The Industry choice

[Alex Martelli]

Bulba! <bulba at bulba.com> wrote:
   ...
> First, even though I disagree with you in places, thanks for 
> this reply - it enhanced my knowledge of the topic in some

You're welcome!

> What you wrote regards especially strong the industries you pointed
> at: fashion, jewellery, esp. I think in those industries those factors
> may be decisive. When this street is renown in the city to have 
> "good jewellers" located there, the choice for a new jeweller in this
> city is almost like this: get located there or die.
> 
> However, I'd argue that industries with those kinds of dependencies
> are actually pretty rare. 

And yet the phenomenon appears in quantitatively similar ways (to
different scales) in many industries where either the supply side or
demand side benefits of clustering apply.

> >Now consider a new company Z set up to compete with X, Y and Z.  Where

Ooops, typo, sorry, call the new company T!-)

> >will THEY set up shop?  Quarter Q has the strong advantage of offering
> >many experienced suppliers nearby -- and in many industries there are
> >benefits in working closely with suppliers, too (even just to easily
> >have them compete hard for your business...).  
> 
> I grant that this model is neat and intuitive - but I'd argue it is
> not very adequate to real world. Read on.

Apart from its being exogenous (the model does not try to explain why X
Y and Z were already located in Q -- only why, given this state of
affair, T has advantages to setting things up in Q, too, which may well
offset the costs such as higher rents), I don't think so -- but, I _am_
reading;-).


> >YOU going to set up shop?  Rents in that piazza are high, BUT - that's
> >where people who want to buy new hats will come strolling to look at the
> >displays, compare prices, and generally shop.  
> 
> That will only be true if the hats from Piazza dell'Orologio are much
> better than elsewhere.

No, this is not necessary (it's not even strictly necessary that they're
_believed_ to be better, say because they traditionally were).  A
sufficient inducement to the would-be customers to go shopping there
rather than elsewhere is that in that Piazza they get more choices for
the same effort, so shopping there is more effective.

Say that the city has ten hat shops of the same quality.  One is in
Piazza dell'Unita`, all the way to the Northern border of the city.  One
is in Piazza Saragozza, all the way to the Southern border.  The other
eight are in Piazza dell'Orologio, smack in the middle of downtown.
Each shop offers hats taken from the same random distribution: we can
say that a normal curve measuring the utility function for the customer
(some function of price, quality, looks, fit, ...), with identical
average and variance, represents the best hat available today from a
given shop from the POV of a given customer.

Say that a customer has one day to shop for their new hat, and the
locations are too far apart for a customer to visit more than one
location within that day.  If a customer chooses to visit the one
southern shop, or the one northern shop, the customer expects to be
presented with the choice of hat from said normal curve.  If the
customer goes downtown, they expect a choice which overall lies along a
DIFFERENT curve -- the best-of-eight samples from the normal curve.
Sorry I can't model that analytically, but both intuitively and from any
little simulation (easy to code in Python) you can see the customer's
expectations are much better if they choose the location where they can
do more comparison shopping!

If you change this handwaving into numbers you can model this
demand-side push of hatsellers towards clustering vs the forces pushing
in the other direction, such as the supply-side issue of higher rents
which may be assumed, all other things being equal, to make each shop's
curve slightly _worse_ in the downtown location (higher costs, all other
things being equal, must mean higher prices).

It's not just fashion.  I can look at the historical location of
fishmongers in Bologna, still to some extent observable today, and I see
them generally spread all over BUT with two obvious centers of downtown
aggregation: a minor one at the Mercato delle Erbe (center-south side of
downtown) and a major one at via Pescherie Vecchie (which just happens
to mean "street of the Old Fishmongers" -- that concentration goes back
many, many centuries).  When I'm just getting, say, some fresh anchovies
to fry for a pasta sauce, I'm likeliest to shop at a fishmonger near my
home (the University side of downtown, northerly) or get them as part of
a generic grocery shopping trip at a supermarket.  But sometimes I'm
preparing for a meal where fish plays a big role, maybe with company,
and then I want the best, and then I head for Pescherie Vecchie.  It's
always very crowded with shoppers there, particularly at the times when
people are likeliest to be shopping for fresh fish.  It's obvious to
even the most casual observer that just the same mechanism is at play:
there are demand-driven centripetal forces because shoppers sometimes
like to comparison-shop and get a lot of choice.

This demand-side process doesn't operate much for "commodities" -- goods
with quality and prices so flat and standardized that comparison
shopping does not matter much.  Much fashion-industry ware is that way,
spewed out at commodity levels by the same factories.  Some
concentration may nevertheless remain for purely historical reasons: I'm
thinking for example of the concentration of bookstores in Charing Cross
Road, and of electronics shops in the nearby Tottenham Court Road, in
London.  Being in London only rarely nowadays, if I'm shopping for books
I'm likely to head for Charing Cross, because I know I'll get choice; if
I was a resident, though, and shopped for books often enough, I guess
I'd find out about other places that may be cheaper or otherwise better.

In any case, this more detailed demand-side sketch already shows that
transaction costs are the key consideration in explaning clustering.
Considering that they're also key in the very EXISTENCE of firms,
according to standard economic theory, it's hardly surprising that they
should play a large role in shaping the firm's foundational decisions,
including, of course, location.  But the demand-side model focuses on
the _customers_' transaction costs, their advantage in comparison
shopping.  Similarly, a supply-side model can focus on suppliers'
issues.  Focusing within the firm, we can get insight on the pluses and
minuses to the firm of being in a single place vs spreading itself
around, but clustering of separate firms is clearly a different issue.


> If the quality and prices of products converged, the gain for the
> customers to go there isn't high. And the prices for customers
> have to be higher, as the high rents drive some suppliers out,
> so the supply of hats is lower, ergo customer prices are higher.

...which is why you know a priori that some equilibrium point is likely:
there are costs as well as advantages to clustering (rents can often be
an important slice of the costs, but other inputs that are costly to
transport could also become scarce, though maybe not to the same point
as land and suitable buildings and other fixed infrastructure).

But there's nothing intrinsic to predict that the amount of clustering
at equilibrium will be low.  Transaction costs are often connected to
acquiring and checking information; "knowing" (or believing one knows)
where the hatsellers are may be enough of an advantage to maintain a
clustering, once it's historically formed, even when you can hardly find
any more present evidence of materials-costs advantage to it... the
information-cost part of the transaction costs may well suffice.


> >That's close to where
> >felt-makers are, since they sell to other hat-makers.  Should your
> >business soon flourish, so you'll need to hire a worker, that's where
> >you can soon meet all the local workers, relaxing with a glass of wine
> >at the local osteria after work, and start getting acquainted with
> >everybody, etc, etc...
> 
> That is true in the model. However, I don't find it very true
> in practice in a lot of industries:
> 
> - "physical proximity" in this city very often means navigating
> through jammed roads, so the transportation costs in terms of 
> time and money may very well increase rather than decrease by 
> locating "in a cluster", even though physical distance may
> be smaller. physical distance != temporal distance & a cost
> of getting there.

Congestion can be a part of transaction costs, sure, just as higher
rents can.  But if your suppliers or customers are clustered in
Hollywood, or the City of London, etc, etc, you don't diminish your
congestion-driven costs much by going elsewhere, as you still need to
see your customers or suppliers often -- you just add the costs of
getting to LA / London / etc to those of moving within it -- in this
sense congestion costs behave very differently from rents.

> - since everything is more expensive, so is labor; even
> if you pay this worker a little bit less elsewhere, he might
> find it more attractive to commute shortly to work and 
> have lower costs of living in a different area, so this very 
> well may work as a motivation to relocate

Sure, you get more competition: just as it's easier for you to get your
competitors' best people, it's easier for your competitors to get yours
(to some extent this applies to other inputs, and customers too, but
highly skilled labor is the key component in many industries).  This
raises general costs, roughly equally for all clustered competitors, but
another factor may be more prevalent because it differentiates among
competitors: if you think you're among the best company in your sector,
a more desirable place for your competitors' best personnel to work,
then you perceive clustering as desirable because of that; if you think
you're an inferior company, so your best people are likely to want to
flee, then you perceive clustering as undesirable.

Thus, seeking clusters can send the message "we're the best"; avoiding
clusters can send the message "we're not so hot".  That's how situations
of asymmetric information tend to work in economics -- and since
Akerlof's seminal work "The Market for Lemons" is 35 years old, and he
got the Nobel prize for it (and related work, of course) in 2001, I
believe we can call this field reasonably classic by now.

Note that I'm not saying there are no forces pushing against clustering:
of course there are, varying by industry etc.  But they're easy to
overstate.  Consider the highly skilled worker who has a choice: they
can stay in some crowded cluster, say Silicon Valley, and keep facing
congestion, high rents, etc, for high salaries and great opportunities
to keep hopping to the highest bidder; or, they can accept an offer at a
lower salary in some more isolated area, say a minor cluster such as
Austin, Tx, and get less congestion, cheaper housing, etc, although also
less opportunity to keep enhancing their careers.  Which kind of worker
will tend to pick which of the two?  Somebody who thinks they may be
past the peak of their career, and won't get many more lucrative offers
in the future anywa, might be more tempted by (say) Austin, while
somebody who's keenly competitive and on a sharp upwards path may keep
braving the congestion for the (to them) attractive lures and challenges
of Silicon Valley.  Of course there are a zillion other factors, but in
as much as we're talking about factors strictly related to clustering,
this is the bias, and therefore (in an Akerlovian model) this is the
message which tends to be sent by such choices.

I.e., seeking clusters send a message "I'm the best", avoiding them
sends "I'm not so hot", for the individual highly skilled worker as much
as for the firm in some highly competitive industry.  In as much as we
live in a "superstar economy" clustering becomes more desirable because
of such competitive pressures and messages.


> - few large businesses nowadays have customers neatly
> clustered in one location, they tend to sell countrywide
> or worldwide; so the premium for getting into this expensive
> place is low and the costs are high

There are exceptions (monopsonies or oligopsonies) but for _large_ firms
this generally holds.  But a _branch_ of a large firm may be set up
specifically to serve a geographically identified market, so the
location decision for the branch need not reflect the overall firm's
issues, but rather only those specific to said market.

> - production nowadays tends to be geographically 
> distributed as well

Not as much as you might think.  For example, people who worry about
programming jobs moving to India may not realize a vast majority of such
jobs go to a SINGLE tiny "spot" within the huge expanse of India, for
all of the usual reasons promoting clustering.


> - communication costs and transportation costs are DRAMATICALLY
> lower nowadays than they used to be (today it costs 1/80 [one
> eightieth] to transport a kg of mass using an aircraft in 
> comparison to what it cost in 1930s; ditto for telecommunication).

...so in the '30s very VERY few goods moved by plane.  Not that it's
exactly _common_ nowadays to move most goods by plane, mind you; but
sure, it's conceivable in some cases.

But there are other transaction costs, mostly connected to the need for
face-to-face interaction as being the most effective form.  I've worked
for a SW development firm which tried to coordinate development
distributed across many locations via cheap video-based teleconferences
spanning timezones from California all the way to India; I've done way
more than my share of telecommuting and airport- and plane-hopping for
development projects geographically distributed and/or mostly located
far from the customers and/or other stakeholders; I know whereof I
speak...


> Consider real world examples: Microsoft in Redmond or Boeing in
> Seattle. Microsoft needs quite a lot of programmers. It would be
> rather hard to find enough programmers born and educated in 
> Redmond, wouldn't it?

Of course, or, at MS's scale, in any other single location.  So, as they
grew, they had to face the usual within-firm tension between the
advantages of a single location and that of many; right now, quoting
their careers webpage, they "employ more than 50,000 people worldwide,
including offices in every major U.S. metropolitan city, and subsidiary
offices in more than 60 countries.".

For example, to attract the kind of highly competitive top-of-career
high-tech worker that just will NOT leave the cluster of Silicon Valley,
they chose to set up a new campus in Mountain View, CA, focusing on the
convergence of entertainment and information technologies and
industries; to attract the kind of European scholar who considers living
in Europe a must and/or has no patience for the silly hassles put up by
the US to keep foreigners out, Microsoft has an excellent Research
laboratory at Cambrige, UK; and so on, and so forth.

Of course, management-wise, there would be some great advantages to MS
to having everybody at the Redmond campus, but even MS must deal with
the real world, and the job market in particular: they're very
pragmatic, so they get things in the Campus where they can, elsewhere
when they must.

How a large firm internally decides to site its various locations is
really a separate issue from that of how different firms cluster, though
of course there are many connections, particularly when you study the
siting of one particular location.  The two examples I give here are
reasonably typical of a focus on supply-side and particularly the
availability of highly skilled workers in high-tech areas... and then
they show, again, the pluses of clustering.  If Microsoft had found it
easier to attract the kind of people they wanted, and build the needed
connections to other firms in entertainment/information merging, outside
the Silicon Valley cluster, there would be no reason for them to site
there -- and instead you find them smack in the middle at Mountain View.


> Boeing: Seattle?! Why Seattle? Doesn't Boeing sell most of its 
> new aircrafts in Washington? 

Obviously they agree with you, since their World Headquarters are in
Chicago since over 3 years ago -- not exactly all that near to
Washington, DC, but way closer than Seattle;-)  "Why Seattle" has
obvious purely historical answers -- they were born there, etc, etc.
"Why Chicago" is a far more interesting question, since clearly the
choice was way more open once they had decided that to keep expanding in
Seattle was too costly or inopportune.  You can find a lot about that
specific decision on the web, of course.


> AFAIK, in Europe Boeing does much of its production in 
> cooperation with Alena in Italy. From this viewpoint it really 

The spelling is "Alenia" (just in case somebody wants to google for more
info).

> doesn't matter much if Boeing is located in Seattle or anywhere 
> else in America really, does it?

Not much, though flying to Chicago is marginally easier than flying to
Seattle, it's not enough to matter;-)

> Right here (Poland), most of the foreign corporations set up their
> call centers in Warsaw, which is totally ridiculous given how
> expensive that city is. Oh I can understand setting up a warehouse
> by HP there because it's close to its biggest customers. But
> I really see no good reason behind HP call center being located in
> Warsaw, too, AFAIK (or at least that's the city code when I have
> to call them sometimes).
> 
> Most of the successful domestic companies I have observed here 
> have started and still operate in some God-forgotten provincial 
> towns, where land and labor is cheap and when it comes to 
> highly skilled professionals, you have to "import" some or all of 
> them from  elsewhere anyway, because not even a big city is 
> likely to have precisely all the kinds of ultra-specialized
> professionals that you may need. And there's less crime, and 
> shuttling kids to school isn't a nightmare, and the costs of living
> are lower. Plus there's less of other things to do, so your workers
> tend to focus more on work. :-)

I remember glancing at some costly booklet called something like "Poland
Infrastructure Report" back when an employer was considering setting up
a branch office somewhere in Poland (selling and customizing SW for the
mechanical industry), and back then issues such as internet and other
telecom access, easy availability of top graduates, ease for expatriates
from Italy to live in the place for a while knowing only English and
Italian, at most German and French, and not Polish or Russian, closeness
to good international airports and other good transportation, closeness
to partner firms and potential customers' decision-makers, all appeared
to point to Warsaw, if I recall correctly.  Mechanical engineers with
some programming experience or viceversa, good translators, and good
salespeople with connections in the mechanical industry, are not as
ultra-specialized as all that, after all.


> >Risk avoidance is quite a secondary issue here (except if you introduce
> >in your model an aspect of imperfect-information, in which case,
> >following on the decisions made by locals who may be presumed to have
> >better information than you is an excellent strategy).  Nor is there any
> >"agency problem" (managers acting for their interests and against the
> >interest of owners), not a _hint_ of it, in fact -- the hatmaker acting
> >on his own behalf is perfectly rational and obviously has no agency
> >problem!).
> 
> I find no justification in most of foreign companies setting up their
> call centers in a ridiculously overpriced capital city - so from
> my viewpoint the  risk avoidance is the only explanation that is 
> left.

The firm I was working for had a consensus decision-making process (even
I was involved) and managers (and other employees) and stockholders were
mostly the same people -- it wasn't all that large a firm at the time.
Nobody needed to practice risk avoidance.  The infrastructure advantages
of Warsaw vs other locations loomed HUGELY large, judging of course from
some consultants' reports purchased for the purpose -- it may look
different to people living in the place, although I'd like to get a
second opinion from Warsaw's Chamber of Commerce since you appear to
have a very specific individual bone to pick (and I can sympathize: even
though Milan may be the economically correct choice for foreign
investors, I'd never want to LIVE there myself, being a Bolognese... but
I must admit that Milan's infrastructure, connections, location, etc,
etc, may in several cases drive a rational decision to set up there).


> Not all corporations do that: in this country Coca-Cola has made
> their big "green field" investment almost in the middle of nowhere 
> in this country. GM, Volkswagen, FIAT, and most of other carmakers
> have made similar decisions. Doesn't seem like those were bad business
> decisions for them.

If their needs were for cheap land for greenfield factories, and cheap
workers for said factories, then they were acting under very different
forces than a midsize company looking to set up a mostly-sales branch
office, not an industrial factory.


> The cluster I've seen - an IT "corporate area" in Dublin, 
> Ireland -  was specifically created not due to "natural clustering",
> but due to govt policy of tax breaks and preparing good
> infrastructure in this place rather than some inter-business and
> inter-customer dependencies, since e.g. Microsoft (where the company
> sent me for  training) neither is able to get all the workers it needs
> just from this city, nor it sells mostly to local customers, but it
> sells all over Europe.

That's addressing only the issue of endogenous vs exogenous original
causes for clustering.  Many attempts to create clusters artificially
have happened ever since the economical advantages of clusters were
discussed in the literature, together with the crucial point that WHERE
the cluster originally happens is in most cases almost random, it's
self-reinforcing once it's properly underway.  Most such attempts have
failed, because governments' powers aren't unlimited; Dublin is a good
example of this strategy succeeding.

No matter WHY the good infrastructure is there, the tax breaks, the
thriving community of high-tech workers, etc, a firm deciding where to
set up may perfectly well find it rational to have all of these
advantages overwhelm the issues of rents, congestion, competition for
good workers.  In other words, my disagreement with your thesis that,
because the government lowered taxes, taking advantage of that is NOT in
the best interests of a firm's stockholders, is now maximal: I find your
thesis not just wrong, but by now outright silly.


> To me, the issue of manager having to face their superiors
> asking him a question - "why on Earth have you decided
> to locate our branch in the middle of nowhere in this country?
> Why not in capital city? Can't you look at the map? Read some
> stats how many inhabitants this city has and what is the
> income level there?" is overwhelming: it would take long 
> time to explain for this manager who e.g. may have already 

You are wrong, because the same decisions get rationally made by
sole-owner sole-manager companies where these considerations cannot
apply.  Deciding where to site an imporant branch is a BIG decision: of
course it takes a long time, *DUH*, and all sorts of factors are taken
into consideration.

> got his hands dirty in this industry in that country to know 
> that say, there's little revenue increase to be gained in 
> locating  the facility in capital city, the needed workers are
> actually hard to find there, etc.

So all of these factors are folded into a huge pile of reports in
companies where such decisions are at all likely to be challenged later,
and sign-off on the folders is carefully obtained for cover-up purposes.

If the local development agencies of the "middles of nowhere in that
country" don't do their job, including showering managers planning such
decisions with supporting materials, that's a bad sign: maybe due to
cultural influences foreign capital, professionals, and managers are NOT
welcome there as they would be in a relative metropolis, for example.
This is (or was not too long ago) surely the case for some rural parts
of the British Islands, and Italy too.  It's perfectly rational and
consone to stocholders' interests to take this into account and site
where you KNOW you'll be welcome.

 
> To me, this is much like decision whether do much of 
> development in  Python or MS VS / VB. "But everybody's
> using VisualStudio / VB" is a simple and compelling argument: 
> "so many people can't be wrong", while explanations that 
> going Python may actually be better decision in this context
> requires long and complex explanations and managers oft can
> not even be bothered to read executive summaries.

The issue of parallels or otherwise is by now totally secondary, to me,
to the main issue that I find your approach to explaining regional
clustering problems across industries totally, irredeemably, and
horribly WRONG.  So, I'm not going to make the post even longer by even
trying to address this part.  Few, besides the two of us, can be left
reading by now, and clearly our disagreements on economics are so total
that it's unlikely we can get anywhere by our discussions anyway.


> I feel econ models frequently are elegantly designed
> abstractions of elegantly designed problems; the problem 
> is how close those are to the real-world problems. At
> the end of the day, it's "human action" that gets all
> of that implemented. I've seen too much resources 
> going down the drain in completely idiotic projects 
> and decisions to believe managers are rational beings.

When you claim managers are acting rationally (if selfishly) to avoid
risk to themselves, you can't then justify this claim by adding that
they aren't rational at all.  Economics does cover, in its modern form,
both issues of agency problems (misalignment of incentives between agent
and owner) AND ones of "bounded rationality" (all the way from
asymmetric information, to transaction costs relating to acquiring and
processing information).  Trying to throw economics overboard because
you can't be bothered to understand it is a kind of behavior that
reminds me closely of the "leftists" you excoriate in your signature.



Alex