Empirical and theoretical reasons why the GFC is not behind us

Flattr this!

Preliminary Remarks

As not­ed in Debt­watch No. 44, I have stopped writ­ing the month­ly Debt­watch Report to focus on my more long term research. I’m still post­ing occa­sion­al blog posts when I feel the need—like the two recent­ly on Aus­trali­a’s new resources tax—but gen­er­al­ly I’ll be work­ing on more tech­ni­cal mat­ters, and post­ing entries based on those here in lieu of the more top­i­cal Debt­watch. This post is a halfway post between the two: it’s a paper that I have just sub­mit­ted to the 2010 Aus­tralian Con­fer­ence of Econ­o­mists, which will be held in Syd­ney in Sep­tem­ber. I have writ­ten it large­ly as a brief­ing paper for main­stream econ­o­mists who would not have come across the analy­sis that I present here before, let alone the vast vol­ume of lit­er­a­ture in Post Key­ne­sian and Aus­tri­an eco­nom­ics.

There is one clas­sic SNAFU in going from Word to the blog for­mat, for the for­mat­ting of a table below;  I wast­ed too much time try­ing to fix it, so I’ll have to live with it for this post!


A major­i­ty of the 16 indi­vid­u­als iden­ti­fied in Beze­mer (2009) and (Full­brook (2010)) as hav­ing antic­i­pat­ed the Glob­al Finan­cial Cri­sis fol­lowed non-main­stream approach­es to eco­nom­ics, with most of them iden­ti­fy­ing as Post-Key­ne­sian (Dean Bak­er, Wynne God­ley, Michael Hud­son, Steve Keen, Ann Pet­ti­for) or Aus­tri­an (Kurt Richel­bach­er, Peter Schiff). The the­o­ret­i­cal foun­da­tions of these authors there­fore dif­fer sub­stan­tial­ly from those of more main­stream neo­clas­si­cal econ­o­mists. In this paper I will restrict my atten­tion to the Post-Key­ne­sian sub­set, which I will here­inafter refer to as the Beze­mer-Full­brook Group. Beze­mer iden­ti­fied the fac­tors that these authors have in com­mon as:

the dis­tinc­tion between finan­cial wealth and real assets… A con­cern with debt as the coun­ter­part of finan­cial wealth… a fur­ther con­cern, that growth in finan­cial wealth and the atten­dant growth in debt can become a deter­mi­nant (instead of an out­come) of eco­nom­ic growth …[the] reces­sion­ary impact of the burst­ing of asset bub­bles… [and] Final­ly, empha­sis on the role of cred­it cycles in the busi­ness cycle… (Beze­mer (2009))

These authors made fre­quent ref­er­ences to the ratio of pri­vate debt to GDP, and the ratio of asset prices to com­mod­i­ty prices—both indi­ca­tors of finan­cial fragili­ty that were empha­sized by Min­sky in his finan­cial insta­bil­i­ty hypoth­e­sis (Min­sky (1982)), which is a com­mon thread in the cred­it-ori­ent­ed analy­sis of the Beze­mer-Full­brook Group. Since these indi­ca­tors are not com­mon­ly con­sid­ered in main­stream eco­nom­ic analy­sis, I repro­duce this key data in the next 2 fig­ures, before con­trast­ing them to those fol­lowed by Ben Bernanke (the acknowl­edged main­stream expert on the Great Depres­sion) in his analy­sis of the Great Depres­sion.

Fig­ure 1: Ratio of pri­vate debt to GDP, USA and Aus­tralia

Fig­ure 2: US asset price bub­bles

The Great Depression: Errant Monetary Policy or the Dynamics of Debt?

Bernanke pre­cedes the sum­ma­ry of his expla­na­tion for the Great Depres­sion with the state­ment that its caus­es must be found in fac­tors that caused a sharp decline in aggre­gate demand:

Because the Depres­sion was char­ac­ter­ized by sharp declines in both out­put and prices, the premise of this essay is that declines in aggre­gate demand were the dom­i­nant fac­tor in the onset of the Depres­sion. This start­ing point leads nat­u­ral­ly to two ques­tions:

First, what caused the world­wide col­lapse in aggre­gate demand in the late 1920s and ear­ly 1930s (the “aggre­gate demand puz­zle”)?

Sec­ond, why did the Depres­sion last so long? In par­tic­u­lar, why did­n’t the “nor­mal” sta­bi­liz­ing mech­a­nisms of the econ­o­my, such as the adjust­ment of wages and prices to changes in demand, lim­it the real eco­nom­ic impact of the fall in aggre­gate demand (the “aggre­gate sup­ply puz­zle”)? Bernanke (2000, p. ix)

His expla­na­tion has two com­po­nents: a con­trac­tion in mon­ey caused by poor mon­e­tary man­age­ment (and a flawed sys­tem based upon the gold stan­dard), and the impact of non­mon­e­tary finan­cial fac­tors:

there is now over­whelm­ing evi­dence that the main fac­tor depress­ing aggre­gate demand was a world­wide con­trac­tion in world mon­ey sup­plies. This mon­e­tary col­lapse was itself the result of a poor­ly man­aged and tech­ni­cal­ly flawed inter­na­tion­al mon­e­tary sys­tem…

I also have ascribed an impor­tant role to non­mon­e­tary finan­cial fac­tors, such as bank­ing pan­ics and busi­ness fail­ures, in chok­ing off nor­mal flows of cred­it and hence exac­er­bat­ing the world eco­nom­ic col­lapse. Bernanke (2000, p. ix)

With regard to the role of poor mon­e­tary man­age­ment in caus­ing the Great Depres­sion, Bernanke favourably cites Fried­man and Schwartz’s iden­ti­fi­ca­tion of “Four Mon­e­tary Pol­i­cy Episodes” where respec­tive­ly a tight­en­ing or loos­en­ing of mon­e­tary pol­i­cy caused a decline or expan­sion dur­ing the Great Depres­sion (Bernanke (2002), cit­ing Fried­man and Schwartz (1963)). These four episodes are sum­ma­rized in Table 1.

Table 1: Fried­man-Schwartz iden­ti­fied peri­ods in US mon­e­tary pol­i­cy dur­ing the Great Depres­sion (Bernanke (2002))

Pol­i­cy Type Start End
Tight­en Spring 1928 Octo­ber 1929
Tight­en Sep­tem­ber 1931 Octo­ber 1931
Loosen April 1932 June 1932
Tight­en 1933 March 1933

Bernanke large­ly dis­missed the debt-defla­tion expla­na­tion giv­en by Fish­er, on the grounds that debt-defla­tion was a dis­trib­u­tive mech­a­nism and not a macro­eco­nom­ic one:

debt-defla­tion rep­re­sent­ed no more than a redis­tri­b­u­tion from one group (debtors) to anoth­er (cred­i­tors). Absent implau­si­bly large dif­fer­ences in mar­gin­al spend­ing propen­si­ties among the groups, it was sug­gest­ed, pure redis­tri­b­u­tions should have no sig­nif­i­cant macro­eco­nom­ic effects. Bernanke (1995, p. 17)

Bernanke then re-inter­pret­ed Fish­er from an equi­lib­ri­um per­spec­tive, where­as Min­sky accept­ed and built upon Fish­er’s explic­it non-equi­lib­ri­um basis:

We may ten­ta­tive­ly assume that, ordi­nar­i­ly and with­in wide lim­its, all, or almost all, eco­nom­ic vari­ables tend, in a gen­er­al way, toward a sta­ble equi­lib­ri­um… But the exact equi­lib­ri­um thus sought is sel­dom reached and nev­er long main­tained…, in actu­al fact, any vari­able is almost always above or below the ide­al equi­lib­ri­um…

The­o­ret­i­cal­ly … there must be- over- or under-pro­duc­tion … and over or under every­thing else. It is as absurd to assume that, for any long peri­od of time, the vari­ables in the eco­nom­ic orga­ni­za­tion, or any part of them, will “stay put,” in per­fect equi­lib­ri­um, as to assume that the Atlantic Ocean can ever be with­out a wave…

in the great booms and depres­sions … [there are] two dom­i­nant fac­tors, name­ly over-indebt­ed­ness to start with and defla­tion fol­low­ing soon after Fish­er (1933, p. 341)

Min­sky adopt­ed this dis­e­qui­lib­ri­um per­spec­tive, and argued that a reduc­tion in debt does have macro­eco­nom­ic effects, because it revers­es the process of debt adding to aggre­gate demand that must occur dur­ing a peri­od of expan­sion:

For real aggre­gate demand to be increas­ing … it is nec­es­sary that cur­rent spend­ing plans, summed over all sec­tors, be greater than cur­rent received income and that some mar­ket tech­nique exist by which aggre­gate spend­ing in excess of aggre­gate antic­i­pat­ed income can be financed. It fol­lows that over a peri­od dur­ing which eco­nom­ic growth takes place, at least some sec­tors finance a part of their spend­ing by emit­ting debt or sell­ing assets. Min­sky (1982, p. 6; empha­sis added)

Min­sky’s thus defines aggre­gate demand as the sum of GDP plus the change in debt, and this demand is expend­ed on both com­mod­i­ty and asset mar­kets, in con­trast to the main­stream macro­eco­nom­ic focus on com­mod­i­ty mar­kets alone. Bernanke’s inter­pre­ta­tion and sta­tis­ti­cal mea­sures of tight mon­e­tary pol­i­cy also implic­it­ly accepts the mon­ey mul­ti­pli­er expla­na­tion of the rela­tion­ship between base mon­ey and broad­er mea­sures of mon­ey such as M1. The Beze­mer-Full­brook Group instead argues that the mon­ey sup­ply is endoge­nous­ly deter­mined, and that changes in base mon­ey fol­low rather than cause changes in broad mon­ey, so that there is no direct causal link between M0 and M1 (Moore (1979); see also Kyd­land and Prescott (1990)). The Bernanke-Fried­man-Schwartz asser­tion is that declines in the rate of growth of the mon­ey sup­ply ini­ti­at­ed by the Fed­er­al Reserve caused the Great Depres­sion; the Beze­mer-Full­brook Group asserts instead that con­tract­ing debt caused the Great Depres­sion. These expla­na­tions can be com­pared empir­i­cal­ly, by con­sid­er­ing the cor­re­la­tions between the rate of growth of the mon­ey sup­ply and unem­ploy­ment, ver­sus changes in the debt-financed pro­por­tion of aggre­gate demand and unem­ploy­ment.

Money and Unemployment in the Great Depression

Fig­ure 3 shows the rela­tion­ship between M0 and M1 over the peri­od from 1920 to 1940. Though the two mea­sures move in con­cert with each oth­er in the peri­od from 1920 to 1930, it is obvi­ous that there was a break­down in the rela­tion­ship between these two aggre­gates in the peri­od between 1931 and 1934. Infer­ences that are based on the appar­ent rela­tion­ship between M0 and M1 in 1920–1930 will thus be sus­pect in the peri­od 1931–1934.

Fig­ure 3: Rates of change of M0 and M1 1920–1940

This is borne out by a com­par­i­son of the cor­re­la­tion of the rate of growth of the mon­ey sup­ply as mea­sured by M1 and unem­ploy­ment in the peri­od 1930–1940, and the cor­re­la­tion using M0. Fig­ure 4 shows that the cor­re­la­tion is mod­er­ate and has the expect­ed sign for M1 (-0.3): an increase in the rate of growth of M1 is cor­re­lat­ed with a fall in unem­ploy­ment, as Bernanke, Fried­man and Schwartz hypoth­e­size (the rate of unem­ploy­ment is invert­ed on the right hand axis, to enhance visu­al inspec­tion of the cor­re­la­tion: when the two series move in the same direction—or high rates of mon­ey growth cor­re­spond to falling rates of unemployment—the hypoth­e­sized rela­tion holds).

Fig­ure 4: M1 Mon­ey sup­ply growth and unem­ploy­ment

The cor­re­la­tion with M0, on the oth­er hand, has the wrong sign (0.42): increas­es in the rate of growth of M0 are cor­re­lat­ed with increas­es in the rate of unem­ploy­ment.

Fig­ure 5: M0 Mon­ey sup­ply growth and unem­ploy­ment

The pos­si­bil­i­ty that this appar­ent para­dox reflects the fact that mon­e­tary pol­i­cy acts with a lag is dis­pelled in Table 2 on page 11, which shows that though the neg­a­tive cor­re­la­tion of changes in M1 does strength­en (peak­ing at ‑0.5 at a 12 month lag), the pos­i­tive cor­re­la­tion between changes in M0 and unem­ploy­ment remains. One way to make sense of this para­dox is to con­clude that, con­trary to Bernanke’s inter­pre­ta­tion, the Fed­er­al Reserve dur­ing the 1930s did try to counter the Great Depres­sion by expand­ing the mon­ey sup­ply, but that not until 1938 were its attempts suc­cess­ful. This infor­ma­tion was masked by sta­tis­ti­cal analy­sis that used M1 rather than M0, since only M0 is direct­ly under the con­trol of the Fed­er­al Reserve. Sub­tract­ing M1 from M0 con­firms this hypoth­e­sis: non-M0 M1 is more strong­ly cor­re­lat­ed with unem­ploy­ment than M1 alone (-0.41, peak­ing at ‑0.52 with an 8 month lag).

Fig­ure 6: Non-M0 M1 Mon­ey sup­ply growth and unem­ploy­ment

A sec­ond way to resolve the para­dox fol­lows from the first. Since M1 (and more expan­sive def­i­n­i­tions of the mon­ey sup­ply) is deter­mined by the actions of the pri­vate finan­cial sys­tem as well as the Fed­er­al Reserve, the pub­lic and pri­vate mon­ey cre­ation sys­tems were work­ing in oppos­ing direc­tions in the 1930s. For the first eight years, the pri­vate sec­tor’s reduc­tions in cred­it over­whelmed the pub­lic sec­tor’s attempts to expand the mon­ey sup­ply. By mid-1938, when the USA’s pri­vate debt to GDP ratio had fall­en to 140 per­cent of GDP (from its defla­tion-enhanced peak of 238 per­cent in 1932), sub­stan­tial increas­es in M0 were able to expand the aggre­gate mon­ey sup­ply and increase eco­nom­ic activ­i­ty by enough to cause unem­ploy­ment to fall.

This inter­pre­ta­tion brings us to the debt-dri­ven analy­sis of the Beze­mer-Full­brook Group, applied in this instance to the Great Depres­sion. From this per­spec­tive, both the boom of the 1920s and the slump of the Great Depres­sion were caused by chang­ing lev­els of debt in an econ­o­my that had become fun­da­men­tal­ly spec­u­la­tive in nature. Ris­ing debt used to finance spec­u­la­tion dur­ing the 1920s made that decade “The Roar­ing Twen­ties”, while pri­vate sec­tor delever­ag­ing when the spec­u­la­tive bub­ble burst caused a col­lapse in aggre­gate demand that ush­ered in the Great Depres­sion in the 1930a. Fig­ure 7 illus­trates both the ris­ing debt of the 1920s and the falling debt of the 1930s.

Fig­ure 7: US Pri­vate Debt and Nom­i­nal GDP, 1920–1940

Fig­ure 8 illus­trates how much the increase in debt dur­ing the 1920s added to demand, and then how much its fall in the 1930s sub­tract­ed from demand.

Fig­ure 8: Aggre­gate demand is the sum of GDP plus the change in debt

Fig­ure 9 cor­re­lates the change in debt with the unem­ploy­ment rate: the cor­re­la­tion has the expect­ed sign, and is sig­nif­i­cant­ly larg­er than that for either M1 or M1-M0 (up to ‑0.85 for a lag of 15 months; see Table 2 on page 11).

Fig­ure 9: Change in pri­vate debt and unem­ploy­ment

Table 2: Lagged cor­re­la­tions between changes in mon­e­tary vari­ables and unem­ploy­ment

This brings us to the cur­rent finan­cial cri­sis, and why—on the analy­sis of those who pre­dict­ed it—this cri­sis is far from over.

The Great Moderation and the Great Recession

As is well known, a major top­ic in main­stream macro­eco­nom­ic debate was explain­ing the sources of “The Great Mod­er­a­tion” (Bernanke (2004); see also Davis and Kahn (2008) and Gali and Gam­bet­ti (2009)). Now the focus­ing is on explaining—and hope­ful­ly escap­ing from—“The Great Reces­sion” that fol­lowed it. From the point of view of the Beze­mer-Full­brook Group, these two events have the same cause, as did the Roar­ing Twen­ties and the Great Depres­sion before them: a debt-dri­ven spec­u­la­tive boom, fol­lowed by a delever­ag­ing-dri­ven down­turn. As an aside, there is no doubt that Ben Bernanke is apply­ing the lessons he took from the Great Depres­sion in his attempts to avoid a sec­ond such cri­sis. Fig­ure 10 shows the annu­al rate of change of M0, M1, M2 and M3 (which the Fed stopped record­ing in 2006) over the two decades since 1990. Growth in M0 is off the scale from late 2008 until ear­ly 2010, as the Fed­er­al Reserve more than dou­bled the lev­el of base mon­ey (the aver­age annu­al rate of growth of M0 over this peri­od exceed­ed 100%). How­ev­er, as dur­ing the Great Depres­sion, broad­er mea­sures of the mon­ey sup­ply are fail­ing to respond as dra­mat­i­cal­ly. While there has been growth in M1, it has been one sixth that of M0 (and mea­sured M0 now exceeds M1); M2’s growth nev­er exceed­ed 10 per­cent and is now close to zero; it is like­ly that growth in the now unrecord­ed M3 is either ane­mic or neg­a­tive.

Fig­ure 10: Changes in US mon­ey stock lev­els since 1990

Changes in debt, on the oth­er hand, were dra­mat­ic. Fig­ure 11 and Fig­ure 12 are the mod­ern equiv­a­lent of Fig­ure 7 and Fig­ure 8 from the 1920s-1940s, and it is evi­dent that the lev­el of debt-financed demand dur­ing “the Great Mod­er­a­tion” far exceed­ed the lev­el of The Roar­ing Twen­ties.

Fig­ure 11: US pri­vate debt and GDP 1990–2010

Fig­ure 12: US aggre­gate demand 1990–2010

How­ev­er, the neg­a­tive con­tri­bu­tion from delever­ag­ing has yet to approach the max­i­mum lev­els set dur­ing the Great Depres­sion. The debt con­tri­bu­tion to demand began at a far high­er lev­el this time than last, peak­ing at 22% in 2008 ver­sus at 8.7% in 1928. On this basis, the fall in aggre­gate demand caused by pri­vate sec­tor delever­ag­ing today is more rapid than in the 1930s, has not reached the max­i­mum rate sus­tained dur­ing the Great Depres­sion, and shows no signs of abat­ing. Fig­ure 13 graphs the debt con­tri­bu­tion to aggre­gate demand—defined as the annu­al change in debt, divid­ed by the sum of GDP plus the annu­al change in debt. On this mea­sure, delever­ag­ing since the peak lev­el of debt in 2008 has been more severe than delever­ag­ing from the com­pa­ra­ble lev­el in 1928, and the reduc­tion in demand from pri­vate sec­tor delever­ag­ing shows no signs of abat­ing.

Fig­ure 13: Com­par­ing the Great Depres­sion and the Great Reces­sion from both pri­vate and gov­ern­ment bor­row­ing

The rela­tion­ship between change in debt and unem­ploy­ment is also much stronger over the peri­od 1990–2010 (a cor­re­la­tion of ‑0.95) than it was for the peri­od from 1930 till 1940.

Fig­ure 14: Cor­re­la­tion of change in pri­vate debt and unem­ploy­ment, 1990–2010

Lags now make lit­tle dif­fer­ence to the results (pre­sum­ably because all data is now quar­ter­ly or month­ly in fre­quen­cy).

Table 3: Lags add no explana­to­ry pow­er to the cor­re­la­tions

On the indi­ca­tor of choice by the group that antic­i­pat­ed the GFC, it there­fore appears that the GFC is far from over. This rais­es one final empir­i­cal issue before I con­sid­er the the­o­ret­i­cal foun­da­tions of the Beze­mer-Full­brook Group: if pri­vate sec­tor delever­ag­ing this time round is falling from a greater lev­el and at a greater rate than in the 1930s, then why has the econ­o­my sta­bi­lized (to some extent) now, ver­sus the almost relent­less decline of the Great Depres­sion? The answer appears to lie in the scale of the gov­ern­ment response to this cri­sis. Fig­ure 13 (on page 14) includes the impact of gov­ern­ment debt on aggre­gate demand. While this added sub­stan­tial­ly to demand dur­ing the depths of the Great Depres­sion (more than 7% of aggre­gate demand between 1931 and 1933 was financed by gov­ern­ment debt), in 1930–2 years after the peak rate of growth of pri­vate debt in 1928—government debt added a mere 1.2% to aggre­gate demand. This time, giv­en the fear of anoth­er Great Depres­sion that pol­i­cy­mak­ers had in 2008, the gov­ern­ment response has been far larg­er and more imme­di­ate. In 2010, gov­ern­ment debt was respon­si­ble for over 12% of aggre­gate demand, and this almost coun­ter­act­ed the ‑15% con­tri­bu­tion from pri­vate sec­tor delever­ag­ing. While this suc­cess is heart­en­ing, the fig­ures indi­cate that growth can­not be expect­ed to con­tin­ue if the gov­ern­ment deficit is reduced. Pri­vate sec­tor delever­ag­ing is still accel­er­at­ing and has some time to go before it could be expect­ed to slow. In the absence of net gov­ern­ment spend­ing, it is high­ly like­ly that the down­ward spi­ral in out­put and employ­ment would con­tin­ue.

Non-mainstream modeling of the GFC

Only two of the econ­o­mists in the Beze­mer-Full­brook Group employ math­e­mat­i­cal macro­eco­nom­ic mod­els in their research: Wynne God­ley (God­ley (1999), God­ley and Izuri­eta (2002), God­ley and Izuri­eta (2004), God­ley and Lavoie (2007)) and myself (Keen (1995), Keen (2000), Keen (2008), Keen (2009a), Keen (2009b), Keen (2009d)). Both authors’ mod­els dif­fer sub­stan­tial­ly from main­stream “Neo­clas­si­cal” mod­els, in that they explic­it­ly eschew con­cepts of opti­miz­ing agents, work in aggrega­tive and gen­er­al­ly mon­e­tary terms, and dis­play non-equi­lib­ri­um-revert­ing dynam­ics.


God­ley’s mod­els employed an account­ing frame­work he described as “stock-flow con­sis­tent” on the basis that account­ing must be cor­rect, and only a lim­it­ed degree of ratio­nal­i­ty is need­ed to accu­rate­ly mod­el eco­nom­ic behav­ior:

Post Key­ne­sian eco­nom­ics … is some­times accused of lack­ing coher­ence, for­mal­ism, and log­ic. The method pro­posed here is designed to show that it is pos­si­ble to pur­sue het­ero­dox eco­nom­ics, with alter­na­tive foun­da­tions, which are more sol­id than those of the main­stream. The stock-flow mon­e­tary account­ing frame­work pro­vides such an alter­na­tive foun­da­tion that is based essen­tial­ly on two prin­ci­ples.

First, the account­ing must be right. All stocks and all flows must have coun­ter­parts some­where in the rest of the econ­o­my. The water­tight stock flow account­ing impos­es sys­tem con­straints that have qual­i­ta­tive impli­ca­tions. This is not just a mat­ter of log­i­cal coher­ence; it also feeds into the intrin­sic dynam­ics of the mod­el.

Sec­ond, we need only assume, in con­trast to neo­clas­si­cal the­o­ry, a very lim­it­ed amount of ratio­nal­i­ty on the part of eco­nom­ic agents. Agents act on the basis of their bud­get con­straints. Oth­er­wise, the essen­tial ratio­nal­i­ty prin­ci­ple is that of adjust­ment. Agents react to what they per­ceive as dis­e­qui­lib­ria, or to the dis­e­qui­lib­ria that they take note of, by mak­ing suc­ces­sive cor­rec­tions. There is no need to assume opti­miza­tion, per­fect infor­ma­tion, ratio­nal expec­ta­tions, or gen­er­al­ized price-clear­ing mech­a­nisms. (Lavoie and God­ley (2000, p. 307))

Fig­ure 15 gives an exam­ple of the God­ley social account­ing matrix.

Fig­ure 15: An exam­ple of the God­ley social account­ing matrix (God­ley (1999, p. 395))

Fig­ure 16 shows a sim­u­la­tion from this mod­el.

Fig­ure 16: Results of a sim­u­la­tion of the mod­el in God­ley (1999, p. 400)

Mod­els are assem­bled from the stock-flow matrix. For exam­ple, house­hold dis­pos­able income in nom­i­nal terms is defined as:

Fig­ure 17: Equa­tion for house­hold dis­pos­able income in God­ley (1999, p. 405)

These terms are tak­en from the House­holds col­umn of Fig­ure 15, and are the sum of income from prof­its, wages and inter­est pay­ments plus changes in house­hold bank bal­ances. On the basis of pro­jec­tions using this frame­work and US data, God­ley observed in 2002 and again in 2004 that the US seemed inevitably head­ed for a seri­ous reces­sion:

So the medi­um term alter­na­tives for the US econ­o­my look pret­ty stark. Either an uncon­venant­ed and sus­tained rise in net export demand pro­vides a motor for expan­sion in a quite new way; or the fis­cal pol­i­cy con­tin­ues to gen­er­ate twin (bud­get and bal­ance of pay­ments) deficits, pos­si­bly grow­ing …, or the US econ­o­my relaps­es into stag­na­tion. (God­ley and Izuri­eta (2004, p. 138))


I con­cur with God­ley on the need for accu­rate account­ing, and the capac­i­ty to mod­el macro­dy­nam­ic behav­ior with only real­is­tic lev­els of rationality—by which I mean that agents ratio­nal­ly react to their own sit­u­a­tion, but nei­ther know nor can pre­dict the char­ac­ter­is­tics or future behav­ior of the entire econ­o­my. They there­fore employ “rules of thumb” in their reac­tions to rel­e­vant eco­nom­ic stim­uli (where these reac­tions tend to be con­sis­tent with­in each social class, but dif­fer between class­es). The most impor­tant rule of thumb is that iden­ti­fied by Keynes in 1936:

The essence of this convention—though it does not, of course, work out quite so simply—lies in assum­ing that the exist­ing state of affairs will con­tin­ue indef­i­nite­ly, except in so far as we have spe­cif­ic rea­sons to expect a change. Keynes (1936, p. 152)

The assump­tion of a con­tin­u­ance of cur­rent con­di­tions in the high­ly non­lin­ear real world (and its sim­u­lacrum in these mod­els) leads to unex­pect­ed out­comes and hence changes in both the mod­el and the state of expec­ta­tions. I add to God­ley’s method an insis­tence on mod­el­ing social class­es dis­tinct­ly (I do not employ an aggre­gate house­hold sec­tor, but instead divide agents into at the min­i­mum cap­i­tal­ists, work­ers and bankers), and I work in con­tin­u­ous time rather than the more awk­ward if sim­pler frame­work of dis­crete time (see Keen (2006) and Keen (2008)). I employ two dif­fer­ent frame­works in my mod­els: an exten­sion of Good­win’s growth cycle mod­el (Good­win (1967)) to include debt (Keen (1995)), and a devel­op­ment of God­ley’s stock-flow con­sis­tent frame­work in con­tin­u­ous time (Keen (2008), Keen (2009a)). The for­mer set of mod­els gen­er­ate debt dynam­ics that, under some ini­tial con­di­tions, qual­i­ta­tive­ly approx­i­mate the debt accu­mu­la­tion data seen in Fig­ure 1. The mod­els are expressed as sys­tems of cou­pled ordi­nary dif­fer­en­tial equa­tions (Fig­ure 18).

Fig­ure 18: Equa­tions for the mod­el in Keen (2009d, p. 353)

In this sys­tem, the state vari­ables are respec­tive­ly GDP, the wage rate, pri­vate debt, the lev­el of spec­u­la­tive invest­ment as a frac­tion of GDP, labour pro­duc­tiv­i­ty, and pop­u­la­tion (g, ?r and ? are vari­ables rep­re­sent­ing the rate of growth, prof­it rate and the employ­ment rate that them­selves depend on the val­ues of the state vari­ables). With­out spec­u­la­tive borrowing—defined as bor­row­ing that finances spec­u­la­tion on asset prices but does not finance the con­struc­tion of new assets—the mod­el gen­er­ates a cycli­cal sys­tem which gen­er­al­ly does not break down; with spec­u­la­tive bor­row­ing, the mod­el almost inevitably approach­es a cri­sis caused by the accu­mu­la­tion of debt to lev­els that exceed the econ­o­my’s debt ser­vic­ing capac­i­ty.

Fig­ure 19: Results of a sim­u­la­tion with the mod­el in Keen (2009d)

Fig­ure 20: Results of a sim­u­la­tion with the mod­el in Keen (2009d) con­tin­ued

My own ver­sion of Stock-Flow con­sis­tent mod­el­ing uses a tab­u­lar approach to build­ing sys­tems of cou­pled dif­fer­en­tial equa­tions. The sys­tem states are bank accounts attrib­uted to spe­cif­ic social class­es, and the rows in each table reflects flows between accounts. The mod­els are derived sim­ply by sym­bol­i­cal­ly adding up the columns of the rel­e­vant matrix. An exam­ple sys­tem is shown in Table 4.

Table 4: A sam­ple table of finan­cial flows (Keen (2009c))

Bank Accounts

Assets (Reserves & Loans)

Lia­bil­i­ties (Deposits)

Actions Reserves (BR) Loans (FL) Firms (FD) Work­ers (WD) Banks (BI)
A Com­pound Inter­est


B Pay Inter­est




C Deposit Inter­est



D Wages



E Work­er Inter­est



F+G Con­sump­tion




H Loan Repay­ment




I Mon­ey relend­ing




J Mon­ey cre­ation



K Res­cue Banks


Res­cue Firms


The dynam­ics of the mod­el are sim­ply giv­en by adding up the sym­bol­ic entries in each col­umn (Table 5):

Table 5: Dynam­ics of the mod­el shown in Table 4

Rate of change of… Equals…
Bank Reserves BR H‑I+K
Firm Loans FL -H+I+J
Firm Deposits FD -B+C‑D+F+G‑H+I+J+K
Work­er Deposits WD D+E+F
Bank Income BI +B‑C-E‑G

Each of the sym­bols A to K is replaced by a suit­able oper­a­tor based on the 5 sys­tem states in the mod­el (Table 6):

Table 6: Sym­bol­ic sub­sti­tu­tions for the mod­el in Table 5

Action Descrip­tion Terms
A Com­pound Inter­est Out­stand­ing debt FL is increased at the rate of inter­est on loans rL.


B Pay Inter­est Accrued inter­est on out­stand­ing debt is paid. This involves a trans­fer from the firm sec­tor’s deposits FD to the bank sec­tor’s income account BI, and the record­ing of this trans­fer on the debt ledger FL.


C Deposit Inter­est Inter­est is paid (at the low­er rate rD) on the bal­ance in the firm sec­tor’s deposit account


D Wages This is a trans­fer from the firm sec­tor’s deposit accounts to work­ers’ deposit accounts WD, using two insights from Marx: first­ly that the sur­plus in pro­duc­tion is dis­trib­uted between work­ers and cap­i­tal­ists (in shares that sum to 1 in this model–so work­ers get 1‑s and cap­i­tal­ists get s); sec­ond­ly that there is a turnover peri­od (?S as a frac­tion of a year) between M and M+ (see Cap­i­tal II Chap­ter 12).


E Work­er Inter­est The deposit inter­est rate times the bal­ance in work­ers’ accounts.


F+G Con­sump­tion This employs the con­cept of a time lag–the length of time it takes work­ers to spend their wages is 2 weeks (say) or 1/26th of a year so that ?W equals 1/26. Wealth­i­er bankers spend their account bal­ances much more slow­ly.


H Loan Repay­ment The rate of loan repay­ment is pro­por­tion­al to the out­stand­ing lev­el of loans divid­ed by the time lag ?L in loan repay­ment (for a stan­dard hous­ing loan this would be shown as ?L =25)


I Mon­ey relend­ing The rate of new mon­ey cre­ation is the bal­ance in the bank­ing sec­tor’s unlent reserves, divid­ed by a turnover lag rep­re­sent­ing how rapid­ly exist­ing mon­ey is recy­cled.


J Mon­ey cre­ation The rate of new mon­ey cre­ation is the bal­ance in the firm sec­tor’s deposit account, divid­ed by a time lag that rep­re­sents the length of time it takes for the mon­ey sup­ply to dou­ble.


K Res­cue Banks This is a ‘Deus Ex Machi­na’ injec­tion of 100 cur­ren­cy units one year after the cri­sis begins, for a peri­od of one year, into either the bank­ing sec­tors reserves BR or the firm sec­tor’s deposit accounts FD.


Res­cue Firms

The mod­el’s equa­tions are as shown in

Fig­ure 21: Equa­tions for the mod­el sim­u­lat­ed in Keen (2009a) and Keen (2009c)

A sim­u­la­tion of a cred­it crunch and two dif­fer­ent pol­i­cy respons­es is shown in Fig­ure 22.

Fig­ure 22: Sim­u­la­tion of a cred­it crunch using the mod­el in Keen (2009c)


The core propo­si­tions shared by the Beze­mer-Full­brook group were that the super­fi­cial­ly good eco­nom­ic per­for­mance dur­ing “The Great Mod­er­a­tion” was dri­ven by a debt-financed spec­u­la­tive bub­ble, which would nec­es­sar­i­ly burst because the debt added to the econ­o­my’s ser­vic­ing costs with­out increas­ing its capac­i­ty to finance those costs. At some stage, the growth of unpro­duc­tive debt had to fal­ter, and when it did a seri­ous finan­cial cri­sis would ensue as aggre­gate demand col­lapsed. The pol­i­cy res­cues since that pre­dic­tion came true have not addressed the fun­da­men­tal cause of the cri­sis, which was the exces­sive lev­el of pri­vate debt. The delever­ag­ing that the Group pre­dict­ed has thus been slowed to some degree by gov­ern­ment action, but the need for that delever­ag­ing has not been removed. As Fig­ure 13 in par­tic­u­lar empha­sizes, the scale of that poten­tial delever­ag­ing appears cer­tain to exceed that expe­ri­enced in the Great Depres­sion.

Bat­telli­no, R. 2007, ‘Some Obser­va­tions On Finan­cial Trends’, Reserve Bank of Aus­tralia Bul­letin, vol. 2007, no. Octo­ber, pp 14–21.

Bernanke, B. S. 1995, ‘The Macro­eco­nom­ics of the Great Depres­sion: A Com­par­a­tive Approach’, Jour­nal of Mon­ey, Cred­it, and Bank­ing, vol. 27, no. 1, pp 1–28.

Bernanke, B. S. 2000, Essays on the Great Depres­sion, Prince­ton Uni­ver­si­ty Press, Prince­ton.

Bernanke, B. S. 2002, ‘Remarks by Gov­er­nor Ben S. Bernanke’, paper pre­sent­ed to Con­fer­ence to Hon­or Mil­ton Fried­man, Uni­ver­si­ty of Chica­go, Chica­go, Illi­nois, Novem­ber 8, 2002.

Bernanke, B. S. 2004, ‘The Great Mod­er­a­tion: Remarks by Gov­er­nor Ben S. Bernanke At the meet­ings of the East­ern Eco­nom­ic Asso­ci­a­tion, Wash­ing­ton, DC Feb­ru­ary 20, 2004’, paper pre­sent­ed to East­ern Eco­nom­ic Asso­ci­a­tion, Wash­ing­ton, DC.

Beze­mer, D. J. 2009, ’ ”No One Saw This Com­ing”: Under­stand­ing Finan­cial Cri­sis Through Account­ing Mod­els’, Fac­ul­ty of Eco­nom­ics Uni­ver­si­ty of Gronin­gen, Gronin­gen, The Nether­lands.

Davis, S. J. and Kahn, J. A. 2008, ‘Inter­pret­ing the Great Mod­er­a­tion: Changes in the Volatil­i­ty of Eco­nom­ic Activ­i­ty at the Macro and Micro Lev­els’, Jour­nal of Eco­nom­ic Per­spec­tives, vol. 22, no. 4, pp 155–180.

Fish­er, C. and Kent, C. 1999, ‘Two Depres­sions, One Bank­ing Col­lapse’, in Reserve Bank of Aus­tralia Research Dis­cus­sion Papers, vol. 1999, Reserve Bank of Aus­tralia, Syd­ney, NSW, Aus­tralia.

Fish­er, I. 1933, ‘The Debt-Defla­tion The­o­ry of Great Depres­sions’, Econo­met­ri­ca, vol. 1, no. 4, pp 337–357.

Fried­man, M. and Schwartz, A. J. 1963, A mon­e­tary his­to­ry of the Unit­ed States 1867–1960, Prince­ton Uni­ver­si­ty Press, Prince­ton.

Full­brook, E. 2010, ‘Keen, Roubi­ni and Bak­er win Revere Award for Eco­nom­ics’, in Real World Eco­nom­ics Review Blog, vol. 2010, ed E. Full­brook, Real World Eco­nom­ics Review, New York.

Gali, J. and Gam­bet­ti, L. 2009, ‘On the Sources of the Great Mod­er­a­tion’, Amer­i­can Eco­nom­ic Jour­nal: Macro­eco­nom­ics, vol. 1, no. 1, pp 26–57.

God­ley, W. 1999, ‘Mon­ey and Cred­it in a Key­ne­sian Mod­el of Income Deter­mi­na­tion’, Cam­bridge Jour­nal of Eco­nom­ics, vol. 23, no. 4, pp 393–411.

God­ley, W. and Izuri­eta, A. 2002, ‘The Case for a Severe Reces­sion’, Chal­lenge, vol. 45, no. 2, pp 27–51.

God­ley, W. and Izuri­eta, A. 2004, ‘The US Econ­o­my: Weak­ness­es of the ‘Strong’ Recov­ery’, Ban­ca Nazionale del Lavoro Quar­ter­ly Review, vol. 57, no. 229, pp 131–139.

God­ley, W. and Lavoie, M. 2007, Mon­e­tary Eco­nom­ics: An Inte­grat­ed Approach to Cred­it, Mon­ey, Income, Pro­duc­tion and Wealth, Pal­grave Macmil­lan, New York.

Good­win, R. 1967, ‘A growth cycle’, in Social­ism, Cap­i­tal­ism and Eco­nom­ic Growth, ed C. H. Fein­stein, Cam­bridge Uni­ver­si­ty Press, Cam­bridge.

Keen, S. 1995, ‘Finance and Eco­nom­ic Break­down: Mod­el­ing Min­sky’s ‘Finan­cial Insta­bil­i­ty Hypoth­e­sis.”, Jour­nal of Post Key­ne­sian Eco­nom­ics, vol. 17, no. 4, pp 607–635.

Keen, S. 2000, ‘The Non­lin­ear Eco­nom­ics of Debt Defla­tion’, in Com­merce, com­plex­i­ty, and evo­lu­tion: Top­ics in eco­nom­ics, finance, mar­ket­ing, and man­age­ment: Pro­ceed­ings of the Twelfth Inter­na­tion­al Sym­po­sium in Eco­nom­ic The­o­ry and Econo­met­rics, ed W. A. Bar­nett, Cam­bridge Uni­ver­si­ty Press, New York.

Keen, S. 2006, ‘The Need and Some Meth­ods for Dynam­ic Mod­el­ling in Post Key­ne­sian Eco­nom­ics’, in Com­plex­i­ty, Endoge­nous Mon­ey and Macro­eco­nom­ic The­o­ry: Essays in Hon­our of Basil J. Moore, ed M. Set­ter­field, Chel­tenham, U.K. and Northamp­ton, Mass: Elgar.

Keen, S. 2008, ‘Key­nes’s ‘revolv­ing fund of finance’ and trans­ac­tions in the cir­cuit’, in Keynes and Macro­eco­nom­ics after 70 Years, eds R. Wray and M. Forstater, Edward Elgar, Chel­tenham.

Keen, S. 2009a, ‘Bail­ing out the Titan­ic with a Thim­ble’, Eco­nom­ic Analy­sis & Pol­i­cy, vol. 39, no. 1, pp 3–24.

Keen, S. 2009b, ‘The dynam­ics of the mon­e­tary cir­cuit’, in The Polit­i­cal Econ­o­my of Mon­e­tary Cir­cuits: Tra­di­tion and Change, eds S. Rossi and J.-F. Pon­sot, Pal­grave Macmil­lan, Lon­don.

Keen, S. 2009c, ‘The Glob­al Finan­cial Cri­sis, Cred­it Crunch­es and Delever­ag­ing’, Jour­nal Of Aus­tralian Polit­i­cal Econ­o­my, vol. 64, pp 18–32.

Keen, S. 2009d, ‘House­hold Debt-the final stage in an arti­fi­cial­ly extend­ed Ponzi Bub­ble’, Aus­tralian Eco­nom­ic Review, vol. 42, pp 347–357.

Keynes, J. M. 1936, The gen­er­al the­o­ry of employ­ment, inter­est and mon­ey, Macmil­lan, Lon­don.

Kir­man, A. 1989, ‘The Intrin­sic Lim­its of Mod­ern Eco­nom­ic The­o­ry: The Emper­or Has No Clothes’, Eco­nom­ic Jour­nal, vol. 99, no. 395, pp 126–139.

Kyd­land, F. E. and Prescott, E. C. 1990, ‘Busi­ness Cycles: Real Facts and a Mon­e­tary Myth’, Fed­er­al Reserve Bank of Min­neapo­lis Quar­ter­ly Review, vol. 14, no. 2, pp 3–18.

Lavoie, M. and God­ley, W. 2000, ”Kaleck­ian Mod­els of Growth in a Stock-Flow Mon­e­tary Frame­work: A Neo-Kaldo­ri­an Mod­el,”, Levy Eco­nom­ics Insti­tute, The, Eco­nom­ics Work­ing Paper Archive.

Min­sky, H. P. 1982, Can “it” hap­pen again? : essays on insta­bil­i­ty and finance, M.E. Sharpe, Armonk, N.Y.

Moore, B. J. 1979, ‘The Endoge­nous Mon­ey Stock’, Jour­nal of Post Key­ne­sian Eco­nom­ics, vol. 2, no. 1, pp 49–70.

Bookmark the permalink.

About Steve Keen

I am Professor of Economics and Head of Economics, History and Politics at Kingston University London, and a long time critic of conventional economic thought. As well as attacking mainstream thought in Debunking Economics, I am also developing an alternative dynamic approach to economic modelling. The key issue I am tackling here is the prospect for a debt-deflation on the back of the enormous private debts accumulated globally, and our very low rate of inflation.