YT YouTube2Text

YouTube2Text → Transcripts → Supplemental lecture: Trade, FDI, and development

Supplemental lecture: Trade, FDI, and development — Transcript

by CEPR & VideoVox Economics · 13,558 words · 2,478 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

1:10my
1:12so let's go ahead and start i'd like to
1:14say good morning good afternoon
1:16good evening and welcome everyone to
1:19this next lecture on fdi
1:23trade and development we're really
1:25delighted to welcome natalia raimondo to
1:27this
1:28steg-sponsored course on key concepts in
1:31macroeconomic development
1:33i think by now probably most of you know
1:35what steg is
1:37it's a program on structural
1:39transformation and economic growth
1:41operated under the auspices of cepr in
1:44london
1:45we gratefully acknowledge the financial
1:47support for the stag program
1:49from the united kingdom's foreign
1:51commonwealth and development office and
1:54the steg program is funded as part of
1:56the uk aid effort
1:58we've been really lucky to have over the
2:00past weeks a series of lectures on
2:03topics in macro development
2:05and we're really delighted to have
2:06natalia with us today
2:09to pick up a topic of
2:12really fundamental importance which is
2:13the role of fdi
2:16in development natalia like the other
2:19lecturers in the course
2:20is essentially donating her time we're
2:23really grateful to you for
2:24for making the effort i'll also thank in
2:27advance the
2:28staff and the team from cepr that have
2:31organized and
2:32made the conference run smoothly so
2:36mandy chan lauren waring kirsty mcneil
2:41nadine clark in past week so i think
2:44she's not here
2:45today just a couple of words on ground
2:47rules
2:48first of all for those of you in the
2:50virtual front row
2:51natalia says she's happy to have you
2:54interrupt her with questions
2:55along the way please don't be shy
2:58i've also mentioned to her that what
3:00we've been doing the past few weeks is
3:02asking those of you in the front row
3:04to put into the chat window just a
3:06little bit of information about
3:08yourselves she may not have a chance to
3:09look at it
3:10but she might catch a glimpse as it goes
3:13scrolling past
3:14if you can just put your name and the
3:16country you're from and the institution
3:17where you're doing your phd
3:19that's a bit of information that gives
3:21her a sense of who's there
3:23but please um if you're willing to keep
3:25your cameras on and to interrupt with
3:27questions when you have them
3:30natalia says that's great if the if the
3:32interruptions get
3:33intense i reserve the right to step in
3:36and
3:37and uh defer things to later for those
3:40of you not in the virtual front row
3:42and you won't control either your
3:44microphone or your camera
3:46the way for you to put questions to
3:48natalia
3:50is to enter them in the q a window she
3:52won't see it if you put your virtual
3:54hand up
3:55but if you enter questions in the q a
3:57joe and i will
3:59do our best to answer the ones we can
4:01and so we'll try to pick that up along
4:03the way
4:05and uh some of those questions if we
4:07think they're better left for natalia
4:09we'll either interrupt her with those
4:11questions or will save them for the end
4:13the last 15 minutes or so i think that's
4:16all i have to say and
4:17on that i'll just say natalia thanks we
4:20really appreciate your
4:21making the time and the effort that goes
4:23into preparing the lecture and
4:24we're really grateful so um over to you
4:28thanks thanks very much for inviting me
4:30no it's like
4:31my pleasure actually so let me um
4:35share here
4:39yep hey
4:42so let me let me start so what this is
4:45going to be a supplemental lecture on
4:47trade foreign direct investment and
4:49development and i would like to start
4:51with
4:51some basic definitions so what is a
4:55multinational income price what i'm
4:57going to call
4:58m e it's a farm with operations in more
5:01than one country
5:02with an ownership stake of 10 or more
5:05so that's the definition so we are going
5:07to talk about parent
5:08in the origin countries of the farms or
5:10headquarters and affiliate farms in the
5:13foreign countries
5:15the other related concept is going to be
5:17foreign direct invest
5:19investment fdi this is a constitute from
5:22the balance of payments
5:24and there is this misleading war
5:26investment this refers to financial
5:28investment so it's a financial flow
5:30and it's differentiated from portfolio
5:32based investment because
5:34we take a threshold of equity stake of
5:3710 percent or more
5:39so you can finance activity of
5:41multinational farms
5:43not only with sdi also with other flows
5:46like debt with local ones
5:48etc and finally the concept that i will
5:51be
5:52carrying along is the one of
5:53multinational production
5:55in the mp that refers to the activity
5:58of parents and affiliates and by
6:00activities i mean like
6:01sales employment output another
6:04like real variables from the from
6:07surveys or financial statements of funds
6:10so it's really the activity that they
6:12perform
6:12in their in the host countries where
6:15they go and operate
6:17so i'm going to be referring by mp by
6:19country i
6:20mean multinationals from an origin
6:23country i
6:24in a country l that is the host
6:26countries where they operate
6:28and for instance in general i will be
6:30using sales
6:32so i will refer to mp by country
6:34inelastics of affiliates belonging to
6:36two parents deny operative so that's
6:39going to be the language
6:42so what's the outline so first i'll show
6:44you some
6:45macro facts facts that we want models to
6:49to replicate then i will
6:52it will be mostly a methodological
6:54lecture on models of trade
6:57and mp so i'm going to go through them
6:59first and melissa and a version of
7:01of that helmet melissipo aer 2004
7:05the ek version with mp this is my work
7:09with andres robbie is claire
7:11then a kind of a synthesis of those two
7:14models together
7:15to talk about innovation production in
7:18the global economy
7:19and that's my work with costa sarcolat
7:22is also undressed for this clear and
7:23steep people
7:24i'll show you some of the subsequent
7:26literature that picked up on these
7:28themes
7:29and i have to say the disclaimer is not
7:31today
7:32we are not going to talk about the
7:33boundaries of the multinational and
7:35contracts that's all
7:36all entrances work and that's the
7:38subject of a
7:40different lecture and a whole lecture
7:42because it's a big literature
7:44then i'm going to talk about spillovers
7:46from an empirical point of view and i
7:48want to just stress the state of
7:50the art in this literature was the
7:52latest on
7:53people trying to measure these
7:54spillovers from foreign to domestic
7:56farms i'm going to talk
7:58briefly about this paper of greenstone
8:00cormac america
8:01and then recent very recent
8:05research on the us by by
8:08ratzen and felix internet and about
8:11costa rica
8:12um by alfa uranium and many lichen
8:15baskets
8:16and a paper also at the end of a pattern
8:18of a historical perspective
8:21if i have time okay what is in the
8:23slides
8:24so let me jump to the facts right away
8:27so the first fact
8:28it's uh one that uh trade economists are
8:31very familiar with is gravity
8:33so mp flows follow gravity
8:37mp shares bilateral empty shares
8:39decrease with distance
8:41that's the left picture and bilateral
8:44lmp flows increase with
8:46gdp in the host country and you can look
8:49at this
8:49pattern many shapes and sizes it's very
8:53robust so we want models to display this
8:57aggregate fact
8:58gravity the second fact is that trade
9:01and np bilateral trade shares and
9:03bilateral
9:04np shares are very strongly and positive
9:06correlated
9:08in the cross-section and also if you
9:09throw the other time
9:11so we want also dc models at the
9:14aggregate level the macro model should
9:15deliver this
9:16this path and finally
9:20there is a surface that comes from firm
9:23level data
9:24and here i show for two countries but
9:26you can do this with
9:28for many countries in the world it's
9:30very robust
9:31it's them any advantage so
9:34if you go to the panel on the right this
9:36is a census data
9:38for the us um the black belt will show
9:42the density on
9:43log employment for domestic funds blue
9:46for exporters
9:47right for multinationals so if you take
9:50the headquarters on multinationals in
9:52the u.s
9:52they're very big in terms of employment
9:56the red belt the spain case
9:59shows the same thing and it's also there
10:01what is there
10:02is also that black belt that shows that
10:06foreign affiliates of
10:07foreign multinationals operating in
10:09spain also
10:11the most productive farms and the
10:12largest in their economy
10:15so their many advantage goes to way in
10:17the home country or also in the host
10:18country
10:21so these three facts and one what i'll
10:23show you next in a minute
10:24will motivate most of the theories
10:26actually
10:29yeah sorry i might miss something so
10:31what's the difference between the
10:33foreign-owned firms and a foreign direct
10:35investor
10:36yeah so basically they are in the same
10:38case
10:39uh are spanish multinationals in spain
10:42or um or like a foreign
10:46multinational that has affiliates in
10:48spain
10:49okay so those are the differences in the
10:52us in contracts
10:53we are only like we have they are the
10:55headquarter of
10:56say general motors but toyota is not
11:00there because it's a foreign affiliate
11:02it's an affiliate of a foreign parent
11:05okay thank you
11:08okay so let me do some taxonomy because
11:11also
11:12this one is going to be useful for what
11:15we will have in the models and what we
11:17won't have
11:18so let me classify affiliate activities
11:21in three
11:22boxes one is what we call horizontal
11:25activities
11:26and these are staged to the host market
11:28where the affiliate operates
11:30that's very important that's one of the
11:32most important um
11:33activities of this many affiliates
11:37the second type of affiliate is export
11:40platform
11:41activities and these are sales outside
11:43the whole
11:44market so general motors goes to brazil
11:48they have an affiliate there and from
11:49there they sell to
11:50all latin america that's export platform
11:53activity
11:55and finally we have vertical activities
11:59that mostly are going to be absent of
12:02the models i'm going to show you
12:03because these are these vertical
12:05activities entail
12:07production being split across locations
12:11and we won't have that actually
12:14and you can see this in the data by um
12:17the sales or trade flows between parents
12:20parent companies and affiliates
12:22in the foreign countries this is
12:25we we document in a previous paper that
12:27for the media and u.s affiliate
12:29this has zero of these flaws with their
12:32parents okay so it's very very important
12:35because they're very concentrated
12:37in very few farms but the majority of
12:39affiliates don't do it
12:41so the models are going to be mainly
12:43about one and two
12:45okay so
12:48this is schematically where we are going
12:50to be so
12:53like really very very schematically what
12:56are the two reasons for
12:57pursuing foreign and international
12:59activities
13:01market access or efficiency we're going
13:03to go the role of market access
13:05in the models that i'm going to show you
13:08in that role
13:10there is a choice between serving those
13:12foreign consumers from home
13:14or relocating production abroad if they
13:17do
13:18it from home we are going to end up with
13:20experts
13:21if they relocate production abroad and
13:24they chose to
13:25own those facilities we are going to end
13:27up with what is called horizontal
13:30mp horizontal fbi i'm going to include
13:33in this
13:34bin also when farms choose to serve
13:37foreign consumers in star markets
13:40operating in a country and then
13:41exporting from there
13:43so this is going to be um this extra
13:46platform activities
13:48um to reach to have market taxes
13:52access to foreign consumers so this
13:55takes us through
13:56one of the first um um
13:59the like seminal theories of
14:01multinationals
14:02that is playing with this trade-off
14:04between choosing
14:06exports or relocating production
14:09and that theory was called the proximity
14:11concentration trails
14:14so the the the dilemma is how to serve
14:18these foreign consumers you can export
14:20from an existing domestic plant or you
14:22can set up a new foreign plant
14:24and in these ways experts and fdi or mp
14:28serve as substitute ways of serving this
14:30foreign market
14:32the trade-off is that mp implies
14:35highs and costs because or fixed cost
14:37because you have to
14:39open a new plan but proximity to
14:42consumer you are there in the host
14:44market where consumers are
14:46exports on the contrary implies um
14:49um are far away from consumers so you
14:52firms have to incur on the trade cost
14:55but they can concentrate production
14:57and take advantage of scale economies in
14:59one location
15:00that's why it's called proximity versus
15:03concentration
15:04trade-offs what's in the data and again
15:07here i'll show you
15:08um it's it's it's a very robust
15:12um evidence at the aggregate level and i
15:16want to stress this so
15:18we observe empirically the proximity
15:20concentration
15:22trade-off at the aggregate level at the
15:24firm level
15:25the readiness is mixed because the data
15:27requirement to test
15:28these these trade-offs as you're going
15:30to see in the theories
15:32are very high so in general
15:35the type of evidence we have is like if
15:38we regress
15:40the log of the ratio of exports to
15:43totals foreign sales
15:44or sales of affiliates in
15:48a foreign market we can do it at the
15:50industry level
15:51for instance we should observe that it
15:54uh decreases with trade costs
15:58right so we we we want to serve foreign
16:01consumers more through affiliates
16:03when trade costs are larger and is
16:06positive related
16:07to some measures of
16:11planned fixed cost in general these
16:13measures are related to production to
16:15non-production
16:16workers or something like that and that
16:19has to be positive
16:20so in general our theories will deliver
16:23at the aggregate level
16:25this trailer
16:29sorry may ask another question so uh now
16:32i
16:32in the index is referring to destination
16:36country right
16:37j is yeah industry yeah only here i'm
16:41going to use that notation
16:43okay so yes because in general the data
16:48most of the evidence is for the u.s we
16:50have detailed data to clean the
16:52to about about foreign affiliates
16:56obvious multinationals that's that's the
16:59va data
17:00has been the more used for the for this
17:03but you can think of if we had all the
17:05big
17:06made bilateral matrix of the war we
17:08could use any
17:10source country and any host country
17:12ideally
17:14and that's why it's been done by
17:16industry
17:17also this test to gain more
17:20data thank you thank you and these
17:23experts are all from the u.s
17:25right yeah yeah in this case
17:28yeah yes so
17:32let me start with the first mod that is
17:34we're in multinational production
17:35and p into the malicious model and
17:38that's the paper i'll tell you about how
17:40mellitt's people are 2004.
17:44so this model we have we'll start with
17:46the proximity concentration trade-off at
17:48the farm and aggregate 11.
17:51and we have another characteristic that
17:53is um also
17:54um from seminal working multinational
17:57from marcus and
17:5884 and is that firms can transfer their
18:01productivity abroad
18:03and in in marcus's work was called
18:06knowledge capital
18:08and it's the sense that there is
18:09something in the multinational that has
18:11this
18:12public good nature and can be shared
18:15across all the units
18:17okay in our case it's going to be the
18:18productivity parameter
18:20there is more recent evidence that in
18:23fact this seems to be the case
18:27um funds are going to be a terror genius
18:29in productivity so this is the melit
18:31flavor of the
18:32of the model so that we can replicate
18:36the fact that most productive farms are
18:38the ones choosing
18:39um mp and
18:43one of the results of incorporating this
18:45heterogeneity is going to be that
18:47um the response is to this trade-off the
18:50proximity concentration trade-off is
18:52going to be different across firms
18:54so the model will have predictions about
18:57these trade-offs
18:58uh across the industries and it's going
19:00to say something like
19:02industries with more heterogeneity will
19:05have
19:05a more mp relative to x
19:09that's going to be one of the the the
19:12testable
19:12implications of the models they're going
19:14to go and test it but i'm not going to
19:16go there i'm going to show you them all
19:18the main elements of the model
19:22so um the setup has end countries
19:25only labors and this i'm not going to
19:28relax these assumptions uh
19:29throughout continuum of varieties and a
19:32cs aggregator with sigma elasticity
19:36now farm productivity as i said is it's
19:38a melee smelly style
19:40environment so it's going to be um
19:46it's going to be random and it's going
19:49to be from a pareto distribution
19:51with a shape parameter remember
19:54higher kappa lower heterogeneity and
19:57vice versa
19:58so we are required here i did a pareto
20:01with a
20:01lower one of one i mean just to make
20:04notation simpler
20:05and we need a condition about parameters
20:08so
20:09everything is finite okay
20:13then there's going to be monopolistic
20:14competition as improvement
20:16and mallets so fars are going to price
20:20um a markup over
20:23marginal cost and the marginal cost is
20:26going to be just the wage
20:28um over uh productivity
20:32now there is going to be valuable trade
20:34cost to trade
20:35here is from country i2j that's how what
20:38power ij means
20:40and then there's going to be a fixed
20:42cost of exporting
20:44to country j and fx and peacocks
20:48to setting up affiliates in this country
20:51and we can have also domestic fixed
20:54costs and an entry cost
20:56um to to to have this first stage of
20:59entry
20:59a la crosman i'm not going to do it here
21:04so profits are ces
21:08so it's just sales over the elasticity
21:10of substitution
21:12and we will have to compute profits from
21:15sales to the domestic market
21:18to the export market each export market
21:21and and the profits from sales
21:24from mp in a in a house market
21:28so a couple of things you notice
21:32productivity is always the same because
21:34um
21:35firms can transfer their productivity
21:37abroad
21:38um for domestic and exports you pay the
21:42salary in your
21:43in your market because that's where you
21:44produce and if you export you'll have to
21:47at the tao in the case of mp you're
21:49producing that for a market
21:51so you keep track of the wage there and
21:54then the respective costs
21:56are added here so i just assume that the
22:00costs are paid
22:01on on in in labor units in the market of
22:05destination
22:08this model these are linear functions so
22:10this model
22:11has a very simple um kind of rule so we
22:14have zero profit conditions and we find
22:17the marginal farms
22:18that are the difference between doing
22:20one activity or the other
22:22so we will have an uh a marginal
22:24domestic firm
22:26and that i denote by a fee b
22:30a marginal exporter vx to each market
22:35and then a marginal multinational vm
22:39okay and graphically and this graph is
22:42done under
22:43symmetry so it's easy to you can
22:46this is kind of um in in in a graph what
22:50i just show in equations so we have this
22:53linear function of profits
22:55of um productivity to a sigma minus one
22:58for domestic firms
23:00for exporters for multinationals
23:04and you can easily see that there is a
23:06point where it's profitable
23:08to export so firms do so
23:11and then at some point it's more
23:13profitable to become multinational but
23:15that happens at the
23:17very um upper end
23:20of the productivity distribution so as
23:23we saw in the data the most productive
23:25farms will
23:25serve multinationals as the foreign
23:28consumers to multinationals
23:30the middle and through exports and the
23:34least productive ones which just stay
23:36domestic
23:38now there is a catch here and this is
23:41under
23:42that symmetry was easy for me to to draw
23:45it
23:46and it's also easy to know the
23:48conditions under which
23:49this is the ranking of cutoffs in
23:53particular
23:54the condition is just a relationship
23:56between
23:57the the variable and
24:00fixed cost the value of the variable
24:02triggers and the fixed cost
24:04okay but you can think like if i
24:08start running a model under asymmetries
24:10it's more difficult how to check the
24:12grip
24:13and there is another one that is also
24:15more difficult to check and in general
24:17you have to impose
24:19that under symmetry again it's trivial
24:21but not
24:22when you start putting some
24:23heterogeneity is that i'm
24:25restricting funds to only horizontal
24:28phase
24:28i'm not allowing extra platforms so i'm
24:32basically assuming an equilibrium where
24:34that
24:34doesn't happen so it's because of these
24:37reasons
24:38that this model
24:42is a bit stark to do um quantitative
24:45analysis
24:46so how the the paper goes
24:50um then is like as i told you they are
24:53going to derive
24:54the proximity concentration trade of
24:56equation
24:57that's going to give them a testable
24:59implication
25:00and they're going to go to the u.s
25:04multinational data and test it with
25:06variation across
25:07industries so i'll show you that um so
25:10the model delivers
25:11indeed the proximity concentration
25:14trade-off export
25:15sales over mp sales follow an expression
25:18that is very easy to solve and the
25:20pareto
25:21and um and exports decrease
25:24relative trump p um with with trick cost
25:29increase with the the kind of the the
25:31fixed
25:32cost and then there is this parameter
25:34kappa
25:35that you remember is the pareto shape
25:37parameter
25:39and the predictions that the authors are
25:40going to test is about
25:42how this capability in lower or higher
25:46uh how it it correlates with the
25:49this ratio of exposure and b and they're
25:51going to choose that
25:52across industry so the intuition is very
25:55simple
25:56um this is a the density of a low a high
26:00dispersion
26:01uh productivity distribution these are
26:03just some of the cutoffs
26:06like the one with high dispersion the
26:08green line you see it has more mass
26:10on the right so you will have more mp
26:12activity
26:13on that type of industry with a lower
26:16kappa okay so as i said
26:20um this is the most you can do i mean i
26:22haven't closed the model it could have
26:25of course it can be solved in general
26:28equilibrium
26:29but has these star conditions that
26:32um parametric conditions that make it
26:34hard to use it to
26:36counter factors maybe so we're going to
26:39move to a framework
26:41that you saw a couple with melanie
26:43morton
26:44and a couple of uh of of lectures ago
26:48that used the ricardian ek model eton
26:50court in 2002
26:52so this is my work with andres rodriguez
26:55claire so basically we introduced
26:57um multinational production in this type
26:59of recovery and framework
27:01and the objective there was to have a
27:04quantitative general equilibrium model
27:06to run
27:06contra factors so we are going to
27:09to allow for forces in the model like we
27:12are going to allow for trading and pitch
27:14we alternate ways to serve a market
27:17and we are going to allow for foreign
27:19affiliates um
27:20importing intermediates uh we're going
27:23to have these
27:24um export platforms we call it bridge mp
27:27so
27:27there is no confusion about like
27:31having kind of production fragmentation
27:33we don't have that
27:34so what we are going to allow for for uh
27:37funds from
27:38a doing mp and b and exporting to c
27:41and then we are going to calibrate this
27:43macro model to
27:44what we observe in the data bilateral
27:46trading and b shares
27:48and then we're going to compute gains
27:50from openness from pm from three and you
27:53can think of other counter factor
27:54lectures
27:56so let me show you how the model
28:00goes
28:03so um let's start with
28:06ek 3 right so let me remind you in a
28:09slightly what the model
28:11that model was about so there was 10
28:14countries
28:14a multi-country model with a continuum
28:17of
28:18goods um b
28:21at cs aggregators i say i'm not going to
28:24go away from this assumption
28:26and productivity here is um
28:30it's this clb sorry it's my dog
28:34um it's going to be independent fresher
28:37over b and across countries else so what
28:41is this crochet
28:42well this is characterized by this tl
28:46the the scale parameter and the shape
28:48parameter
28:51so the innovation in their work was
28:54really that this
28:55was a random variable and had this
28:58particular
29:00extreme value distribution that i'm
29:01going to talk about a bit more
29:03later then um we have iceberg trade cost
29:07to to ship goods from l to n
29:10and the unit cost of producing this good
29:13v
29:14in l and ship it to n and
29:17rich consumers in l was just the unit
29:20cost
29:21which over reproductivity times the
29:24trade cost
29:26and then what was the the rule there
29:29there was this head-to-head competition
29:30it's a recurvian model
29:32so we all know how to produce each bee
29:36some are more efficient than others so
29:38who's the lowest cost producer of good
29:40fee
29:41for country and well the most
29:45efficient location okay
29:48um and it's because of this had to have
29:52competition
29:54feature of the model that fresher is
29:56going to
29:57be important for aggregation
30:00okay so at the end we want
30:03this distribution here because we want
30:07nice aggregation that's the whole
30:10point so i'm going to talk more about
30:12that
30:14so now have a picture is small
30:17okay now allow for a good v
30:20can be produced in country l with
30:23technologies from i
30:24and so in n okay so that's going to be
30:27the notation
30:28i for source country l for location of
30:31production
30:32and for location of consumers
30:36so now the unit cost
30:39of would be producing l with
30:41technologies from i
30:43should be sold in n has this form it's
30:46basically the same but it's analogous to
30:49what i show you so
30:50we produce that that you have a unit
30:52cost of inputs i'm going to tell you
30:54more
30:54about that maybe it's il specific
30:58then we have a productivity for that
31:01good fee that is not only i specific but
31:04might
31:05have something from the location we
31:08produce
31:09and then we need to ship that good from
31:11l to n that's why we have the tab
31:12so i'm going to have this triple index
31:15uh price okay
31:19so i need to tell you these two objects
31:22where
31:23where are they um and what we have
31:26inside here
31:27but again we are going to have this head
31:30to head competition
31:32feature so we all know how to produce d
31:35it's just we need to choose the lost
31:37cost producer
31:38for good d to serve and i was going to
31:42be that
31:43the minimum between the technologies and
31:46locations
31:48okay that's it that's how
31:51you see really how we move a degree away
31:54from
31:56from eton corton that only trade model
31:58with a model where
32:00now and p are these technologies and we
32:02keep track where these technologies are
32:04coming from
32:07so um so let me talk about these two
32:10objects
32:10okay i'm going to to do more regarding
32:14the distribution here
32:16of this this random draw and we i can do
32:19a lot of things also with that
32:22so i call that cil the international
32:25multinational input
32:27bundle and you can introduce sourcing
32:30through that object
32:31basically so what you can do is like
32:34say okay i'm an affiliate now for an
32:38eye well i can source
32:41labor from or input from from
32:44l i have to pay a cost i'll tell you in
32:46a second
32:47and then i can bring inputs from all
32:49over the world
32:51including home okay and then imagine
32:54there is vcs aggregator
32:56for that now you can restrict that as
33:00you we do it like this in the paper just
33:02bring
33:04home sourcing just bring um
33:07intermediate from your parent okay so
33:10you have
33:11you have to ship them to the host
33:14country
33:14and then you can use also um some
33:18inputs locally labeled and then and this
33:21is the case i'm going to go
33:22with today unless i say the contrary
33:26is just imagine you just use labor in
33:28your location
33:30okay so if you do that um then
33:33this pil lnn iln
33:37for would be is just the the
33:40wage the right little t in that location
33:42and then
33:43you pay an empty cost of saying
33:46i say something else in the in a second
33:48and the
33:49tau the three costs to ship frontal to
33:52end
33:53so this mp cost is introduced
33:56as an ideal cost and think about it as
33:59an efficiency
34:01loss of using a technology
34:04in a location that is different from the
34:06one it was created
34:08okay so it's gamma il is bigger than one
34:12like uh also trickles but it's equal to
34:14one
34:15if you use that productivity drop
34:18in the location where it's originally
34:22from
34:22so we will call it empty cast
34:27and just previewing you you remember i
34:30show you that
34:32this um these flows follow gravity well
34:35this is the way
34:37the same way the same way trade models
34:39introduce gravity through tau
34:41we're going to introduce gravity through
34:44gamma
34:46that's that's that's why we need we use
34:48all these wedges
34:50okay so um as i said i'm going to go
34:53through
34:54with this simpler case because i can go
34:56a long way solving
34:58the model um
35:02um with closed forms so as i said i talk
35:05about c
35:05now i'm going to talk about c
35:11so again let me stress that this
35:13distribution assumption is for
35:14aggregation proper
35:16proper purposes in closed form
35:19okay if you don't care about close form
35:21you can assume whatever you want
35:23but we really wanted these mods to have
35:26these uh
35:27aggregation properties and then be very
35:29clear how
35:30identification when we calibrate the
35:33model etc
35:34so in in the original paper we use
35:38this what symmetric multivariate
35:40thresher
35:41and basically think about putting some
35:44heterogeneity
35:45across locations of production
35:48okay so productivity is like location
35:52specific not only
35:56sorry um so um
36:00i will have there is a characterization
36:02of it we are going to
36:04is going to have a scale parameter that
36:06is il
36:07specific origin production location
36:11specific
36:12and then we're going to have this row
36:14that is the one that introduced
36:15correlation
36:16if rows equals to zero growth are
36:19completely independent across locations
36:22if draw both shu one we are in the melee
36:25civil war where productivity is the same
36:27within the
36:29in any location for that particular
36:31technology
36:32pronoun for that could be okay
36:35so and then you can think that we can be
36:38in cases in between
36:41so the theta is going to be the shape
36:42parameter
36:44now the assumptions that i'm going to
36:46keep is productivity is going to be iid
36:48overv
36:49for the continuous groups and it's going
36:52to be
36:53independent across origins eyes
36:57so let me talk more this is my work with
37:00nelson lane
37:01about these functions these extreme
37:03value functions
37:06so basically that symmetric thresher
37:10that i just
37:11show you it's a special case of
37:15a multivariate max table thresher with
37:18some correlation function
37:21so the form of that is the one up here
37:24so the joint distribution of
37:25productivity across
37:27countries it's this is the form of a
37:31multivariate
37:32max table friction so it has this g
37:35function
37:36and then uh we have the the t's that are
37:39this case and the theta are the shapes
37:42so what's that this g function this is a
37:44way it's a type of copula tree
37:47to incorporate correlation
37:50so the key property of the g function is
37:53that is homogeneous of the degree one
37:55and that gives you max stability so
37:58what's max stability i've been using
37:59this term
38:01so max stability says that if you take
38:04fresh air
38:04and you take the max of fresh air
38:08you get fresher so and which pressure
38:11you get well one with the same g
38:14that aggregates the underlying scales
38:18and then it preserves the shape
38:19parameter
38:21so this is a crucial um property for
38:24these
38:25head-to-head competition models or any
38:28role model for the math
38:30so you start with fresher you take the
38:32max
38:33or the mean if you're talking about cost
38:36and you'll get friction
38:38so what this property implies
38:42is that the conditional and conditional
38:44probability of the maximum point size
38:47and if you go to your lecture on ek
38:50that's crucial to have this max
38:52stability
38:54gives you a key implication it means
38:56that the probabilities
38:58are equal to expenditure shares is is
39:02thanks to this property the max
39:03stability of fresher
39:05that we can do that jump there you
39:08remember that
39:09the probability of a a country
39:12in the lowest so uh source is
39:16we can also say that is the the share of
39:19goods at the ex
39:20input from that country and that's the
39:22expenditure share
39:25so i can talk more in the q a about this
39:27but
39:28but that's why this this this type of
39:31functions are
39:33probability functions are really key for
39:35this type of model
39:36for aggregation gene is to have other
39:40properties um um
39:42[Music]
39:44it has to follow a pattern of
39:46differentiability
39:47and i'll i'll tell you in a second why
39:50that's important
39:51so now look there are special cases so
39:54we've been talking about independence
39:56if gst here you put like a summation
40:00if it's additive that's independence
40:02that's the ak case
40:04if it's symmetric say the cs is a cs
40:08function we are back on the case
40:10i just show you that we used in the mp
40:13case
40:14now what is what is imp what
40:18using these max tables per share with
40:20the g function implies
40:22is that you know where we're going
40:25in terms of expenditure shares so
40:28basically
40:29a multivitamin pressure distribution
40:32for productivity is equivalent to
40:35winning the
40:36general extreme value gv expenditure
40:38this is a mcfarland result
40:41so you can always give me a import
40:44demand system that is
40:45gv and i can always work it out
40:49with a ricardian model where
40:50productivity
40:52has this form with a given g
40:55so you can and again this is like it on
40:58corto
40:58okay so trace shares
41:01have this form where you have the the
41:04the scales here
41:06um the unit cost and then you have the g
41:10function um aggregating those
41:14those those unit cost index and then
41:17this gl is the first derivative of that
41:19function so this is the form of
41:21geb think about ek
41:25when we have independence so we have an
41:27additive g
41:28so g here is summation that's what we
41:31have
41:32there and the derivative is one
41:35right so this term goes away so you get
41:38your cs
41:39uh import demand system as in ek
41:43as in alcohol at this post
41:47so this is a this takes you to a more
41:50general
41:51um class
41:54of expenditure so
41:58in our case with multinational
42:01production we introduced correlation
42:03of course now farms can be uh from one
42:06place can be in many places so the
42:08the g function that corresponds to this
42:10space is
42:11what is called a cross nested cs and it
42:13has this form over here
42:15so we aggregate over we have um
42:18this outer nest that tells us
42:23that each nest corresponds to a home
42:26country
42:27so knowing the general properties of
42:29this g
42:30allows me to relax some assumptions okay
42:33so for instance i can allow for this
42:35road to be
42:36home country specific so i put a sub i
42:39there then i can have more general
42:43trade and then because they don't have
42:44to be iceberg okay they can
42:47be embedded on these t's so just truly
42:50precise
42:51the unit cost of producing would be in l
42:54with technologies from eye to eleven to
42:56n
42:56it can be written this way wl over
43:00a productivity draw that has the triple
43:02triple
43:03uh subscripts instead of already
43:06assuming that
43:07those of if there is like a iceberg
43:10type cost what is key
43:14in all this is that i will cut the
43:17hardwire property
43:19here is that productivity has to be id
43:21over
43:22the continuing of goods b and that's a
43:25property that is
43:26uh difficult to relax and
43:30also i kept the property that
43:32productivity is iiv
43:34across i we have cases where we
43:37we know how to relax it that's easier
43:39okay we do it in the background
43:41with nelson so yeah
43:46yeah yeah joe yeah i don't
43:49i don't teach this but i want to teach
43:51this uh um
43:53but with the two setups um it seems to
43:56me like the
43:57uh the one with the trade costs and the
44:00platforming
44:01the gamma and the tau uh is
44:05i mean the other the the left is more
44:07general but can you identify the left
44:09because i thought usually we put the
44:11parameter parametric restrictions to
44:13identify
44:14yeah we do that um um
44:18we can you're right and we can we
44:21we need some we are going to go to
44:24gravity
44:25at some point again to identify some of
44:28these things
44:29but i want to point out that it's not
44:30necessarily that you have to
44:33to assume that okay it can be
44:36all embedded on on the on the margins
44:40basically of the distribution this is
44:43just the way if you were talking about
44:45i want to be more general because you
44:47can use this for any role model
44:51so you can make a migrant choosing where
44:53to go
44:54and occupations or somebody choosing
44:56occupations
44:57everything that has had to have
44:58competitions you can use
45:00um this type of distributions if you
45:03care about aggregations again
45:07so um but now um and
45:11we didn't know this when we wrote that
45:13paper um
45:14this this came after so now i know how
45:17expenditure has to look like in my model
45:20of
45:20multinationals so um
45:24so again what i'm after why i'm doing
45:27this
45:27uh at the end what i want to take to the
45:29data are
45:30flows expenditures and some i'm going to
45:34observe them some
45:36not not because they are not there just
45:38because the data are limited
45:40um so that's why i want to have these
45:42subjects characterized
45:44and and because i want to characterize
45:46the gains
45:48the welfare gains so expenditure
45:51in in an end of what's producing l
45:54with technologies from i will have um
45:57it will come from this crossness ccs so
45:59you will have a familiar expression
46:01expression if you're familiar with
46:03sectoral models this is
46:05this is if you replace i with s for
46:08sector
46:09this is an expression say in caliento
46:11para okay
46:12so we have a within i expenditure
46:16or within sector if you were in the
46:17sectoral mode and then a between nests
46:20expenditure between eyes in this case
46:23and
46:24um and then we have this um we call it
46:28pricing this is they are defined here so
46:31um this is this is nice because this
46:34will take us to gravity so at the
46:39home country level we have like a
46:41gravity expression that
46:43we can potentially use to estimate
46:47um the elasticities now
46:51in in the paper with unrest we were in
46:54the special case as i told joe where we
46:57decompose these t's in in the index for
47:01iceberg trade and then because and then
47:03we also
47:04are a bit less general on this rows
47:08so now it's easy to to know to we have
47:11expressions
47:12and we we can compute bilateral trade
47:15that's this sum over i of this x iln
47:18bilateral and p with sum over n and we
47:21have np frontal to l2l
47:24those are things that we observe in the
47:26data and then we have total expenditure
47:28that we also observe in the week
47:31but now what i can do is i'm going to
47:33use a sufficient statistic approach
47:35to characterize gains that's why i
47:37wanted to characterize the flaws
47:40so what i'm going to do is like what
47:44um is done in arcola is fostering
47:47clear the acr formula i'm going to wait
47:50for this
47:51more general class the gv class instead
47:53of cs and then i will apply it to my
47:55model
47:56with multinationals so
47:59um and this is in in our paper will with
48:02with nelson gains from
48:03trade for gv class can be characterized
48:06as the the tree share
48:11and something related to correlation is
48:13the
48:14derivative of these g's okay
48:17so under independence this derivative is
48:20one you are you are back to
48:22acr cs but in any in
48:26other cases every time you you have some
48:30correlation across countries within
48:31technologies you are going to have
48:34less gates from trade because partners
48:36are more similar to you so there is less
48:38scope for comparative advantage
48:40okay so the case of independence
48:43given the elasticity is an upper round
48:46for games from trade so
48:50let's apply that to our model i mean i
48:52know exactly how
48:53i just i know this function so i can i
48:57can do that
48:57so for this crossnet ccs subclass
49:01but if my mp model the gains from trade
49:05are again the these these
49:08these domestic shares um
49:12divided by the population and i can
49:14again this is this is
49:16the algebra i can characterize that
49:17correlation in terms of flows
49:20and i will have around flows that are
49:23observable okay again if rho is equal to
49:27zero zero
49:28um only this time will survive
49:31right and here what i wanted to make
49:34explicit
49:35is this summation so where i i didn't do
49:37it here
49:38okay so when i talk about the gains from
49:42trade
49:42i'm talking about a situation where i go
49:45from trade to
49:49trade and mp sorry i'm going through a
49:51situation from only mp
49:53technologies are allowed to flow around
49:56the world
49:57but not trade and then i moved to also
49:59three
50:00okay so i'm going to call that the games
50:02from trade
50:05and i didn't say that but the gains are
50:06always compute as differences in real
50:08wages
50:10okay so um again if you
50:13have data a lot of data you will kind of
50:16fill in the blanks here and compute
50:18these gains
50:20we don't have so much data um
50:23so we are going to go around this this
50:26i have a question on this yeah
50:29yeah speaking of data so uh so
50:32understand that
50:34for your ar white paper with costas and
50:36also
50:37the arp mp15 paper so you use this 1999
50:40mp flow data
50:42so i'm curious is there any updates for
50:44the mp flow data
50:46uh currently yeah other people did it
50:49not task
50:50okay like that update yeah
50:54but um here the problem is the
50:57the you see that it's a triplet that's a
50:59problem so sometimes we don't have the
51:01triplets
51:02for one country a very important one the
51:04u.s we have them
51:05so um that's good news i think japan
51:08also had them and
51:10korea so let me just define a couple of
51:13things
51:14um where we are going again this paper
51:16was about and the model was
51:17just about computer games quantifying
51:20games and i just wanted to review the
51:22metallurgology of that i'm not going to
51:24tell you the numbers because you know
51:26like uh data
51:28were all and now we have better data etc
51:32so i'm going to call the games from
51:33trade star
51:35when i go to isolation total isolation
51:38to three
51:39and gains from mp star when i go from
51:41total isolation to mp
51:44and those are very easy games to compute
51:46it's just an acr type of
51:48formula okay you need the domestic
51:51shares of
51:52trade or mp and some elasticity the
51:55gains from trade
51:56the ones that i just show you is i have
51:58only mp and i go to trillion mps
52:01and then the instrument and p are just a
52:03combination of
52:04things that i calculated before and then
52:07i have the gains run open
52:09okay i go from total isolation to open
52:11up to these two
52:13uh flows these gains are always
52:16positive basically and you can make the
52:20the special cases where the rows are
52:22zero we refer to
52:24to a case with with only like a cr type
52:27formula
52:28this independence case etc here we are
52:30contemplating the fact
52:32that mp allows us to trade stuff around
52:35et cetera but in this paper we want you
52:38to kind of say okay what's the
52:39relationship between these
52:40these games well we say if we have
52:44if the games from trade tips are bigger
52:46than
52:47that the gains from trade star that they
52:50want
52:50from isolation to trade it um
52:54it means that that that you you ain't
52:56more when you open up to
52:58mp so that's we call it
53:01mp complements if they were less we call
53:04it empty substitutes
53:06and independent if they were the same
53:08and then
53:09i'm not gonna i don't have time so i'm
53:11not going to do it but basically
53:13research-wise how we proceed we we work
53:16a case under symmetry
53:19so we could have all the objects of
53:22interest
53:22these g's these g gains in terms of
53:26parameters
53:27instead of those the sufficient
53:29statistic approach
53:31and and we will see extreme cases where
53:34you know we could characterize these
53:36gains and put signs
53:38and that's what was done in the paper
53:40and then the calibration of the model
53:42tell us was telling us where we were
53:45okay but how we would i mean
53:49there was like no wells and whistles so
53:51we have
53:52a bunch of bells and whistles to the
53:54model to make it more quantitative
53:56so we added intermediate we have only
53:59three double boosts we have
54:00non-treatable groups
54:02we added the input output loops that are
54:04very common in the tree literature
54:06etc so i want to talk a two minutes
54:10about um calibration
54:13because i think it's more general than
54:15just these models
54:17so as i said gravity is important this
54:19bilateral and trade and empiricals is
54:21where we are going to kind of sneak in
54:24grab it
54:25so one way to go is to assume there are
54:28functions of observables
54:30distance and that's what joe question
54:32was going
54:33so you target observe bilateral trade
54:36and mp flows and just
54:38find the the distance elasticities
54:42another way of going is like just find
54:46these wedges the thousand the gammas are
54:48the wedges that
54:49just match exactly what you see in the
54:51date so you have this big matrix of
54:54of of parameters and you just match
54:56exactly
54:59the third material that is particular to
55:01this model is that
55:03if you have the aggregate shares
55:06the bilateral shares uh we show in the
55:10in the in the
55:10follow-up paper that there is a unique
55:13sense of these trilateral flows
55:16so if you are willing to assume that uh
55:18trade costs
55:19um and because are symmetric you can
55:21compute them
55:22uh from these trilateral costs okay
55:26of course you don't have them in the
55:27data so what we did is we impose
55:29consistency at the aggregate level
55:32the model speed up uh uh trilateral
55:36uh flows and we just use those
55:37simulators to calculate our tiles and
55:40them
55:40papers came thereafter that had more
55:42data and show
55:44that the model doesn't do very well on
55:46this side so that's a
55:48matter of maybe future research
55:51so um okay so
55:54that those are the cost then we have the
55:57troop
55:58elasticities and this is always like a
56:00big deal in these three models and np so
56:03we have the theta we have the row
56:05and again this was not for this paper
56:08but the subsequent one
56:09we could uh target what we call an
56:11unrestricted gravity
56:13uh regression to map this beta
56:17coefficient here and then if we
56:20restricted
56:22this gravity equation to one origin the
56:24us that we had
56:26well the coefficient on the taus was
56:29something
56:30that it was informative about the row so
56:32if we have teta we will
56:33follow we call that restricted gravity
56:35so you need
56:36data at least from one origin and
56:39be able to follow up the route
56:42okay so that's in our uh our collab is a
56:46tall paper
56:47and then there is always this these
56:50things how you calc
56:51how you calibrated the scales
56:55there is many ways to go we in gel or
56:58go assuming that is proportional to some
57:01measure of size
57:02with labor and proportionally given like
57:04some
57:06around the employment or something like
57:08that so
57:10we set it up basically from from data
57:14and then you will be left in say if you
57:17incorporate sourcing you should
57:19target some intermediate imports
57:22of a village to calibrate that cs
57:26function for the for the sourcing and
57:29maybe intra firm
57:31that's how we did it and then some other
57:33remaining parameters we have them from
57:35the glitch
57:37so this is the how we
57:40we went with this paper what we weren't
57:42like
57:43it really we didn't have any proven
57:46force i mean
57:47things were um good
57:50in this paper i mean games were always
57:52possible etc um
57:55and so we we started thinking how to
57:58to talk about multinationals you know
58:01in a helma melliput type of setup but
58:03making it more flexible
58:05to kind of being able to again quantify
58:08gains and other effects that maybe these
58:11perfectly competitive recurrent models
58:14didn't have
58:15so that's why we moved to a model i call
58:18it program malicious
58:20ek and then we
58:23we had a motivation for that and the
58:25motivation is that we realized that
58:27we could talk about innovation and
58:29production in this type of setups
58:32um the idea is that um
58:36goods are produced in one place and
58:38ideas are produced in another place
58:40and what made that separation possible
58:42are multinational farms
58:45so what's the problem some countries
58:47specialize in innovation like the u.s
58:49another specialized in production
58:52but there is this kind of um concern
58:56and you saw it play all the time all
58:58these years in
58:59in politics is that the ones that are
59:01specialized in production
59:03are worried about low innovation rates
59:05and the ones that specialize in
59:06innovation and
59:07worrying about losing jobs in production
59:10so we wanted to see this type of models
59:13where we
59:14put krugman time forces and ek type
59:17forces
59:18uh will inform this question
59:22and we are fighting with it
59:25so what we did is this again a
59:28quantitative and when i say quantitative
59:30i mean
59:31free geography many countries and
59:33generally believe
59:35um with innovation and production though
59:37those are this is the two sector model
59:39right and it has three and it has
59:41multinationals
59:43and the specialization patterns in
59:46innovation and production
59:47will reflect comparative advantage
59:50and home market effects fail reflect
59:53that
59:54and from locating large markets
59:57so comparative advantage so innovation
1:00:00will be
1:00:00pulled to places that are relatively
1:00:03more productive
1:00:04in doing innovation than production and
1:00:07home market
1:00:08effects will tell us that innovation
1:00:10will will be pulled to
1:00:12places where farms are good at producing
1:00:15anywhere okay so multinationals
1:00:18and then production is going to be
1:00:21pulled to
1:00:21places that are like um
1:00:26good at uh they have um absorption
1:00:28potential in the sense that
1:00:30they have high um they are centrally
1:00:34located
1:00:34etc so those are of course
1:00:38specific parameters in the model and
1:00:42and there's going to be some things that
1:00:44are older in the literature
1:00:46um for instance in venables in the 80s
1:00:49and that these are and we can have um
1:00:52losses from um openness
1:00:57so that's that's that's where the new
1:01:00mechanisms that we
1:01:01introduce in in this paper and some of
1:01:04the
1:01:05the the the stuff they write the weight
1:01:08i wrote it the ricardia model was taken
1:01:11from
1:01:11from this paper so what what's what we
1:01:14have
1:01:15so we have um workers that are allocated
1:01:18to innovation or production
1:01:20okay and that's a refresher
1:01:23um standard there then we are going to
1:01:27france will pay a fixed cost to innovate
1:01:29and that's the krugman entry cost okay
1:01:32so this
1:01:33wfe so i'm i'm going to work with a case
1:01:37where there is endogenous centrifugal so
1:01:39wages are going to equalize across
1:01:42across
1:01:43um sectors and basically think about the
1:01:46france owning the idea
1:01:48to produce a good okay so
1:01:51you pay the cost this entry cost and
1:01:54then you
1:01:55get your mail its productivity okay so
1:01:58the outcome of innovation is stochastic
1:02:01and here the methodological contribution
1:02:05was that we introduced these
1:02:06multivariate pareto distributions
1:02:08in front productivity so that's going to
1:02:11be
1:02:12farm location specific um
1:02:15productivity it's different from the
1:02:17fresher this is not a head-to-head
1:02:18competition mode
1:02:21and then countries this we are going to
1:02:24have countries that
1:02:24differ in their productivity across
1:02:26goods and then
1:02:28countries that differ in the quality of
1:02:30these ideas okay all these
1:02:33these blueprints okay and that will give
1:02:36some variation in comparative advantage
1:02:38in innovations versus production
1:02:41okay so that we are going to have
1:02:45geographic barriers to trade an mp and
1:02:48then
1:02:48firms will choose where to produce
1:02:50according to these barriers so we are
1:02:52going to have taos we are going to have
1:02:54gammas and we are going to have also
1:02:56fixed export costs
1:02:57you know like in the melee small
1:03:00and then typical continuous varieties
1:03:04cs aggregator and here monopolistic
1:03:06competition
1:03:07that's in helma melissa so
1:03:11what is this multivariate pareto i mean
1:03:16intuition is similar to the one the
1:03:18distributions i've been talking about
1:03:20i mean pareto is good for this um
1:03:23for for krugman type melis type sorry
1:03:26models okay so we are not there is
1:03:30each farm produce a variety and there is
1:03:32monopolistic competition
1:03:34it's not like the ricardian model that
1:03:36we are competing
1:03:37and whoever is more efficient kicks the
1:03:39other out
1:03:41okay so as long as i
1:03:44we produce different varieties well the
1:03:47more inefficient we have
1:03:48a lower market share but it will be
1:03:51there
1:03:53as long as profits are profits are
1:03:55positive right
1:03:57so this multivariate pareto that's the
1:03:58trick as we did with this
1:04:00multivariate thresher to introduce some
1:04:02correlation
1:04:04across locations for a farm
1:04:07okay so um
1:04:11basically here it is regulated
1:04:14across locations and data regulates
1:04:17still
1:04:17the heterogeneity over varieties that
1:04:20here correspond to farms
1:04:22and then we introduce this this this
1:04:25comparative advantage in production
1:04:27versus innovation through this
1:04:29piece we are going to assume that it has
1:04:32a
1:04:32an e component and a t component okay so
1:04:36a country that has comparative advantage
1:04:38nowadays has relatively higher the te
1:04:40than the tp
1:04:43so that's that's the setup i'll i will
1:04:46tell you an isomorphism about this
1:04:48distribution again
1:04:50in a little bit in a couple of lights
1:04:52okay
1:04:53um so um
1:04:56um again why we want this multivariate
1:04:59pareto
1:05:01because it will aggregate this model
1:05:03really nicely
1:05:04okay so we are going to have the unit
1:05:06cost from from
1:05:08c uh from country r serving l
1:05:11and from l this is going to be a
1:05:14a a random variable and it will have a
1:05:17a nice distribution that's the thing
1:05:20okay
1:05:21and because the problem will have it
1:05:24will be a
1:05:24bit different from before so the firms
1:05:27will choose
1:05:28the cheapest location for pro usage knl
1:05:32so we will choose the argument of
1:05:34disease
1:05:36and then we see when that happens
1:05:39it will kind of see if profits are
1:05:41positive
1:05:43or not so and there's going to be a
1:05:47a a threshold
1:05:51and they're going to produce if this c
1:05:54such that is below okay and gives
1:05:57positive profits so remember they have
1:06:00to
1:06:00to pay this export so what the
1:06:04multivariable
1:06:05give us is these closed form
1:06:07probabilities
1:06:09these conditional probabilities and we
1:06:11have a closed form and that will
1:06:13translate in having close one for
1:06:15aggregate flows again
1:06:17we are taking the model to the data so
1:06:18we really want to have transparency on
1:06:21that
1:06:22so at the firm level going back to the
1:06:25proximity concentration traders
1:06:27i didn't say anything about the previous
1:06:29model that model delivers the proximity
1:06:31concentration trade-off at the aggregate
1:06:33level
1:06:33the one we saw in the data this one too
1:06:36but it doesn't have it at the
1:06:38firm level okay i don't have any
1:06:42any force of that so we call it
1:06:44proximity
1:06:45advantage trade off a different level so
1:06:50um
1:06:54as i said the model has two sectors
1:06:56production and innovation
1:06:58so we can we can see what happens with
1:07:01innovation share we want to see
1:07:02what happened with innovation when
1:07:04countries open up or
1:07:05or or or have higher barriers to trade
1:07:09etc
1:07:09all of that so we can compute it in this
1:07:12model
1:07:12and it's just the fraction of the wage
1:07:15bill to innovation workers over
1:07:17total expenditure now what is this share
1:07:20well this share um is composed of
1:07:24a parameter that captures the sales
1:07:26shares of profits but it's also the
1:07:29the innovation share in authority so
1:07:31when you open up
1:07:33and you can go up or down for another
1:07:36key
1:07:37you go up if x is bigger than y
1:07:40and that means if x is bigger than y
1:07:43it means that you have a trade deficit
1:07:45because remember this model
1:07:47has multinationals house profits
1:07:49floating around so
1:07:51you need current account balance not
1:07:53trade a trade balance
1:07:55okay so you can have trailing balances
1:07:57here so
1:07:58r is going to be bigger than eta
1:08:01innovation on the router key as long as
1:08:04x is bigger than y
1:08:06and as i said that's a tray deficit okay
1:08:10so these countries will specialize in
1:08:12production in innovation
1:08:14so it's like the u.s so you are they're
1:08:16going to import goods and they're going
1:08:18to repatriate profits from multinational
1:08:20so the us is an innovation hub
1:08:24other countries will specialize in
1:08:25production and they're going to sell
1:08:27goods
1:08:28to the rest of the world so they will
1:08:30have y bigger than x they will have a
1:08:31trade surplus
1:08:33okay they're going to be production have
1:08:36they're going to export goods and then
1:08:37send profits of multinationals
1:08:39abroad think ireland for instance
1:08:43okay and why their production has well
1:08:46they're relatively more productive in
1:08:47production but also they are centrally
1:08:49indicated they choose a value that's why
1:08:52they have that
1:08:53okay so all these forces are in play and
1:08:56they shape this
1:08:57innovation trade
1:09:00um the gains from openness now
1:09:03they're exactly as before but they have
1:09:05this extra term that comes from this
1:09:08entry channel this innovation channel
1:09:11and
1:09:11it has to do with also the deficit
1:09:15okay so only because we have this extra
1:09:19term is that now
1:09:20the gains from openness right and also
1:09:24front end and or from trade can be
1:09:26negative
1:09:27it can be less than one okay so
1:09:30um is really in this group model from
1:09:33with these crew mentality forces we open
1:09:36the door to these
1:09:37losses from openness and from
1:09:40multinationals and from tree
1:09:42how plus we release in the in the data
1:09:45when we quantify the model
1:09:48it wasn't the case that this was big
1:09:51enough to overturn
1:09:52this direct effect but you could imagine
1:09:56exercise where it
1:09:57is okay also because we have these
1:10:01workers underneath
1:10:02we can also compute who wins who gains
1:10:05who losses every time we
1:10:07we shout the economy
1:10:10so i'm i think i have five minutes right
1:10:14um so i wanna
1:10:17talk about um uh
1:10:20four isomorphisms that will work and i
1:10:23think they're very useful but they're in
1:10:24the online appendix of the paper nobody
1:10:27never reads them
1:10:28but i think this is really useful for
1:10:31for when crafting these models so the
1:10:34first one is that this multivariate
1:10:36pareto productivity
1:10:38is isomorphic it's equivalent to have
1:10:42a male its core productivity fee
1:10:46times a farm location specific
1:10:49independent pressure
1:10:50routine yeah so
1:10:53it's the same it's the same model you're
1:10:57going to end up with the same stuff
1:11:00so um um
1:11:03the multivariate pareto gives you maybe
1:11:05more flexibility so
1:11:07here the thing introduces correlation
1:11:09because you know it's common across
1:11:11within the farm across locations and
1:11:13then you have this cl
1:11:14parameter this for instance is the setup
1:11:17in felix intel nodes
1:11:18uh job market paper but they are
1:11:21equivalent
1:11:23then the other thing that is there and i
1:11:25think this is um
1:11:28interesting is like this chroma meet ck
1:11:31model of mp and trade
1:11:32is equivalent equivalent to a two sector
1:11:36model where one sector is krugman
1:11:38free entry monopolistic and the other
1:11:40factor is
1:11:42ek head to head so if you work that
1:11:45model
1:11:46you are going it's like talking about a
1:11:48model of np because mp at the end of the
1:11:50day
1:11:51is entry of farms okay
1:11:54so uh that's one thing the other thing
1:11:57that we show in this appendix is like
1:11:59i was talking about innovation and
1:12:02innovation it's like creating products
1:12:04it's like the
1:12:05it's like mellits or recruitment right
1:12:08so that's that's what it really is but
1:12:10we show
1:12:10in the appendix that is equivalent to
1:12:13have process innovation where processing
1:12:15innovation is just
1:12:16um you make some investment
1:12:19um based mechanic to lower unit cost of
1:12:22production
1:12:24and finally and and this is maybe
1:12:28this one took us a long time if if you
1:12:31are willing to assume the model under
1:12:33independence the multivariate pareto
1:12:36with zero correlation and this model
1:12:39that doesn't have um production fixed
1:12:43cost is equivalent with a model with
1:12:46production fixed cost
1:12:48okay and there's some very mild
1:12:50conditions
1:12:51so just to say a word about that a model
1:12:55with plan level
1:12:57uh where you open a multinational you
1:12:59pay again
1:13:00you have to pay this plan fixed cost in
1:13:03each location
1:13:05that will make the problem very
1:13:07difficult right because it's
1:13:08kind of a combinatorial problem
1:13:11and it's the problem like felix has in
1:13:14his
1:13:14job market paper um and
1:13:18so we show that if you are willing to
1:13:20assume independence this free equal to
1:13:22zero
1:13:23actually there is a an equivalence
1:13:25between the two months
1:13:27it's good news it saves you a lot of um
1:13:30computation okay
1:13:34so let me talk in the last three minutes
1:13:36i think
1:13:37uh about spillovers um
1:13:41and and here i just um
1:13:45i'm i'm like just putting papers that
1:13:49kind of expand or these things like um
1:13:53with trade these three and mp models
1:13:56that um
1:13:57that that expanded the models as they
1:13:59were but they are like
1:14:00some results are surprising right so
1:14:04um i think like uh it's it's it's
1:14:07they're very nice papers in general
1:14:10so i can talk more in the q a so let me
1:14:14tell you something about
1:14:15i want to finish with these spillovers
1:14:17the state of the art
1:14:18in empirics because um
1:14:22what i what i mean by this i mean that
1:14:25we mean by these spillovers that you
1:14:27have an incumbent from
1:14:29and they increase productivity when a
1:14:31more productive foreign firm
1:14:34locates nearby nearby in quotes
1:14:38so traditionally um the
1:14:41the empirical analysis went that we have
1:14:43the dependent variable some outcome
1:14:45for a domestic form and the
1:14:48control or the independent variable was
1:14:51some measure of nearby
1:14:53multinational activity and this there
1:14:56were many papers
1:14:58in the past the results were mixed even
1:15:01with
1:15:01some level data it was very difficult
1:15:03trading
1:15:05so more recently in the 2000s um
1:15:08people start refining the channel of the
1:15:11spillover
1:15:12for instance backward linkages who are
1:15:15your suppliers
1:15:16okay instead of like overall no just
1:15:19let's look at that particular
1:15:20channel that is javascript 2004 paper
1:15:24and a rich literature director
1:15:28now the state of the literature is that
1:15:30the data became
1:15:31so detailed that we can really
1:15:34identify channels
1:15:38really uh clearly so
1:15:41we have fun to film beta because
1:15:44countries that have the va t can have
1:15:46this type of data
1:15:47we have employer employee data and
1:15:50some very cleverly than natural
1:15:52experiments
1:15:54so let me talk briefly about the three
1:15:56and i don't want you to read the whole
1:15:57slide so i think the paperwork is really
1:15:59important
1:16:00and they're being follow-up for for
1:16:02countries like ethiopia with
1:16:04multinational is this large plan opening
1:16:06paperwork
1:16:07um greenstone and cultures so where they
1:16:11use
1:16:11the opening of new plants of large new
1:16:14plants one million dollar plants across
1:16:17u.s counties and they use this
1:16:20information from this site selection
1:16:22magazine that is
1:16:23really good if you're interested in
1:16:24multinationals through to look once in a
1:16:26while
1:16:28where they report counties that were
1:16:30chosen but
1:16:31by these plans and countries that were
1:16:34almost chosen
1:16:35but they at the end they lost the bid
1:16:38okay
1:16:39so the identification strategies just
1:16:41compare those
1:16:42those two um incumbent plans in these
1:16:46two type of countries
1:16:48and there is um there is
1:16:52uh a bunch of uh exercise there to make
1:16:55sure that we're identifying the
1:16:57right the right thing but there were
1:17:00follow-up papers for
1:17:01for instance as i say ethiopia where the
1:17:03government was uh
1:17:04was allocating across regions these
1:17:07large plants
1:17:08and the authors used that variation for
1:17:10instance
1:17:11so um i think this is a promising way of
1:17:15identifying
1:17:16spillovers then recently there is this
1:17:19paper for the u.s
1:17:22now brad zelda and felix that they use
1:17:25employee
1:17:26employer employee data these are for tax
1:17:29filings this
1:17:30data are amazing they identify
1:17:32multinationals
1:17:34surprisingly enough we knew little about
1:17:36multinationals in the us
1:17:37um they are not identified in the u.s
1:17:40census so you have to go another way
1:17:42and the va is separate from the u.s
1:17:44census census now there is an effort to
1:17:47put all the data together so now we'll
1:17:49be able to identify them
1:17:51um and they can use modern econometric
1:17:54techniques
1:17:55from the labor glitcher to this two-way
1:17:57things effects
1:17:58to identify um direct effects and
1:18:01indirect effects of
1:18:03of multinationals what i like
1:18:06a lot about this paper is that we all we
1:18:08forget about the trade-off
1:18:10and they bring it to the table very
1:18:12clearly so basically you gain from
1:18:14having multinationals but
1:18:16governments state governments local
1:18:17governments spend a lot of money trying
1:18:19to attract them
1:18:20so they kind of put together those two
1:18:23forces it's like
1:18:24these megadeals right so i think that
1:18:27has to be
1:18:28brought to a table when analyzing these
1:18:31spillovers
1:18:32and finally there is um the case of
1:18:36costa rica i'll call it because all
1:18:37these papers are
1:18:38are run thanks to very rich data from
1:18:41costa rica
1:18:42employer employee data and function fund
1:18:45data from costa rica and again these
1:18:48studies like some channels um
1:18:52to like the the supplies channel to
1:18:54identify
1:18:56a type of spillover the paper are re is
1:18:59very rich on the empirical design i
1:19:01encourage you to read i think it's very
1:19:04interesting and finally
1:19:05there is a new paper a history with a
1:19:08historical
1:19:09experiment by vienna van patten the
1:19:11united fruit company the infamous united
1:19:14food company
1:19:15that have plantations in costa rica and
1:19:17other central american countries
1:19:19and she exploits a new channel monopsony
1:19:22power
1:19:22of this farm on how effective
1:19:25development
1:19:25at that point and maybe thereafter and
1:19:28she exploits a quasi-random assignment
1:19:30of land they changed the assignment of
1:19:32line to this company
1:19:34and she she focused on that um
1:19:38so what's next what i think where the
1:19:40big
1:19:41big topics for multinationals are well i
1:19:44think dynamics is a
1:19:45understudy topic um not less because
1:19:49data are difficult to get and
1:19:51models are difficult to work with
1:19:54taxation is another one
1:19:56um there again there is some
1:19:59restrictions on the theory part
1:20:01but i think it's very important to try
1:20:02to understand these
1:20:04issues one way of
1:20:08that fdi takes are mergers and
1:20:10acquisitions
1:20:12we in general don't don't distinguish
1:20:14mergers when acquisitions from
1:20:16greenfield fbi
1:20:18that they have important differences and
1:20:21maybe there is a big leader to
1:20:23incorporate finance but not
1:20:25in the tradition also what's the role of
1:20:28enemies in automation and the adoption
1:20:31of robots and how
1:20:32that will play with offshoring i think
1:20:34that's that's a growing area
1:20:37and finally um this global value change
1:20:40not only in production but also in r d
1:20:43so m
1:20:44a slice sliced guaranteed chain across
1:20:47countries
1:20:48the same way there is like the
1:20:49production chain and
1:20:52again not all around is the same
1:20:55and there might be more gains
1:20:58of doing what some but not the other
1:21:01well we need to see it in
1:21:03maybe with the theory and the data
1:21:07so thanks very much
1:21:10yeah i'm on time fantastic thank you so
1:21:14much natalia
1:21:15um let's open things up for questions
1:21:18um i'll start with the front row if
1:21:20there are people who have questions
1:21:22please just go right ahead
1:21:28i have a question uh thank you natalia
1:21:33i love your presentation and
1:21:36i was thinking throughout all the
1:21:38presentation just as
1:21:39this first point about the dynamics uh
1:21:44i'm trying to understand how this uh
1:21:46entry fixed cost of
1:21:48multinational works because this
1:21:50multinational is actually like entering
1:21:52the market but it's
1:21:53static decision right i think
1:21:57uh i i guess it's super difficult to
1:22:01to make this a dynamic problem inside
1:22:03this very complex
1:22:04static problem but uh
1:22:07i don't know what i don't know if you if
1:22:09you have thought about this problem or
1:22:11or or
1:22:12yeah so we we have a paper
1:22:15with dynamics and we need to sacrifice
1:22:19some stuff right um always the
1:22:24the you want some trick
1:22:27to make the problem separable that's
1:22:30kind of
1:22:31if if not i mean i don't know even how
1:22:33to start solving it even in the computer
1:22:36like um in a dynamic setup so we
1:22:39we have um we have a trick where
1:22:42basically if you assume independence
1:22:45across markets
1:22:48you you can you can solve them all like
1:22:50if you were
1:22:51looking when i say independence i mean
1:22:55and avoid the the complex problem means
1:22:57that you are looking at
1:22:58profits here right in country a
1:23:02and you say oh i entered or not and
1:23:05and it's not linked to what you do in
1:23:06country b
1:23:08okay so of course in static setups we
1:23:12have already
1:23:13people working this problem there is the
1:23:16entrance
1:23:16um fort and intel for sourcing
1:23:20and and there are algorithms that allows
1:23:23us to do that
1:23:24maybe at some point we need to put them
1:23:26together we haven't come up with a way
1:23:27of
1:23:28using the the complex combinatorial
1:23:31problem in
1:23:32dynamics to be honest but that's very
1:23:35interesting and fruitful
1:23:38area i would say thank you so much great
1:23:42presentation again
1:23:44thank you natalia i have a question for
1:23:45you from maria
1:23:48says how do these models capture
1:23:50productivity spillovers
1:23:51for the countries where multinationals
1:23:53locate that's obviously something that
1:23:56people think about a lot in the policy
1:23:57side
1:23:58i know that lots of policymakers are
1:24:00convinced there must be big productivity
1:24:01spilling yeah so
1:24:03there is no spillovers okay
1:24:06i don't know if you can hear me though
1:24:07but he waits till the end
1:24:10um so the
1:24:14there is no esp lovers in the sense that
1:24:17uh
1:24:17the domestic farm cannot get the
1:24:20productivity from
1:24:22the the the the foreign firm
1:24:26in that sense there is no spillover
1:24:29of course there is gains from the
1:24:30country they go because when you measure
1:24:32aggregate productivity now
1:24:33we have these more productive actors
1:24:36than in there
1:24:38okay there may be some channels for the
1:24:40labor market so prices everything
1:24:42happens to prices
1:24:44in this mod but
1:24:47yes i mean it's easy to put them maybe
1:24:50yes
1:24:51the problem is that that's why i
1:24:52restrict my my presentation of
1:24:55spillovers to
1:24:56to the empirics because i don't think we
1:24:59have like
1:25:00the model i mean to talk about
1:25:03spillovers
1:25:05i think that most promising channel
1:25:09if you allow me to say one more thing
1:25:11about that
1:25:12is more about the suppliers those we can
1:25:15work with
1:25:16right so there are suppliers and you
1:25:18know
1:25:19they serve better firms um there is some
1:25:22learning
1:25:23so maybe like introducing learning on
1:25:25those relations
1:25:27that's that's promising it's been doing
1:25:30people are doing it in trade
1:25:36other questions from the front row
1:25:44um hello yeah
1:25:47thank you natalia hi yeah yeah hi uh
1:25:50my questions are in uh twofold uh number
1:25:53one
1:25:54uh do you really think uh this concept
1:25:57of uh
1:25:58breaking down production process uh from
1:26:01country to country i mean gvcs basically
1:26:05uh do you really think it has had uh
1:26:08impact on uh economic growth and
1:26:11development
1:26:12in in i mean with respect to developing
1:26:14countries number one
1:26:15because the argument has been um
1:26:19neither here nor there and and then how
1:26:22do we
1:26:23factor in the role of uh variation in
1:26:25institution
1:26:27into the model uh i mean the particular
1:26:29model
1:26:30i mean as a determinant of when
1:26:33and how the direction of investment
1:26:37will flow into the model i i don't
1:26:40really see
1:26:42the mention of that and i think it's
1:26:44very critical thank you
1:26:45yeah so that's more as i said that's
1:26:48more than research
1:26:49uh for um like polandra season
1:26:54and thereafter research i mean um
1:26:57there are for instance um davinciour's
1:27:01work has been more related to exactly
1:27:04what you say
1:27:06here as i said there is not a global
1:27:08value change really the
1:27:09the main motive for multinationals to go
1:27:12into countries is
1:27:14market access to consumers um
1:27:18some sourcing can be introduced uh
1:27:22but not in the way you are thinking i
1:27:23think they are to talk about
1:27:25institutions and all that in fact the
1:27:28the work of
1:27:29the insurers it kind of tried to put
1:27:31these
1:27:32um institutional contracts um
1:27:35forces into
1:27:39these more quantitative models so maybe
1:27:42that's a a
1:27:43researcher that you can can can look for
1:27:47for more answers on that aspect um
1:27:52so there is this
1:27:55theme right you want to go up on the
1:27:57global value chain value
1:27:59the global value chain high value right
1:28:05there is something about that at least
1:28:07in the data it seems to
1:28:08we prepare like um for the asian
1:28:12development bank a report on this and
1:28:15some of the correlations go on the
1:28:19on the way you expect right it's better
1:28:22to have some
1:28:23an activity with more with more value
1:28:26added in the global value chain
1:28:29than that less etc but
1:28:33um yes but
1:28:36not in these models i mean
1:28:40all right i think it would be great to
1:28:42have all in one
1:28:43i mean and i think there are efforts on
1:28:45that direction
1:28:47natalia along the same lines perhaps
1:28:49thinking of again the policy side of
1:28:52things
1:28:53um there's a lot of conversation about
1:28:56distributional impacts
1:28:58of of fdi um
1:29:02where is the literature sitting on this
1:29:05so in real in this in this
1:29:08one of the concerns for the second paper
1:29:11we wrote
1:29:12was about that so
1:29:16we started like time ago was trump
1:29:18wasn't even president at that point but
1:29:20there was like
1:29:21uh these concerns about um
1:29:24well who's gaining who's losing that's
1:29:27why we
1:29:28we even in this if you want
1:29:31simpler mode where there is these
1:29:33workers choosing to go to
1:29:35innovation production sector and they
1:29:38are already
1:29:39rich enough that you can have
1:29:42losses even for the innovation work
1:29:45so um people are using a lot this type
1:29:49of
1:29:49framework and maybe in previous lectures
1:29:53um others talked about this this royal
1:29:56fresher type of
1:29:57setups where workers are heterogeneous
1:30:01and they choose where to go
1:30:03that's the the state of the literature
1:30:06in terms of
1:30:07of modeling but the concern is they are
1:30:09almost
1:30:11it's introduced um in
1:30:14like in in the models like really
1:30:18and as i said this um i didn't highlight
1:30:21that in this
1:30:22in this talk because um um
1:30:27is there and and and was one of the main
1:30:30concerns
1:30:31why we wrote the paper so yes
1:30:35what about like multi-factor models
1:30:38where the
1:30:38factor intensity might be different for
1:30:40domestic production
1:30:42versus uh you know multinational or fdi
1:30:46do those
1:30:47[Music]
1:30:51in my in that literature review
1:30:54there is one paper um uh
1:30:57changsung she wrote a paper on that with
1:31:00capital and how uh multinationals are
1:31:03different from the
1:31:04from the local farms okay
1:31:07so so yeah there is something and in
1:31:11general yes you see that three bars only
1:31:12have layer
1:31:15but um but there is some and there was
1:31:18an old literature in multinationals
1:31:20about this can you repeat who was the
1:31:21who's the paper that you just mentioned
1:31:23um was the one by
1:31:25chang soon in my s-u-n in my list
1:31:29okay mp with non-neutral technologies
1:31:33slide 35
1:31:37yes and i think there is a growing
1:31:40interest in
1:31:41automation and the role of
1:31:44multinationals and how they are going to
1:31:46do their offshoring
1:31:52there are a couple of very quick
1:31:54questions well some of them are quick
1:31:55and some of them not so quick that have
1:31:56popped up in the last
1:31:58the closing minutes here people sneaking
1:32:00in questions before the bell
1:32:02one i think is a clarification of a
1:32:04point you made earlier
1:32:06from francisco he says quick question
1:32:08regarding r d
1:32:10and trade so the domestic firm does not
1:32:12benefit from productivity from the
1:32:14multinational
1:32:15because is that because research is not
1:32:18done in sight
1:32:19or because it lacks the capital human or
1:32:21physical to do so or
1:32:23it's just not there in the model it's
1:32:25not there it's like by
1:32:27by assumption if you have your blueprint
1:32:29it's yours
1:32:31um aren't these done in the in the home
1:32:34country
1:32:36um and that's it right and
1:32:40and then the effect that is negative if
1:32:43the
1:32:43when the multinational comes in it can
1:32:45wipe out
1:32:46domestic farms during rmb and that's
1:32:49where that negative effect
1:32:51there might be the negative effect
1:32:52coming from that and that's a price
1:32:54effect
1:32:55yeah it's already saved again that's a
1:32:57price a price effect
1:33:00yeah yeah so there is a paper
1:33:04um again if you look at that list fan
1:33:08fam he has rnd optionally
1:33:11not spillovers but offshoring meaning
1:33:14that the multinational can
1:33:15choose not only where to produce but
1:33:18where to
1:33:19around it so that might be a starting
1:33:23point
1:33:25great um one one more and it's probably
1:33:28the last from kevin williams he says
1:33:31it's a
1:33:32very open-ended question he says i'm
1:33:34thinking what role does corruption play
1:33:35in these models is
1:33:37is it important to consider corruption
1:33:39um
1:33:40it's a big question yeah
1:33:45so we i think a simple correlation tells
1:33:49you that this
1:33:50this multinationals though the higher
1:33:52the corruption the least the
1:33:54multinational press
1:33:55right um there is some
1:33:59literature it's not model explicitly but
1:34:03that goes like this if you wish the
1:34:05shell team is like
1:34:07why you see so much south south
1:34:11fbi well maybe it's because
1:34:14you have a competitive advantage in
1:34:16working in those type of environments
1:34:19with weak institutions so i saw papers
1:34:23on that
1:34:24the theories are difficult i have to say
1:34:26i mean but but still i think like
1:34:28again the work by paul andres and
1:34:31having sure that's the way um
1:34:34maybe to go and again
1:34:38um there is this idea like multinational
1:34:42till very recently
1:34:43was a phenomenon of the developed
1:34:44countries there were very few flows
1:34:47to the south right and so
1:34:50and again the the same thing for trade
1:34:53the weaker the lower the rule of law the
1:34:56less multinational
1:34:57so um and in the rule of law
1:35:00it includes corruption right so that's
1:35:04that's a fact and it's a it's a stylish
1:35:06correlation if not causality but
1:35:09um but still in the models can be
1:35:13difficult yeah so
1:35:16thanks i think maybe that's a good point
1:35:18to stop and simply to say
1:35:20it's a it's clearly a field where
1:35:22there's lots of work
1:35:24yet to be done and so for all of you out
1:35:26there
1:35:27who are students and looking to looking
1:35:30to carve out spaces
1:35:32there's a ton of important questions
1:35:34important real world questions
1:35:36there's a set of models that offer a
1:35:38framework to begin thinking about these
1:35:40things
1:35:41and the you know it's not a blank slate
1:35:44but there's a lot to build on
1:35:46so on that note i i think we should
1:35:48really just thank you natalia for
1:35:50um for taking the time and again for the
1:35:52the contribution out of the
1:35:55kindness of your heart and the
1:35:56commitment to the
1:35:58to sharing knowledge i wonder if those
1:36:00of you who are in the front row would do
1:36:02what we've been doing and
1:36:03unmute yourselves um briefly so we can
1:36:06give natalya just a round of applause
1:36:08so thank you natalia thank you
1:36:12thank you very much yes the feel is open
1:36:15um
1:36:15work on it it's great
1:36:19thank you so thanks very much to
1:36:21everyone and
1:36:22um we hope you'll come back tomorrow we
1:36:25have our last
1:36:26lecture in the whole in the whole course
1:36:29klaus desmet is going to be talking
1:36:30about urbanization so
1:36:32please do join us and we look forward to
1:36:35seeing you all there
1:36:36many thanks again to natalia many thanks
1:36:39to the cepr team
1:36:40and uh have a good afternoon evening
1:36:44or whatever time of day it is for you
1:36:48so long thank you bye natalia you
1:37:34you

About this transcript

This page contains the full transcript of Supplemental lecture: Trade, FDI, and development by CEPR & VideoVox Economics, generated from the public captions YouTube serves with the video. The transcript has 13,558 words across 2,478 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.