YT YouTube2Text

YouTube2Text → Transcripts → Lecture 12: Applications to development

Lecture 12: Applications to development — Transcript

by CEPR & VideoVox Economics · 15,466 words · 2,658 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:28sorry should i take it away
0:30sorry oh no we'll wait just uh
0:33a few more seconds david for people to
0:35tell me yeah of course sorry i
0:36suddenly feared that pause was because i
0:39was meant to be talking
0:46great well let me go ahead and start us
0:49off um
0:50just say welcome to everyone for this
0:53next lecture
0:54in the in this series on
0:58key concepts in macroeconomic
0:59development this
1:01is for many of you not the first lecture
1:03and probably many of you have joined us
1:05already for a number of lectures we're
1:07really delighted
1:08to have david atkin here today to give
1:10the next lecture in the sequence
1:12just a word about ground rules and i
1:14think everybody's probably familiar with
1:16this by now
1:17those of you who are in the virtual
1:20front row here
1:21um have control of your over your
1:23microphones
1:25and your cameras if you can leave your
1:26cameras on that's great
1:28david has said he's happy to take
1:29questions along the way
1:31um and then we'll also ask david to
1:34leave a bit of time
1:35in the last 15 minutes or so to
1:38field questions from all of you and from
1:41and from the broader group
1:42for those of you not in the front row um
1:46david won't be able to see if you put
1:47your hand up so what we ask you to do is
1:50to put your
1:51questions in the q a where i and
1:54possibly others will do our best to
1:55answer as many as we can
1:57and the ones we can't answer will will
1:59make david answer
2:00at the end um so those are what i've got
2:03to say on the ground rules
2:05i guess beyond that what i'd really like
2:06to do is to take a moment just to thank
2:09the united kingdom's foreign
2:11commonwealth and development office
2:13which funds the steg program and which
2:16is therefore the
2:17the source of the whole operation but
2:21although they fund the program
2:22i would just say again that the the
2:26lecturers who are taking part in this
2:28series are really doing this
2:30out of their own kindness and
2:32intellectual generosity
2:33we really appreciate david taking the
2:35time today and i want to thank him
2:37i also want to thank the team at cepr uh
2:41nadine clark um christy mcneil
2:44lauren waring mandy chan who's been
2:47sending out emails
2:48even when she's doing other things and
2:50so we really appreciate all the help
2:51that we get from them
2:53so david on that note without further
2:55ado i'll turn it over to you
3:01great thank you much i can everyone see
3:03my screen okay
3:05perfectly
3:08oops all right so
3:12um thank you very much uh for inviting
3:14me and giving me this uh
3:16opportunity uh uh to lecture to you guys
3:20so uh you know what i wanted to do today
3:23was
3:23really tackle two topics uh measuring
3:26spatial frictions and then
3:28uh thinking about some of the
3:29implications of those spatial frictions
3:32and i was wildly uh uh uh optimistic
3:35when um
3:36writing the slides and so there's uh way
3:38more material than i'll
3:39possibly be able to get through uh so
3:41i'll go relatively quickly
3:42in in places but the slides are online
3:45if you want to
3:46uh follow up on some of the things later
3:49okay so uh you know broadly speaking
3:53spatial frictions affect developing
3:54countries in many many uh
3:56uh uh different ways and uh in you know
3:59an hour and a half
4:00i i certainly don't have time to cover
4:02uh all of them
4:04and so instead i'm going to focus on one
4:05particular aspect today which is
4:07uh spatial frictions that impede the
4:09movement of goods and services which we
4:11often call trade costs
4:12and so i won't be talking much about uh
4:14migration costs which i melanie talked
4:16about
4:17in detail uh uh last week now of course
4:20these
4:20two may interact and there's uh some
4:23recent economic geography models where
4:25you have frictions on both labor and
4:27goods moving between locations
4:29and there's interesting uh implications
4:32of those but i won't
4:33i i i talk too much about that today
4:37okay so as i mentioned the kind of two
4:39questions we'll be talking about is
4:41first
4:41you know how large are these spatial
4:43frictions uh
4:45why do we care how large they are well i
4:46mean the obvious reason is
4:48if we're writing down models and often
4:50quantitative models trying to understand
4:52uh uh uh the impacts of these frictions
4:54or how things might change if you change
4:56these frictions
4:56we kind of need some measures of the
4:58frictions and so they're important
5:00inputs
5:01in that way but also from a positive
5:03perspective it's important to
5:05understand uh the trade data i
5:08understand why it has the features it
5:09does
5:10uh that's not just for uh uh
5:13trade flows within or across countries
5:15uh it's also true
5:17um for international macro uh observed
5:20and rogoff have a very influential paper
5:22noting how possibly uh all of the six
5:25big puzzles of international macro
5:27uh can be explained at least partially
5:30uh
5:30through a more realistic treatment of
5:32trade costs in those models
5:35okay so that's why i think it's
5:37important to measure these things if
5:38they are large and they do appear to be
5:40uh at least in developing countries then
5:43what are the effects of reducing these
5:44frictions that is kind of the second
5:46question i
5:46will talk about there's obviously a vast
5:48amount of money uh spent on reducing
5:50trade costs this is a big policy
5:52question
5:52as well as a question of just pure
5:55academic interest uh but trying to
5:58understand for example
5:59uh what was the benefit of a large uh
6:01road building project
6:02uh uh uh we need to kind of know the
6:04answer to these questions so
6:06i'll try and i i i talk about both of
6:08these uh
6:09uh in the lecture today
6:13okay so it has a measuring trade cost so
6:16what do we mean by trade costs well you
6:19know roughly
6:20uh broadly speaking we're talking about
6:22all the costs that are going to impede
6:24uh our trade from moving between an
6:26origin oh and a destination d
6:28and that covers a range of things the
6:30most you know well-known types of trade
6:32barriers or things like
6:34taxes or tariffs or transportation costs
6:36like the cost of going across roads
6:38but of course uh there are many other uh
6:41impediments
6:43to the kind of movement of these goods
6:45everything from kind of the need to pray
6:46bribes or corruption
6:48to contractual frictions which are much
6:50more severe across
6:52space and this is avenue graves
6:53fundamental problem of exchange
6:55you kind of have to put your trust that
6:57someone will deliver
6:58some goods to you after you've paid them
7:02uh and that adds a complexity
7:04particularly institutionally uh poor
7:06environments
7:08um there's also information frictions
7:10from doing business over space not
7:12knowing
7:13you know prices pertaining uh or
7:15preferences pertaining in
7:16in various locations uh which are less
7:19severe when you're doing business
7:20locally or the need to secure finance
7:23while those goods are in transit and
7:26obviously uh
7:26the further you're shipping them the
7:28longer they're going to be in transit so
7:31all of these types of uh uh costs we
7:33group together
7:34uh into uh kind of trade costs
7:38okay so you know the first question is
7:41are trade costs large uh in developing
7:44countries and you know there's some
7:45debate about this question and we'll
7:47kind of try and shed some light onto it
7:49in the next few slides you know a common
7:52argument is that you know a fundamental
7:54problem in in the developing world is
7:56very high trade costs everything from
7:58you know very poor road quality uh to
8:01the
8:02need to pay lots of bribes at various
8:04points
8:05on these journeys to poor contracting
8:08institutions
8:09there's a nice economist article from
8:11more than a decade ago now
8:13entitled the road is hell to hell is
8:15unpaved uh which just kind of
8:16follows a beer truck uh on its uh
8:19deliveries
8:20uh in cameroon as it has to travel
8:22across the country with
8:23a lot of insight on on some of these
8:25costs
8:26so you know these types of stories would
8:28tell you that your trade costs are very
8:30high in the developing world
8:32but when you actually go and try and
8:34look at say shipping rates they're not
8:35that high
8:36uh if we look you know within africa or
8:38on some of the more expensive land
8:40routes
8:41at least in international terms uh say
8:44mombasa to kampala
8:45or uh uh
8:49prices are around 10 cents per tonne so
8:52you know
8:53uh per ton kilometers so you know while
8:55those things can certainly add up
8:57if you're shipping a container full of a
8:59kind of reasonably high value product
9:02these are going to be a kind of a small
9:04fraction of those costs
9:06also if we look at kind of costs such as
9:08tariffs for international trade
9:10uh uh or or taxes internally those
9:14aren't necessarily that big uh and you
9:16know where they are big
9:18uh we often see in the developing world
9:20that they're massively evaded and we'll
9:21talk
9:22more about that in due course and so
9:25it's kind of de facto they're not uh
9:27potentially such big barriers and
9:29similarly these kind of uh
9:30contracting problems uh there's been
9:32some nice recent work by people like
9:33rocco machiavello showing
9:35how repeated games and reputations uh
9:38may
9:38circumvent a lot of these contractual
9:40issues uh
9:42and we'll talk a bit about that uh uh
9:44later
9:45and so you know potentially these aren't
9:47uh as problematic as we might think they
9:49don't impose such large costs
9:51so how can we go about actually trying
9:54to put some numbers on this well
9:56broadly speaking there's kind of three
9:58ways we can measure spatial frictions
9:59and i'll kind of touch
10:01on all three uh the first is kind of
10:04direct measurement that's kind of the
10:05brute force approach
10:07uh where we're just going to go out and
10:09try and measure some components of these
10:11transport costs now some are quite easy
10:13to measure
10:13things like transport prices tariffs
10:17many aren't and i'll talk about uh kind
10:19of some more creative ways
10:21of trying to estimate trade costs uh in
10:23developing countries
10:24uh directly in the next uh a few slides
10:29so there's kind of some nice work done
10:31at the world bank
10:32uh where they essentially do uh
10:34something like they're doing business
10:35surveys if people are familiar with
10:37those
10:37uh but for transportation costs and in
10:40particular
10:40they look at how easy it is to try and
10:43do things such as
10:44uh export products from from our country
10:47so here's for burundi you have to go
10:50through 17 different steps to export uh
10:52something from burundi everything from
10:54getting letters of credit
10:56to passing various technical controls
10:59and they not only do they measure how
11:01many steps there are but they also
11:03measure the time taken
11:04for each of those steps on average and
11:06these are through surveys of freight
11:08forwarding firms
11:09around the world and so here's the case
11:11of burundi again
11:12and you can see these 17 steps take a
11:15total of 70 days 7-0
11:17so that's that's an enormous number of
11:19days and obviously this is one country
11:21but this kind of tells you that these
11:23uh uh uh uh uh kind of uh measures of
11:25these
11:26uh uh uh freight forwarding from freight
11:28forwarding funds
11:29uh can give us uh you know very uh uh
11:32large uh uh
11:33uh numbers at least if we converted
11:36these uh time costs
11:37uh into uh um uh uh sort of
11:40uh monetary terms so you know they don't
11:43just do this for burundi they do this
11:45for many countries and as you might
11:46expect these costs
11:47are much higher in the developing world
11:49than in developed countries
11:51um this may be a little misleading in
11:54the sense that even if you have a mean
11:56for say africa in the middle east of 41
11:59days as opposed to europe of 22 days
12:02you know maybe we should care more about
12:04the minimum and the maximum here
12:06and there's kind of some work
12:10here by lance pritchard who kind of
12:13points out that
12:14you know when you look at these types of
12:16surveys the kind of big firms who do a
12:18lot of trade
12:19they've often worked out ways to do this
12:21quickly if that involves
12:23uh you know having very good
12:24relationships with government officials
12:26and such like
12:27that they pay the cost to do that and so
12:30you know maybe this isn't as big a deal
12:31as it seems just because
12:33uh the the in terms of quantities of
12:35people who are exporting a lot
12:37uh they're able to get uh uh much more
12:39favorable uh
12:40uh uh conditions than if i want to just
12:42do one freight export once
12:44uh and so uh this kind of a standard
12:47deviation these minimums and these
12:48maximums tell us
12:49that if you see kind of the minimum here
12:52ranges between
12:53you know ten and a hundred and sixteen
12:56uh there's a lot of variation there
13:01okay that's uh kind of by interviewing
13:03freight forwarders you can also kind of
13:05survey
13:06uh people who travel along these routes
13:08uh ben
13:09alkin and co-authors have done work like
13:12this thinking about
13:14trucking in indonesia and thinking
13:16particularly about corruption
13:18and they've just gone and worked out uh
13:20uh through surveys how much
13:22uh uh different costs add up to the kind
13:25of cost of moving goods between those
13:26locations
13:27and they show that kind of bribes
13:29whether that be extortion or protection
13:31payments
13:32uh um uh i i i come to a substantial
13:36share
13:36uh uh of uh the costs of the trip uh
13:39it's about fifteen percent here
13:41uh uh of the cost uh with uh you know
13:43fuel being about fifty percent
13:45um and those are everything from you
13:48know payments to checkpoints
13:50uh and a way stations to protection fees
13:52or or convoy fees
13:54and they can go and uh uh uh uh
13:57not only um uh look at these averages
14:00but also see
14:01how those payments vary along the trip
14:03and they have this
14:04kind of nice uh uh model and result uh
14:07where they show at least
14:08in one of the two routes that they uh uh
14:10look at
14:11that these kind of uh bribe payments the
14:13extraction that's going on
14:15is increasing uh as you move closer to
14:18the end of the trip
14:19and that's consistent with the idea that
14:21there's you know more surplus to extract
14:24as you're close to the end of the trip
14:25and so uh some optimal
14:27extraction by uh some
14:32bribe taker uh would charge a higher
14:34price i
14:35uh demand a higher bribe are close to
14:37the end of the trip
14:40okay as i mentioned uh you know when it
14:43comes to international
14:44trade uh then um
14:47you know often we see very relatively
14:50high tariffs in developing countries
14:52but again that may be misleading if a
14:54lot of people are evading those uh
14:56uh tariffs and sandra sequeira has a
14:59nice paper showing exactly that from
15:00mozambique
15:02uh essentially uh as she compares before
15:04and after
15:05a um reform
15:08which um uh liberalized trade
15:12with south africa and massively lowered
15:14uh the tariffs with south africa
15:16and she shows that prior to this uh a
15:18kind of massive drop in tariffs
15:20about 80 percent of shipments uh were
15:23paying a bribe
15:24uh to customs officials and
15:27most of the time that bribe was for
15:29tariff evasion which
15:31typically in this context means i'm
15:33shipping in something which should pay a
15:34high tariff
15:36i pay the customs official through an
15:38intermediary
15:39uh to go and write that import down
15:41rather than an import of
15:43steel that has a high tariff to write it
15:45down as an import of
15:46of iron which has a low tariff and so
15:49this way they end up paying a lower
15:51rates
15:53and once you put in the kind of lower
15:55tariffs then people stop paying bribes
15:57and they're willing to pay the tariffs
15:59so the proportion paying a bribe falls
16:01from 80 to 26
16:03and the bribes that are still being paid
16:05are primarily to deal with congestion to
16:07essentially get your stuff to move
16:09through faster
16:10uh rather than to pay a a lower a rate
16:15um you know one thing that's been noted
16:17in in many uh
16:18uh uh other settings as well is just how
16:21cheap bribery is
16:23the ratio of bribe amount to tariff duty
16:25saved is about seven percent
16:28so uh this has kind of been noted in
16:30many settings
16:31uh these officials are not extracting
16:33all of the surplus they're extracting
16:34surprisingly little of the surplus uh uh
16:37uh you're
16:38getting a a kind of uh
16:41big big discount on your tariff savings
16:44uh for
16:44a bribe that's only uh seven percent of
16:47the discount you're getting
16:49um okay so
16:52you know the kind of implications of
16:53that is is uh uh that we should you know
16:56think about twice about these official
16:58barriers uh
16:59uh uh uh how large they may be uh uh
17:02when there is a substantial uh
17:04evasion there's the final
17:07uh kind of direct measurement i'll talk
17:09about comes from meredith starts and she
17:11has a
17:11a very uh uh uh interesting paper
17:15uh where she goes and surveys uh
17:17nigerian traders
17:18uh are mainly uh in goods such as
17:20clothing uh electronics etc
17:23and she highlights that uh that distance
17:25creates uh
17:26essentially uh uh two uh uh uh
17:30uh problems uh the first is kind of a
17:33search problem
17:34which is i want to know what goods are
17:36available you know say i'm a trader in
17:38nigeria i want to know what the new
17:39products are in china
17:42and it's hard for me to do that i are
17:44from nigeria
17:46and so that's kind of a surge problem
17:47then there's also a contracting problem
17:50which is you know it's hard for me to
17:51trust uh uh
17:53that you know once they send the money
17:54they'll send the goods uh uh
17:56or they'll send the right goods rather
17:58than uh some broken ones or not exactly
18:00what i ordered
18:01and there again that's much easier if
18:04you can go
18:04and you know hand over the money as you
18:06see the goods being transferred locally
18:08much harder if you're trying to do this
18:10from nigeria and get them to
18:12kind of ship the goods are from china
18:16so what uh uh uh meredith starts
18:19uh tries to do is is try to you know get
18:21a real handle on this
18:22just through uh uh setting up her own
18:24surveys and measurement
18:26and there's this pretty stark finding
18:28which is just how much travel is going
18:29on
18:30uh amongst these nigerian traders so
18:32this is kind of the
18:34fraction of shipments uh that they that
18:36they that they uh
18:37uh purchase uh where they end up
18:40traveling
18:42and um as you can see when they're kind
18:44of getting shipments from togo ghana
18:46been in countries nearby
18:48it's cheap to travel there and almost
18:50all the time they travel there
18:52before getting a shipment meanwhile if
18:55they're going to
18:56you know these far-off countries to buy
18:58from such as the uk or china
19:00then they're much less likely to travel
19:03only maybe 30 percent of the time for
19:05china
19:05um it also costs a lot more to get there
19:08about two and a half thousand dollars
19:10per trip uh including accommodation and
19:13uh visas etc
19:14uh and so unsurprisingly they're less
19:16likely to go
19:17but the thing that's kind of surprising
19:19here at least uh
19:21uh compared to my priors is just how
19:23many people are going
19:24even about 30 of of of transactions from
19:27china
19:28there's kind of a trip by the trader to
19:30china associated with it
19:32and these trips are expensive uh
19:34compared to the transportation or tariff
19:36costs on one of these trips where they
19:38actually travel rather than just order
19:40directly
19:41the travel costs actually are more than
19:43the transportation and tariff costs
19:45and so we you know worry a lot about uh
19:48you know how big are these terriers how
19:49much it costs to move goods and trucks
19:51uh we worry much less about how much
19:53does it cost for a trader
19:55uh uh and how easy is it for a trader to
19:57fly to china
19:58spend a few days there and come back
20:01and i should say these trips are in for
20:03people to
20:04to uh you know stick stuff in their
20:06suitcase and fly back with it
20:08uh uh the stuff typically travels in
20:10containers uh
20:11uh these are really uh uh trips to work
20:14out what they want to buy
20:15uh and maybe to uh actually do the deal
20:18um okay uh so these costs are quite
20:22large
20:24the the third thing which is maybe even
20:26more surprising
20:27is that they keep on traveling even many
20:30years
20:30uh after they've started trading with a
20:32supplier so
20:33just looking at this third column there
20:35are trader and country fixed effects
20:38and there's essentially no relationship
20:40uh
20:41for say you know a nigerian trader who
20:44buys from
20:46china they're just as likely uh uh to
20:49buy from they're
20:50just likely to travel to china when
20:52they're buying from someone they've
20:53bought from
20:54uh for many years versus someone they
20:56haven't bought from before
20:58okay and this is kind of very surprising
21:00in terms of uh
21:02uh we would think that these contracting
21:04problems are alleviated substantially
21:07uh uh uh the longer your relationship is
21:09you've built up a reputation
21:11uh uh uh and uh there's kind of repeated
21:13game here
21:14and so you'd think that you would uh uh
21:17you know be able to trust them more and
21:18you travel less
21:20over time and so that's a pretty stark
21:23uh
21:24finding finally
21:27when you do travel what are the benefits
21:29of traveling well
21:31you get a lower unit cost you're paying
21:33less for each unit that you buy
21:36so you're better able to kind of pick
21:37the cheaper products
21:40you're more likely to get new styles you
21:43know especially when it comes to things
21:44like clothing uh
21:45you're more likely to pick up kind of
21:47the latest fashions you're more likely
21:49to add new suppliers
21:51you're also spending a longer time
21:53between purchases
21:55and you're buying bigger batches so
21:57typically when you travel you go
21:59less frequently and buy bigger bigger uh
22:02uh
22:02uh shipments as you'd expect given the
22:04large fixed cost uh
22:05of traveling so these are all uh kind of
22:08suggestive facts
22:10uh uh kind of uh talk towards these two
22:14frictions we talked about earlier
22:15of they're both being kind of a search
22:17problem where i i want to go and see
22:19what the products available are
22:21and that's hard at distance and a
22:24contracting problem and it's also a hard
22:26distance
22:27and so uh you know starting to kind of
22:30think about a
22:31a a kind of measurement here going
22:33beyond direct measurement
22:35um meredith writes down a model that's
22:37going to allow her
22:38to kind of back out the relative
22:39importance of these two frictions
22:41and to do that she's going to take a a
22:43kind of monopolistic com
22:46competition framework of trading firms
22:49and add a stocking problem how much do i
22:51want to buy
22:53there's a fixed cost of each purchase so
22:55how much stock should i hold
22:57a search problem where i'm looking for
22:58new varieties
23:00and a contracting problem where you're
23:02going to have this repeated game
23:03and this moral hazard where they may
23:05default and just take your money uh
23:08and not send you the stuff or send you
23:09crappy stuff
23:11and so uh you're trying to uh kind of
23:14play a repeated game
23:16and what she shows as as as in many
23:18repeated games is
23:20uh in equilibrium the trader is going to
23:22pay a premium to make sure that the
23:23seller doesn't cheat
23:25right by making this relationship uh
23:27sufficiently valuable to them
23:29uh it's in the interests of of the the
23:31uh kind of seller
23:32uh in china uh or in whatever country to
23:36to uh uh you know send the goods after
23:38the payment
23:39and you know what how travel interacts
23:42with this
23:43is that travel both allows more
23:45up-to-date vintages of goods to be
23:47purchased you're able to see the latest
23:49things and buy them
23:50it also mitigates the temptation for the
23:52seller uh
23:53uh uh uh to cheat okay it's now a
23:56a kind of less worthwhile for the cell
23:59to treat if you're if you're traveling
24:00regularly
24:01uh since these order sizes are gonna be
24:03smaller and so they're optimally
24:04deciding uh
24:05kind of how much uh uh uh to travel
24:08and intuitively what's going to kind of
24:11pin down
24:12how large the contracting friction is
24:14relative uh
24:15out to the search friction is you know
24:18the probability of me buying from the
24:19same supplier again
24:21is is telling us about how important
24:23this contracting friction is
24:25it's it's that continued uh relationship
24:28which is allowing
24:29uh uh uh uh us to solve this uh uh
24:32uh contracting friction problem uh uh by
24:35uh
24:35having a repeated game meanwhile the
24:38choice of whether you
24:39uh at travel versus order
24:42is pinning down the search friction you
24:45know i can keep on ordering from you to
24:46solve that repeated game
24:48uh but traveling versus ordering uh
24:51remotely
24:52that's what's pinning down how important
24:53the search is for me
24:55uh and so uh you know using that
24:58intuition
24:59and the structure of her model actually
25:01shows that kind of both these frictions
25:03are
25:03are important and removing them together
25:06and there's some complementarity between
25:08them
25:08uh increases welfare in the sector by
25:11about 14
25:12which is about two-thirds of the gain uh
25:15from eliminating physical or regulated
25:17costs that we normally think about
25:19as a kind of primary uh distance related
25:22uh spatial frictions
25:25the other thing that's uh uh uh
25:27interesting here
25:28thinking about development more broadly
25:30is uh
25:31you know is this something unique to
25:33developing countries or not and and our
25:35answer is
25:36uh uh probably yes and in particular
25:40if nigerian consumers acted like
25:42american ones where
25:43they go to kind of very large uh uh
25:46shops like walmart rather than
25:47go to markets and small traders and when
25:50they do go to those shops they spend
25:52much more because they're wealthier
25:55then these types of travel costs become
25:58inconsequential the travel costs of the
26:00walmart buyer going to china
26:02doesn't really matter i i you know it's
26:04one
26:05one two trips for millions of dollars of
26:07sales
26:08it really does matter for nigerian
26:11traders
26:11uh where it's one trip uh for fifty
26:14thousand dollars of sales or one trip
26:16for thirty thousand dollars of sales
26:17uh uh when that trip costs three
26:19thousand dollars
26:21the other reason this matters a lot for
26:22developing countries there's been a
26:24little more work on this since then
26:26since emeritus original uh paper came
26:28out which was uh
26:29four or five years ago is that um
26:33international travel is particularly
26:36expensive uh
26:37for people from the developing world
26:39firstly there are a few flights
26:41and so you often don't have direct
26:43flights uh to say you know
26:45uh between where you live in nigeria and
26:47the place you want to go in china
26:49and so that adds to the cost but uh the
26:52biggest cost is often visas where it's
26:53often very expensive
26:55and and takes a lot of time to obtain
26:58visas to travel to these countries
26:59uh in order to do this trade and uh
27:02there are potentially very large uh
27:03uh gains uh to welfare uh from countries
27:07uh working out ways to reduce uh the
27:09hassle uh
27:10uh and cost of of traveling uh
27:12internationally
27:14uh can i ask you a question hello david
27:16yes okay
27:18oh so sorry um it seems
27:22a little bit that uh that this cost of
27:25traveling will be especially
27:28uh true for importers so nigerians
27:30traveling from the from
27:31to the united states to import something
27:34is it so
27:35this is also the case for like exporters
27:38and how
27:39that matters in this setting yeah so we
27:42don't have a
27:43similar study to this as detailed with
27:46micro data um
27:48on exporters but uh surely that's that's
27:50the case for exporters too
27:52i guess uh um exporters tend
27:56uh at least african exporters tend to be
27:58pretty large
27:59uh and so uh potentially the scale of
28:02of exporters is quite different to the
28:04scale of importers
28:06um but that's going to be an empirical
28:07question my guess is that would be true
28:09in nigeria but may not be true in the
28:11u.s
28:11but many of these same issues would
28:14occur i mean there's a question of
28:16you know who's doing the traveling if
28:18you know why aren't the american firms
28:20traveling to nigeria
28:21to buy from from the nigerian firms
28:24and so uh you know in some cases they
28:26might in other cases
28:28you may need to go to trade fairs to
28:30make people aware
28:31that nigeria has these products and can
28:33export them
28:34and obviously uh not obviously but uh it
28:37is true that
28:38uh developing country export promotion
28:40agencies spend a lot on doing exactly
28:41that
28:42we may think that there are particular
28:43information frictions which
28:45which make rich countries unaware of the
28:47product makes in poor countries
28:49and poor countries well aware of the
28:51product mixed in rich countries
28:53or much better aware and so you get this
28:55asymmetry
28:56and a policy may be needed to alleviate
28:59that
29:00um but i i i uh it's an interesting
29:03question
29:04uh uh on uh kind of thinking explicitly
29:06about this asymmetry
29:07typically papers have looked at one or
29:09the other rather than
29:11than both of them at the same time
29:15excuse me hello david uh sorry
29:19my question is with respect to the
29:23the study we're reviewing is that does
29:26the study really account for
29:28a possible aggregation with respect to
29:33uh the importation of the goods because
29:36it's possible that the importer uh
29:40actually does the import for more than
29:43uh for other other uh
29:46people so thus reducing the cost of say
29:49because they now do
29:50larger importation uh one importer
29:54uh taking care of like 10 small right
29:57so so i didn't give a lot of details
29:59about the study i recommend
30:00you read the paper it's a very good
30:02paper so uh most of these are
30:04wholesalers
30:05so yes they are de facto importing for
30:08for many people because they're then
30:09selling to many shops uh
30:12and so these are actually the orders
30:13these guys do
30:15they may not be they're typically not
30:16the final seller of the goods
30:18uh so yes these guys are already
30:20importing de facto for many retailers
30:23um uh uh but um you know
30:27these are the actual uh uh flows uh uh
30:30uh
30:30uh uh that purchases made by these guys
30:34even if they're grouping together orders
30:36that
30:36they're then going to go and sell on to
30:37other people
30:39okay thank you
30:43great should i push on any other
30:45questions
30:49okay so that's direct measurement we've
30:52already seen a little bit of using the
30:54model structure to kind of back out
30:55spatial frictions although not in detail
30:59so easily the kind of most uh
31:02common way to uh kind of measure trade
31:05costs
31:06uh by backing them out uh uh uh uh from
31:09uh uh uh uh other information is
31:13backing them out from trade flows uh and
31:16you know the attractiveness of this is
31:18that we're only ever going to get so far
31:20by um uh trying to um
31:25measure the cost yes we may do a decent
31:27job at measuring a handful of them
31:30but it's just so hard to kind of get the
31:32full costs uh for example we may get the
31:34cost of
31:35of you have to pay the the shipper to
31:37move the good from a to b but not the
31:39time costs and the hassle costs
31:41uh that involves at your end and filling
31:42out the forms
31:44so the kind of most commonly employed
31:46way for getting the kind of full extent
31:48of trade costs
31:49uh is essentially exploiting a
31:51relationship we call the gravity
31:52equation
31:53uh and that's basically a relationship
31:56between
31:57our kind of trade flows and
32:00trade costs to some elasticity
32:04condition on some fixed effects and it's
32:06a particular way
32:08of inferring trade costs from trade
32:10flows where the idea is very simple
32:12which is
32:13we see how much trade there is in the
32:15real world
32:17and then we compare that to how much
32:19trade at least our model says there will
32:21be if there were no frictions
32:24and the difference between the trade we
32:25actually observe and how much trade
32:27there should be with no frictions
32:29uh must be the trade costs and then
32:31we're gonna
32:32i i i take that difference and call it
32:34uh our trade costs
32:36and the kind of particular structure
32:38gravity puts on that
32:39is it gives us this particular
32:41specification typically you derive this
32:44from cs preferences
32:46uh uh identical across everyone in the
32:49world or everyone within the country
32:51depending if you're doing internal or
32:52external
32:53trade costs uh it's less restrictive on
32:56the supply side there's a range of uh
32:58supply side micro foundations
33:00uh but in almost all cases we're going
33:02to need uh the cs preferences
33:04and that's going to give us a functional
33:06form as follows which says
33:07the amount of trade or the log
33:11value of trade between i and j of good k
33:15is equal to something that depends on
33:18the origin country
33:20uh uh and good k something depends on
33:22the destination country
33:23and good k so that might be you know how
33:25productive i am
33:27this is going to be related to you know
33:28how big the demand is
33:30in be plus every other country's access
33:32to those
33:34locations and it depends on the trade
33:37costs between
33:38i and j ah for that good uh to some
33:41elasticity uh
33:42epsilon okay and if you can write if
33:45your model gives you this relationship
33:46uh uh we we say the you know the model
33:49uh uh uh
33:50generates uh gravity and you know what's
33:53nice about this structure
33:55uh is that essentially uh uh you know
33:58all the bits which are kind of a pain
34:00here
34:01uh about you know your own productivity
34:04uh
34:04and uh everyone else in the world's
34:06productivity uh as well as demands
34:08around the world are all captured in
34:10these two fixed effects
34:11we're getting a lot of mileage from the
34:13ces preferences here
34:14which is essentially telling us that you
34:16know i'm going to
34:18spend the same way uh uh on on on
34:21different goods irrelevant of where they
34:23come from just based on
34:24uh kind of how cheap they are it's gonna
34:26give us essentially a separability
34:28between
34:28my consumption and production decisions
34:30uh and the distribution of trade
34:32across the world but if if we're willing
34:34to buy this and there's strong empirical
34:36support
34:37that this this relationship holds
34:39reasonably well
34:42then uh we can actually go and quite uh
34:45simply back out trade costs and head and
34:48reese
34:48kind of have the uh kind of simplest way
34:50of doing this which is
34:52let's make an assumption that
34:54international trade i
34:55trade with myself tau iii in good k
34:58is free so that's making an assumption
35:01that there are no trade costs
35:03for trading between myself if i'm
35:05thinking about
35:06you know uh trade costs within countries
35:09then this would
35:10instead be an assumption that trade
35:12within your location is
35:13is free and it's costly to trade outside
35:16your location
35:18then we can get a lot of mileage as long
35:21as we further assume that it's just as
35:23costly to send something from i to j as
35:25it is from j to i
35:27so these costs are symmetric that's not
35:30quite true there's back cool issues if
35:32everyone wants to move stuff in one
35:33direction it's often
35:34expensive to move it in one direction
35:36and cheap in the other way as you have
35:37to get the trucks back
35:39but we think it's a reasonable
35:41assumption in many settings
35:43then just very simple algebra just
35:46essentially taking double differences of
35:48this expression
35:49for the gravity equation tells us that
35:53uh we can extract the tau i j
35:56to some elasticity just from
35:59uh this product of how much i trade
36:03with you how much you trade with me over
36:06how much i trade with myself how much
36:07you trade with yourself
36:09okay and essentially the logic here is
36:12if you trade as much with other people
36:14as you trade with yourself then we're
36:16down to
36:17there being no frictions so you can
36:19think of this as maybe there are
36:21frictions for trading with yourself and
36:23this is just a normalization
36:24how much higher uh how much more costly
36:27is it to trade with other people than it
36:29is to trade with myself
36:31okay so you know this doesn't come for
36:33free but
36:34uh uh it is very uh uh uh relatively
36:37simple to calculate what do we need well
36:40we need to know how much i trade with
36:42others that's typically recorded uh
36:44certainly international trade data
36:46domestically there's a number of
36:47countries and situations where we have
36:49trade flows within countries so those
36:52are these two objects
36:54and then we need to know how much i
36:55trade with myself that's often a bit
36:57harder
36:58essentially you're going to need to know
36:59how much you produce and how much you
37:01export but if there's decent data on how
37:03much i produce
37:04then i also know how much i trade with
37:05myself and then finally i'm going to
37:08need to know in elasticity
37:10you know this is a hard object this is
37:12an obvious object to need
37:14essentially we're trying to back out you
37:16know prices here
37:17uh uh uh from from our quantities and to
37:22do that we're going to need to know
37:24some elasticity and so if you think you
37:26can estimate that elasticity
37:28or you have some outside source for it
37:30then
37:31you can go and back out these tau ijs
37:35okay so a lot of people have done this
37:37in many settings internationally
37:38internationally
37:39uh the world bank has even produced a
37:41trade cost database
37:42based on uh this inverse gravity as they
37:45call it
37:46uh sorry i cut off the log gdp per
37:48capita
37:50a label from the x axis here and uh
37:53uh here you can see uh kind of all of
37:55the estimates these are aggregates
37:57for each of these countries for trade
37:59different products to different
38:01countries they break it up between
38:02manufacturing and agricultural products
38:04and they're all bilateral but these are
38:06just aggregates for each country
38:08and as uh you might have expected uh
38:10trade costs us do appear to be higher
38:13for less developed countries there's a
38:14pretty a tight
38:16negative relationship these countries in
38:18grey are all kind of small island
38:20nations
38:21uh they seem to be kind of off the curve
38:24which may not be so surprising
38:26these are all landlocked countries in
38:27blue which also may be a slightly off
38:29the curve
38:30a uh but you can see um
38:34this kind of general uh uh uh
38:37pattern um maybe these aren't all
38:41landlocked
38:42countries um
38:47yes i guess they probably are so um
38:51uh and so this is kind of some
38:54some evidence for that early conjecture
38:56our trade costs particularly high in
38:57developing countries
38:58well if you back them out in this way
39:00then yes they they do relatively more
39:02trade with themselves than with others
39:03compared to rich countries
39:08okay so that's kind of uh the extreme
39:12version
39:13of estimating trade costs and gravity
39:14where i want to get the full trade costs
39:16from gravity
39:17more common is i'm going to kind of take
39:20these gravity relationships
39:22and then rather than trying to extract
39:23the full trade costs which will
39:25involve know that symmetry assumption we
39:28wrote down
39:29as well as assuming that tau i i i is
39:32one
39:33instead we can just kind of parameterize
39:34this function and then
39:36you know project this kind of
39:39residual on various observables like
39:42distance
39:43and we find as you might expect that
39:45trade costs
39:46uh in terms of you know how much less do
39:49i trade with each other with with other
39:50countries
39:51than in a frictionless world uh we do
39:53indeed find
39:54uh that declines with a distance
39:59um so you know people have looked a lot
40:02of different barriers in this way
40:04whether we have kind of the same
40:05currency whether we're both members of
40:07the wto
40:08uh uh you know roads port access etc
40:12language barriers colonial ties
40:14information barriers based on
40:16uh uh kind of ethnic uh uh uh uh
40:19shared ethnic networks uh etc
40:23um and um
40:26even a paper on whether or not a
40:29cia-sponsored coup
40:30uh reduces uh uh uh kind of trade
40:33frictions
40:34uh the answer to that seems to be yes uh
40:36the cia was kind of helping out
40:38american commercial interests
40:41it's not necessarily that easy to go and
40:43aggregate all of these although anderson
40:44and neri uh
40:45provide a way to do that and there's a
40:48nice anderson van wing coupe
40:50general of economics literature trade
40:52cost chapter which kind of summarizes
40:54some of the information
40:55uh on uh uh how important these various
40:58uh um uh barriers are uh you know
41:01language barriers currency barriers
41:03national borders
41:04etc okay
41:07so you know before moving on to uh kind
41:10of
41:11inferring trade costs from price gaps
41:14you know there's a big issue here with
41:15identification so
41:17this kind of gravity approach assumes
41:19that the reason
41:20that uh we don't trade with each other
41:22so much say
41:23say let me take an example the reason
41:25why you know
41:27a country trades a lot with more with
41:28its neighbor than it trades with the
41:30country 500 miles away
41:32is because of trade costs but there may
41:35be other reasons why
41:36those two countries trade more with each
41:38other and the two obvious ones which are
41:40kind of
41:41ruled out by the kind of inverse gravity
41:44approach we just talked about
41:45is one of the integrated supply chains
41:47are choosing to locate near each other
41:50right in that kind of simple gravity
41:52structure that
41:53that tau ijk what i'm calling trade
41:55costs
41:56uh is potentially you know any
41:59combination of of
42:00true trade costs plus a technology shock
42:03that's specific
42:05to that ij pair right so if it's true
42:08that the kind of the
42:09things they produce are particularly
42:11good matches for the things
42:12i i i i i need and there's kind of
42:15submit some input output
42:17uh uh uh uh uh um flows here
42:21uh then uh we're going to call those
42:23trade costs
42:25and i'll show you uh uh very briefly
42:27that we have strong evidence
42:28certainly in developed countries that
42:30there is a lot of this uh
42:32supplier location decisions you know i
42:35build my big engine factory next to the
42:37car plant
42:38if i'm supplying the car plan to
42:40minimize trade costs and
42:42we're going to say oh well it must be
42:44because of the low trade costs that's
42:45why you're trading so much with them
42:47well no it's because you know this is a
42:49specific input
42:50and they've decided to locate the
42:52factory next to mine
42:54similarly this tau ijk is also going to
42:56pick up any any
42:58demand shocks that are specific to an ij
43:01pair so if it's true that you know uh
43:03people in chat particularly like the
43:05products made in ghana
43:07or the the product k made in ghana the
43:09the soups made in ghana
43:11then we're going to load that into
43:15there being low trade costs for soup
43:17within ghana
43:19if we kind of use this separable uh
43:21gravity
43:22equation which assumes everyone has
43:24identical preferences i want to buy
43:26evenly from all the countries
43:28in the world conditional on prices
43:31so that's another assumption which is
43:34unlikely to hold
43:36particularly if you think tastes are
43:37spatially correlated and i i have some
43:40work showing that indeed they are
43:42through a very simple mechanism which is
43:44endowments are spatially correlated
43:46you know the climate and the land is
43:48very similar in neighboring countries
43:50and that means that historically they
43:52were producing similar types of
43:54things uh because they had similar
43:55endowments uh and
43:57uh uh they developed tastes uh uh for
43:59those types of things and so
44:01you know i'm more likely uh as someone
44:04who grew up in england
44:05uh uh to want to eat uh you know the
44:06type of food a french person produces
44:09uh then the type of food uh maybe a
44:11russian produces
44:12uh because i'm more used to those types
44:15of ingredients
44:17um so we have some evidence for both of
44:21these as i mentioned
44:22this is for the u.s i haven't seen a
44:23similar study in a developing country
44:25but uh
44:26my guess is you'd find some similar
44:27patterns maybe not as extreme
44:29this is uh from commodity flow surveys
44:32in the us
44:33this is uh how much trade occurs across
44:36different distance bins
44:38using zip code to zip code our truck
44:40movements
44:42and you just see a colossal fall-off in
44:45distance
44:46uh in in the value of trade as soon as
44:49you move essentially out of the zip code
44:52right so most trade is occurring within
44:54the same zip code in the us
44:56and this is really strongly pointing
44:58towards you know that engine plant
45:00locating next to my factory
45:02uh maybe i even own the engine plant uh
45:04um
45:05and uh most of the trade is going on in
45:07these very small different distances
45:09uh and this doesn't seem like uh uh this
45:12is truly a transportation
45:14issue once it's loaded on the tracks you
45:16know the difference between
45:1710 10 versus a 300 isn't so great that
45:21uh you'll see this just absolute uh full
45:24off a cliff
45:26and that sorry sorry can i ask something
45:28about that graph
45:30so is that a taking away non-tradables
45:33or services
45:34or that includes like those type of
45:36goods yeah so these are just
45:37this is just looking at uh uh truck
45:40flows
45:41and values moving in these trucks uh
45:43betw between zip code uh
45:46zip code to zip code uh or within zip
45:48code
45:49from the commodity flow survey in the us
45:51so they go and do it basically a survey
45:53of shipments uh so it's just goods
45:56moving
45:56moving in trucks
46:00um maybe on trades as well okay um
46:04how do we also account for possible uh
46:06trade agreements
46:08uh among countries i see so
46:11so that would be um in uh
46:15typically what people have done is
46:16they've regressed these say gravity uh
46:19uh residuals on say being a member of
46:22the same trade agreement or something
46:23like that and that matters
46:24that's another one of these determinants
46:26of of these uh
46:28our trade costs uh and so uh uh
46:31absolutely that matters and there's a
46:32bunch of work looking at that most
46:34notably by andrew rose
46:36this is obviously this is all movements
46:37within the u.s so there's no
46:39trade agreement i'm actually talking
46:41about the model we're
46:42looking at for uh international trade
46:46so one would assume that if two
46:48countries have certain
46:49um agreements then maybe that could be
46:52a possible driver or an incentive for
46:56uh trade to uh improve
46:59between between the two countries
47:01absolutely so if those trade agreements
47:03say lower tariffs or lower
47:04administrative burdens that's going to
47:06lower the
47:07the uh tau ij
47:11here um and that's gonna and that's
47:14gonna and we're gonna recover that
47:16that'll be something we capture and then
47:18some people have then projected
47:20those recovered tau ijs or or or or the
47:23residual at least
47:24on uh on whether or not you have a
47:27regional trade agreement with that
47:28country and the answer is yes
47:30indeed uh we are recovering those that
47:33is part of what we're picking up
47:34okay thank you okay as i mentioned
47:38uh these endogenous preferences you know
47:40if you kind of just look at
47:42it uh you know say rajasthan uh
47:46in in the west of india the kind of
47:48price per calorie
47:51between rice and wheat is not that
47:53dramatically different
47:54uh uh yet in rajasthan they're spending
47:5731 percent of their budget on wheat
47:59in in the east of india and west bengal
48:01they're spending 57 of their budget on
48:03rice
48:04once i condition on relative prices uh
48:06then uh
48:07the residuals are strongly correlated
48:10with kind of historic endowments
48:12uh uh that kind of determined what these
48:14regions were eating
48:15uh uh many centuries ago and so there we
48:18would get this spatial correlation
48:20which would we would be loading in in
48:22this case to trade costs
48:25okay so the third uh method of measuring
48:28spatial frictions is going to be uh
48:30using price gaps
48:32so there's a very large literature uh
48:34that does this
48:35a lot of it in agricultural economics
48:36and the basic idea is
48:38instead of using uh our values and
48:41quantities
48:42instead we can actually do something
48:44quite non-parametric uh we basically
48:46just need one key assumption
48:48to back out trade costs from the
48:49difference in prices if we have good
48:51price data
48:52and so in particular you know there's
48:55this arbitrage idea which says
48:57that the price difference between
48:59location j
49:00and location i for some good k
49:04if there are goods flowing between those
49:07locations
49:08must be equal to the cost of moving it
49:10between those locations
49:12if the price was larger the price gap
49:14was larger then people would come in
49:16uh and uh uh uh uh uh go and uh compete
49:19away
49:20that uh uh uh um price gap
49:23until it equals the trade costs where
49:25this trade cost is a very expansive
49:27definition including
49:28you know the cost of time of all the
49:29traders etc
49:32okay so if we're willing to buy that
49:33arbitrage assumption which is a simple
49:35one and a single uh assumption then
49:37you know we can potentially just back
49:39out these trade costs uh from price gaps
49:42so there's kind of three challenges here
49:43the first is we have to observe the same
49:45product in both locations it's no use
49:48observing
49:49you know rice in one location of rice in
49:50another uh if
49:52the rice and location i is low quality
49:54and the rice and location j
49:56in a big city is high quality uh then
49:58we're not really comparing like what's
50:00like and some of that price difference
50:01is the quality difference
50:03secondly we need to know if the
50:05countries if the two locations are
50:06actually trading the product
50:08if they're not trading then this isn't
50:09an inequality this an
50:11it isn't an equality it's an inequality
50:13we know that the price gap is less than
50:15or equal
50:16to that trade cost uh uh the price gap
50:18uh could be
50:19uh sufficiently small uh uh that there's
50:21actually no
50:22no one uh wants to go uh pay those costs
50:26uh of moving goods between the two it's
50:29not that easy to know if two locations
50:30are trading and the reason is
50:32if you want to have a really homogeneous
50:34product the trade data is typically not
50:36at that level we don't see
50:38you know trade at the level of uh uh you
50:41know
50:41cans of 330 milliliter cans of coke
50:44we see trade at the level of beverage uh
50:47uh or carbonated
50:49beverage with sugar uh and so we're not
50:51going to be able to
50:52see that and if we're looking at
50:54domestic trade costs uh then we don't
50:56often even have trade flows
50:58within locations third challenge is we
51:01have to believe in perfect arbitrage
51:02which also means believing in perfect
51:04competition that there's kind of
51:06enough people coming in here competing
51:07away are those uh
51:09any rents from arbitrage all we have to
51:12correct for this so
51:13i'll talk in a few minutes about uh dave
51:15johnson's paper on railroads in india
51:17that kind of solves
51:18one and two uh problems one and two
51:20using salt uh
51:22and now i'll talk a little bit about my
51:25work with dave donaldson which attempts
51:26uh to solve all three of these at least
51:29make progress on all three of these
51:30i should just mention that there's some
51:32ag econ work in econ history work
51:34that's tried to use the inequality
51:36itself
51:37and make some progress there
51:41okay so what do we do in our paper so
51:44how do we solve these challenges so the
51:46first is we have to observe homogeneous
51:48products so
51:48what we're going to do is we're going to
51:49collect a sample of barcode level
51:51products
51:52from cpi data in nigeria and ethiopia
51:55and we'll have a benchmark of the us
51:58um and this is going to give us kind of
51:59bar uh it's going to give us price
52:01quotes for things like that
52:03milliliter can of regular coca-cola in
52:06different locations
52:08we also have to know where the country
52:09whether locations are trading that
52:11product and we're going to be looking at
52:12trade costs within nigeria trade costs
52:14within ethiopia
52:15and so what we're going to do is we're
52:16going to have price quotes of various
52:18different cities or markets
52:21and we're also going to know the factory
52:23location
52:25of our those uh uh uh our products
52:28and so that's going to tell us you know
52:31whether or not there's actually trade
52:32occurring where we know that trades you
52:34know
52:34only occurring on on or almost
52:37exclusively occurring on routes where
52:39one of the uh where the origin is the
52:41factory location
52:44okay so you know does it matter us
52:46restricting attention just to trading
52:48pairs well the answer is yes so
52:50here are three countries ethiopia and
52:52nigeria the u.s
52:53all i'm plotting here is the price gap
52:55the difference in price between
52:57uh either i and j the absolute
52:59difference if i use all pairs
53:01or if i'm using trading pairs the orig
53:03the destination price minus the origin
53:05price
53:06and i'll plot this against the distance
53:08between those two
53:10and if we look at the dash green line
53:12these are for all pairs we can see that
53:14it is indeed the case
53:16that uh the price gap goes up uh and the
53:19further away you are from
53:20uh uh uh uh the further apart the two
53:23locations are
53:24if i'm doing the origin destination the
53:26price gap goes up
53:27the further away i am from
53:30from the origin and the slope of this
53:33curve is telling us the marginal cost of
53:35distance
53:35the derivative of this uh uh uh trade
53:38cost function with respect to distance
53:42and as you can see it matters quite a
53:44lot that we restrict attention to these
53:46informative pairs these trading pairs
53:48uh rather than these uh pairs which
53:50aren't trading which
53:51uh uh uh uh aren't uh uh uh very
53:56informative
53:57uh beyond potentially uh uh adding to
53:59the moment inequality
54:00uh and indeed the marginal cost of
54:03distance at least in ethiopia and
54:04nigeria approximately doubles
54:06uh uh once we uh account for the fact we
54:08should only be focusing on these
54:09informative pairs
54:11uh that arbitrage equality doesn't tell
54:13us about
54:14the non-trading pairs okay what about
54:17that third problem
54:18the imperfect competi the perfect
54:20competition so
54:21that's a strong assumption given that we
54:23hear a lot of anecdotes about
54:26oligopolistic traders or market power
54:28amongst traders
54:30and so we kind of try and tackle that uh
54:32head-on by writing down
54:33a relatively simple model where there
54:36are oligopolistic traders who are buying
54:38at the factory gate
54:39and then moving the good across space
54:41and selling it to the final consumer at
54:43a markup
54:45and so in that world the price gap the
54:47destination price minus the origin price
54:49is equal to the trade cost of going
54:51between o and d
54:52as well as the markup being charged by
54:54these traders when they sell at d
54:57and this kind of gives us a way to
54:59potentially correct
55:01uh for the bias that's caused when we
55:05in that simple plot i showed you before
55:07in these blue curves
55:08where we don't account for the fact that
55:09there may be imperfect competition in
55:11the training sector
55:12and the logic is as follows so
55:16the markups are are if they're constant
55:19then that's not going to be problematic
55:20but
55:21what is going to be problematic is if
55:22these markups uh
55:24respond to higher trade costs i.e if
55:27markups are differential
55:28in in in more remote or more proximate
55:31locations to the origin
55:33uh and in general they will be and the
55:36reason they will be
55:37is that other than in knife fed cases uh
55:40uh
55:41markups respond to marginal costs either
55:44markups are higher or markups are lower
55:45when marginal costs go up depending on
55:48something called the super elasticity
55:49the second derivative of the demand
55:51function
55:52and um uh higher trade costs that's
55:56that's part of the trader's marginal
55:58cost it's one of the kind of two
56:00components along with
56:01buying it the factory gate the price
56:02they pay in their marginal cost
56:05and so uh if it is the case
56:08uh uh that we're regressing price gaps
56:11uh on a distance for example
56:13then we're confounding both the fact
56:15that trade costs are going up
56:16with the fact that markups are also
56:18changing with these higher trade costs
56:21and the way we kind of can deal with
56:22this is this kind of idea that
56:26if we can observe another marginal cost
56:28shifter
56:30and we're going to particularly in
56:31particular be able to observe the price
56:33pass-through rate which is how
56:35kind of the prices in the destination
56:37respond when there's a change in the
56:39price in the origin
56:41that's basically another the other
56:42component of the traders marginal cost
56:45how much they're paying for the good uh
56:47the factory gate
56:49if we observe those price pass-through
56:50rates and we can do that just from price
56:53data by seeing how shocks
56:54at the price in the origin feed through
56:56to shocks in the destination
56:58that's going to tell us exactly what we
57:00need to know about how the markup
57:02responds in any particular location for
57:04any particular product
57:06okay and so we don't need to go and work
57:08out for every location every product
57:10uh the super elasticities those demand
57:12parameters or the market structure
57:14uh both of which are going to matter for
57:16this pass-through rate
57:17the pass-through rate is exactly what
57:20tells us how that markup is going to
57:22change
57:22uh uh uh uh when marginal costs change
57:25and so we can use that positive rate to
57:27essentially
57:28correct our estimates of the change uh
57:31in
57:31our prices with distance uh from what is
57:34the change
57:35uh in the markup and the the remaining
57:37derivative
57:38uh is the change uh in trade cost with
57:40respect to distance
57:42okay and so that's basically what we do
57:44and there's not many more details
57:45in the paper on how we deal with uh
57:47other potential confounders
57:49but once we do that we find another
57:50approximate doubling
57:52and so i i uh uh these uh uh uh this
57:55imperfect competition
57:56uh essentially uh uh uh uh leads us to
57:59underestimate
58:00uh the uh um uh uh slope of
58:04of distance uh the marginal cost of
58:06distance in developing countries
58:07once we account for that uh the marginal
58:10cost of distance appears to be about
58:11three to five times higher
58:13in ethiopia and nigeria than in the us
58:16again supported of there being these
58:17much higher trade costs
58:18uh within developing countries uh than
58:20within developed countries
58:23uh we also can say a little bit more
58:25about implications
58:26of these high trade costs in developing
58:28countries uh something we'll turn to uh
58:31uh in in the rest of the lecture today
58:33uh when i talk about uh
58:35the impacts of infrastructure and
58:39essentially what we can say is two
58:40things
58:42uh you know thinking about this thought
58:44experiment whereby
58:45imagine there's a kind of a big increase
58:47in globalization
58:49how does that affect our people in
58:51nigeria how does that affect people uh
58:53in ethiopia
58:55and so the kind of idea here is imagine
58:57prices fall at the port because
58:59you know lower tariffs uh better
59:02logistics whatever it is
59:04now in in lagos uh uh or or the entry
59:07point into ethiopia
59:08uh you can now buy uh cheaper uh um
59:12imports how does that benefit people
59:15across ethiopia across nigeria well
59:17if i live in a remote location in
59:19ethiopia or nigeria
59:21then there's kind of a smaller increase
59:23in the pie
59:24when there's kind of improvements at the
59:26border why well i'm not i'm not buying
59:28very much
59:29from abroad anyway because it's so
59:31costly to get it to me
59:32uh uh that uh uh you know a change in
59:35the price on the border has
59:36relatively little effect in the extreme
59:38if i'm you know
59:39so so isolated i i don't it's not
59:43worth me buying any of these imports
59:44because they're so expensive
59:47then i'll be not affected at all by a
59:49price change at the border
59:51the second fact is that you know because
59:53these remote locations have these high
59:55costs of trade
59:56to get goods to them that means they're
59:59not very profitable
1:00:00uh for the uh intermediaries because
1:00:02they have to charge these really high
1:00:04prices
1:00:04uh just to cover their trade costs uh
1:00:07and so
1:00:08um because they're not very profitable
1:00:10few traders enter
1:00:11and because few traders enter they're
1:00:13also not very competitive trading
1:00:15markets
1:00:15and that means the markups are really
1:00:17high as well and for that reason
1:00:19uh that that means that a smaller share
1:00:22of the surplus
1:00:24that is going to these locations is
1:00:26going to consumers rather than going to
1:00:28traders or deadweight loss
1:00:30and we find kind of both of those
1:00:32effects in our paper
1:00:36okay so let me kind of change tack a
1:00:39little bit now and in the last
1:00:4115 minutes let me talk about estimating
1:00:43the impacts of spatial frictions
1:00:45on outcomes such as welfare
1:00:48and here i'll take a a a deep dive into
1:00:51dave donaldson's paper
1:00:52as a kind of very nice example that
1:00:54matches up very nicely with
1:00:56a melanie's uh lecture last week
1:00:59who kind of introduced kind of the basic
1:01:01structure of these models and i'll talk
1:01:02about
1:01:02uh the implementation of one of those uh
1:01:05if i have a few minutes at the end we'll
1:01:06also talk about
1:01:07a reduction in information frictions
1:01:10uh and how that affects uh uh uh uh
1:01:13trade uh
1:01:14and the fund size distribution uh in
1:01:16india
1:01:19okay so uh you know maybe the primary
1:01:22way that
1:01:24governments intervene to reduce trade
1:01:26costs is through infrastructure
1:01:27investment
1:01:28whether it's building railroads or roads
1:01:31or better ports
1:01:32or cell phone networks but it's not easy
1:01:35to identify the effects of
1:01:36infrastructure
1:01:38and that's for two potential uh
1:01:40substantial compounds the first is that
1:01:42you know where government puts
1:01:44infrastructure uh
1:01:46may be in exactly the places which were
1:01:48expected to grow the most
1:01:49so you know in a booming country uh you
1:01:52know think about china in in the 1980s
1:01:55or early 90s
1:01:56they may have decided let's go and you
1:01:58know build roads in
1:01:59in to connect the kind of most important
1:02:01cities
1:02:02those are the places we expect are going
1:02:04to you know really grow anyway
1:02:06and anticipating that you're you're
1:02:08trying to kind of put
1:02:09infrastructure in response to that there
1:02:12what you'll do is you'll kind of
1:02:13overestimate the effect of
1:02:14infrastructure a lot of this would have
1:02:15happened without the infrastructure
1:02:17the opposite may also be true which is
1:02:19the government may use infrastructure as
1:02:20a tool
1:02:21uh to boost poorer areas the eu spends a
1:02:24lot of
1:02:25of its money uh on boosting uh on giving
1:02:28infrastructure to poor areas so they can
1:02:30catch up
1:02:31so in that case you'd get the opposite
1:02:33you'd underestimate
1:02:34uh the impacts of infrastructure since
1:02:37it's going to the areas that are kind of
1:02:38least likely to grow
1:02:42okay so dave donaldson tries to tackle
1:02:45these questions by looking at the
1:02:46expansion of railroads
1:02:48in colonial india and asking if it kind
1:02:51of promoted development particularly
1:02:52agricultural development which was most
1:02:53of the economy at that point in time
1:02:55and uh he's able to uh uh kind of
1:02:58capture very well with the data
1:02:59available
1:03:01and if it did benefit uh
1:03:04uh uh agricultural development which is
1:03:06something that uh
1:03:07it can be answered at least in part by
1:03:09reduced former uh approaches
1:03:11uh were those gains caused by reduced
1:03:14trade costs
1:03:15and here he's going to lean heavily on
1:03:16on the kind of eaton courtroom style
1:03:18uh version of a spatial model that
1:03:21melanie
1:03:22uh talked about in the last lecture uh
1:03:24and i i use a kind of inferential
1:03:26technique where essentially
1:03:28what he's going to do is estimate the
1:03:29reduced form impacts and see whether or
1:03:31not
1:03:32those review source impacts i can be
1:03:35mostly or entirely explained by a model
1:03:38based sufficient statistic
1:03:40for what the benefits of the
1:03:43infrastructure should have been
1:03:44uh uh uh through the lens of a of a
1:03:47model where all of those benefits are
1:03:48purely coming
1:03:49uh are from trade uh and so uh
1:03:52uh uh we're going to uh kind of talk
1:03:54through this as a
1:03:55a kind of a bunch of insight at various
1:03:58points in this paper
1:04:01okay so here's kind of the uh
1:04:03transportation network in 1853
1:04:05uh uh in india you can see there's a
1:04:07tiny railway line in bombay
1:04:10and very rapidly uh uh the railway line
1:04:12started growing 1860 1870 1880 1890 1900
1:04:171910 1920 1930 by that point in time
1:04:21it's effectively looks like it does
1:04:23today there hasn't been that much
1:04:24investment since then and it now covers
1:04:27most of
1:04:28of india so what we can do is try and
1:04:30explore what the impacts are of this
1:04:32massive expansion of the railway road up
1:04:34until
1:04:34the 1930s uh uh in uh
1:04:37uh uh uh to various uh indian uh uh
1:04:40outcomes particularly uh things that try
1:04:43to map into welfare
1:04:44and trade flows okay so a bit of
1:04:47background
1:04:48you know what is india at this point in
1:04:50time how do we think about it so
1:04:52it's primarily an agricultural economy
1:04:54it's about two-thirds
1:04:56of its gdp in agriculture almost no
1:04:58factory-based manufacturing
1:05:00most of the agriculture is rain-fed
1:05:02there's very very little irrigation
1:05:04and so that's going to be a very nice
1:05:06setting which is we have good data on
1:05:08agriculture typically
1:05:09and that's going to cover most of the
1:05:11economy and we'll be able to use
1:05:13rainfall as
1:05:14observable and exogenous
1:05:18shocks to productivity
1:05:22okay so you know what exactly is this
1:05:25railway shock so
1:05:26it's basically a new technology that's
1:05:28massively better than the existing
1:05:30technology previously of some
1:05:31combination of bullocks traveling on
1:05:33roads
1:05:34it may take two to three months to reach
1:05:36the port rivers which is very seasonal
1:05:38and slow
1:05:39and coasts which have kind of limited
1:05:41port access
1:05:42for the newer ships meanwhile railroad
1:05:46is massively fast
1:05:47uh compared to the alternatives about
1:05:49600 kilometers per day
1:05:51it's safe there's limited piracy uh
1:05:53which was a big issue on many of these
1:05:55other uh
1:05:56uh uh routes and it's much cheaper so
1:06:00you know
1:06:00in order to get your good to market i uh
1:06:03say it to port
1:06:04uh it's gonna be about four to four
1:06:06point five times cheaper than roads
1:06:08uh about uh two times cheaper than the
1:06:10coast which obviously can only go to
1:06:11certain places
1:06:12um at least based on uh some outside
1:06:16estimates
1:06:18this is really a massive change and
1:06:20we're going to try and see
1:06:21uh uh uh the impacts of this change so
1:06:24as i mentioned
1:06:25you know the the kind of empirical uh
1:06:28exercises are going to be very
1:06:30tightly guided by a general equilibrium
1:06:33spatial model and in particular it's
1:06:35going to be a multi-sector extension of
1:06:37eaton and cotton
1:06:38so it's going to be many regions
1:06:41many goods think rice think wheat and
1:06:45trade costs uh iceberg trade costs and
1:06:48uh each of these goods there's a
1:06:50continuum of varieties a different
1:06:51variety produced by each location
1:06:55uh and um we're gonna have
1:06:59a um uh sorry a continuum of
1:07:03of of of varieties uh uh um even within
1:07:06locations
1:07:08and um
1:07:11uh we're gonna have origins destinations
1:07:14uh oh origins to your destinations
1:07:18okay so the production technology is
1:07:20quite simple there's constant returns to
1:07:22scale
1:07:22there's a single factor here it's just
1:07:24going to be labor uh and so
1:07:26that means uh that uh kind of the only
1:07:29uh incomes here are gonna be rental
1:07:31uh rates from labor that's our think of
1:07:33it like
1:07:34uh uh uh sorry one factor land and
1:07:37there'll be rental rates on land r
1:07:39you can think of this as uh very similar
1:07:41to a model which only has one factor
1:07:42labor
1:07:43where wages are other returns to that uh
1:07:46factor um
1:07:50the uh productivity of a kind of unit of
1:07:53land
1:07:53is is the z which is different depending
1:07:57on
1:07:59k and across
1:08:02the different uh varieties j
1:08:08um and the price because of perfect
1:08:11competition here is simply going to be
1:08:12equal to the rental rate over the
1:08:14productivity
1:08:16okay like uh in in melanie's lecture
1:08:18last week
1:08:20we're going to have an extreme value
1:08:21distribution that you're drawing
1:08:23productivities from
1:08:24that extreme value distribution is going
1:08:26to depend on some productivity that's
1:08:29specific to the origin for that good
1:08:31this is going to be where rainfall is
1:08:32going to enter uh that's going to
1:08:33essentially
1:08:34i change the probability of getting a
1:08:36high or low draw
1:08:38and then this theta k which is affecting
1:08:40the dispersion of productivities
1:08:42uh across locations that's what's going
1:08:44to be generating
1:08:45a comparative advantage that's going to
1:08:47be generating kind of gains from trade
1:08:49is is this dispersion parameter the more
1:08:51dispersion uh the bigger benefits
1:08:53are from trade
1:08:57okay what our preference is so we're
1:08:58going to have cobb douglas across
1:09:00different goods across wheat across rice
1:09:02that's that mew here
1:09:03and then ces within these goods uh as in
1:09:06a lot of these uh modern spatial models
1:09:08where epsilon is this elasticity of
1:09:11substitution
1:09:12uh across uh other varieties uh uh
1:09:15within uh uh
1:09:17a rice and there'll be some iceberg
1:09:19costs of trading
1:09:20uh t where if i want to consume a good
1:09:22produced uh
1:09:23you know in location d i have to pay uh
1:09:26this uh
1:09:27iceberg cost a tkod on top of the
1:09:30price in the origin
1:09:34okay so the kind of magic of these
1:09:35extreme value distributions uh
1:09:38uh in this uh a particular model is
1:09:40consumers
1:09:41are facing many potential suppliers they
1:09:43consume uh the cheapest variety uh
1:09:46the cheapest version of each variety and
1:09:48what this is going to give us
1:09:50is a very tractable expression thanks to
1:09:52the extreme value
1:09:54for the expected price in each location
1:09:57it's going to depend on these rainfall
1:09:58shocks
1:09:59this is a function of the gamma
1:10:00distribution that also depends on
1:10:02on theta this lambda term and it's going
1:10:04to depend
1:10:05on these trade costs and rental rates
1:10:08i know and the trade costs we know
1:10:10indeed and this dispersion parameter
1:10:12okay and this is going to make things
1:10:14very tractable and generate gravity
1:10:16relationships
1:10:17uh as well okay so
1:10:21there's kind of four main predictions uh
1:10:23that uh
1:10:24dave's going to bring to the data the
1:10:26first uh is that
1:10:28if a good o can only be made in one
1:10:30region but this good is consumed
1:10:32elsewhere
1:10:33then because of the kind of perfect
1:10:34competition in this model
1:10:36the price gap between the two is just
1:10:38going to equal those trade costs of
1:10:40moving the good between the two
1:10:41locations
1:10:42and we're going to leverage this uh for
1:10:45the a particular product salt and i'll
1:10:47explain why in a minute
1:10:48uh to kind of back out the function
1:10:51underneath
1:10:52underlying t if we know you know these
1:10:55price gaps we know t
1:10:56then if i know the full road network and
1:10:58rail network and coastal shipping
1:11:00network
1:11:01i can then parameterize the relative
1:11:03cost of those different means of
1:11:05transport
1:11:06uh uh uh by essentially uh uh uh uh
1:11:09uh working out kind of the best fit to
1:11:11hit these price gaps
1:11:13the second prediction is is exports take
1:11:16the gravity form that we kind of
1:11:18already talked about uh uh earlier in
1:11:20this lecture
1:11:21in particular the kind of uh imports
1:11:25of good k uh from o into location d
1:11:30uh depend on both the uh kind of range
1:11:33shocks uh in
1:11:34uh uh uh uh location o how productive
1:11:37they are
1:11:37uh in location o because of the good
1:11:39rainfall this dispersion parameter
1:11:42that's kind of
1:11:42uh telling us how much these things vary
1:11:44across our locations
1:11:46uh plus these trade costs and what this
1:11:49gravity equation is going to allow us to
1:11:51do is if we have data on these trade
1:11:53flows and we will have data on these
1:11:54trade flows
1:11:56and if we're able to back out this t
1:11:58function
1:11:59which we will be able to from the
1:12:01previous regression
1:12:02then we can also back out theta which is
1:12:05this kind of key elasticity
1:12:07which is this dispersion parameter
1:12:10by essentially regressing uh the log of
1:12:12this object on the log of the trade cost
1:12:15we can also back out the mapping between
1:12:17rainfall and this a
1:12:18the a our term this residual is going to
1:12:21be a residual
1:12:22uh uh which we can then i see how rain
1:12:25affects that a
1:12:26and then we're going to be able to use
1:12:28that
1:12:30very shortly then in terms of
1:12:33you know both of those are useful in
1:12:35terms of allowing us to get the key
1:12:36parameters of this model
1:12:38then in terms of actually thinking about
1:12:40welfare effects
1:12:41we can look at the impacts on real
1:12:43income levels
1:12:45uh welfare in this model because we just
1:12:47have a single factor welfare simply
1:12:49my wage or the rental rate in this case
1:12:51because of the single factor being land
1:12:53divided by the price index uh and uh
1:12:57dave shows at least around the symmetric
1:12:58equilibrium that when i get a railroad
1:13:02to my origin that's good for me it means
1:13:04i can now sell my stuff more cheaply so
1:13:06my sales go up
1:13:08uh to others and i can buy stuff from
1:13:10others more cheaply
1:13:12and that helps my price index
1:13:15meanwhile if others get a railroad but
1:13:17it doesn't connect to me that's bad
1:13:19and why is that bad well now another
1:13:21location that previously bought for me
1:13:22will buy from other locations
1:13:24because it's kind of cheaper to buy from
1:13:27those other locations
1:13:28and so we can go and test this in a very
1:13:30kind of reduced form way
1:13:33and then kind of the key test in the in
1:13:35in the paper of whether or not
1:13:37you know what we find in those
1:13:38regressions uh is indeed coming about
1:13:40through
1:13:41trade flows uh is uh uh this kind of
1:13:44model derived
1:13:46expression for for our welfare
1:13:49which says that despite all of these
1:13:51complex g interactions
1:13:53uh welfare can be written as a function
1:13:55of a constant
1:13:56of something that depends on these
1:13:58productivity shocks which i'm going to
1:14:00have a way to measure thanks to rainfall
1:14:02and the fact i have a mapping
1:14:04between rainfall and a and i know the
1:14:07theta and i can easily get the mu which
1:14:09is just a cob douglas
1:14:10share spent on different goods and this
1:14:13pi
1:14:14o okay term which uh dave calls the
1:14:17authoric-ness which is
1:14:19essentially how much i trade with myself
1:14:22and the more i trade with myself the
1:14:24more toxic i am
1:14:25and and the lower is my my welfare the
1:14:28more i trade with others
1:14:30uh uh this number goes down and my
1:14:32welfare goes up
1:14:33and so at least through the structure of
1:14:35this model
1:14:36uh despite the complex uh a spatial uh
1:14:39uh uh
1:14:40uh ge model essentially everything
1:14:44that's happening due to trade
1:14:45is going to be loaded in how much it
1:14:47changes like how much i trade with
1:14:49myself
1:14:49if i still trade as much with myself
1:14:52before as i did after
1:14:53then essentially uh the railroads
1:14:56haven't given me any benefits through
1:14:57trade
1:14:58uh if i trade more with others uh unless
1:15:01with myself
1:15:03then this tells us exactly how that
1:15:05should affect welfare
1:15:07and what we can do is we can see if
1:15:08indeed our welfare changes that we can
1:15:11measure from rents over
1:15:12over over price indices do respond to
1:15:16water quickness as the model predicts
1:15:21okay so in the interest of time i should
1:15:23i should go fairly quickly and i'll try
1:15:25and wrap up in a couple of minutes doug
1:15:26if that's okay
1:15:28um how does uh dave actually go
1:15:31and uh implement this well as i
1:15:33mentioned
1:15:34these price gaps we need a product as we
1:15:37talked about that
1:15:38you know we know where it's made
1:15:41and is homogenous and it turns out in
1:15:43india uh
1:15:44uh the colonial government of the
1:15:46british was uh attempting to control the
1:15:48population by
1:15:49essentially they made it illegal for you
1:15:51to make salt yourself gandhi famously
1:15:53violated that as a protest and salt was
1:15:56only made in eight different locations
1:15:58in india
1:15:59and was very tightly controlled and it
1:16:00was taxed heavily it was a major source
1:16:02of revenue
1:16:04and so what you can eventually do is you
1:16:06can see the price of
1:16:07different types of salt in different
1:16:09locations because the salt
1:16:10uh was essentially tagged with where it
1:16:12came from uh
1:16:14for these regulatory purposes and so
1:16:17here's where the salt mines are in red
1:16:19here's where we have price data on the
1:16:21salt which is tagged by the location
1:16:23where the salt came from and then we can
1:16:25see how over time
1:16:27those prices change when railroads were
1:16:29added which made it cheaper or less
1:16:32expensive
1:16:33are to uh uh kind of uh uh
1:16:36move assault between those different
1:16:38regions and so
1:16:40you know essentially what we'll do as i
1:16:41mentioned is we'll kind of project those
1:16:43price gaps which come from
1:16:45the kind of price we observe in a
1:16:47particular destination
1:16:49and a fixed effect uh which is specific
1:16:52to that variety of salt in that location
1:16:54that's
1:16:55sweeping out the origin price
1:16:58and then the residual we're going to
1:17:00project on
1:17:01uh the road network and the rail network
1:17:04and the coastal
1:17:05river network and a bunch of parameters
1:17:08which
1:17:09say how costly different means of
1:17:11transport are
1:17:12relative to each other and we're going
1:17:14to find the alphas the parameters that
1:17:16best fit this relationship
1:17:18that's going to tell us how costly road
1:17:19is relative to rail
1:17:21and the numbers you get line up
1:17:23relatively well with the kind of figures
1:17:25i showed you before that came from
1:17:26uh kind of observers at the time
1:17:30um we can also do kind of more reduced
1:17:33form things like see whether
1:17:34these price gaps uh decline uh when
1:17:36there's a railroad between
1:17:38the origin and the destination and a
1:17:41railroad gets built and indeed they do
1:17:43they fall by about 11 when uh the salt
1:17:46price falls by around 11
1:17:48when i connect to these locations
1:17:51there's a nice placebo i guess it got
1:17:53cut from the final version of the paper
1:17:55but it's maybe my favorite column in the
1:17:56paper
1:17:57where uh dave shows that uh when you're
1:18:00trying to move camels elephants carts
1:18:02and inland boats
1:18:03we don't see such a impact of of the
1:18:05railways
1:18:06and the reason for that is all of these
1:18:08things are bigger than railway carts are
1:18:10and so uh uh uh you can't stick them on
1:18:12railway cards
1:18:15okay so you know with those estimates of
1:18:18of t in hand which i've kind of backed
1:18:21out from those price gaps
1:18:22uh i can now go and see whether uh this
1:18:25gravity
1:18:26regression uh uh uh from scratch
1:18:28regression i can go back out
1:18:30uh this parameter theta this key
1:18:31dispersion parameter uh
1:18:33in the fresh air inside the extreme
1:18:34value distribution uh by projecting the
1:18:37gravity
1:18:38uh the uh uh import shares uh on these
1:18:40trade flows
1:18:41and then i can take the residuals and
1:18:43work out how a
1:18:45uh ko these are the productivity shocks
1:18:48uh depend
1:18:49on rainfall and that that mapping we're
1:18:51gonna need shortly
1:18:54and then the kind of key reduced form
1:18:56regression here is
1:18:57i'll regress real income uh in each
1:19:00location
1:19:01on whether or not they had a railroad
1:19:03link in their district
1:19:05and whether your neighbors had a
1:19:06railroad link and as
1:19:08a kind of at least first order a
1:19:11prediction from the model
1:19:12i would say i benefit when i have a
1:19:15railroad in my location
1:19:16in fact my my real income goes up by 16
1:19:21and i lose out uh condition on me having
1:19:24a railroad when my neighbor gets a
1:19:25railroad
1:19:26uh as more people are purchasing from
1:19:28them rather than from me
1:19:29so this kind of support uh uh for these
1:19:32model predictions
1:19:34uh and then uh what we're going to want
1:19:37to see
1:19:38is whoops hold on a second uh what we're
1:19:40going to want to see is
1:19:41how much of this is driven by the
1:19:43mechanism uh uh proposed in the model
1:19:45i.e
1:19:46uh by trade uh itself and here we'll add
1:19:49those two extra terms the model predicts
1:19:52uh and the model says just these two
1:19:54things should matter
1:19:55it's it's something to do with a's which
1:19:57are rainfall times
1:19:58uh uh this kappa coefficient we've
1:20:00estimated uh
1:20:02uh plus uh your authoric-ness which
1:20:04we've also uh i i i've been able to
1:20:06uh take out of the data and uh uh uh
1:20:10uh uh once we are kind of uh uh
1:20:13used uh these model predicted uh impacts
1:20:17of the railroads we find there's no
1:20:18longer an effect of me having a railroad
1:20:20in the district
1:20:21or a railroad in the neighboring
1:20:22district all that matters is
1:20:25uh what the model says should matter
1:20:26which is your targetness so
1:20:28this is kind of a very elegant way to
1:20:30show that indeed
1:20:32these substantial increases in welfare
1:20:34are indeed coming about through the
1:20:35mechanism
1:20:36of of greater trade and uh being able to
1:20:40exploit this comparative advantage that
1:20:41the model
1:20:42i would predict and these gains are
1:20:44large and the final thing you can read
1:20:47the slides or read the paper
1:20:48is there's a whole bunch of placebo
1:20:50checks that are very compelling
1:20:51involving lines that weren't built that
1:20:55were planned and we don't find effects
1:20:57on those instrumental variables
1:21:00using rainfall which predicted the
1:21:02famine
1:21:03and the famine commission built railway
1:21:05lines uh afterwards
1:21:06in response to the famine as well as
1:21:09bounds
1:21:10using lines which were labeled as
1:21:12commercial military or redistributive
1:21:14which should have very different biases
1:21:16but we find similar effects
1:21:18in all those cases okay so
1:21:22in the interest of time i i should
1:21:24probably wrap up as i mentioned there's
1:21:26a
1:21:26nice paper by rob jensen and co-authors
1:21:28on information frictions
1:21:29showing what happens to the boat
1:21:30building industry in various towns in
1:21:32india
1:21:33uh when uh now there's a kind of free
1:21:35flow of information
1:21:36uh of uh the various uh attributes of
1:21:40the firms in different locations
1:21:42and uh uh how we get very mellots like
1:21:44adjustments to that
1:21:46but i'll leave that uh for you guys to
1:21:47read up in the slides
1:21:49so just to conclude we were really only
1:21:52able to touch the surface of the vastus
1:21:54show on the effects of spatial frictions
1:21:55in developing countries
1:21:57there's a couple of nice review articles
1:21:59none of which uh
1:22:00maybe cover all of this but i i which
1:22:03gets at this in parts
1:22:04uh uh there's a very nice annual view of
1:22:06economics uh
1:22:07article on the gains from market
1:22:08integration not development
1:22:10specific uh but very relevant here i
1:22:13have a review article on
1:22:15trade and development which touches on a
1:22:17bunch of these issues or although
1:22:19also on other issues i think about kind
1:22:22of distortions
1:22:22not just spatial ones uh more generally
1:22:25and then there's a couple of recent
1:22:27papers
1:22:27think about economic geography either
1:22:29urban economics
1:22:30or spatial frictions within cities uh as
1:22:33well as uh
1:22:34uh uh spatial frictions across locations
1:22:37uh
1:22:37in that trade in geography again uh the
1:22:40latter one's not
1:22:40development specific but as some of
1:22:43those papers are
1:22:45um okay so thank you very much uh uh
1:22:48uh uh for for listening and i'm very
1:22:51happy to
1:22:52open the floor to questions for the rest
1:22:55of the time
1:22:56fantastic thank you so much david um
1:23:00let me open this to the front row
1:23:03yeah i would have a quick question um
1:23:06which was on about um your 2015 paper
1:23:11where you were um saying that it was
1:23:14very important that you could
1:23:16show that there was actual trade going
1:23:17on you said in that case you
1:23:20you collected data on the factory i was
1:23:22i was wondering whether you could maybe
1:23:24repeat that like
1:23:25exactly what what you did there to
1:23:27ensure that
1:23:28and maybe share any any ideas on
1:23:32uh on a good good approaches for us to
1:23:35see
1:23:36whether trade is actually occurring
1:23:38maybe strategies to find that out
1:23:40most easily um yeah or any learnings you
1:23:42might have had from that project
1:23:44okay so in that case we just did it by
1:23:46brute force we had ras call up
1:23:49uh all of the companies or or find
1:23:52newspaper articles
1:23:54about factory openings uh et cetera
1:23:57and in africa they're actually pretty
1:23:59willing to hand over the information in
1:24:00the u.s
1:24:02more more were skeptical of why do we
1:24:04need to know exactly where my
1:24:06garlic and onion ragu pasta sauce is
1:24:09made i told you
1:24:10where we make pasta sauce but
1:24:13uh we managed to at least get about 100
1:24:16products that way
1:24:18um uh obviously
1:24:22you know there is some reselling going
1:24:23on and so there's still some measurement
1:24:25error
1:24:26attached to that and maybe some
1:24:27smuggling across the border where they
1:24:29come in from other factories
1:24:31uh uh but we think those effects will be
1:24:32somewhat minimal so
1:24:34um that is on um uh um
1:24:38uh what we did there laurent bergquist
1:24:41has a nice paper
1:24:43with um
1:24:47with craig mcintosh
1:24:51i guess they haven't published the paper
1:24:52yet or haven't got a working paper yet
1:24:54but has the
1:24:55data collection which they use in
1:24:56another paper now with with
1:24:59ben faber and ted miguel and others
1:25:03where she went and did surveys where she
1:25:05basically asked
1:25:07traders i believe where the flows were
1:25:10occurring within the country so
1:25:12where did you get you know this stuff
1:25:13you're selling where did it come from
1:25:16uh there's also many countries have have
1:25:18trucking surveys
1:25:20um maybe sub samples uh which may have
1:25:22information on that
1:25:24and the final source uh i i hear would
1:25:26be
1:25:27uh sometimes market surveys also ask
1:25:30uh for the origin uh of locations and so
1:25:34you know in my mind the you know you
1:25:36know
1:25:37if you have a setting which you've
1:25:39already got lots of other data on you've
1:25:41already written the paper on
1:25:42you may not be able to find this data
1:25:44but if you're
1:25:45if you're if you're trying to do
1:25:47something from scratch
1:25:49uh there are plenty of places you can
1:25:50get this data sorry there may be a
1:25:52little bit of an invasion as my son's
1:25:53just come home from the park
1:25:56and um he likes people so it should be
1:25:59okay
1:26:00um uh so yes i think you
1:26:04there are many countries you can find
1:26:06this type of data but it's not
1:26:07necessarily
1:26:08available for the products you already
1:26:09have in the country you're already
1:26:10working on
1:26:11and so it depends on how important this
1:26:13type of thing is
1:26:14uh uh uh whether you kind of look in
1:26:17those settings or not
1:26:21thank you
1:26:25other questions
1:26:31yeah i have a small question just out of
1:26:33curiosity when you presented
1:26:35the like how trade costs change by
1:26:38distance
1:26:39or how the price gap changed my distance
1:26:42in the us
1:26:43the the price gap becomes negative on
1:26:46like the trading peril me
1:26:47thing yeah
1:26:51right i guess why do you think
1:26:54that is that this
1:26:57um so i'm forgetting that's the trading
1:27:00power of the non-trading pair
1:27:07uh obvious explanation or remaining
1:27:10explanation is
1:27:10the markups must differ across space and
1:27:13there so happen to be higher markups in
1:27:15proxima locations and more distant
1:27:17locations
1:27:18the the thing that's unusual about the
1:27:20us compared to nigeria and ethiopia
1:27:22is nigeria and ethiopia basically all
1:27:24the factories are in the big cities
1:27:26uh often in the hub of the country
1:27:29either in lagos or
1:27:30or or close to addis and so
1:27:33uh there uh we have a very different
1:27:36relationship compared to the us
1:27:37where maybe you have factories spread
1:27:39out across the us and then the big
1:27:41cities are like
1:27:42west coast chicago east coast and so you
1:27:45get very different
1:27:46kind of uh uh uh uh relationships and
1:27:50these urban versus rural price
1:27:51differences or bigger cities versus
1:27:53smaller price
1:27:54uh uh smith city price differences which
1:27:57are related to markups
1:27:58i seem to be generating that pattern
1:28:01thank you
1:28:11it's a quiet group at the moment which i
1:28:14think may just
1:28:15also reflect the fact that it's the for
1:28:18people on this side of the atlantic it's
1:28:19the end of the day on a friday
1:28:21um so if i'm not hearing any more
1:28:25questions i think maybe what i would
1:28:28just do is ask those who are in the
1:28:29front row
1:28:30to unmute and join me in giving david a
1:28:34round of old-fashioned applause
1:28:41thank you so much david for taking the
1:28:42time and for for
1:28:44sharing your your knowledge with us
1:28:47and um again for doing this out of the
1:28:50goodness of your heart
1:28:51we really appreciate it it's great
1:28:53lecture and
1:28:54uh lots of useful links for people to
1:28:56follow so
1:28:58i think um these review articles are
1:29:01available i think before your your young
1:29:05company comes back into the room
1:29:07um we might just want to let you go
1:29:09david and again to thank you enormously
1:29:11for taking the time
1:29:13so um yeah and say and thank you all for
1:29:16uh quick for great questions and also
1:29:18for for uh
1:29:19sticking around i think this is the last
1:29:21official lecture i serve the course
1:29:23uh so congratulations if you've attended
1:29:25all of the official
1:29:26uh non-bonus lectures um uh
1:29:30and i'm happy if anyone has any further
1:29:32questions just reach out to me
1:29:33at uh adkin mit.edu
1:29:38and i'm happy to reply or wanted to
1:29:42ask about related things so thanks a lot
1:29:45and just to say though that we do have
1:29:48um
1:29:49we do have two supplemental lectures
1:29:52for next week on thursday and friday so
1:29:55continuing the spatial frictions theme
1:29:58in a sense
1:29:59so natalie raimondo will be talking
1:30:00about trade fdi and development
1:30:02and then klaus desmond will be talking
1:30:04about urbanization
1:30:06so i think as you'll all appreciate the
1:30:09the partialing out of the lectures into
1:30:12core lectures in supplemental was kind
1:30:14of arbitrary but i think those will very
1:30:16much follow and fit with
1:30:18what you've seen from melanie and from
1:30:20david and so we look forward to seeing
1:30:22people back
1:30:23next week so maybe on that note i'll
1:30:26just
1:30:27again say say thanks steve
1:30:32thank you

About this transcript

This page contains the full transcript of Lecture 12: Applications to development by CEPR & VideoVox Economics, generated from the public captions YouTube serves with the video. The transcript has 15,466 words across 2,658 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.