The Value of Simple Models, with Examples of Economic Dynamics

• Tweet

• Tweet

Guest post by Geoff Davies

Many peo­ple, includ­ing many het­ero­dox econ­o­mists, under­stand that the neo­clas­si­cal equi­lib­ri­um approach to under­stand­ing economies is futile and mis­lead­ing [1], because mod­ern economies are far from equi­lib­ri­um. The neo­clas­si­cal pre­dic­tion of equi­lib­ri­um or near equi­lib­ri­um requires a string of patent­ly absurd assump­tions. How­ev­er the devel­op­ment of bet­ter the­o­ries seems to be sig­nif­i­cant­ly hin­dered by a feel­ing that any super­sed­ing the­o­ry has to be thor­ough­ly quan­ti­fied before it can be use­ful, and a feel­ing that the neo­clas­si­cal the­o­ry has set a bench­mark for sophis­ti­cat­ed math­e­mat­ics that must be matched before anoth­er the­o­ry can be respectable. Less fun­da­men­tal­ly there seems to be a com­mon per­cep­tion that empir­i­cal insights can only be gained through elab­o­rate sta­tis­ti­cal treat­ments of obser­va­tions.

Here I offer some dis­cus­sion from my expe­ri­ence as a nat­ur­al sci­en­tist, and some exam­ples regard­ing the Glob­al Finan­cial Cri­sis, to counter these hin­drances. Use­ful and rel­a­tive­ly sim­ple mod­els can be con­struct­ed that can imme­di­ate­ly over­come major neo­clas­si­cal lim­i­ta­tions, for exam­ple by per­mit­ting non-equi­lib­ri­um behav­iour. The solu­tion of the math­e­mat­ics can be done using very stan­dard numer­i­cal inte­gra­tion meth­ods that are read­i­ly avail­able in com­mer­cial pack­ages. Math­e­mat­i­cal machis­mo is not required. There are also sit­u­a­tions in which the empir­i­cal les­son is obvi­ous with no analy­sis, as will be not­ed here.

I should be clear that there are cer­tain­ly many mod­ellers who oper­ate out­side neo­clas­si­cal con­fines, report­ed for exam­ple in Bein­hock­er’s excel­lent sur­vey of “com­plex­i­ty eco­nom­ics” [2]. The lessons offered here will not be news to them. Also some of them are con­struct­ing quite com­plex mod­els that are nev­er­the­less very instruc­tive, such as mod­els with many inter­act­ing adap­tive agents. This arti­cle is prompt­ed by my read­ing of some het­ero­dox blog dis­cus­sions, and is addressed to any­one who may have some dif­fi­cul­ty see­ing how to move beyond the neo­clas­si­cal approach. Nor are the mod­els here are offered as orig­i­nal inves­ti­ga­tions, though they may lead to such.

Read the whole piece on Geof­f’s blog.