:04:29
once again think about the death here of
00:04:30
a cartel you know Cardinals born it
00:04:33
could just internally collapse as we've
00:04:35
observed with sometimes lawful cartels
00:04:37
like OPEC which is right now now that
00:04:40
you know not very functional of and so
00:04:42
it could just die that way it could be
00:04:44
discovered and its conviction would lead
00:04:47
to its collapse so we can think about
00:04:49
cartels is differing in these two key
00:04:51
characteristics what's the likelihood of
00:04:53
collapsing at say in a year what's
00:04:55
likely to being discovered in a year if
00:04:57
cartels tend to vary really largely in
00:05:01
terms of the likelihood of collapse what
00:05:04
we find is that the measured duration of
00:05:06
discover cartels is an over estimate of
00:05:09
the true duration and the reason the
00:05:11
intuition is that if you have a cartel
00:05:13
that's really very stable with a low
00:05:15
chance to collapse it just hangs around
00:05:17
for a long time and has plenty of
00:05:20
opportunities to be discovered cartels
00:05:22
that are short live will tend to die and
00:05:25
avoid discovery so so the bias in that
00:05:28
case will work towards saying that well
00:05:30
if we measure for example the average
00:05:33
duration discovered cartels to be lets
00:05:35
say six years that that probably is an
00:05:37
over estimate now there's other
00:05:40
assumptions you can make whereby you
00:05:42
give buys go in the other direction so
00:05:44
but but one contribution and takeaways
00:05:46
is to be able to kind of frame these
00:05:48
issues so we can understand okay what
00:05:50
drives bias the other contribution and
00:05:54
more where I think the the key takeaway
00:05:55
is is to then uses framework to get a
00:05:58
measure about the extent of this bias
00:06:00
how big is it so we take this framework
00:06:03
and we use data on convicted cartels by
00:06:07
the Antitrust Division from say 1961 to
00:06:11
cartels that were born from 61 to 80 4
00:06:13
and we can use that data to get a map
00:06:17
first of all if you just you could just
00:06:18
measure the average duration of these
00:06:20
convicted cartels and it's a five point
00:06:22
eight years but we can use this
00:06:24
framework then get a sense of how big is
00:06:26
the bias with regards to the duration of
00:06:29
action all the cartels those also those
00:06:32
that are undiscovered and what we find
00:06:33
is that and I think one of the
00:06:35
surprising kind of results is that the
00:06:37
bias is actually is not that large we
00:06:38
find kind of a kind of high likelihood
00:06:41
the duration
00:06:43
average duration of all cartels probably
00:06:45
ranges from about five point two to six
00:06:47
point eight years right so what are some
00:06:50
practical implications of your research
00:06:52
I think probably the most subsidy one is
00:06:55
regards to evaluating policy you know
00:06:58
it's a I think there's been a lot of
00:07:00
policy innovations by various
00:07:02
competition authorities one of the big
00:07:04
challenges is determining well are they
00:07:07
helping in the fight against cartels you
00:07:09
know are they resulting in fewer cartels
00:07:11
reducing the overcharge resulting in
00:07:13
lower duration and I said the challenge
00:07:16
here is that we don't directly measure
00:07:17
those things so it's not like with some
00:07:20
types of crime where we can measure the
00:07:23
crime rate because all the crimes are
00:07:25
reported in case the cartels it is
00:07:28
criminal but those who are harm don't
00:07:31
necessarily know that they've been
00:07:32
harmed they don't know that they paid
00:07:34
excessively for this good so so that's a
00:07:37
real policy challenge and they could be
00:07:39
more concrete about this probably the
00:07:42
biggest innovation in policy in fighting
00:07:45
cartels in the last 30 years has been
00:07:46
the revision of the corporate leniency
00:07:48
programme by the Antitrust Division so
00:07:51
this is kind of an age-old idea with
00:07:53
conspiracies which is that if someone
00:07:57
from the conspiracy comes forward
00:07:58
cooperates with the authorities they'll
00:08:00
be absolved the government penalties
00:08:01
okay so this was a program that was in
00:08:03
place starting in nineteen seventy eight
00:08:05
but it had certain design flaws to it
00:08:07
was rarely used 93 it was revised I was
00:08:11
made kind of you know kind of
00:08:12
structurally sound and immediately the
00:08:14
leniency application started coming in
00:08:16
so it's been certainly successful as
00:08:19
measured by the number leniency
00:08:21
applications it's been adopted by many
00:08:23
other jurisdictions probably more than
00:08:26
50 countries and unions have a leniency
00:08:27
programme but still the question is well
00:08:31
but is it affecting the actual
00:08:33
population of cartels and that's a
00:08:34
difficult question so one of the things
00:08:37
we want to do here is with this kind of
00:08:40
approach is it can provide a method to
00:08:43
try to measure the impact of a program
00:08:46
such as the leniency programme to
00:08:48
specifically address the question of has
00:08:51
it for example resulted in cartels be
00:08:54
enough shorter duration that would be an
00:08:56
immediate
00:08:56
benefit from that and so so that's
00:08:59
something we plan to do kind of in the
00:09:01
next step of this research project we've
00:09:02
looked at using this model to measure
00:09:07
the underlying duration of cartels with
00:09:10
data prior to the revision leniency
00:09:12
programme and that was somewhat do to
00:09:14
kind of data limitations but a new data
00:09:16
set has recently become available and
00:09:18
that will encompass cartels that were
00:09:20
born and died after the advent of
00:09:23
leniency programme and so we can redo
00:09:25
the analysis and then see well what has
00:09:27
happened to the duration of cartels kind
00:09:29
of before and after the leniency
00:09:31
programme this won't be able to tell a
00:09:33
kind of causality story but I think it
00:09:35
will provide some kind of kind of needed
00:09:38
data to speak to the issue of well what
00:09:41
impact are these programs having what
00:09:44
conclusions of any surprised you I think
00:09:47
the biggest one was that there wasn't as
00:09:48
much bias as I would have expected I
00:09:50
think that's been a a running concern of
00:09:53
of economists and policymakers in this
00:09:56
context but the bias is definitely there
00:09:59
we definitely find evidence of it but
00:10:01
it's not as large as one would have
00:10:03
thought so so let's Viswa yup I think
00:10:06
some of the conclusions drawn based upon
00:10:09
what were known to be biased estimates
00:10:12
probably most of them are robust because
00:10:14
of the fact that we find the device
00:10:15
isn't actually that large so what sets
00:10:18
your research apart from prior work in
00:10:21
this area well I think this is a topic
00:10:24
which people kind of talked about and
00:10:28
mentioned as they engage in the analysis
00:10:30
saying well okay our data is probably
00:10:33
biased because it's just discovered
00:10:35
cartels and and those could be very
00:10:37
different from the universe of cartels
00:10:39
you could easily imagine that well we're
00:10:42
just discovering the ones that are you
00:10:44
know that very ineffective and that's
00:10:46
why they're being discovered of or as I
00:10:49
said before we could be discovering the
00:10:51
ones that are really stable and
00:10:52
effective and so they're around for a
00:10:54
long time with lots of opportunities to
00:10:56
be discovered that was kind of
00:10:59
recognized and mentioned and then people
00:11:01
just went forward and just kind of did
00:11:04
their work as if the the data was not
00:11:06
biased this is really the first study to
00:11:08
say well let's see what we can actually
00:11:10
say
00:11:10
the extent to the bias and is it really
00:11:13
that large and to what extent can we
00:11:14
kind of compensate for it you know in
00:11:16
our conclusions so how will you follow
00:11:19
up this research well I think right now
00:11:22
what we're going to be doing as I said
00:11:23
it's going to be a follow-up empirical
00:11:25
analysis kind of looking at date using
00:11:27
data from the post leniency programme
00:11:29
era so C would get some sort of
00:11:31
assessment about whether duration has
00:11:32
been impacted by Lindsay program I think
00:11:36
where one can also make advances in
00:11:39
terms of the theoretical framework right
00:11:42
now the framework itself is kind of
00:11:44
sparse in terms of how it models the
00:11:46
birth death and discovery of cartels but
00:11:49
there's kind of a developing literature
00:11:50
in economics on how to how to model
00:11:54
birth you know what determines whether
00:11:56
or not a cartel is created you know what
00:11:57
in kind of inspires managers to say well
00:12:01
let's not compete anymore let's collude
00:12:03
what causes cartels to collapse there's
00:12:06
a kind of a rather large does body work
00:12:09
on that and unless so in terms of what
00:12:11
leads to discovery but I think trying to
00:12:14
take it into account those those
00:12:15
structures get a richer structure to
00:12:18
embed that within our framework so we
00:12:20
can tell kind of a richer story about
00:12:22
the kind of the life cycle cartels from
00:12:24
when they're born and when they die well
00:12:27
thank you very much for joining us it's
00:12:29
a pleasure been in being here
00:12:46
you
