A great deal of chatter on the blogosphere on a paper by Negro, Giannoni and Schorfheide of the NY Fed that DGSE can predict the great recession. Noahopinion discusses it the day after he mused about what use was DGSE.
What they do is take the most well known New Keynsian model Smets-Wouters (2007) New Keynesian model and add on the the “financial accelerator” model of Bernanke, Gertler, and Gilchrist (1999). In the Financial Accelerator model credit shocks transmit through the real economy through amongst other reasons undermining the value of collatoral.
Both Noah and Mark Thoma’s reaction is ‘pah’ we could have predicted the Great Recession all along, we knew how we just didnt put two and two together’. But the Negro et al. is not a ‘forecast’ had they had the model in 2007 they would not have predicted the Great Recession. This quite apart from the criticism made by some commentators that they have engaged in post-hoc calibration of parameters to fit the result. I don’t make that accusation simply that the result forecasts nothing because their baseline data
begin[s] by estimating the DSGE model… based on macroeconomic data from 1964:Q1 to 2008:Q3. ..A forecaster in the fourth quarter of 2008 would have observed the Lehman collapse and have had access to current-quarter information on the federal funds rate and the spread (pg 13)
That is they take a pre-existing credit shock and project its results forward. If the Credit Accelerator theory is correct then it is no surprise that the results are broadly accurate. What the model does not do and cannot do is predict the causes of the Lehman collapse, the financial conditions that caused the ‘shock’ and overall fragility in the system (implicit in the baseline data). In other words as a predictor of the recession it has no value, as a predictor of events AFTER the commencement of the great recession it has some value.
So we are back in the mathematical dark ages introduced to Economics by Friusch, where steady cycles can only be disturbed by ‘Shocks’ – such as too many people going on holiday at once. The rest of the applied mathematical modelling community has long abandoned this approach for a systems dynamics based approach where modelling of state can create limit cycles and chaotic breakdown of stable systems (see this paper for the mistake Frisch made in criticising Kalecki’s dynamics).
Moreover even if the Financial Accelerator theory is correct the DGSE approach cannot tell us about the direction and strength of causation. This Bayoumi and Darius paper argues convincingly
instead of changes in the value of wealth driving bank lending, our empirical results imply that in general changes in willingness of banks to lend (often associated with changes in bank capital used to provide a cushion against lending) drive wealth.
So get over DGSE and properly model mathematical state, including net credit and debt of agents. Mark and Noah, get around to reading some Wyne Godley and begin to understand why such stock-flow consistent models did predict the great recession.