Sunny with a chance of a hous­ing bub­ble

The Pak Banker - - OPINION - Mark Buchanan

AT the time of World War I, many me­te­o­rol­o­gists had all but given up on the idea of ac­cu­rate and sci­en­tific weather fore­cast­ing. Then a physi­cist and am­bu­lance driver by the name of Lewis Richard­son, in spare mo­ments be­tween ter­ri­fy­ing bouts res­cu­ing the in­jured, un­der­took a mo­men­tous project.

His aim was to cal­cu­late, by sim­u­lat­ing the ac­tual physics, the devel­op­ment of the weather over a lo­cal zone of Europe over eight hours. He failed, but only be­cause of a small arith­meti­cal er­ror. His ba­sic idea was cor­rect, and weather fore­cast­ing cen­ters around the globe now use vari­a­tions of the tech­nique with im­pres­sive pre­dic­tive success.

In Read­ing, Eng­land, for ex­am­ple, the Euro­pean Cen­tre for Medium-Range Weather Fore­casts runs two su­per­com­put­ers sim­u­lat­ing a vir­tual at­mos­phere. The model fore­casts the wind, the tem­per­a­ture, and the hu­mid­ity at more than 20 mil­lion points from the earth’s sur- face up to a height of about 40 miles. In the U.S., the Na­tional Cen­ters for En­vi­ron­men­tal Pre­dic­tion does much the same thing.

The sim­u­la­tions lie be­hind the daily and weekly weather fore­casts re­ported on the nightly news, as well as more speci?c pre­dic­tion ser­vices for farm­ers, the air­line and ship­ping in­dus­tries, the mil­i­tary, and any­one else whose projects de­pend se­ri­ously on the weather. When an oil com­pany charts a path for a tanker jour­ney of sev­eral weeks, it saves tens of thou­sands of dol­lars by rout­ing away from strong winds and storms.

Forces much like those that de­ter­mine the weather also drive the most im­por­tant and dis­rup­tive events in eco­nom­ics and fi­nance — bub­bles, debt crises, bank runs, even waves of cor­po­rate cor­rup­tion. With ideas and tech­niques from other parts of sci­ence, it’s pos­si­ble to ex­plore mar­ket feed­backs and in­sta­bil­i­ties in de­tail never be­fore pos­si­ble.

In the not-too-dis­tant fu­ture, it’s easy to imag­ine a U.S. or Euro­pean Cen­ter for Fi­nan­cial Fore­cast­ing. Thou­sands of re­searchers would over­see mas­sive sim­u­la­tions prob­ing the de­vel­op­ing net­work of in­ter­ac­tions among the world’s largest ?nan­cial play­ers, fol­low­ing the vast web of loans, own­er­ship stakes and other le­gal claims that link banks, gov­ern­ments, hedge funds, in­surance com­pa­nies and rat­ings com­pa­nies.

The com­put­ers would test sce­nar­ios and cal­cu­late hun­dreds of in­di­ca­tors of sys­temic lever­age, the den­sity of in­ter­con­nec­tions, or the con­cen­tra­tion of risk at sin­gle in­sti­tu­tions. Ex­perts would probe models of the fi­nan­cial sys­tem, look­ing for weak points and test­ing re­silience, much as engi­neers now do with models of the elec­tri­cal grid or other com­plex sys­tems.

What’s cur­rently miss­ing, aside from the will­ing­ness of the eco­nom­ics pro­fes­sion, is data. To en­sure the safety and sta­bil­ity of a nu­clear re­ac­tor, engi­neers need ac­cess to ev­ery de­tail of its op­er­a­tions and the abil­ity to ex­am­ine ev­ery com­po­nent and its links to oth­ers. The same should be true of any agency try­ing to sup­port the sta­bil­ity of the fi­nan­cial mar­kets.

Au­thor­i­ties

have

his­tor­i­cally col­lected fi­nan­cial data on an in­sti­tu­tion by in­sti­tu­tion ba­sis, be­ing less con­cerned by the links be­tween them. Such a piece­meal ap­proach ob­vi­ously makes it im­pos­si­ble to say any­thing about in­ter­con­nec­tions and the feed­backs they cre­ate.

The cri­sis has spurred moves to col­lect much greater amounts of data on ?nan­cial net­works. In the U.S., for ex­am­ple, the Dodd-Frank Act cre­ated the new Of­fice of Fi­nan­cial Re­search to bring bet­ter fi­nan­cial data to pol­icy mak­ers. Pri­vate hedge funds will soon be obliged to report in­for­ma­tion on their funds’ ex­po­sure to dif­fer­ent as­set classes, their use of lever­age, and their vul­ner­a­bil­ity to liq­uid­ity short­ages.

A real data rev­o­lu­tion might go much fur­ther. Mod­ern sen­sor sys­tems — as com­put­er­ized com­po­nents find their way into al­most ev­ery ob­ject we use and own — will prob­a­bly gather as much data in the next 10 years as we have gath­ered in all of hu­man his­tory. One can only imag­ine how all th­ese data might feed into fore­cast­ing models. If ex­ces­sive op­ti­mism or pes­simism drive many mar­ket crises, th­ese col­lec­tive ex­cur­sions from re­al­ity al­most cer­tainly show up in the phys­i­ol­ogy of the peo­ple in­volved. Think of a patch worn by vol­un­teers that gath­ers phys­i­o­log­i­cal in­for­ma­tion and up­loads it di­rectly to some data­base.

Of course, none of the fore­casts based on such data will meet the ideal of per­fect knowl­edge of the fu­ture. Weather fore­cast­ers don’t aim for this ideal, as they al­ways have in­com­plete data on the at­mos­phere and can only work with ap­prox­i­mate equa­tions. They make a strength of this un­cer­tainty by run­ning thou­sands of sim­u­la­tions, chang­ing the data ran­domly to re­flect their ig­no­rance and so gen­er­at­ing thou­sands of pos­si­ble fore­casts about the fu­ture. The re­sult is a cloud or “en­sem­ble” of guesses about where the fu­ture will lie.

En­sem­ble fore­cast­ing in fi­nance and eco­nom­ics might work sim­i­larly, us­ing slightly dif­fer­ent pos­si­bil­i­ties for how peo­ple and com­pa­nies be­have, also en­abling those el­e­ments to have their own in­de­pen­dent in­tel­li­gence to try things and learn tricks the modeler may never see. The re­sult would be not a sin­gle pre­dic­tion but a swarm of pos­si­bil­i­ties.

In pon­der­ing this fu­ture, del­i­cate is­sues loom into view. As we de­velop large com­pu­ta­tional sys­tems packed with masses of data mon­i­tor­ing the fi­nan­cial and eco­nomic sys­tem and pro­ject­ing its likely fu­ture, this knowl­edge be­comes ex­tremely valu­able. It should be treated as a pub­lic good, akin to clean air and water. Deal­ing with such ques­tions is the price we pay for mov­ing be­yond the myth of a per­fect self-reg­u­lat­ing equi­lib­rium — the par­a­digm that has dom­i­nated eco­nomic think­ing for the past sev­eral decades and that has done such a poor job of pre­dict­ing the eco­nomic weather.

Newspapers in English

Newspapers from Pakistan

© PressReader. All rights reserved.