Macroeconomic Schisms

There has been a lot of breast-beating in the press and in the blogosphere about how economists failed to discern the possibility that not all was going well in the years leading up the current financial and economic crisis [1]. I think the notion that all economists were blithely optimistic has been dispelled (well, okay, here’s a couple of exceptions: Dan Gross h/t Free Exchange, A. Kaletsky). At the risk of some gross simplifications, I will speculate that there was — until recently — less optimism among academic macroeconomists than Wall Street economists. There was probably less anxiety among say finance professors who focused on asset pricing (as opposed those who worked in banking) than macroeconomists (Dani Rodrik highlights the diversity). One divide that I think is not particularly relevant in locating the source of the crisis is the most well known one — specifically whether prices are sticky.

In my opinion, the big divide in thinking relates to how economists conceive of financial markets working. This is a divide that cuts across other divides. For instance, the Hicksian decomposition (IS-LM), in its simplest incarnation, treats the financial world as one wherein bonds are identical, and the only means of borrowing; there is no separate channel for lending, say via bank loans, to influence aggregate demand (see this post for the many channels of monetary policy). In the real business cycle literature, and many New Keynesian DSGE models, there is a representative bond (and lending rate) which summarizes the asset markets (see Camilo Tovar’s survey of DSGEs for a discussion).

I think these types of models — both New Keynesian and RBC — are useful for thinking about many aspects of macroeconomic behavior. In fact, typically, the models I work with in my academic research on for instance exchange rates incorporate such assumptions. However, in thinking about the causes of the current financial and economic crisis, and the appropriate policy response, it seems to me it is essential to utilize models that either incorporate, or allow for, the lending channel via the financial system to be important. In other words, models that do not have some sort of financial accelerator and role for banks are likely to provide highly misleading policy conclusions (so the debate is not “merely academic”).dbcifig7.gif
Figure 7 from M. Biggs & T. Mayer, “Credit impulse: Lessons from the Great Depression,” Deutsche Bank Global Macro Issues, 13 March 2009 [not online]. Credit Impulse = (ΔCt/GDPt)-(ΔCt-1/GDPt-1). See M. Biggs, “The impact of credit on growth,” Deutsche Bank Global Macro Issues, 19 Nov 2008 [not online]Who’s been thinking about these sort of issues? In one of his recent columns, David Warsh mentions the work of Geanakoplos, as if the issues of leverage and credit cycles are new. It’s true that some of Geanakoplos’s ideas are new (especially the focus on leverage), but the theoretical modeling of how borrowing in the presence of agency issues has been around for a long time. This assertion can be easily verified by consulting the reference list in this (recent) review in the 2007 NBER Macro Annual, essentially a set of notes by Kiminori Matsuyama, “Aggregate Implications of Credit Market Imperfections”. This point validates Dani Rodrik‘s point about how many economists saw impending trouble: “Bob Shiller, Nouriel Roubini, Raghu Rajan”. To this list, I’d add George Akerlof, who foresaw in his 1993 Brookings paper with Paul Romer, the “looting” that would occur as regulations were relaxed in environment of pervasive asymmetric information (as I discussed in this April 2007 post).
How can one incorporate the idea of the importance of the banking system (separate from its importance to the money creation process)? One very simple static model is provided by a twenty year old paper by Bernanke and Blinder.Loan demand and loan supply:Ld = L(ρ, i, y)

∂L/∂ρ < 0, ∂L/∂i > 0, ∂L/∂y > 0

Ls = λ(ρ,i, Z)D(1-τ)

∂λ/∂ρ > 0, ∂λ/∂i < 0, ∂λ/∂Z < 0

Where ρ is the lending rate, i is the bond interest rate, y is income, τ is the reserve/deposit ratio, Z is the riskiness of the marginal investment project.

LM curve:

D(i,y)=mR

∂D/∂i < 0, ∂D/∂y > 0

Where D(.) is deposits, m is the money multipier and R is reserves.

Commodity-Credit (CC) curve [composite IS and credit equilibrium]:

y=Y(i,ρ)

∂Y/∂i < 0, ∂Y/∂ρ < 0

To solve for equilibrium, set loan demand equal to loan supply, after substituting the LM in for D:

L(ρ,i,y) = λ(ρ,i,Z)mR(1- τ)

Without knowing the parametric form of the functions, one can’t solve out analytically for ρ, but one knows in the end that:

ρ = φ (i, y, R, Z )

Substituting this solution for ρ into the expression for the revised IS curve incorporating a role for ρ and taking a total differential leads to the solution for dY:

dY = (1/Δ){[((Yρρi+Yi))m/Di )+YρρR]dR + YρρZdZ}

Where Δ ≡ [1-YρρY+((Yρρi+Yi)DY)/Di] and Yρ is ∂Y/∂ρ, and so forth.

So, the change in income is, given the posited parameter values, a positive function of R and m, and a negative function of Z. More math and context here [pdf]. Some of these effects are summarized in Table 1.

bbtable1.gifTable 1 from Bernanke and Blinder. I like to think in terms of graphs. Below I show how the CC and LM curves are shifted in due to increases in risk (Z going up), if the money multiplier falls (m declines).

schismpix1.gif

Figure 1: Shifts due to increase in project risk (Z rises) and decrease in money multiplier (m declines) (white arrows).An increase in reserves (or somehow decreasing excess reserves so m rises) will shift out both the CC and LM curves, as shown in Figure 2.

schismpix2.gif

Figure 2: Shifts due to increase in project risk (Z rises) and decrease in money multiplier (m declines) (gray arrows).This type of analysis also highlights how monetary policy can affect both the LM and CC curves; it also illustrates why fixing the banking sector is so important (increasing D increases lending).

The financial accelerator is not explicit in this interpretation, but it could be introduced in the following fashion. If Z is not exogenous, but depends on the level of income — so that declines in income raise Z, and vice versa, then a positive (in a technical sense) feedback loop can be introduced.

This modification would then illustrate why stimulating the economy (increasing GDP relative to what it otherwise would be) is important to fixing the banking sector. If the CC curve is shifted out, and Z depends upon the level of output, then the CC curve will be shifted out even further out, as shown in Figure 3.

schismpix3.gif

Figure 3: Holding monetary policy constant, increase government spending (gray arrow). Additional shift if Z is endogenous to output level (black arrow).The foregoing has been a static, extremely ad hoc, discussion. I’ve eschewed sophistication in favor of working in a framework that is somewhat familiar to those who’ve worked with IS-LM. For those who want a more sophisticated, dynamic, interpretation, see Chapter 21 of the Handbook of Macroeconomics, “The financial accelerator in a quantitative business cycle framework,”, by Ben S. Bernanke, Mark Gertler and Simon Gilchrist.

A logical conclusion from the foregoing discussion is that there were indeed many people who were concerned about the apparent buildup of financial sector imbalances. And these economists were typically not those steeped in the macro literature which assumed smoothly working (neoclassical) financial markets. They were the ones concerned with how some markets are not continuously cleared by the Walrasian auctioneer.

Hence, as we consider how to get out of this mess, we should pay more attention to the economists who warned of trouble, than those who viewed the economy as running smoothly (e.g.). So I’ll finish off by quoting from Akerlof and Shiller’s Animal Spirits (p.96):

…Standard macro models are fairly accurate regarding the monetary-fiscal stimulus necessary to achieve full employment. But financial markets must also be targeted. The financial system is not the same as it was just a few months ago, fore the fall of Humpty Dumpty. Only a portion of its prior self is now operating.The aggregate demand target will indicate, on the one hand, the fiscal stimulus and interest rate policy needed for full employment. The credit target wil show what judicious application of methods 1, 2 and 3 [discount window, direct injections of capital, direct credit from GSEs] must achieve: together they must create the financial flows — the issuance of commercial paper, bonds, and other instruments — that are also associated with full employment.


Originally published at Econbrowser and reproduced here with the author’s permission.