DebtWatch No 29 December 2008

flattr this!

What's Really Going On? or...

Why Did I See it Coming and "They" Didn't?

Part 2: The Models

"But this long run is a misleading guide to current affairs. In the long run we are all dead. Econ­o­mists set them­selves too easy, too use­less a task if in tem­pes­tu­ous sea­sons they can only tell us that when the storm is long past the ocean is flat again.” (Keynes, A Tract on Mon­e­tary Reform, 1924)

In last month’s Debt­watch, I explained why the data side of why the “Finan­cial Insta­bil­ity Hypoth­e­sis” enabled me to pre­dict this cri­sis, long before con­ven­tional “neo­clas­si­cal” econ­o­mists had any idea it was approaching.

This month I explain why the mod­els neo­clas­si­cal econ­o­mists build are hope­lessly inadequate–as mod­els of the econ­omy in gen­eral, as guides to what is likely to hap­pen in the future, and as sources of pol­icy rec­om­men­da­tions to end this crisis.

In par­tic­u­lar, the Aus­tralian Treasury’s pre­dic­tion that Aus­tralia will avoid reces­sion sim­ply can­not be trusted.

Neo­clas­si­cal Mod­els: Cri­sis? What Crisis?

The focus of neo­clas­si­cal econ­o­mists on mis­lead­ing indi­ca­tors is com­pounded by the mod­els they build, which–as well as omit­ting cru­cial data like the debt to GDP ratio–are con­gen­i­tally inca­pable of iden­ti­fy­ing seri­ous turn­ing points in the economy.

This are sev­eral rea­sons for this, but first and fore­most is their belief that the econ­omy is fun­da­men­tally sta­ble, and will always return to a long-run equi­lib­rium growth path after any shock. The mod­els they con­struct have this expec­ta­tion of a return to long-term growth paths after any short term diver­gence from trend “hard wired” into their results.

An instance of this for the Aus­tralian Treasury’s macro­eco­nomic model (TRYM) is shown in Fig­ure One, which shows the impact on busi­ness invest­ment in the model of a sim­u­lated mon­e­tary shock in 2010. The shock ini­tially pushes busi­ness invest­ment above the long term trend, to which it then returns after eight years.

This is not a pre­dic­tion by the model as such, but a prod­uct of its struc­ture, which assumes that the econ­omy will always return to a supply-side dri­ven equi­lib­rium in a rel­a­tively short time frame.

Fig­ure One

TRYM’s supply-side behav­iour is deter­mined sim­ply by the assump­tion that, in the long run, the econ­omy will return to an equi­lib­rium rate of growth, given by the sum of assumed trends in pop­u­la­tion growth and labour pro­duc­tiv­ity, at an assumed equi­lib­rium rate of unem­ploy­ment called “NAIRU” (“Non-Accelerating Infla­tion Rate of Unem­ploy­ment”). As the Treasury’s doc­u­men­ta­tion of TRYM puts it (see

The model could be described as broadly new Key­ne­sian in its dynamic struc­ture but with an equi­li­brat­ing long run.  Activ­ity is demand deter­mined in the short run but sup­ply deter­mined in the long run… The model will even­tu­ally return to a sup­ply deter­mined equi­lib­rium growth path in the absence of demand or other shocks.” (THE MACROECONOMICS OF THE TRYM MODEL OF THE AUSTRALIAN ECONOMY, p. 6; empha­sis added)


the aggre­gate sup­ply curve is ver­ti­cal in the long term at a level of employ­ment and pro­duc­tion con­sis­tent with the NAIRU. (Or more pre­cisely the econ­omy grows along a steady state growth path con­sis­tent with the NAIRU.)” [AN INTRODUCTION TO THE TRYM MODEL APPLICATIONS AND LIMITATIONS p. 6]

Neo­clas­si­cal mod­els like TRYM are thus vari­able in the present–and have some capac­ity to pre­dict the very short term, if their guesses about the size of any shock are rea­son­ably accu­rate. But they are anchored to some point in the (not too dis­tant) future when it is assumed that “equi­lib­rium” will once again apply, and they are there­fore use­less as guides in the medium term.

They are also use­less for long term pre­dic­tion because the model’s long run equi­lib­rium is unaf­fected by the short term dis­tur­bance: if the fig­ure assumed for the NAIRU in the model remains unchanged, along with the esti­mates for pop­u­la­tion and pro­duc­tiv­ity growth, then the model will aver­age the rate of growth those assump­tions imply, regard­less of how severe a shock the short-term dis­tur­bance causes.

In the case of the RBA’s main model, this is a real growth rate of 3.25 per­cent per annum (see Tables 7 & 8 of RDP2005-11)–so the econ­omy is assumed to con­verge to a tran­quil future path after any dis­tur­bance, with no residue from the shock itself (apart from a change in the price level for per­ma­nent increases in the money supply).

Iron­i­cally, this means that mod­els like TRYM pro­duce medium term pre­dic­tions of an accel­er­a­tion in growth after the impact of a shock like this finan­cial crisis–otherwise the model could not get back to its “long run equi­lib­rium growth path”.

There was thus no prospect that Neo­clas­si­cal mod­els could pre­dict the cri­sis, and their guid­ance on what will happen–with or with­out pol­icy intervention–are irrel­e­vant.  Unlike these mod­els, the actual econ­omy does not have a point of bal­ance in the future to which it is teth­ered. It is there­fore no won­der that these mod­els gave no warn­ing of the impend­ing crisis–indeed the won­der would be if they had done so!

This is why sup­pos­edly author­i­ta­tive bod­ies like the OECD could claim “our cen­tral fore­cast remains indeed quite benign” just two months before all hell broke loose (as noted in my last Debt­watch). If eco­nomic data have been appar­ently tran­quil, these mod­els will pre­dict tran­quil­ity ahead; if the data have been depressed, they will pre­dict a bit of a down­turn, fol­lowed by a return to equi­lib­rium some years hence.

Nei­ther pre­dic­tion is worth a pinch of salt.

To have any hope of pre­dict­ing the future using an eco­nomic model, it has to be one with gen­uine dynamics–not a model that sim­ply assumes that “when the storm is long past the ocean is flat again”, as Keynes satir­i­cally remarked. Such a model has to spec­ify what it sees as the main causal fac­tors in the econ­omy, and then let those fac­tors inter­act. The medium and long term out­comes are thus a prod­uct of the inter­ac­tion of the causal vari­ables in the model, just as the short term is.

Mod­els of this nature are com­mon­place out­side eco­nom­ics, and sci­en­tists, math­e­mati­cians and engi­neers have designed an impres­sive range and vari­ety of com­puter sim­u­la­tion pro­grams to sup­port this gen­uinely dynamic approach to modelling.

I devel­oped such a model of Minsky’s Finan­cial Insta­bil­ity Hypoth­e­sis in the early 1990s.

A Min­sky Model: Finance and Eco­nomic Breakdown

The basic prin­ci­ples in Minsky’s finan­cial insta­bil­ity hypoth­e­sis are extremely sim­ple. A cap­i­tal­ist econ­omy is nec­es­sarly cycli­cal. Dur­ing a boom, investors will take on debt to finance invest­ment, but because the econ­omy is cycli­cal, they will later find them­selves in a reces­sion when they have to repay that debt.

There­fore their repay­ments don’t quite can­cel all the extra debt, and debt lev­els tend to ratchet up over time. These debt cycles with an over­all sec­u­lar trend towards increas­ing debt can lead to an ulti­mate cri­sis where the debt over­whelms the economy–a Depression.

This is not an inevitable out­come of Minsky’s the­ory, but he empha­sises that since mar­ket economies have expe­ri­enced Depres­sions in the past, to be valid a model of the econ­omy must…

make great depres­sions one of the pos­si­ble states in which our type of cap­i­tal­is­te­con­omy can find itself”  (Min­sky, 1982, Infla­tion, Reces­sion and Eco­nomic Pol­icy, p. xi)

In the model I devel­oped in 1993, under some cir­cum­stances, the econ­omy could taper to equi­lib­rium; but under oth­ers, a series of debt-driven finan­cial cycles would lead to an even­tual cri­sis where debt over­whelmed the econ­omy. The fol­low­ing graph­ics set out the model in flow­chart for­mat. It can also be sum­marised in three very sim­ple propositions:

  1. Firms bor­row to invest dur­ing booms;
  2. Work­ers’ capac­ity to secure wages rises is affected by the rate of employ­ment; and
  3. Banks lend money to finance investment;

and four very sim­ple “stylised facts”:

  1. Wages share of out­put will rise if wage rises exceed productivity;
  2. The employ­ment rate will rise if the rate of growth exceeds the sum of pop­u­la­tion  and pro­duc­tiv­ity growth;
  3. The debt to GDP ratio will rise if invest­ment exceeds prof­its; and
  4. an increased rate of eco­nomic growth will reduce the debt to GDP ratio.

As a flow­chart, the model is as shown in Fig­ure Two (the blue boxes con­tain math­e­mat­i­cal sub-systems).

The sim­u­la­tion below and in Fig­ure Three are with no debt in the model–in which case the model gen­er­ates sim­ple cycli­cal growth.

Fig­ure Two

Fig­ure Three explodes the “Graph” sub­sys­tem of the model. The same set of graphs is used in sub­se­quent Fig­ures to dis­play the behav­iour of the more com­plete mod­els, where debt and Ponzi invest­ing are added.

Fig­ure Three

When the debt switch” is flicked to include bor­row­ing to finance pro­duc­tive invest­ment only–so all bor­rowed money leads to an increase in the capac­ity to pro­duce output–then one of two sit­u­a­tions will apply.

Fig­ure Four shows the first such sit­u­a­tion: when the model begins close to its equi­lib­rium val­ues, it con­tin­ues to con­verge towards it. Employ­ment and income dis­tri­b­u­tion (prox­ied here by the wages share of out­put) taper to equi­lib­rium val­ues, as does the debt to out­put ratio (which is neg­a­tive, imply­ing pos­i­tive net finan­cial assets for firms).

Fig­ure Four

How­ever, if the sys­tem starts fur­ther away from equi­lib­rium, then the system’s behav­iour is rather like that described by Fisher in his Debt Defla­tion The­ory of Great Depressions:

There may be equi­lib­rium which, though sta­ble, is so del­i­cately poised that, after depar­ture from it beyond cer­tain lim­its, insta­bil­ity ensues, just as, at first, a stick may bend under strain, ready all the time to bend back, until a cer­tain point is reached, when it breaks.

This sim­ile prob­a­bly applies when a debtor gets “ broke,” or when the break­ing of many debtors con­sti­tutes a “ crash,”  after which there is no com­ing back to the orig­i­nal equilibrium.

To take another sim­ile, such a dis­as­ter is some­what like the “ cap­siz­ing”  of a ship which, under ordi­nary con­di­tions, is always near sta­ble equi­lib­rium but which, after being tipped beyond a cer­tain angle, has no longer this ten­dency to return to equi­lib­rium, but, instead, a ten­dency to depart fur­ther from it.” (Fisher, 1933)

With this far from equi­lib­rium start­ing point, the sys­tem goes through a series of cycles in which the debt to out­put ratio ratch­ets up, as Min­sky sur­mised, until such time that the next boom leads to such an accu­mu­la­tion of debt that it can­not be repaid–debt ser­vice con­sumes all avail­able revenues–and the econ­omy falls into a per­ma­nent slump.

Fig­ure Five

I have com­mented fre­quently that econ­o­mists are pris­on­ers of their models–rather than see­ing the econ­omy, they see their model of it. Though I dif­fer from the neo­clas­si­cal main­stream in the type of model I see, on this front I was not very dif­fer­ent. I there­fore expected to find a pat­tern like that shown for the debt to out­put ratio in Fig­ure Five in the Aus­tralian data, when I pre­pared an expert wit­ness case for the NSW Legal Aid in Decem­ber 2005: a grad­ual hump-like increase in debt to out­put ratios.

Instead what I saw was the pat­tern shown in Fig­ure Six.

Fig­ure Six

That was an almost purely expo­nen­tial increase in the debt to GDP ratio over time–disturbed only by the growth and burst­ing of two obvi­ous super-bubbles (one in the early 1970s that was asso­ci­ated with the demise of the Whit­lam gov­ern­ment, and the other that drove Keating’s “reces­sion we had to have” in the early 1990s).

It was obvi­ous that a key aspect of Minsky’s the­ory that my model omit­ted had to be intro­duced: Ponzi invest­ing, in which indi­vid­u­als take out debt to spec­u­late on asset prices, but don’t actu­ally build any assets in the process. I intro­duced this into the model by adding a spec­u­la­tive debt com­po­nent, where bor­row­ing for spec­u­la­tion rose when­ever the rate of growth exceeded a min­i­mum level. With that mod­i­fi­ca­tion, the pat­tern shown in Fig­ure Seven resulted.

Fig­ure Seven

This model gen­er­ates a gen­er­ally expo­nen­tial increase in debt lev­els, with super-bubbles in spec­u­la­tive bor­row­ing occur­ring reg­u­larly, and bor­row­ing to finance spec­u­la­tion grad­u­ally accel­er­ates to ulti­mately dwarf bor­row­ing for pro­duc­tive investment.

At some point the debt bur­den becomes too great for the econ­omy to finance, and debt accu­mu­lates faster than it is repaid, lead­ing to a sec­u­lar cri­sis and not merely a finan­cial cycle. Guided both by Minsky’s hypoth­e­sis and my math­e­mat­i­cal mod­els, I felt that we were at such a sec­u­lar turn­ing point in the real world when I saw the data in Fig­ure Six (three years ago, in Decem­ber 2005).

I feared that Australia–and prob­a­bly the rest of the world–was in for a seri­ous debt-induced down­turn. Know­ing that there was lit­tle if any like­li­hood that this dan­ger would be per­ceived by the neoclassically-trained econ­o­mists who dom­i­nate Trea­suries and Cen­tral Banks (and Uni­ver­sity Eco­nom­ics Depart­ments) around the world. I decided to go pub­lic with my analysis

This was more than con­firmed when RBA Deputy Gov­er­nor Ric Bat­tellino pub­lished a graph show­ing Australia’s long term debt to GDP ratio dur­ing a speech in Sep­tem­ber 2007 (see Fig­ure 8, which is aug­mented to include esti­mates of non-bank credit prior to 1953). 

Fig­ure Eight

The model I’ve out­lined above is extremely sim­ple, and would need to be sub­stan­tially embell­ished to cap­ture the main dynam­ics of a mar­ket econ­omy. But it is already streets ahead of neo­clas­si­cal mod­els by not mak­ing an arti­fi­cial dis­tinc­tion between the short and long term. To para­phrase Keynes, “in the long run we are still in the short run”.

Avoid­ing Recession?

The Aus­tralian Gov­ern­ment is almost unique amongst OECD nations in pre­dict­ing pos­i­tive growth dur­ing 2009. Indeed, only ten coun­tries are hold­ing out for pos­i­tive real growth next year in the OECD’s recent Eco­nomic Out­look are Aus­tralia (1.7%), the Czech Repub­lic (2.7%), Greece (1.9%), Korea (2.7%), Mex­ico (0.4%), Nor­way (1.3%), Poland (3%), the Slo­vak Repub­lic (4%), Swe­den (0%), and Turkey (1.6%) (see Fig­ure Nine, taken from the OECD Eco­nomic Out­look No. 84 for Novem­ber 2008, p. 82). The other nine­teen coun­tries all expect to record neg­a­tive growth.

These “pre­dic­tions” should be seen for what they are–not so much pre­dic­tions as assump­tions of a class of eco­nomic mod­els that has lit­tle con­nec­tion with the real world. The scale of the down­turn “pre­dicted” for 2009 largely reflects the judg­ments of national Treasuries–including Australia’s–as to how severe a shock the finan­cial cri­sis represents.

On that scale, the only coun­try giv­ing this finan­cial cri­sis seri­ous weight is Ice­land, which is esti­mat­ing its dam­age as equiv­a­lent to about 14% of GDP–represented by the change between the growth rate recorded for 2007 and that expected for 2009. Australia’s Trea­sury has appar­ently per­suaded the OECD that this cri­sis will knock only a cou­ple of per­cent off growth.

Fig­ure Nine

Notice also that the OECD expects growth to dra­mat­i­cally improve in 2010–even Ice­land is expected to almost return to pos­i­tive growth for the cal­en­dar year as a whole, while in the 4th quar­ter it is expected to record pos­i­tive growth at an annual rate of 2.6% (see Fig­ure Ten). Aus­tralia is expected to rebound to 3.1% annu­alised growth by the last quar­ter of cal­en­dar year 2010.

Why? Because by that stage into the future, the “long run” in the OECD’s neo­clas­si­cal model of the econ­omy starts to reassert itself, and every econ­omy is pre­dicted to boom away, to erase the impact of the “tem­po­rary” shock of the 2008 finan­cial crisis.

I’d love to see the OECD’s pre­dic­tions for Ice­land for 2011 and 2012–they should show mas­sive growth to over­come the impact of the –9.3% fig­ure for next year, and return to the long run equi­lib­rium assump­tion the model makes for Iceland–which is prob­a­bly above 5% p.a. Aus­tralia is also likely to be shown to enjoy a rosy 4% or above rate of growth.

These pre­dic­tions, and those of any other neo­clas­si­cal model–including the Treasury’s TRYM model, and the RBA’s small model–are no more than a state­ment of faith in the long run sta­bil­ity of a mar­ket economy.

I expect that faith will be sorely tested in the com­ing years.

Fig­ure Ten


Com­ments on the Data

It appears that Australia’s debt to GDP ratio has peaked at 165% of GDP, and it is now start­ing to fall. It could still turn up once again if defla­tion takes hold, but for the mean­time, this seems to be the top of the bubble.

Now as debt lev­els start to fall–firstly rel­a­tively to GDP and then, ulti­mately, in absolute terms as well–the macro­eco­nomic effect of the bubble’s burst­ing will be felt. This trend appears in the next two graphs, which show the annual rate of change of debt and GDP, and the con­tri­bu­tion that change in debt makes to aggre­gate demand (which I define as the sum of GDP and the change in debt).

One intrigu­ing aspect of the next two charts is the fact that the rate of growth of debt has fallen over time–from a peak of over 34% year on year for the 1973 bub­ble, to 25% for the 1989 bub­ble, and 17% for the bub­ble that has now peaked in 2008–but the con­tri­bu­tion that ris­ing debt makes to demand has risen–from just over 10% in 1973, to 14% in 1989 and 19.5% in 2008.

This appar­ent para­dox is the result of the increas­ing scale of debt com­pared to GDP. Back in 1972, when the first debt super-bubble began, debt was equiv­a­lent to only 33% of GDP–and there­fore a 1/3rd increase in debt, while dra­matic, only increased the ulti­mate debt to GDP ratio by 11%. In mid-1984, when the sec­ond super­bub­ble took off, debt was already 54% of GDP, and there­fore the slower rate of growth of debt resulted in the debt to GDP ratio ris­ing vt 57% by the time the bub­ble burst.

This time round, though debt grew by a max­i­mum of only 17% in one year, the debt super­bub­ble which began in mid 1993 increased the debt to GDP ratio by a stag­ger­ing 109%, from 79% at its com­mence­ment to 165.43% by its peak.

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.
Bookmark the permalink.

52 Responses to DebtWatch No 29 December 2008

  1. Steve Keen says:

    That’s my cur­rent research project hmt,

    and the model in the “Thim­ble” paper on the Research page is a first pass at this. I’ll be going much fur­ther in the book I hope to get back to writ­ing next year, once this crazy semes­ter is over.

  2. Pingback: An interesting challenge | Centre For Economic Stability

Leave a Reply