YT YouTube2Text

YouTube2Text → Transcripts → Lecture 2: Firm Upgrading with Eric Verhoogen

Lecture 2: Firm Upgrading with Eric Verhoogen — Transcript

by International Growth Centre · 18,267 words · 2,737 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:04thank you
0:18hello everyone
0:20um so welcome everyone to the second
0:22lecture of our bread igc virtual PhD
0:24course on firms today we have the great
0:27pleasure of having Eric verhugen uh talk
0:30to us uh about uh upgrading and uh uh
0:33the kind of component of aggregate
0:36productivity uh that we kind of uh put
0:39in the a in the previous lecture uh just
0:41a reminder after each uh lecture um the
0:45lecture notes uh the lecture slides will
0:48be posted on the course website as well
0:51as a reading list and uh the second
0:54thing to note is that remember if you
0:56have any questions please put them in
0:57the Q a and we'll stop on occasion and
1:01ask Eric uh some of those questions uh
1:04as well as have a set of 20 live
1:06panelists will be able to ask questions
1:08in person so I welcome everyone and uh
1:11take it away Eric thank you so much for
1:13joining us
1:15okay thanks so much great to see you
1:16great to see everyone yes for the
1:18panelists if you want to ask the
1:19question just raise your hand put your
1:20hand up in Zoom um and then Caleb will
1:23unmute you and then I can I can call on
1:25you and I'll try to stop periodically to
1:27to listen questions
1:29um okay so let me I wanted to start out
1:32talking about how this lecture today
1:34relates to what uh David and Dave talked
1:37about yesterday right so the I think
1:39they phrase it very well the big
1:40question for the courses and why are
1:42some countries rich in some countries
1:43poor and sort of a more operational way
1:46praising that is why do some countries
1:47have low
1:48um aggregate output per person uh and
1:51they discussed
1:52two broad types of answers to that
1:55question they're probably more than
1:56non-exhaustive list but two broad types
1:58uh one type would is that individual
2:01firms are not reaching the production
2:04Frontier are not producing as much as
2:07they could with a given amount of
2:08difference so in their case this is
2:09their notation Qi was output at the firm
2:12level X I where the inputs being used by
2:14the firms and then a I guess was the set
2:16of Technologies or the states the state
2:18of knowledge you can say uh in The Firm
2:21that's helping the firm convert inputs
2:23into outputs and so the idea one idea is
2:26that the the AI of the firm currently
2:28has it may be below some potential a and
2:31so maybe that a could be increased or
2:33expanded in some way the other big type
2:38of answer was that somehow the X's are
2:41misallocated across firms the inputs is
2:43like labor capital and other things may
2:46not be flowing to their their most
2:47productive use okay and so in this
2:49lecture I'm going to focus entirely on
2:51number one okay and that's and I guess
2:53Jing guys uh lecture next time also will
2:57also mean leave me out about that about
2:59firm's relation there that you know the
3:01repression Frontier for them given there
3:04given their x's and then subsequent
3:05lectures by deep clean out and others
3:07we'll talk about you know reasons why
3:09the X's might not be flowing to the most
3:10productive
3:11okay so that's how that's one definition
3:14of upgrading people have different
3:15definitions I'm going to be a little bit
3:17more precise and Technical later on
3:19about what upgrading is but I want to
3:21think about this as this process of
3:22reaching the production Frontier within
3:24the firm so Shifting the firms AI more
3:26towards it sort of globally Global
3:28potential uh hey
3:30okay now in principle in developing
3:32countries the fact the knowledge exists
3:34elsewhere has been developed in rich
3:36countries should be a boon to those to
3:39those firms okay so again famously maybe
3:42somewhat undiplomatically referred to
3:44this as the advantage of backwardness so
3:46it's not very PC but I think it's useful
3:47it's useful useful concept right there
3:50being a latecomer and development
3:51process should be helpful in the sense
3:52that all these Technologies and new
3:54products and new ideas have been
3:56developed elsewhere and could in
3:58principle be adopted off Shelf but I
4:00think it's not controversial to say that
4:04for many developing country firms these
4:05events ever made those something seems
4:07to be getting in the way of adopting to
4:10a frontier technologies that you're
4:11producing Frontier uh products Etc and
4:15so that's you can see the motivating uh
4:17question at least for my part of the
4:18lecture for this part I guess of this of
4:21this course is what is that that's
4:23getting in the way
4:24um I uh I think that can be turned
4:27around also restated in a positive way
4:28anytime you identify a barrier you're
4:30also implicitly identifying a driver of
4:32upgrading if only something that's
4:34reducing a bear here right so because
4:35think this that's kind of the question
4:37is what are the drivers of upgrading at
4:39the firm level uh I've got I've just
4:41written a review paper for the Journal
4:43of economic literature close to my
4:45website uh and so that you can refer to
4:47also there's more details there I'm
4:48going to rely on that fairly heavily in
4:50the in the lecture and then as uh as as
4:53David said I guess the the slides will
4:55be posted right after right after this
4:57okay uh so just a little bit of a road
5:00map because after the introduction very
5:01short I'm going to talk some do a
5:04conceptual framework if you think of it
5:06almost as an extension of the framework
5:07that that David and Dave
5:09um put up yesterday I'm going a little
5:11bit more micro uh to to try because I
5:14think it's actually important to
5:15understand concretely what upgrading is
5:17to go a little bit more micro on that so
5:18I'll do that and I'll talk about sort of
5:20measurement issues how do we measure
5:22upgrading in the light of that
5:23conceptual framework uh and then the
5:25meat of the of the talk today is it's
5:27really about examples of
5:29uh great you know high quality papers on
5:32evidence on drivers of upgrading what is
5:34it what do we know about what's what's
5:35driving upgrade or not and then some
5:37conditions
5:39Okay so
5:41essentially what I want to do in this
5:42first part of the of the lecture is can
5:44you see my my cursor moving around yes
5:47okay is think about you know what is
5:49this AI thing right it's it's kind of
5:51conceptually it's common to put it in
5:54that a production function like the one
5:55the ones that they they did but but I
5:57think we could unpack it even even more
5:58for that we need to go even more uh
6:00micro I'm gonna this is slightly
6:02confusing and I apologize but I'm going
6:03to use the notation that's in my
6:05literature review so that you can that
6:06is going to help you if you want want to
6:08go uh refer to my literature review uh
6:10but it means a slightly different from
6:11what Dave and David
6:13um did yesterday okay but I want to go
6:15micro in the sense I want to go all the
6:17way down to the firm product technique
6:19level production solution okay so why is
6:21firm sorry a lie's output I is firm J is
6:25product K is technique T is time okay so
6:28you can think of that as a particular
6:29production process you know a pricker
6:31technique for predicting for producing a
6:33particular product okay and that's
6:37um so that's that so production function
6:39is going to be specific to a firm and a
6:40product and a technique it's going to
6:42use a vector of inputs which can also be
6:44specific to a product and technique and
6:47then it'll depend on these lambdas
6:49uh which may also be specific to a
6:51product technique okay which I'm going
6:53to call that following uh John Sutton
6:56and others call that capability
6:58what some people would call it
6:59productivity but I think productivity is
7:01used in lots of different senses so I
7:02think it's clear to call it
7:04um uh call it capability okay so now a
7:07key this issue about capabilities we
7:10want to think of this as some this is
7:12really something that has to be
7:13homegrown it can't really just be
7:15purchased on a market okay so the
7:17Homegrown that's that comes from Gibbons
7:19that language or uh and Andrea Pratt
7:22have a recent paper developing this this
7:24idea more that that if it were something
7:27that you could just go out and buy on
7:28the market then arguably it could go and
7:30use them back right but so capability is
7:32something that sort of has to be
7:33developed uh within the firm okay I'm
7:37going to talk about quality I'm going to
7:39have just a single parameter
7:41characterizing the out this is output
7:42quality and then a vector of input
7:45qualities for the different inputs uh
7:47it's a it's a strong assumption to say
7:49there's only one dimension of those
7:51there may be multiple Dimensions but I
7:52think we can communicate many ideas just
7:54just one dimension and so we can think
7:56of the the with this notation think
7:58about products of different Quality
8:00Systems having different JS you can
8:01think of those as different products but
8:03they also have they're also
8:04characterized by a fee or an inputs are
8:06characterized by a by now
8:08okay then I I'm gonna let into a little
8:11bit of notation having this just I
8:12promises two slides of notation and then
8:14some some more General discussion
8:17um so think of J as the set of products
8:20uh and K as the set of techniques which
8:23are specific to those products
8:24potentially
8:25um that for which The Firm knows this
8:27production works the firm you know may
8:29or may not know about how to produce
8:30something if it knows it it knows that
8:32you know improves that product using
8:33that technique then it would they would
8:34go in these
8:36um in these steps okay and then you can
8:37think of the sort of the superset of
8:39these lambdas I've had a little uh
8:42controls pop up here I hope that you're
8:44not seeing those but
8:46um you can think of the superset of the
8:48of the of the Lambda capabilities I
8:49think that's a big set of capabilities
8:51and then you have the J and K I'm going
8:53to refer to those together as know how
8:54all those things are components of
8:56know-how
8:57okay a little bit more notation if you
9:00can think of destination but there may
9:01be different destination markets index
9:03those by B there might be a bunch of
9:05fixed costs at different levels and
9:06fixed costs of a particular product
9:08technique and fixed costs and producing
9:09a product for a given Market if the
9:11substitution apart a product that's not
9:13specific to given Market a cost that's
9:16specific to the market product unless it
9:18costs that specific to the firm and then
9:20also I think it's important sometimes
9:22lost that we can think about firms as
9:24investing in know-how they may invest in
9:26game capabilities they may invest in
9:28learning how to produce more products
9:29they may invest in learning about about
9:31new techniques okay I'm going to fold in
9:34all of the sort of external to the firm
9:36stuff that's going to be folded into two
9:38functions wanna an inverse output demand
9:40curve and an inverse input Supply so
9:42it's kind of a it's a it's a it's a
9:45shortcut in a way uh there's lots of
9:46interesting stuff going on outside the
9:48firms I'm just going to fold it into
9:49these functions but you can think of so
9:51this is an inverse output demand curve
9:52so I'm relating this is the price that
9:54the firm can get for its products so
9:55there's a function D of how much they
9:58produce
9:59the quality of the product and that a
10:01bunch of exotions to the firm stuff okay
10:03this Capital gamma is sort of a whole
10:05bunch of excess of the first stop
10:06external external firm and then similar
10:09on you can imagine them saying it's
10:10facing an inverse input supply curve it
10:13does nothing doesn't have to be you know
10:14perfect competition either on the output
10:15side or the input side of this Frame
10:17will kind of can incorporate imperfect
10:19competition Market power on either side
10:21so this is this would be you know input
10:23prices are the W's this is the supply or
10:26inverse apply function okay it's going
10:28to depend on how much of each input
10:29they're using the qualities of those
10:31inputs and then some submits and some
10:33external to this term to the firm's die
10:35okay and then the first problem
10:37okay is going to be as total profits
10:41it's going to be they're going to be
10:42maximizing the present discounting value
10:44okay that's this is this um Delta is
10:48just a just a discount Factor B Star is
10:52the set of uh markets they choose to
10:54they choose to enter J star is the
10:57capital J Stars a set of products that
10:59the firm choose to produce those are
11:01going to be those are those are those
11:02are choices
11:04um this would be revenues okay so output
11:07times times the prices minus that's
11:09that's input costs of variable inputs we
11:12have a bunch of these fixed costs
11:14okay also the firm level and then they
11:15can decide to you know invest in
11:17know-how investing capabilities invest
11:19in products or invest in techniques okay
11:21so the firm's uses again I'm having this
11:23little I don't know how to
11:25get rid of those controls
11:28I don't know can you all see those
11:29controls that are blocking my bottom of
11:31my screen is next
11:34okay uh so the firm's gonna you could
11:37choose definitions products techniques
11:39um for for a technique for each product
11:41you know what inputs to use and how much
11:42to invest in these things okay so that's
11:45a lot of notation not a lot of
11:47falsifiable predictions yet but I think
11:49it helps to clarify a few things
11:52um one is just what being more precisely
11:55what we mean by upgrading okay and I
11:57think that the literatures use the term
11:58in four relatively specific uh ways one
12:02is probably the most common would be
12:04gaining know-how so one form of
12:06upgrading is you know you you gain
12:08lambdas versus it possibly for some
12:10subset of these product techniques
12:12or maybe for all product techniques you
12:16gain knowledge of products regain
12:17knowledge of techniques okay so that's
12:19that's typical learning
12:21another form of upgrading would be
12:23quality upgrading which is we want to
12:25think of it as just increasing the
12:27average quality this of this average of
12:29our of our fee so that just means this
12:31is just an output weighted average of
12:33the fees of the products that you
12:36produce
12:37okay
12:39um I'll come back to a point about that
12:41in a second product Innovation would
12:43just be producing a product that you
12:45haven't produced before that wasn't in
12:47your jstar in the past
12:49okay uh for for you for for some for
12:52some for some Market that would be your
12:55product innovation technology adoption
12:57in a similar sense would be just use of
12:59a technique that you as a firm have not
13:01used before
13:01we uh we you might like to think of sort
13:04of Technology adoptions adoption about
13:06some new and improved product and
13:08product Innovations as production but
13:10some kind of a new improved product but
13:12what new and improved mean is actually
13:14complicated and so I think we're better
13:16off just thinking of any new product and
13:18any new technology as going into going
13:20to upgrade okay so one first point here
13:23is that these dimensions of upgrading
13:25are they're related often they go
13:27together but they're distinct okay so
13:30you could have quality upgrading like
13:32you could just among the products that
13:33you're already producing you could start
13:35producing more about the high quality
13:37one and less of the low quality one that
13:39would lead your average quality to
13:40increase when we count that as quality
13:42upgrading even though you haven't you're
13:43not producing any new products maybe you
13:45haven't learned anything about how to
13:46produce this product you're just respond
13:47to Market this is Shifting the weights
13:49on the different those different
13:50products right so you could do that
13:52without without product Innovation you
13:54can have uh technology adoption uh
13:57without without learning
13:59et cetera so you could you could you
14:00could adopt a new products sorry you
14:02could adopt a new technology that you
14:04already knew about it was already near K
14:06you just had never had never used it
14:08okay and so that could be that wouldn't
14:10constitute learning but it would
14:11constitute uh technology okay let me
14:14pause it I see that Caleb already has a
14:15hand up so you want to say something
14:17there
14:18hi yeah we have a few questions in the
14:21chat regarding this
14:22um one from Keith the Christian uh can
14:25one in four year drive two and three
14:28and then another question from Victor
14:30zuluaga asking or saying friends can
14:34increase the quality of their inputs for
14:35example they can provide training to
14:37their workers I'm wondering how that is
14:39captured in the current framework
14:42yes either yes yes yes yeah okay so I
14:45don't want to take yes so good question
14:47I don't want to take a strong position
14:49on what's driving what among these four
14:51I think that's true you could have think
14:53you know that uh well and often it's
14:55sort of hard to know you're going to
14:57have some external shop the firm is
14:58going to react in a bunch of ways one
15:00two three four might all move so it's
15:02hard to say you know one is driving four
15:03two is driving three I feel more
15:05comfortable with statements like you
15:07know on a product Market shock is
15:08somehow inducing The Firm to to uh
15:11to to react on different in different
15:13ways but in principle you could you
15:15could like if you could just drop some
15:16capability on people's heads right which
15:18is argument anyway you know the projects
15:20will help I'm happy to talk about we're
15:21trying to do that we're trying to drop
15:22information which is the way drop
15:23learning and then seeing how firms
15:25platforms react right so that could
15:26happen but also could happen that like
15:28I'm gonna um
15:31they could go the other way like the I'm
15:33going to talk about a paper of
15:34Davidson and Anaconda wall and Alamosa
15:37which is arguably what's happening is
15:39that there's you know firms are being
15:41reduced to improve fire quality stuff
15:42and that ends up generating learning so
15:44it could go you know two to one would be
15:46a possibility okay so any any of those
15:48things
15:49um uh could be could be could be
15:51possible and then the second question
15:52was what about a a firm increase in the
15:55quality of its inputs yes so often the
15:58upgrading process often for instance
16:00output quality upgrading involves input
16:02quality operators I'm trying to argue
16:04that in the industry papers
16:06um I guess you could say the question
16:08maybe is wouldn't input quality
16:10upgrading by itself counts as
16:14um count as upgrading
16:17I guess what I would say about that is
16:20which is not particularly
16:23um
16:24anyway it's not particularly insightful
16:26but one could think of remember the the
16:29using a different mix of inputs I would
16:32consider using a different technology
16:34that's a different technique so that
16:36some sense the mix of inputs is included
16:38in that in the what I mean by technique
16:40and so if you just you know use the
16:42different input so different quality
16:43input I would call it different input
16:44and so that would be using a different a
16:46different uh you know different
16:47technique so in this framework at least
16:49I would go into technology but I agree
16:51that that often goes directly with
16:53quality upgrading so that's a little bit
16:54of a little bit unsatisfying
16:57okay all right let me uh let me move on
17:00thanks for the questions I'm glad to
17:01have
17:02um some questions happening okay now
17:06uh
17:09normally
17:11when we're talking about firm Behavior
17:12we don't go to this firm product
17:14technique level and right foreign
17:16product technique building Productions
17:17right the standard thing to do is to
17:19posit that there's a firm level
17:21production function now David and Dave
17:23left that as pretty General it's just
17:25you know f of x i comma a I right
17:27without specifying too much a more
17:30typical thing would be to specify
17:31something like a cop that was production
17:33functionality like this might be trans
17:35log it might be you know there are a
17:36couple of other popular forms that are
17:38used would be that use something like
17:39that there's there's an important and I
17:42think underappreciated question and one
17:43that might be good for research to be
17:45honest if for those who are fishing
17:46around for for research topics is can
17:49production product with technique level
17:51Perfections like the ones I wrote before
17:53do they aggregate nicely into you know a
17:56firm level production function like like
17:58this one okay and this kind of mirrors
18:02or Parallels an old debate about whether
18:04firm level production functions can be
18:06aggregated nicely into an aggregate
18:08level production okay so so does that
18:11you know could you write down this
18:13product technique level functions then
18:15do they does does a function like this
18:17still an adequate description of how the
18:19firm is converting you know inputs k and
18:21l and m into Intel
18:23and the answer to well and there's a
18:25couple answers there so one is under
18:27certain conditions so who attacker had a
18:29famous paper this is again when from
18:31firmative aggregate level you know if if
18:34the firm level production functions are
18:36Leon TF and there's a distribution of
18:38those uh the possible Technologies and
18:40you aggregate up then you get a complex
18:42function at that
18:43level
18:44um but that there was a whole debate
18:46about this Franklin Fisher probably the
18:48best known sort of critic of that of
18:50this kind of uh this kind of approach
18:51saying you know it's only under very
18:53special conditions very special
18:54assumptions about distributions Etc
18:56where that works
18:58okay
18:59um now one could make a similar argument
19:02about you know going from The Firm
19:04product technique level to the firm up
19:06that it's it's under special condition
19:08so here are a couple of recent papers we
19:09want to look at these papers I recommend
19:10them Chad Jones has a nice paper again
19:13uh you know um we got the abduction
19:16functions at the at the at the
19:19product or the technique level and they
19:20aggregate up to a Cobb Douglas at the
19:22firm level or in Bowman Overfield never
19:24even worked learning something along
19:26similar lines but again it's pretty
19:28pretty under pretty special conditions
19:30that happens
19:31um so
19:33let's say let's say we don't have those
19:35special conditions forwarding right and
19:37what we do is we say we take this
19:38reality which is sort of varying at the
19:40firm product technique level and we
19:42assert that there's this firm uh firm
19:44level production function that looks
19:45like equation two there
19:47in that case a lot of stuff is going to
19:50end up buried in the a
19:52right so that a is going to end up
19:54capturing basically all those four
19:55dimensions of upgrading I was talking
19:56about will end up showing up largely in
19:59this in this a rather than in the the
20:01choices of techniques okay so it's I
20:04guess why am I saying that when I asked
20:06in the beginning like what is a
20:08I'm going to then talk about way you
20:10know empirical papers that are all
20:12sheddy line on one of these four
20:13dimensions of of upgrading okay but in
20:15some sense those are all going to feed
20:16into this a
20:18okay that's a little bit of an
20:19unsatisfying answer it's about I think
20:20that you know there'll be more work to
20:22be done here about you know tractable
20:24ways of going from these firm products
20:25technique level to the to the to the
20:28firmware I hope that's I hope that's
20:29clear someone please ask a question if
20:31uh if that point is uh is not clear okay
20:35so now let's I'm gonna let's step aside
20:37from this from this frequent X group
20:39within firm aggregation issue
20:41um and back to some broader issues that
20:43I think the conceptual framework
20:44highlights
20:46um one is I think it's important to keep
20:48in mind that LDC firms face very
20:50different conditions than firms in rich
20:52countries okay in particular often
20:54they're selling they're selling largely
20:56to local markets which are going to have
20:57pretty different demands but Bankers
20:59they face locally are going to be pretty
21:00different than the demand currency firms
21:02face locally in in rich countries you
21:04know typically they'll be facing poor
21:06consumers those consumers will be less
21:07willing to pay for Quality may make less
21:09sense to produce high quality stuff for
21:11instance you know facing poorer local
21:13consumers same thing different input
21:15Supply curves and prices so often high
21:18quality inputs and high school workers
21:19are relatively expensive in in uh yeah
21:23in the United States in developing
21:25countries they may not be more expensive
21:27than rich countries but relative to say
21:28low skill Labor uh High School labor is
21:31is relatively expensive okay and then
21:33also these firms and developers
21:35developers often have lower levels of
21:36know-how they have lower less capability
21:38you know fewer capabilities and maybe
21:40there's don't know about products
21:43related to that to this point about you
21:46know conditions are very very
21:47heterogeneous
21:49is that upgrading as we defined it may
21:52or may not be optimal for a firm okay so
21:54often the way I post it was the question
21:57in advance of backness why isn't this
21:58happening I was almost sort of
21:59implicitly saying well we think it's
22:00optimal to use Frontiers to do the art
22:03technology and to produce you know
22:04state-of-the-art uh products but that's
22:06not necessarily the case right because
22:08it firms facing very different demand
22:10conditions very different Supply
22:11conditions they have different know-how
22:12it may be optimal for them to produce
22:14you know products using techniques that
22:16are that are below different too here's
22:18a nice quote from from Andy Foster and
22:21my Grocer's hike which has had a nice
22:23review paper in 2010 on technology
22:25adoption
22:26um thing you know it cannot be inferred
22:28from the observation that farmers using
22:30high levels of fertilizer earn
22:31substantially higher profits than than
22:33farmers who use a little fertilizers the
22:35more performers should use for
22:36fertilizer right the conditions facing
22:38those Farmers may be different right and
22:39so the optimal choice of fertilizer may
22:41be the amount of fertilization
22:43now know how there's a little bit of
22:45when you're talking about know-how
22:47know-how is in itself like having more
22:49land is good for the firm
22:51so in that sense you might say well it's
22:52a no-brainer I should really try to try
22:54to acquire our capability but remember
22:56that we had to pay for that keyboard I
22:58was assuming in the model that in this
22:59framework that there's this capital I
23:01Lambda which was the investment you have
23:03to make in know-how so if you have to
23:05pay the game know how right then that
23:08payment may or may not be worth it the
23:09amount of know how you get uh may not be
23:12worth the amount you have to invest in
23:13large in order to receive
23:15okay
23:16um Jing is going to talk next Thursday I
23:19guess it is on About Management more
23:22detail but I think this framework I'd
23:23like to just highlight a couple of quick
23:24points and then I'll uh I won't talk in
23:26in a lot of detail but highlights and I
23:29think there's some fuzziness in the
23:30literature out there about about these
23:31books
23:33um one is that management uh people say
23:35oh you know why are our country firms
23:38not doing well Batman
23:40so management
23:42um though means three different things
23:44in this at least in this framework I
23:46think it's useful to kind of distinguish
23:47them one is that you know firm may not
23:50have a lot of know-how they have low
23:51know-how and that would include you know
23:54I know how to include entrepreneurial
23:55ability entreprenewability you might
23:57think of as like a Lambda that applies
23:58to all products and techniques for a
24:00given entrepreneur
24:01getting myself well
24:03we'll know how another thing uh that's
24:07often warmed under management is the
24:09skill of employed managers right which
24:11is that's in the that M Vector of inputs
24:14okay so a firm may have may not have
24:16hired a very good managers uh this is
24:19not the owner or the entrepreneur but
24:20this is the employee managers and so
24:22that would also might show up as you
24:24know what was typically thought it was
24:25bad management and then the other thing
24:27is that it might be management you know
24:29poor management practices
24:31now but practice is here I want to think
24:33of practices as components of techniques
24:36right or they are techniques or those
24:37are component of a k for a human product
24:40I mean good company and and think of
24:42things that way John Van green has a
24:43paper making that argument since they're
24:45an early paper of the blue medal in the
24:48experiment also make makes the statement
24:49very very clearly so we want to think of
24:51the technology so the practices
24:52themselves or technology choices which
24:54is made on the basis of know-how and on
24:56the basis of the you know the skill of
24:58the employee managers
24:59um okay why am I why am I sort of going
25:02into into detail here I because I think
25:05there's a little bit often there's a I
25:06think people are too quick to jump to
25:08the conclusion that oh it's just a
25:10problem of bad management if the you
25:11know managers in developed countries
25:13went to business school in the US all
25:15you know then everything would be better
25:16and and they reach the frontier more
25:17quickly and they'll be upgraded
25:19um
25:20so that view is kind of basically
25:22underlying that view is the notion that
25:24that management practices can be ranked
25:27in a vertical sense of labeled the
25:29vertical view that some practices are
25:30better than others across context so
25:33that would be in a way that thing that
25:36that the you know proctoral profit at a
25:38given you know base year you know town
25:41for a k given all the other stuff okay
25:44is higher than profit four
25:46um using a k Prime some other technique
25:48for all possible sort of demand curve
25:50Supply curves extended know-how and and
25:53uh and and and products now there may be
25:56some technology and techniques or
25:57technologies that have that
25:58characteristic like you know probably
26:00like letting your inventory rot in the
26:01warehouse is probably not a good
26:03practice uh any you know at any time but
26:06uh but for many anyway so
26:09the other view that I'm more sympathetic
26:11to what we often call the horizontal or
26:13the contingency view which is that you
26:15know which practice is best depends on
26:17the context depends on marketing
26:18conditions it depends on know-how right
26:20so so you know which management practice
26:22is the best if you depending on the you
26:24know the market positioning of the firm
26:25what they're trying to produce who
26:26they're trying to cater to
26:28okay that's ultimately an empirical
26:30question whether you know the vertical
26:32view or the horizontal view makes makes
26:33more sense I would say
26:36controversial I think sorry but I would
26:39say so normally if you had
26:41don't think about a management practice
26:42think about just some other sort of
26:43Technology right like uh I'm going to
26:45talk about a cutting technology we think
26:47about I don't know Sewing Technology or
26:49a type of machine and you looked out of
26:51the world you saw different firms are
26:52using different sewing machines or you
26:54know in different in different places
26:55you wouldn't necessarily say aha the
26:58guy's using this the non-state of the
26:59art sewing machine they're making
27:00mistake you think about you know what
27:02are the constraints that perms are
27:04facing that lead them to use that type
27:06of machine right you wouldn't
27:07immediately assume that like well the
27:08state of the art one the more modern one
27:10is is better in all in all situations
27:12and therefore they're just they're just
27:13they're just making mistake
27:17okay
27:18um
27:19okay so that's that that's what I had to
27:20say about so let me pause there I don't
27:23know Caleb are there are there q a
27:24questions that uh seem like it uh would
27:28be useful to to discuss
27:31um yeah there are a few ones that I'll
27:34read that have just been upvoted uh one
27:37is from Adrian saying are the Homegrown
27:40capabilities here tied to a firm or can
27:43they also be tied to workers and then
27:45indirectly be purchased through the
27:47market by hiring such workers this is
27:49going back a little bit yeah yeah good
27:52question good question yeah so I I guess
27:56um
27:56yeah I would think of where did I have
28:00it here yeah I would think of if it's a
28:02worker capability I I could think of
28:04that and you know it's possible to go
28:06out on the market and buy a worker who
28:07has this capability to pay for one then
28:09I would think of this as part of the
28:10part of the vector of inputs right
28:12whereas the the Lambda is something
28:14that's that really is that you can't be
28:16bought it can't be bought on the market
28:18that really is sort of somehow homegrown
28:19or embedded embedded in The Firm now
28:21it's there's a little bit of a gray area
28:22here because often capabilities and
28:24workers end up being transmitted to the
28:26firm right and that's that's an
28:28important process so uh and it's it's
28:30not always you know black and white to
28:32what extent that that has occurred but
28:34but as far as the framework goes I would
28:35think I would tend to think of it as
28:36worker skills in the is the m and then
28:39the Lambda stuff that that really is
28:40sort of embedded in The Firm that's
28:42that's not available online
28:45and we do have a Live question here from
28:48Leticia
28:49um do we have time for that yep
28:56hi
28:57um good morning
28:59I was just wondering where does
29:00political risk fall in your conceptual
29:03framework if it's something you're
29:05thinking about in terms of conditions
29:07that develop in country firms face as
29:10opposed to
29:11other firms in other countries
29:14yeah yeah good question I mean this is
29:16and it's sort of a sort of a fudge I
29:18recognize so I'm sort of embedding all
29:20that you know wanting all that stuff
29:22into these Capital gamma terms right so
29:24that stuff that's happening outside the
29:25firm which might affect you know how
29:27much demand there's going to be for
29:29their products you know given the
29:31quality or it might affect how much you
29:32know Supply there is going to be of
29:34their of their of their inputs right so
29:36it's I think it's I think it's it's
29:37relevant and and I I think an
29:39interesting research program is to look
29:41at things that are like shocks to these
29:42gammas
29:57and similar Point could apply to
29:59political risk um yeah
30:03maybe let's do one more I mean I I've
30:05left something Eric um yes yeah story I
30:10I sorry to interrupt I I think you got
30:12cut off there right you froze when you
30:14said an interesting research program
30:16would be shocks to these Gammas and then
30:17we lost you for about 20 seconds oh okay
30:21sorry about that sorry about the I'm in
30:23I'm in Mexico City uh and uh so the I
30:26guess the internet connection is not
30:27optimal sorry about that yeah so I
30:28started saying I I think it's an
30:30interesting research program think about
30:31trucks to different Gammas and how firms
30:33respond so uh Juan ma Castro vicenci
30:37this is coming up from Princeton on the
30:39telemarket this year has is sort of
30:41thinking about that with firms and
30:42environmental stocks and now you know
30:43firms have to head and maybe they do
30:45things that are so optimal that end up
30:47you know the output is reduced because
30:49they're spreading out the production
30:50across different places adjust to these
30:52environmental risks and so you can
30:54imagine a similar thing with political
30:55risk right that you want to you have to
30:56hedge your bets you don't want to have
30:57all your all your all your factors in
30:59one place that you know there could be
31:00uh in some uh government could come to
31:04power that doesn't doesn't like you and
31:05so you spread things out and that has a
31:07negative consequence for for output and
31:08so that's that's that I think that's an
31:09interesting program
31:12okay but let's maybe do was there one
31:14more question from Muhammad uh
31:17you want to do it quickly and then and
31:18then yeah sorry I don't know if you can
31:21hear me
31:24yeah I could I could hear you no I could
31:26hear you
31:28now you've muted yourself
31:30oh okay sorry I just wanted to ask a
31:33question about uh maybe cross products
31:35or cross techniques learning so is there
31:38this framework away that if you invest
31:41in one product
31:42uh you know in learning one product that
31:46induces learning
31:48or if that can import the channel in
31:51your opinion
31:54yeah yeah that's a good that's a good
31:57question so I haven't I haven't
31:58explicitly modeled that here I guess in
32:00the sense I mean it could I haven't
32:01ruled it out right in the sense that
32:04there'll be some you know process
32:06evolution of each of these lambdas right
32:09which could in principle depend on
32:11what other products are being produced
32:13right I guess I was I was mainly think
32:15of this you know these Investments but
32:17you could think of it you know I might
32:18decide today to produce what product a
32:20even though it's not that profitable but
32:22it's helping me learn about how to
32:23produce product B which I think is
32:24possible I agree that's a very
32:26interesting uh that's a very interesting
32:29possibility and I think that I think
32:30that could be you know built into that I
32:33haven't said anything about how these
32:34you know uh it's kind of a Lacuna here
32:37in the framework about how these these
32:38these lambdas uh evolve so you could
32:41build in a process for evolution that
32:42has that kind of those kind of process
32:46okay thanks thanks very much for having
32:48for the few questions and let me uh so
32:50let me let me move on to the to the
32:52measurements
32:54okay so so conceptually we have an idea
32:57of what we mean by by upgrading now
32:58think about you know how people try to
33:00try to capture these things
33:03um there there's a whole field of
33:04innovation mainly focused on rich
33:06countries uh and mainly using patents
33:09and r d expenditures as indicators of
33:11upgrade in terms of innovation
33:15um in developing countries arguably
33:17things like patents and r d expendures
33:19are less informative right one is that
33:21they're manipulative this is a paper by
33:23uh so it's Daniel Shu and others uh
33:27basically in China they gave incentives
33:29for having more r d spending and so
33:30firms relabeled other spending that they
33:32were doing anyway as as r d spending
33:34okay so that's that's one example of the
33:36of the you know r d spending maybe not
33:38being that informative but also launches
33:40what firms are trying to do they're
33:41trying to catch up they're not trying to
33:42necessarily push the world Frontier
33:44they're trying to catch up to the world
33:45Frontier and that won't necessarily show
33:46up in patents and won't even necessarily
33:48show up as r d spending and so in that
33:50sense arguably you know spending or not
33:53are not great as as measures I'd say by
33:56far the most common thing that people do
33:57to try to measure upgrading is they
33:59construct measures of total Factor
34:01productivity which I'm going to talk
34:02about in some detail so conceptually
34:04there's a real advantage to that it's
34:06really trying to get at directly these
34:07lambdas or some you know aggregation of
34:10these Landers into an a right uh you
34:13know in a way that maybe is a little bit
34:14unspecified
34:16um so that's that's a plus the minus is
34:19that there's lots of problems with cfp
34:21estimation I'm going to talk a little
34:22about that I think those problems are
34:24underappreciated there's sort of a
34:26strange kind of division of labor in the
34:28profession which is there's a
34:28productivity literature and then there's
34:30the non-practicumulator that uses those
34:31productivity measures and people just
34:33sort of accept whatever the state of the
34:35art is in productivity literature and
34:36just apply it and I think we need to be
34:38more critical um than than people often
34:41are
34:43um and so part of what I you know what
34:45I've been arguing and which I'm going to
34:46talk to you about today also is that I
34:48think there's a real benefit to going to
34:50more direct measures of Technology
34:52adoption quality upgrading and product
34:54Innovation so these uh you know data on
34:57these Dimensions is becoming
34:59increasingly available right it's now
35:00possible to get actual product level
35:02data see what products performs you're
35:04producing it's possible to see there are
35:06increasing numbers of uh indicators of
35:09quality
35:10um and increasing increasing all those
35:12you know not much that uh not everywhere
35:14but increase of
35:17indicators of Technology Technologies
35:19okay now but a minus of this
35:22there's two mices so one is as I said
35:24before you if you see someone adopting
35:27or not adopting a new technology you
35:28can't necessarily say they're that
35:30that's good for them or not right it's
35:32not clear it's optimal to adopt a a new
35:34technology
35:36so that's one as opposed to like
35:38basically think increasing capability is
35:40good and so that
35:42there's a little more ambiguity there
35:43about the optimality issues
35:46um the second is that uh which is
35:48important practically is that you know
35:49this big standard a nationally
35:51representative firm data sets almost
35:52never have quality or uh technology
35:55adoption and then increasingly having
35:57product Innovation that may be at a
35:58pretty pretty course level
36:00um and so really we could only do this
36:02at uh in relative typically so far in
36:04relatively specific sectors okay now I I
36:08think that's okay I think that that's
36:10the the way we should go is let's build
36:12up it takes we understand very well
36:14what's going on and kind of
36:15incrementally built from there for
36:16deductively built from there
36:18um but it means that you have to sort of
36:19give up on making random things
36:21as a whole
36:23um you know the the you know big plus
36:26though is you don't require the same
36:27sort of theoretical assumptions in order
36:29to constructive measures for instance
36:30that are going to be true in the in the
36:32tfp measures I'll talk about I'll talk
36:34about now
36:36um there's there's some literature using
36:37sort of indirect kind of residual
36:39measures of quality I'm gonna uh that's
36:41active and I I don't if some of it's
36:43good work but I'm not going to talk
36:44about it so much today
36:46okay so let me talk about briefly I'm
36:48going to try and shorten uh or you know
36:50I've got three slides basically on on
36:52issues with cfp estimation
36:54um so the standard uh you know thing to
36:57do if you've got a carb Douglas
37:00production function like we had it was
37:01equation two if you can remember if you
37:03write that in logs I'm going to use
37:04lowercase letters for the to represent
37:06you know log values uh and then I'm just
37:08going to fold these things that you know
37:10the capital labor and the materials into
37:12into a vector Z and then I've got a you
37:14know the coefficients into Vector beta
37:16and then you can think of typically we
37:19think of you know the productivity so
37:20this would be the a together right this
37:22thing in bracket should be the a but it
37:24has two components one of which is an X
37:25anti-shock so the firm wakes up in the
37:27morning and sees you know its
37:29productivity that day and then can make
37:31decisions about variable inputs knowing
37:33that that component of productivity and
37:36then you might call it an expo shock
37:38that would just be a weather shock or
37:39something that that
37:41um that that rises after the the input
37:44decisions have been made
37:45okay so and then you know pfp is simply
37:48just you you like run this regression uh
37:51using some corrections which I'm going
37:53to talk about uh and then tfp if you
37:55find y minus the you know the the fitted
37:57value for for y um okay now there's a
38:01famous problem known since the fifth
38:031950s transmission bias so since the the
38:05variable inputs in this case uh
38:08materials and labor chosen after the
38:10Omega is observed you might get you know
38:12positive correlation there between the
38:13bank and the inputs and that's going to
38:15lead that's going to bias you know OLS
38:17coefficients so less estimates of the of
38:19the betas
38:20the still the dominant strategy for
38:22doing tfb estimations would be proxy
38:24variable strategies the original one was
38:26only in Pecos 1996 which is basically
38:29saying okay well if you wake up in the
38:31morning you're very productive you're
38:32going to expect to be productive in the
38:34future there's a compliment there's a
38:36complementarity between your
38:37productivity and and capital in the
38:39future therefore you're going to invest
38:41more therefore we can infer something
38:43about your your ex-anti productivity
38:45today by the amount that you're
38:47investing in okay so that's essentially
38:49like inverting this investment demand
38:51function and to say okay Omega I don't
38:54really observe it but I know it's going
38:55to be a function of this is the of my
38:57today's Capital stock and investment
39:00okay so then you take that proxy and you
39:02plug it in you put in like a flexible
39:03polynomial for K and I here for Omega
39:06and then that should absorb the
39:08endogenous part in the production
39:09function
39:10um Levinson Petra and ackerberg instead
39:13use materials this dolly Pickens has the
39:15Practical problem in these data sets
39:16have you know many firms have zero like
39:18more than half a firms have zero
39:19investment you have to throw them out
39:20that you know creates a sample selection
39:22so though it's infection I recommend
39:24using materials instead something
39:26similar after or do something similar
39:28again uh I won't go into great details
39:30about you know how that literature is
39:32exactly evolving but the basic idea is
39:33the same
39:34okay so the first one I want to make
39:35about this is that this mod you need a
39:38monotonicity in order to do these
39:39inversion invert these these functions
39:41you need a monetization assumption you
39:43need a monotonic relationship between a
39:46scalar
39:47Omega and investment
39:50okay so so that that uh
39:53and that's going to be that
39:56monotonic relationship is going to be
39:58invalidated by uh you know other sorts
40:00of heterogeneity across the firms so for
40:02instance if firms are heterogeneous and
40:03credit constraints in a way that doesn't
40:05line up with these productivity shots
40:06that they if they have that's going to
40:08invalidate the monitor this decision
40:10okay also if their head face
40:12heterogeneous input Market frictions
40:14okay or even if there's just firm fixed
40:17effects which are different from the
40:19different from the Omega shocks okay all
40:21those things are actually going to
40:22invalidate that the proxy variable
40:23strategy okay and so I think you know in
40:26development we think credit constraints
40:28are endemic right and there's very
40:30little think they're homogeneous across
40:31across all firms and so basically very
40:33quickly you start getting into deep
40:34water you know I guess you'd say hot
40:36water here in in these proximatable
40:38strategies you know in picker they've
40:39been applied in particular settings so I
40:41think I think it's fine but in most
40:42settings especially in developing
40:44countries I think uh there's there's
40:46some there's some problems
40:48okay
40:49second issue uh price biases so
40:54um it's it's this is changing a bit but
40:56typically we don't uh it's rare to
40:58observe physical quantities right so
41:00typically in our local data sets we
41:02observe revenues on the output side
41:03individual expenditures
41:06okay so remember this is in log so the
41:08revenues would be you know log output
41:10log physical output plus log prices and
41:13the expenditures will be log
41:15physical inputs plus plus log input
41:17prices okay now but suppose that in fact
41:21there's idiosocratic variation in output
41:23prices and input costs that might come
41:25for it you know one firm has a friend in
41:27the government who's willing to pay a
41:28lot for their for their products they
41:30have to either might have some Market
41:31power there might be something else you
41:32know they might have specialized access
41:34to some some input okay so there's a
41:36there's you can think of as a often a a
41:39an economy-wide a price index or or it
41:42might be just sector-wide output price
41:44and input price index but then there's
41:45firm specifically Socratic components
41:48okay you plug that into the you know the
41:50production function we had in you know
41:52written down the equation three uh and
41:55this is the regression that people
41:56typically run which is they take
41:58revenues they deflate it by a sector
42:00level output price index they regress
42:03that on on
42:04uh input expenditures deflated by a
42:07sector-wide input price index okay and
42:11then uh so so and then what happens is
42:13that these infocratic components which
42:15are not observed and about out here in
42:17the energy along with the Omega and the
42:19Epsilon okay now uh so that it's uh the
42:22local and Goldberg have a nice review
42:24paper in the annual view of economics
42:27arguing that uh you know that generates
42:30output and input price biases they're
42:32not the first phenosis but it's very
42:33clear
42:34um explanation of that fact so basically
42:36you know if if you it is practically
42:38have a get a high price some you have
42:40some you know uh captive Market that's
42:42willing to pay high pressure your
42:43products and then you produce more and
42:44you use more inputs that's going to give
42:46you a bias correlation between the
42:49plated expenditures and the output price
42:50which is the air term and that's going
42:51to buy as Oscars okay and similar thing
42:54if you you know for some reason you have
42:56a high
42:57input price you do syncretically you
43:00face some some constraints in the supply
43:02Market in that input Supply market then
43:04uh you know that might you might choose
43:06fewer inputs and so you're going to get
43:08that's going to give you an input price
43:10pass
43:10okay
43:12um which can be uh that would look ring
43:14over paper arguments that it can be can
43:16be significant okay now I'd say current
43:18sort of moment we're in in this
43:20productivity literature is people are
43:21very excited about data sets that do
43:23have physical quantities
43:24okay so there's some nice papers this
43:26Foster rosensweigen Foster halterweight
43:28and rosalike 2008 using U.S data in
43:31homogeneous sectors where you do see
43:33reducing physical quantities after that
43:35has a follow-up also also in U.S data so
43:37in principle then you wouldn't have to
43:38do this even deflate just by the you
43:40know a firm level of flavor you can just
43:41put the physical quantity of output on
43:42the left-hand side physical Monument but
43:44on the right hand side that works great
43:46in homogeneous sectors with single
43:49single single product single output
43:51homogeneous sectors that works fine
43:55um when
43:57products can Varys or outputs and or
44:00inputs can vary in quality and or
44:03variety across firms then uh that could
44:06be problematic this is this has been
44:08known people have observed this of Jim
44:10Thai boot at all have a have a uh
44:13earlier early paper 2009 paper uh sort
44:16of pointing this out um David
44:18again they would I can has it has a
44:20paper pointing this out because I'm
44:21going to talk at more length about
44:23um I have a recent paper this is
44:25basically reminded with co-authors with
44:27Nicholas de Rule marceloslava and
44:29Santiago Franco where
44:31um uh I've skipped in details and so you
44:33can when when you get the slides if you
44:35like you can click on this on this link
44:36and then it'll take you to some of the
44:38algebra or you can look at the paper
44:39itself
44:40uh what do we do basically we make some
44:42assumptions about you know how about the
44:44aggregation within the firm we aggregate
44:46materials we assume materials are have
44:48constant elasticity substitution among
44:51each other we assume that products have
44:53constant losses have substitution for
44:54you know multiple products in The Firm
44:56we use existing aggregation or CES
44:58aggregation results okay we simplify
45:00things and we get something that looks
45:02like this where you can have it like a
45:04quantity index of output the SV stands
45:06for stop to ovar2 which made it some of
45:07you may have you've heard of but I want
45:09more detail but anyway basically have a
45:11quantity index of output a quantity
45:13index of input labor Capital we got a
45:16firm effect year effect and then in the
45:18air term here we're going to have the
45:19standard X anti-effects the X the X
45:23anti-shock to expose shock but then also
45:25some unobserved quality in bright this
45:28is input quality output quality input
45:31variety and output right okay and so
45:33that can give rise to to uh to biases so
45:37basically if you know your choices of
45:39Labor or capital or material or material
45:41quantities is correlated with the
45:43quality or input or quality of your
45:45output or the variety of input or
45:46variety of output then you're going to
45:48get you know and you just run this
45:49regression and unless you're going to
45:50get you're going to get biases which we
45:52call Quality variety devices so one
45:54example was you know if producing high
45:56quality I don't know high quality
45:58watches requires more labor you have to
46:00be more careful okay that's going to
46:01induce a correlation between L here and
46:04that 2y here and that's going to give
46:05you a negative rise from the in the
46:07coefficient on labor in your in your
46:10regression okay so it's just basically
46:12at this point just I mean we do have in
46:14this paper we have a method that doesn't
46:16try to solve this problem the resale
46:18album whether that's going to whether
46:20that's going to be accepted and how
46:21influential that's going to be so my
46:22message today is uh you have to be
46:24careful you have to watch out these are
46:25these are problems that I think existing
46:27methods
46:28um don't don't deal with well and I
46:30think it's a special Salient in
46:31developing countries in the same that
46:33most sectors are most sectors we look at
46:34certainly the manufacturing sector but
46:36even the more you look at sort of a
46:37particular agricultural sector often the
46:39products are not that homogeneous right
46:41so they're and and also typically firms
46:43are not single product single input
46:46they're multiple product multiple
46:48okay and so I think these problems are
46:50are are reasonably uh reasonably severe
46:52and need to be need to be worried about
46:55um
46:56okay let me do this one last slide and
46:58then I'll stop again maybe for for a
47:00couple of questions
47:01um
47:02okay so what there are some promising
47:04things that I think are happening in the
47:05literature uh where that the growing
47:08number of Studies have pain the cost of
47:11going out and getting better data where
47:13you get some technology adoption
47:14inequality I'm going to talk about some
47:16of those
47:17um I won't talk a great at Great length
47:19about the measurement of management
47:20practice but I think that's a very
47:21promising uh you know development that
47:23that is led by Nick blumens lots of
47:28people have been going out and actually
47:29listening what management practices are
47:30being used which you can think of that's
47:32gathering information on on techniques
47:34on what technologies firms are using the
47:37nice thing about management practices is
47:38that they apply across different sectors
47:39many times studies and Technology have
47:42to you know a technology is only useful
47:43in one particular sector whereas
47:45management practices arguably it can be
47:46applied across a functions that's part
47:48of the reason why there's been so like
47:49so much excited about this about this
47:51literature David uh mentioned yesterday
47:53in this in his comments that the World
47:54Bank has also cutting a bunch of
47:55Technology surveys that's very promising
47:57we're getting a lot more information
47:58than we had before and so I think that's
48:00likely to have a good effect we make it
48:03possible to do research on on these
48:04topics in a way that wasn't available
48:06before so in terms of how do I think the
48:09what I think the way forward is I said
48:10before I think we want you know we want
48:12direct measures of upgrading that seems
48:13very promising to me uh and I still
48:15think we you know look for incredibly
48:16exotic variation which might be you know
48:18the question came up about you know
48:19political factors and these things in
48:21these capital gains that are external
48:22firms look for shocks for those Capital
48:24Gauss and then look at how firms are
48:26reacting in terms of their technology
48:28choices are in their quality choices in
48:30their products
48:32um okay uh yeah one last point in the
48:35just that as I've already said though
48:36that you know upgrading on these
48:37dimensions of quality technology and
48:39product Innovation aren't necessarily
48:40optimal and so we have to be careful
48:41about how we interpret you know what's
48:43happening and we can't just say you know
48:44should have been doing this and they're
48:46not we have to be thinking about you
48:47know what is it really the case this is
48:49something different
48:50okay good let me pause for a couple
48:52questions I see that has a question that
48:55may be uh okay I'll be going to launcher
48:57the Q a else
48:58hi I'm here thanks Eric I just wanted to
49:01follow up on your point about the
49:03agricultural sector you mentioned that
49:04uh even there we can see some of these
49:07issues and I was wondering like in that
49:09scenario where you don't need to use
49:11prices you could use like physical
49:12output and uh in many cases like there
49:15are attempts to measure quality like of
49:17cocoa beans or coffee or many other
49:19crops to what extent do you think
49:22um agriculture could be like maybe the
49:25the applications of these problems could
49:27be maybe not that bad if you could just
49:29comment on that that'll be great thanks
49:31yeah no no I I agree I agree and even
49:34quality differences per se are not
49:36necessarily a problem right so you can
49:37have come on if markets are sort of
49:39commodity-like in the sense that you
49:41know once you you know you said so I I
49:44have coffee I've worked in other
49:45projects I'm thinking about coffee but
49:46certain coffee you know coffee beans
49:48thumb eyes which is sort of an
49:49observable thing and you say okay you
49:51know like you there's kind of a under on
49:54certain segments of the coffee Market
49:55there's basically a commodity like
49:57coffee of this size or coffee of that
49:58size and so for that I think through the
50:00standard productivity estimate
50:02estimation can can work so that's partly
50:04why you know that the productivity
50:06spaces are runs we grill kids and they
50:07were originally we're doing it in
50:08agriculture and I think partly because
50:09it has that that attractive characters
50:11but now but even in coffee there's like
50:13a segment where now you're having
50:14specialty coffees and you're having
50:16direct relationships and you know the
50:17Joe coffee on the corner wants to say is
50:19from this estate in Ethiopia is where
50:21these beans are from and so then you get
50:22all this differentiation and sort of uh
50:24you know and then you know bilateral
50:26bargaining power between this between
50:27the suppliers and the intermediaries
50:29much more kind it starts to look much
50:31more like a standard you know
50:32differentiated product sector and so so
50:35just to say we can't you know naively
50:37say well it's agricultural therefore
50:41okay thanks yeah
50:45a couple of questions and then the chat
50:47or in the Q a box here one from Muhammad
50:50says how does incentive to upgrade a
50:52market power fit into the model uh
50:56going back maybe eight minutes he says
50:58in a sense there's less incentive to
51:00upgrade if a firm in the developing
51:01country is the dominant firm in the
51:03sector
51:05uh if you could speak to that
51:09uh sure yeah I mean there's the so one
51:13of the what I'll talk about I'll come
51:14back in a few slides I'm going to come
51:15back to like competition is often seen
51:17as a driver of upgrading right that and
51:20I'll give you a give a nice example from
51:22an Indian rail firm that was basically
51:23producing rails for the Indian railroads
51:25kind of a quasi public uh firm and then
51:27the government uh suggested that uh
51:30they're gonna allow entry right that the
51:32proximity wasn't high if we're in a
51:33library and productivity showed up right
51:35so that's That's a classic example Jim
51:37Schmitz has examples in the in the steel
51:38sector in the in in the United States
51:40also so it's clear the case that like
51:43Monopoly power can be a uh a deterrent
51:47or can lead to lower you know to less
51:49upgrading on the other hand I think uh
51:52it's not there are other factors like
51:54it's a little complicated competition
51:56like if if there's lots of competition
51:57and there's you know so there's the
52:00returns to innovating are maybe not that
52:03high because you know the gains are
52:04going to be going to be competed away
52:05almost immediately then you may get less
52:07Innovation and less and less and less
52:09upgrading and so there are a couple of
52:10examples I'll talk briefly if I have
52:12time I'll talk briefly about another
52:13example also uh machiavello and uh and I
52:17meet more Daria have a nice paper about
52:18that about how when there's more
52:19competition between Ron and coffee
52:22Growers and sellers then you get fewer
52:24relational contracts being developed and
52:26that tends to reduce productivity and
52:28reduce operating so I think competition
52:30is a complicated one there's lots of you
52:32know conflicting ideas out there in the
52:33literature
52:34um so it certainly can be the case I
52:36guess I'm kind of a skeptic that
52:37competition per se has much of a
52:40positive effect in in most Industries
52:41but it certainly can in in some
52:43Industries
52:48uh okay so okay we should move on
52:51because I just have another one you were
52:52gonna
52:52share her uh it just it's uh whatever
52:56you need uh there's I mean questions
52:59yeah okay maybe I am seeing the time
53:01maybe I should move on a little bit no
53:03and then uh so apologies to those who
53:04asked questions and maybe we can try and
53:06they can be saved I can try and uh
53:07answer them maybe afterwards okay so now
53:10I'm going to move on to that was sort of
53:12a long thing on what how do we think
53:14about upgrading how do we measure it now
53:15let's think about you know evidence on
53:17on drivers
53:19um so one and here I was trying to also
53:21run in this review paper I was trying to
53:23draw a kind of General conclusions uh
53:25that seems to me like you know lots of
53:27different papers uh you know it seems to
53:29be sort of robust evidence that
53:30particular thing seem to be driving
53:31driving upgrading on one of the mentions
53:33that we've we've highlighted uh and one
53:36uh Dimension that I think there's you
53:38know it's pretty solid I think I feel
53:39pretty confident that that that that we
53:41know this is that selling to richer
53:44consumers which might be selling to them
53:45directly through exporting it could be
53:47selling indirectly through sort of
53:49selling into value chains that
53:51eventually end up selling temperature
53:52consumers that selling device consumers
53:55uh in that way is associated with
53:57upgrade
53:59um oops oh no okay here I just
54:01luckily I have these these these uh
54:03these circles okay
54:04um so this is this is not a brand new
54:07idea so they're they're early papers
54:09actually the tfp papers about you know
54:11just exporting more so this paper
54:14clearing's last entire group was
54:15Colombia Mexico and Morocco and I'm very
54:19mixed results they didn't find strong
54:20evidence they're like tfp went up when
54:22when firm started sorry I started
54:24exporting here there were some this is
54:26also my original my dissertation basic
54:29page and some other papers have found
54:30pretty strong I think you know robust
54:32evidence that like selling to richer uh
54:34countries than you know prices that that
54:36firms charge go up probably pay for it
54:39for inputs go up uh and then you know
54:41you can make some inferences about about
54:43quality there but the there's been more
54:45recent papers that I think are
54:46particularly convincing or very micro
54:48pretty convincing and have raised some
54:50important new questions and so one of
54:52those I'm going to talk about
54:53um uh is and I I I sort of had been
54:56talking about this neighbor before I was
54:57on a on a program that was organized by
55:00you know co-organized by by David Eck
55:02and I I'm really a fan of this paper
55:04um on Egyptian rugs so again I'm a
55:07kettle Adam oseman so essentially what
55:09they did is there's in the Egyptian rug
55:11sector which are typically very small
55:12firms that they're sort of a guy in a
55:14wound selling selling Rogues they might
55:15have it might have played two or two not
55:17very large but they randomized initial
55:20export orders so working with an NGO
55:22they they and an intermediary in Egypt
55:25they were actually randomized in just an
55:27initial order first first export order
55:29okay selling to uh risk country markets
55:33the European and I guess some also the
55:36US market in a little Vegas uh and then
55:39they track detailed quality indicators
55:40they kept trackable conversations
55:41between the buyer between the
55:43intermediary and the producers uh and
55:46then they've got a nice thing where they
55:47they brought producers into a laboratory
55:48foreign
55:52okay so what were the what what were the
55:55results there so here this is okay we're
55:57just going to look at these aren't just
55:58ITT is just the basic like what an
56:00assignment to having one of these
56:01initial export contracts what did it
56:03what effect did it have on the these are
56:05different uh dimensions of quality of
56:07drugs okay I'm going to read the you
56:10know as much as I know about Egyptian
56:12rucks very quickly but basically how the
56:13corners are done how much waviness there
56:15is the weight the touch the packiness
56:17you know more practice is better warp
56:19thread tightness tighter than you have
56:20the threads that's also better firmness
56:22design accuracy etc those are all
56:23dimensions of
56:25you know how good were the inputs which
56:26is like the I guess the weft thread you
56:28know how high quality was where that the
56:30inputs that were that were being used
56:31all those things quite clearly either if
56:33you do intentionally the sort of the
56:34reduced form or you do the the IV using
56:38um uh using a finance treatment for you
56:40know for taking out the initial export
56:42off the initial export offer uh very
56:45very strongly robustly correlated with
56:47quality upgrade
56:48okay so the export offer they produce
56:50higher quality
56:52um also profits clearly went up measured
56:54in a bunch of different ways that's not
56:56that's not a surprise I guess I'll come
56:58in a nice big order and a network
57:01connection to to buy our connected to
57:04International markets okay profits also
57:07clearly go up there's different ways you
57:08can just ask firms about their profits
57:10you ask them about their revenues and
57:11cost calculate profits you ask them
57:13about you know physical quantities and
57:15then calculate profits on those bases or
57:16you could ask them a hypothetical
57:17question if they use this in the input
57:19or the profit fee everything goes in in
57:21a similar Direction okay that's not
57:22surprised I think it's consists of what
57:23we would think
57:25um you know professor
57:26and and this is related to what I think
57:28what I think about productivity before
57:31um
57:32uh
57:33tfp goes down okay so you could do this
57:37as you know basically physical so it's
57:39essentially what's happening is that
57:40these guys are taking they're taking
57:42longer to sew a square meter rug so if
57:45you if you measure of tfb is quantity
57:46tfp
57:48right in terms of uh you know square
57:49meters are rug per day or per hour uh it
57:53looks like it's going down even though
57:54in all every other dimension looks like
57:55they're they're uh looks like uh they're
57:58doing better and so that's just a that's
58:00sort of a caveat about uh
58:02about about tfp measures to reinforce
58:04what I said before I think more
58:06importantly when they went to these
58:08identical specification domestic drugs
58:10in under laboratory conditions
58:13um they they found that these guys who
58:16got the an export offer uh did much
58:20better on all different quality
58:21Dimensions this is you know uh a value
58:24by Master artisan and a local local
58:25Professor but basically they agreed that
58:27these guys were producing uh better rugs
58:30and using no more time that's basically
58:32what it's saying here they're not using
58:33they're not significantly more time and
58:35they're doing more federalism so that
58:37looks like these guys gain capabilities
58:40that makes sense so it looks like the
58:42quality upgrade is not just the canopy
58:44these are selling conversion consumers
58:45you start producing a higher quality
58:46versions of the product you already know
58:47how to produce you're actually learning
58:49something in the process and so I think
58:51that's I think that's a very nice very
58:52nice finding my previous work I sort of
58:54modeled it the formal way kind of for
58:56convenience supposed to say firms
58:58already know how to produce a high
58:59quality low quality version and it's
59:00just the mix that gets affected by
59:03by you know trade costs or by by
59:05exchange rate movements where it's here
59:06they're saying no no actually the it's
59:08not just that you're selling to a
59:10different Market it's that you're
59:11actually you know gaining some
59:12capability I think it's an interesting
59:14question which maybe is something to be
59:15explored just you know do the game
59:17capability uh for all product techniques
59:20across the board or is it just you're
59:22learning how to produce the kinds of
59:24things that rich country consumers are
59:25going to want right or are there rugs
59:27you know that the domestic Market also
59:29better so that's that's maybe something
59:31that is kind of important question
59:33okay
59:35um
59:35let me pause there but I you felt going
59:38to read that paper if you haven't yeah
59:40let me keep going just to get through
59:42this is one little section
59:44um still on the idea that who yourself
59:46matters the device or the demand side
59:48matters there's never another very nice
59:50paper this is Chris hansman you know
59:53shorts Matthew tishu and uh John Marco
59:56Leon
59:57um here this is Peruvian fish meal
1:00:00factories okay or they're basically
1:00:02taking fish grinding it up which then
1:00:04becomes part of the FED Fanta livestock
1:00:08um but essentially what they're doing
1:00:09they're arguing that there was
1:00:11International shots to different
1:00:12qualities of fishing right and the
1:00:14quality fishing basically refers to the
1:00:16protein content of the fish meal the
1:00:18other producers of uh high protein fish
1:00:21meal are Denmark based on the chilling
1:00:23one of the biggest ones who have
1:00:25sometimes binding quotas so when the
1:00:26quotas were binding on Denmark I saw the
1:00:28Chile then there's more demand for high
1:00:30protein fish meal from Peru
1:00:33what they found is that These Guys these
1:00:35Brewing personal firms when there was a
1:00:38high demand for high quality went out
1:00:39and bought boats also I haven't told you
1:00:42much about the the sort of how it is
1:00:43organized but you know there's two
1:00:45there's two certain modes one is that
1:00:46there's a there's a factory and they buy
1:00:48fish from independent boats the other is
1:00:51sometimes factories they're buying their
1:00:52own boats and so it turns out when
1:00:54there's you know
1:00:55after they're trying to produce more
1:00:57high quality they buy more votes why is
1:00:59that it's basically because they can
1:01:01control what the boats do in that case
1:01:03they can uh the the boats don't go as
1:01:06far right they carry more ice and so the
1:01:09the fish that they bring into Port are
1:01:12of higher quality than the guy than the
1:01:14independent Builder are just going far
1:01:16away wherever they can find the most
1:01:17fish okay but that that quality the
1:01:19protein con is not observable right at
1:01:21that transaction and so the way that
1:01:23that uh the firms can ensure that
1:01:25they're getting high quality fish is
1:01:26they I can go buy the boats and then
1:01:28they can control it they can control
1:01:29when they go so that's um
1:01:32you know let's suggests that the
1:01:34supplied quality is an important
1:01:36constraint that's related to questions
1:01:37someone had earlier about you know what
1:01:38about input quality how's that happening
1:01:39so that but I think it's nice because
1:01:41these decisions about like the the
1:01:43extent of vertical integration within
1:01:45the firm is going to be sensitive to you
1:01:47know the demand conditions
1:01:49and and generally this is opening up you
1:01:52know later in the course you're going to
1:01:53have Rocco machiavello and and uh Chris
1:01:55Woodruff and David McKenzie and others
1:01:57talking about you know Contracting
1:01:59problems uh and I think there's a long
1:02:01history of work in development on
1:02:03Contracting problems and I think that
1:02:04that's a promising area to bring that
1:02:06long history work on like a Contracting
1:02:08problem works together there's questions
1:02:09of how you know firm upgrading or how
1:02:11how firms are uh you know what's what's
1:02:13driving from from productivity
1:02:17um and I think that the you know the
1:02:19information is argue information about
1:02:20quality is arguably particularly
1:02:22asymmetric in development and so hence
1:02:24you know in addition to the fact that
1:02:26development has its own history of
1:02:27analyzes problems I think that this is
1:02:29this is a fertile fertile ground for for
1:02:31working on you know you have this
1:02:33contract problems and input markets
1:02:34matter for for for for performance
1:02:38okay let me pause there for a second if
1:02:40there's I don't know if there's a
1:02:41question or if uh Caleb you want to try
1:02:43and pick out one of the
1:02:45one of the uh Q A's sure one that has a
1:02:49few upvotes is uh is the know-how
1:02:52increase or the increased motivation and
1:02:55uh improve future expectations how can
1:02:59we disentangle these
1:03:03I know can you just entangle like
1:03:04motivation from know-how
1:03:08is to know how to increase or the
1:03:10increased motivation
1:03:12is one of these improved future
1:03:15expectations
1:03:16how do we disentangle these and another
1:03:19question is regarding
1:03:21um
1:03:22whether upgrading is optimal if you
1:03:25imply that just the decision might be
1:03:27independent of profit
1:03:31uh okay well so the second one I think I
1:03:34can deal better with so typically I'm
1:03:36thinking of your profit that is the
1:03:38objective of firms uh that's what firms
1:03:40are trying to do I guess I'll come back
1:03:41at the very end about you know that you
1:03:43know the behavioral firms question but I
1:03:45think the first order approximation
1:03:46saying firms are profit maximizing I
1:03:48think is reasonable and I think uh you
1:03:50know so I'm comfortable with
1:03:53um
1:03:54uh with using that assuming that that's
1:03:56the objective
1:03:58um and that and then therefore like
1:03:59whether you know upgrading is awesome or
1:04:01not it depends on whether it raises
1:04:02profits or not so it's very comfortable
1:04:04with that there was a question about
1:04:05yeah that's kind of it was a good
1:04:07question about
1:04:09um like let's say if I go all the way
1:04:11back so I'm going to go all the way back
1:04:12to my framework I see
1:04:14um
1:04:16you know let's say
1:04:19you have a given worker that you're
1:04:20paying given wage for maybe that's goes
1:04:23in the M but somehow you're this
1:04:25entrepreneur is good at getting that
1:04:27worker to supply effort
1:04:28right they're nice to them or Jing and
1:04:31and Chingy Wang have a nice paper that's
1:04:33coming out of qge about uh you know the
1:04:35Chinese Auto firm changed its basically
1:04:38allowed workers to give feedback on
1:04:40their managers that managed from manager
1:04:42salaries uh and that raised productivity
1:04:45of the of the teams right so that's in a
1:04:47way that's kind of like that's something
1:04:48about you know the capability of the
1:04:50firm or know how the firm that's able to
1:04:52uh induce workers to to
1:04:55to supply more effort or not to quit or
1:04:57somehow contribute to productivity so
1:04:59it's kind of basically it's a little bit
1:05:00of a gray area I agree you could go you
1:05:02could go either way uh in terms of
1:05:04interpreting that as part of capability
1:05:05or interpreting that as just maybe
1:05:06effort in the case where effort is
1:05:08contractable then you'd say well effort
1:05:10is just you know an input that you're
1:05:11paying for
1:05:12foreign
1:05:33about the following from the example of
1:05:36uh Egypt can preferential access to
1:05:40foreign markets through the regional
1:05:42trade agreements act as a catalyst for
1:05:45Quality upgrading
1:05:48uh I I do think so I think the basic
1:05:51answers to that is yes yeah so I think
1:05:53uh you know some of this work
1:05:55this early work yeah it's basically
1:05:57using
1:05:58uh
1:06:00you know uh in my case it was using you
1:06:02know exchange rates moved and that gave
1:06:04Mexican producers more access to the US
1:06:05market okay but you know uh trade
1:06:09liberalization or you know reduction of
1:06:10terrorism because Market would have
1:06:11would have a similar effect no so I
1:06:13think that is a great benefit to ferns
1:06:15to to having the fact that you know that
1:06:17to excellent markets in rich countries
1:06:19especially it's likely to have so it's
1:06:21likely to lead them to upgrade their
1:06:23quality but also then following from the
1:06:25Egyptian Rush paper that's likely to
1:06:26generate learning at the scene so I'm
1:06:28very Pro you know getting for allowing
1:06:30firms to export more there's obviously
1:06:32other issues that they might then face
1:06:34more competition to the domestic market
1:06:35and so it's you know there are there are
1:06:37complicated distributional issues uh in
1:06:39trade literalization but in terms of
1:06:40just this sense of like what is for
1:06:42upgrading I think I think yes
1:06:46okay
1:06:47uh
1:06:49let me okay let me let me let me move on
1:06:51another very nice paper which I'm
1:06:53talking about apart because I think it's
1:06:55a new data Frontier so it seems like a
1:06:56very promising area and this is not now
1:06:58not about exporting directly it's about
1:07:01exporting indirectly so this is a study
1:07:03by Alfonso
1:07:05Isabella and Jose Vazquez
1:07:08um and qg
1:07:11so there what they have in Costa Rica
1:07:13you have information from the vat system
1:07:15so you can see firm to firm transactions
1:07:18in particular they can see who which
1:07:20domestic suppliers are selling to
1:07:21multitask in Costa Rica that's a picture
1:07:24of the Intel plant in Costa Rico that's
1:07:26been there for me I think it arrived in
1:07:272007. so and then what we're going to do
1:07:31is is compare suppliers to
1:07:32multinationals to it's basically a
1:07:34matching design it's not super fancy
1:07:36there but comparing to A supplier step
1:07:38multinational suppliers to other other
1:07:40types of firms
1:07:42um okay so here you can see
1:07:45um I'm just picking up one graph of many
1:07:47but basically year zero is the year that
1:07:49a given supplier starts that starts
1:07:52starts applying to a multinational those
1:07:54are going to be the green circles and
1:07:55then they come they can compare that to
1:07:57when a domestic flyer starts selling to
1:07:58a to a non-multi-national um
1:08:01a large firm okay and what you can see
1:08:03is uh that initially when what and this
1:08:07and then the nice thing about this LGBT
1:08:09is they can track sales to other firms
1:08:12that's not the multinational say what's
1:08:13the effect of selling the multinational
1:08:15on your salesman and what they find is
1:08:17not surprisingly in the short term okay
1:08:18there's capacity constraints firm sales
1:08:21to other firms go down in both cases but
1:08:23then
1:08:25excuse me in the slightly longer term
1:08:27the the ones that the for instance start
1:08:30selling to Mountain Nationals have more
1:08:31success selling to others even in the
1:08:34domestic Market okay and they do a lot
1:08:36of qualitative work basically showing
1:08:37that it seems to be that those guys
1:08:39partly they're learning something from
1:08:40selling to the multinational and partly
1:08:42they're getting reputation so now they
1:08:43can say I sell the Intel they use that
1:08:45sort of as part of their branding and
1:08:46they sell more sell more domestic
1:08:49um so and and then the other buyers of
1:08:51the new of the of these new
1:08:53multinational suppliers they tend to be
1:08:54larger they have higher export import
1:08:56chairs they have longer relations
1:08:57requires it seems like they're able to
1:08:58sell those for better firms so in a way
1:09:01it's sort of it's sort of upgrading
1:09:02that's driven by the fact that they were
1:09:04able to sell this at the multinational
1:09:05elements okay and there so it appears
1:09:07like there's a warning going on going on
1:09:09there
1:09:10okay
1:09:12um I'm I'm a little bit conscious at the
1:09:14time with about 20 minutes let me go
1:09:15I'll go fast this is this is just a
1:09:17graph that I've mentioned thing I showed
1:09:19I mentioned about this Indian firm
1:09:21producing you know kind of basically had
1:09:22a monopoly on producing rails the
1:09:24government threatened to introduce
1:09:26competition they clean up their act very
1:09:28important okay so essentially this was
1:09:29like productive right in the window when
1:09:31the when the when the entry was
1:09:32threatened relative to that's a rails
1:09:34relative to other kinds of things that
1:09:36this this big firm was was producing
1:09:39um you know the competition always has
1:09:41to have an argument for some reason
1:09:42firms were not optimizing before why did
1:09:44they start optimizing so that's often a
1:09:45little bit a little bit unclear it's not
1:09:47impossible that could happen but it's
1:09:48something sometimes unclear
1:09:51um here's this is the slide on kind of
1:09:52the skeptic my my kind of skeptical side
1:09:54about competition one is empirically
1:09:56it's often a difficult to distinguish
1:09:59you know if you see an aggregate product
1:10:00aggregate average productivity going
1:10:02going up in the sector is that because
1:10:04firms themselves will become more
1:10:06productive or is because the low least
1:10:09productive firms were killed off okay
1:10:11it's often empirically hard to
1:10:12distinguish between those two Dynamics
1:10:15um Holmes and Schmitz have to have a
1:10:16that's a review paper again again or
1:10:18real economics I think uh that making
1:10:21that point
1:10:22um there's another issue you have so
1:10:24you're going to have speaking on on the
1:10:25last lecture eight in this series has a
1:10:27nice paper this is an historical book
1:10:29setting in France but basically
1:10:30competition May reduce scale effects and
1:10:32so there may be less incentive for firms
1:10:34to invest those eye Lambda and I J and I
1:10:38K investing and learning they may be
1:10:40less willing to do that if there's a
1:10:42smaller Market at the end of the day for
1:10:43them to to profit from and then this is
1:10:46what I was mentioning this is
1:10:50um about these Rwandan coffee uh Mills
1:10:54so they have a nice use of geographic
1:10:56geography as an instrument for like how
1:10:58many
1:10:59um uh how much competition there is
1:11:02basically four farmers who are growing
1:11:03coffee are there several Mills like the
1:11:06given Pharma could sell to or is there
1:11:08just one okay and that's based on sort
1:11:10of the you know the record geography in
1:11:12Rwanda but basically they're finding
1:11:13again as quickly they're finding there's
1:11:15more relational Contracting between like
1:11:18relational country I mean they're like
1:11:19favorite trading essentially favorite
1:11:22trading where you know the the mill
1:11:24gives gives inputs to the farmers or the
1:11:26mill paid you know uh
1:11:29uh the the farmer allows for a late
1:11:32payment or you know they they provide
1:11:35help uh or loans um the military there's
1:11:39more of that happening when there's less
1:11:40competition relative I think a
1:11:42provocative uh provocative so assessment
1:11:46today it's true that there could be
1:11:47competition couldn't matter but but
1:11:48sometimes he goes through I think uh
1:11:50very very quick to say oh yeah
1:11:52competition is going to going to be
1:11:53great for upgrading and that's not
1:11:54allowed
1:11:55okay uh let me talk a little bit about
1:11:58uh input side drivers
1:12:02um
1:12:03okay so so far I've been talking about
1:12:05like demand total effects like demand
1:12:07side matters could it be that like
1:12:09greater supplies they have higher
1:12:11quality inputs affects
1:12:14upgrade
1:12:16um there's some work on this so uh a lot
1:12:19of it is in the trade literature so for
1:12:21instance there are a couple nice papers
1:12:22showing that when uh import tariffs into
1:12:25China go down uh export prices go up for
1:12:29those firms that are affected right and
1:12:30the idea there is that you know you're
1:12:33they have the firms have to the tariffs
1:12:35go down so they're paying less on inputs
1:12:37especially from rich countries
1:12:39the us or others countries and then they
1:12:42use that to upgrade the quality of their
1:12:43products and so then export prices rise
1:12:45as a consequence okay so that's uh to be
1:12:49a very sort of intuitive
1:12:50um story it could also be shown in terms
1:12:52of input varieties not just not in the
1:12:54quality but just having a greater supply
1:12:56of different types of inputs could allow
1:12:58you to produce a greater support you
1:13:00know more different types of outcomes
1:13:01and so this is Penny Goldberg I'm a
1:13:04condo wall uh you know Patrick and
1:13:08um I have a nice paper on that uh that's
1:13:11been reinforced uh by in recent work by
1:13:13Maria Boston
1:13:14[Music]
1:13:16okay
1:13:17uh yeah I was very quickly summarize I'm
1:13:21gonna I want to talk a little bit more
1:13:22about our soccer ball project uh coming
1:13:25up in the next in the know-how section
1:13:26we have sort of a new talking about
1:13:28paper yet it's just just a project
1:13:29basically what we did so in in talking
1:13:31about producers uh Seattle Club Pakistan
1:13:33has a has a big cluster of soccer ball
1:13:35producers we randomly randomly
1:13:37subsidized high quality versions or high
1:13:40quality imported versions of the main
1:13:41material which is artificial leather
1:13:43called brexit and then we're looking at
1:13:45how firms responded on other dimensions
1:13:47okay and essentially this is uh so we're
1:13:50we're gonna you know instrument uh
1:13:54whether you're you're using one of these
1:13:55high quality rexels with whether you've
1:13:57got a subsidy uh for for that for that
1:13:59rexin and interacting potentially with
1:14:01firm size because there's only those are
1:14:02smaller terms that are our subsidy is
1:14:04only large enough to make a difference
1:14:05for the smaller firms but it turns out
1:14:08essentially that perms they they improve
1:14:10the quality not only of their accent but
1:14:12also all the other dimensions
1:14:14so
1:14:16um if you do that as an index but you
1:14:17the latex is the kind of glue they use
1:14:19to glue the layers of of leather to sort
1:14:22of layers of cotton or polyester
1:14:24um the number of layers that'll kind of
1:14:26do more of those typically higher
1:14:27quality associated with having more
1:14:28layers it gives more heft to the ball
1:14:31oh there's the positive coefficients on
1:14:33the what kind of bladder you use and
1:14:35then in terms of cost the sort of the
1:14:37cost of the glue cost of printing all
1:14:40these things are uh you know you have
1:14:42positive positive coefficients if you
1:14:44put them in an index
1:14:45it looks like you know firms have in
1:14:47upgrade quality on other dimensions
1:14:50okay so that's preliminary evidence I
1:14:52think that this you know input input
1:14:54quality upgrading can push upgrading up
1:14:56push upgrading of other inputs and then
1:15:00okay one brief thing I guess again in
1:15:02the spirit of trying to highlight areas
1:15:04where I think there's there's maybe more
1:15:05work to be done is in my review I was
1:15:08looking for
1:15:10uh you know impacts of Greater capital
1:15:13greater Capital availability on like
1:15:15upgrading outcomes
1:15:17and and I I just didn't find very much
1:15:19right so I think a couple of the best
1:15:21papers so uh Martin rodenberg has a nice
1:15:24paper using
1:15:26um basically a change you know that any
1:15:28government would like force Banks to
1:15:30channel a certain amount of loans to a
1:15:31certain size of the firms and then the
1:15:33limits changed over time and so we can
1:15:35that's like a capital shock in a way to
1:15:37different sizes of firms over time he
1:15:39looks at tfpq which you can do in the
1:15:41Indian data but he basically doesn't
1:15:42from you know little little effect on on
1:15:44tfpq of that of those Capital shocks uh
1:15:47naturally about incrementary how about
1:15:48uh paper form coming in chronometrica
1:15:52where they had a loosening of
1:15:53restriction on foreign investment that
1:15:55in a way you can interpret that as a as
1:15:57a Capital stock also they're doing other
1:15:59stuff they're mainly talking about Miss
1:16:00allocation which is coming later in the
1:16:01course but they also look at like within
1:16:03firm changes in in tpq and again they
1:16:06don't buy very much and certainly uh
1:16:08Jamie has a paper van Harrison uh
1:16:11looking at there was a VP reduction on
1:16:13capital goods in China which are really
1:16:15made from investment products uh less
1:16:17expensive but they don't find uh
1:16:20okay so now
1:16:22uh maybe that's because these issues
1:16:25that I raised about tfp measures right
1:16:27maybe that we're just not measuring tfp
1:16:29very well that's one possibility but
1:16:31maybe uh maybe there really isn't really
1:16:33much of an effect of capital supply per
1:16:35se on on uh on upgrading and so that's
1:16:37that's I think a good area to try to try
1:16:40to think about
1:16:42okay how am I doing on time I've got 12
1:16:44minutes
1:16:45um let me maybe before the last know-how
1:16:48I'll take maybe one more question if
1:16:50someone has Caleb do you have a do you
1:16:51have one for the Q a that's come up
1:16:57no I think it's all right for now no
1:16:59I'll keep going I'll keep going okay
1:17:01okay good
1:17:02um all right so now drivers of know-how
1:17:04right so
1:17:05I I didn't say but I can organize these
1:17:07output side input side drivers you know
1:17:09in part like based on what happened
1:17:11based on the framework that that uh that
1:17:13I put put in the beginning and so
1:17:15dropped you know of the things that
1:17:16affect the lambdas or maybe the capital
1:17:18j or the cap
1:17:21okay so first you know you can imagine
1:17:24or we have cases where we think it seems
1:17:26like the output doing conditions and
1:17:28input Supply conditions are favorable
1:17:29would favor upgrading like they're not
1:17:31they don't seem to be a big constraint
1:17:32on firms and yet firms are still failing
1:17:34to to upgrade you're still feeling to
1:17:36adopt what seems to be a better
1:17:38technology or you know produce that
1:17:40seems to be a better product
1:17:42um okay so then the framework would lead
1:17:44us to suggest that the that uh the
1:17:46reason for that is a lack of know-how
1:17:48right so it's not the not the banners
1:17:49Supply conditions but it's just what the
1:17:51firm knows how to do that's the that's
1:17:53that's the constraint
1:17:55um there's a bunch of factors that might
1:17:56affect that level of know-how
1:17:59um I'm going to talk about right now
1:18:01about agency issues there's it may be
1:18:03you know entrepreneurial ability uh
1:18:06related to you know how how controlled
1:18:08or passed within families for instance
1:18:09that could be an issue uh we'll see how
1:18:11much time I have to get through these
1:18:12things
1:18:13um you know learning from others is
1:18:15might be an important mechanism for for
1:18:17learning but it's hard basically there's
1:18:19learning stuff especially learning
1:18:20facilities it's very hard to identify so
1:18:22they're gonna there are a few very good
1:18:23uh papers on that but not for any and
1:18:26then I'll by time I'll talk a little bit
1:18:28about behavioral issues okay I'd like to
1:18:30talk about the the soccer balls paper
1:18:32it's fun for me to talk about
1:18:34um it's Amazon uh and also David Atkin
1:18:36and Ahmed uh condo well and awesome
1:18:39challenge we're in lapor um but also I
1:18:42think it does illustrate a few of these
1:18:43of these of these issues part of the
1:18:46reason I like that I I like it is
1:18:48because we came up with it so it's uh
1:18:50soccer balls we came up with a
1:18:51technology which really does look like
1:18:53one of these no you know vertical
1:18:54Technologies like basically everyone
1:18:55should be using it right it doesn't have
1:18:57a horizontal cortisol Dimension and yes
1:18:59and yet still many firms didn't didn't
1:19:02adopt it and so it's it's part of I
1:19:03think it raises a lot of interesting
1:19:04questions about about why they didn't
1:19:07uh we were expecting so I'm going to
1:19:09tell you about what the technology was
1:19:10but uh we were expecting it to spread
1:19:12like wildfire right we were you know
1:19:13plan on setting spillovers not on
1:19:15whether firms adopted in the end but
1:19:17what we discovered someone unexpectedly
1:19:18that actually even firms that we given
1:19:21to gave it to many of them didn't adopt
1:19:22it so uh you know there's a question why
1:19:24not and our argument is that it was a
1:19:26misalignment of the Sabbath Within
1:19:30so you can see here that the soccer
1:19:32balls have hexagons and pentagons
1:19:34this is a cutter he's got these are the
1:19:36sheets of Rex in this artificial leather
1:19:38this is a press he moves a die I'm going
1:19:41to show you a picture or anything with a
1:19:43metal here he's cutting hexagons he he
1:19:45moves he translate he moves it by hand
1:19:47and then brings the press down on top of
1:19:49it okay so now they're pretty good at
1:19:51cutting and the Rex is by far the most
1:19:52expensive thing it's like half of cost
1:19:54wants to be treated uh and and uh
1:19:58the hexons are pretty good at cutting
1:19:59hexagons in a way that minimizes waste
1:20:01in part because hexagons testimony like
1:20:03you can tile the floor with hexagons uh
1:20:05pentagons don't template it's a harder
1:20:07problem
1:20:08necessarily there's going to be more
1:20:09waste I don't know if you can see it
1:20:11here it might be a little bit small but
1:20:12essentially they were imitating The
1:20:13Cutting pattern of the hexagons with the
1:20:16pentagons do I have here yeah here I
1:20:17have the dyes
1:20:19so
1:20:20that's the hexagon guide that the guy
1:20:22had in the picture this is the Pentagon
1:20:24die that was in existence before we
1:20:25showed up in town which is you know you
1:20:27know sharing a full side here and and so
1:20:30kind of imitating the hexagon
1:20:32okay uh uh what was the origin of the of
1:20:36the new of the of the new technology was
1:20:39this was a video of a Chinese Factory
1:20:40producing a ball for World Cup uh I
1:20:43noticed that the pentagons were laid out
1:20:45in a slightly different way that we knew
1:20:46that was being laid out in CL code
1:20:48already
1:20:49uh it turns out that this was something
1:20:52known to mathematicians there's a paper
1:20:53in 1990 in the Journal of discrete and
1:20:55computational geometry that I identified
1:20:57that
1:20:58um you could also go to the Wikipedia
1:21:00Pentagon page you could still do that
1:21:01and they'll show a different you know
1:21:03layout of layout of pentagons anyway but
1:21:05for us it came for it came from this
1:21:07came from this video
1:21:08um uh I'm Lisa coutini I I don't know if
1:21:11it's appropriate page we're married uh
1:21:13so my wife she's an architect and we
1:21:15then made a blueprint using this using
1:21:17this design in this optimal packing of
1:21:19pentagons to uh to get more pentagons
1:21:23per feet okay and here we made a new die
1:21:25you can see the new die doesn't they
1:21:27don't share full size they just share
1:21:28houses okay so essentially you can just
1:21:31get there's just less weight and that
1:21:32rexin you know it wasn't being used for
1:21:34anything uh so this seems like a just a
1:21:37pure reduction in costs
1:21:39uh what happened this is going to
1:21:40summarize very briefly there are 135
1:21:42firms we gave it out to 35 of them
1:21:4515 months later a grand total of six uh
1:21:47had a doctor which was surprising to us
1:21:49right we were expecting many more much
1:21:50much more adoption
1:21:52um we were encouraged that it was
1:21:54actually working because one of the
1:21:56firms which was actually a silver what
1:21:59if silver we saw
1:22:00was an enormous firm I can tell you now
1:22:02they were producing her Nike they have
1:22:03more than a thousand employees they were
1:22:05using you know state of the art
1:22:06Technologies in lots of Dimensions but
1:22:07they had they've been cutting the way
1:22:08that the traditional way and they bought
1:22:11over 30 dies and then 100 of their
1:22:13pentagons were being cut you know using
1:22:14our our technology so it sure looked
1:22:16like our technology was working we asked
1:22:18uh owners why they didn't hadn't adopted
1:22:21and what they said was their workers
1:22:23didn't like it you know what they said
1:22:24were worker resistance that was the
1:22:25number one answer
1:22:26in retrospect it turns out that the
1:22:29reason was these Cutters so I showed you
1:22:30the picture of them they're paid per
1:22:32ball
1:22:35and our dye was pulling them down like
1:22:38they've been many of these guys have
1:22:39lots of experience they've been doing
1:22:40this for 20 or 30 years they're going
1:22:42really fast right in the in the cutting
1:22:44and then all of a sudden now they have
1:22:45this new fancy dye they've got to figure
1:22:47out how to how to locate it and it was
1:22:49slowing them down it turns out that if
1:22:51you know it doesn't it without within
1:22:53about a month they could get up to speed
1:22:54but they didn't they didn't know that
1:22:56right and all they realized was if the
1:22:58labor contract doesn't change and I'm
1:23:00being paid peace rates my salary is
1:23:02and they were worried that might be
1:23:04forever
1:23:05right and so uh essentially what they
1:23:07did is they and we have lots of actual
1:23:09levels about this they told they like
1:23:11told their owners terrible this doesn't
1:23:12work it's dangerous you're like you
1:23:15don't get more pentagons they relying to
1:23:16you whatever and so the owners didn't
1:23:18really know the owners were sort of
1:23:20delegating decisions to the Cutters and
1:23:21then there was crazy people from you
1:23:23know from elsewhere coming outside and
1:23:25telling them there's this new technology
1:23:26and uh and so they didn't adopt okay
1:23:30show that then we did the second
1:23:32experiment we said that we're going to
1:23:34give the Cutters actually and the
1:23:35printers let me kind of ignore that it's
1:23:37a related issue but we're gonna give the
1:23:39Cutters A month's salary as a bonus not
1:23:41very much money it was like 150 dollars
1:23:44bonus if they can demonstrate to the
1:23:46owner that the technology works and
1:23:48demonstrate the hospital technology
1:23:49works in the presence of the owner that
1:23:51work that workers are excited about that
1:23:53right and then we got a specifically
1:23:55significant uh share of firms adopted as
1:23:57a result
1:23:58right so that's what this is showing
1:23:59basically the here this is the the you
1:24:02know the second experiment you can think
1:24:03of this I guess you can focus on column
1:24:05three if you're assigned to group a
1:24:06which is the one that got this got this
1:24:08bonus then you know because they went
1:24:10share of firms subsequently adopted
1:24:12right so it seems like we we found the
1:24:14the bottleneck right and that 150 we
1:24:16gave out was very small relative to the
1:24:18amount that the firm could gain from the
1:24:19doctors
1:24:22okay uh so what is that what are we
1:24:25learning then about
1:24:26um
1:24:27uh about this so I a few things so one
1:24:30is I think these organizational
1:24:33features matter if agency issues matter
1:24:35a lot
1:24:36um there seems to be some kind of an
1:24:38inertia in labor contract so so maybe
1:24:39before we came to town like the piece
1:24:41rates were the alcohol contract many
1:24:43people that was here for his career was
1:24:45amazing peace or the art optimal
1:24:47contract
1:24:48um maybe they were but then you know
1:24:50once the technological environment
1:24:53changes then those those existing
1:24:55contracts may be hard to change right so
1:24:56that's kind of the story that we have
1:24:58coming out of out of this
1:25:00um concepts of times I'm skipping a
1:25:03little bit uh there are complementaries
1:25:04between technological and organizational
1:25:05Innovations in this case changing to you
1:25:07know internalize the for the workers a
1:25:09bit to adoptions technology helped a lot
1:25:11to get the adoption there's going to be
1:25:13those sorts of
1:25:14complementaries and
1:25:16sort of the op-ed version is that
1:25:18workers need to expect to share in the
1:25:19gains
1:25:20to adoption in order for adoption to be
1:25:22successful
1:25:23question about why firms didn't adopt on
1:25:26their own maybe this didn't understand
1:25:27the game or didn't understand the
1:25:29process or maybe they did but they
1:25:31thought the hassle of recontracting you
1:25:33know might have been too costly that
1:25:34didn't want someone workers coming back
1:25:36from the next week and they want a bonus
1:25:37for something else or you know
1:25:42okay uh what let me
1:25:44uh I've got I've got three minutes so
1:25:47I'm going to and some of this a lot of
1:25:48this stuff is going to be talked about
1:25:49in the management lecture I'm presuming
1:25:51by Jim next week I I think what I'd like
1:25:54to do is talk about the
1:25:56so let me I'll just summarize what this
1:25:58is about so there's issues about why
1:26:00family is family control bad there's
1:26:02evidence that like inherent control is
1:26:04bad for productivity
1:26:06um but family control is very persistent
1:26:07so I think there's an interesting
1:26:08question about it's so families girls
1:26:10like putting your sons and having them
1:26:12be managers for you and your firm seems
1:26:14to solve some of these Agency problems
1:26:16like how do you control the managers
1:26:17they you know the the sun's interests
1:26:20are aligned with the owners and so that
1:26:21sells but then that may have negative
1:26:23consequences once uh you know once the
1:26:26owner dies and then one of those gets
1:26:28promoted they may not be the best best
1:26:29match
1:26:31there's lots of work on learning from
1:26:33other firms some very influential uh you
1:26:36know interventions have shown that keep
1:26:38providing more information has positive
1:26:40effects I think Gene is going to talk
1:26:41about this this first paper is this
1:26:43India in the Imagine practices
1:26:45um experiment Liberty Cardinal and so
1:26:47are also again intensive Consulting
1:26:49seems to have a positive effect on
1:26:51purpose
1:26:52um let me
1:26:53I do kind of want to say something about
1:26:55this behavioral for instance one big
1:26:56conceptual point which is
1:26:59um
1:26:59so
1:27:01if you took my I kind of presume this
1:27:03related to a question that came before I
1:27:04presume profit maximization in the in
1:27:06the framework and so I'm kind of
1:27:09presuming that
1:27:10if firms are not adopting
1:27:13uh and it's not we can't the attribute
1:27:15to National block conditions there must
1:27:16be a lack of know-how okay so that but
1:27:19it could be that the pro the Assumption
1:27:22of profit maximizations
1:27:23right either because entrepreneurs hold
1:27:25goals besides profit maximization they
1:27:27want a quiet life they want to enhance
1:27:29their reputation you know look for
1:27:30non-economic reasons you know outside of
1:27:32the outside of the market or it may be
1:27:34that they're making mistakes okay so
1:27:36those would be sort of like behavioral
1:27:37firms and there's now a lot of talk
1:27:38about Behavioral Services
1:27:40uh who might have you know might failed
1:27:42profit maximized for one of these uh
1:27:46um
1:27:47I put myself I'm kind of a skeptic about
1:27:49the behavioral firms um uh partly
1:27:52because I guess uh it's partly in my
1:27:55instinctive you know people used to talk
1:27:57about
1:27:58like irrational peasants
1:28:01on agriculture developing countries and
1:28:04then uh Schultz the results came along
1:28:05and said no we should think about these
1:28:07producers as poor but rational if we see
1:28:09them doing things that look at rational
1:28:10we should understand the constraints of
1:28:12this and so I think I I would make a
1:28:14similar case about how we should think
1:28:16of uh firms in developing countries that
1:28:19they they may we should think of the
1:28:21individuals as rational but they may
1:28:22lack know-how and then or there may be
1:28:24aware things that prevents you know a
1:28:27collection of rational individuals from
1:28:28implementing the best decision without
1:28:30having to presume that there's you know
1:28:33that they're not uh rationed
1:28:35right or that they're not profit
1:28:37maximizing or even if they're making
1:28:38mistakes now they may be making mistakes
1:28:40but I don't think we can assume that or
1:28:42I don't think we can jump to that
1:28:43conclusion just on the basis
1:28:45foreign
1:28:53I've said basically what um
1:28:56most of the things that I intended to
1:28:59say I think let me just read I'll take
1:29:00one more minute just
1:29:01so so I think we should de-emphasize tfp
1:29:04as a metric for upgrading
1:29:06that's one point I've been trying to
1:29:07make some drivers that seem to be
1:29:09important selling your consumers
1:29:10directly or indirectly the available
1:29:12high quality input seems to matter so
1:29:14let's relate to the trend mobilization
1:29:15one thing that trade mobilization does
1:29:17is you know increases availability of
1:29:19high quality inputs in developing
1:29:20countries and so I think that's valuable
1:29:22the ability to resolve Contracting
1:29:24frictions within and across the firm
1:29:25seems to matter a lot and so that's an
1:29:27area you know for a little for research
1:29:29and exposure to information from other
1:29:31firms seems to matter more evidence that
1:29:32I'm sure counting on to talk about those
1:29:34things
1:29:35the key challenge in this whole
1:29:36literature is how to promote learning I
1:29:38do think it's out of remote learning
1:29:39inference and I've had some kind of some
1:29:41notable caution on on competition on
1:29:43Capital on behavioral first
1:29:45okay let me I'll just leave this here uh
1:29:48but you guys can there are some ideas
1:29:50for for future research but even maybe
1:29:52in the in the review paper you can look
1:29:54at those and and think about those on
1:29:55your own
1:29:56um I I was told okay so I need to stop
1:29:59uh David wanted me to remind you that
1:30:01next Thursday I believe it's at the same
1:30:04time at 10 A.M Eastern please jump in
1:30:07David or Dave or Caleb if that's
1:30:09incorrect for the next lecture by Jin
1:30:11Kion management practices
1:30:13um thanks very much for your questions
1:30:14I'm available to answer any of them I'll
1:30:17try to um before things get close out
1:30:19I'll try and copy the Q a to see if
1:30:21there's something I can I can respond
1:30:23quickly to but it was a pleasure
1:30:26thank you so much Eric
1:30:31and Eric I will download the Q a and I
1:30:34can send it to you later as well with
1:30:35people's emails attached to their
1:30:37questions if that's helpful that would
1:30:39be that would be great that would be
1:30:40great was I correct on the time next
1:30:41week for Jinx uh that's correct 10 a.m
1:30:44next Thursday 10 a.m Eastern next
1:30:46Thursday okay
1:30:48great
1:30:50thank you so much
1:30:56bye-bye

About this transcript

This page contains the full transcript of Lecture 2: Firm Upgrading with Eric Verhoogen by International Growth Centre, generated from the public captions YouTube serves with the video. The transcript has 18,267 words across 2,737 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.