Arvind Krishnamurthy on QE: What have we learned?

Video & Timestamps

On Thursday, March 24, Arvind Krishnamurthy joined Markus’ Academy for a lecture on QE: What Have we Learned? Krishnamurthy is a Professor of Finance at the Stanford Graduate School of Business.

You can also download the presentation slides and watch all Markus’ Academy webinars on the Markus’ Academy YouTube channel.

Timestamps:

[0:00] Introductory Remarks. 

[13:46] Identification challenges in asset pricing event studies.

[29:11] Cases: UK Gilt purchases

[35:42] Cases: Fed reinvesting asset portfolio

[39:12] More unconventional narrow-channel studies

[49:58] Asset pricing theory with narrow channels

[1:00:19] Macroeconomic effects

[1:11:52] Further modeling, policy implications, and research wishlist

Executive Summary

• [0:00] Introductory Remarks. Monetary policy comes in many forms, including Quantitative Easing through the purchase of various assets: government bonds, mortgages, or corporate bonds. For government bonds, the Central bank swaps fixed interest rate bonds for floating interest paying reserves. Alternatively, the Treasury can shorten its debt maturity. Has to be a balance between the Central Bank and the Treasury. There is also some signaling effect from the Central Bank as a future interest rate hike generates losses. For corporate bonds, the I Theory suggests that QE affects risk premia and with it redistributes wealth possibly to a financial constrained sector. Similarly, the purchase of mortgage products, lifts house prices and redistributes wealth to a possibly financial impaired household sector.   

• [13:46] Identification challenges in asset pricing event studies. Looking at a specific case in 2009 when the Fed had a significant QE announcement, looking at the hourly trading volume as well as treasury yields shows that they are both strongly affected by this announcement. Identifiable because tight event windows make it unlikely that these reactions are random or raise reverse causality concerns. The main challenge is in identifying possible different channels for QE. There are both conventional channels (such as signaling path of policy rate, news about economy) and unconventional channels (such as impacts on liquidity premia, risk premia, and safety/scarcity premia). 

• [29:11] Cases: UK Gilt purchases. Looking only at the yield changes by maturity for UK Gilts, the announcement can be interpreted as a conventional monetary policy shock, not QE. However, when additionally considering the OIS (swap curve) yields, which would be equally affected by a conventional monetary policy shock, we see Gilts move much more than the OIS curve, indicating that there is also an unconventional channel at work. Result here is that QE affects the purchased assets (Gilts) more than others.

• [35:42] Cases: Fed reinvesting asset portfolio. We can see this effect of policy on purchased assets even more starkly when looking at an announcement from the Fed regarding reinvesting their existing asset portfolio.  A first announcement indicates that the Fed will purchase long-term bonds, without maturity specificity in the guidance, and as a result all long-term bond prices rise.  The purchased-asset effect jumps out when upon the Fed’s clarificatory announcement that they were only buying bonds under 10 years, after which any longer maturity bond prices came back down, but the prices of bonds under 10 years remained elevated. 

• [39:12] More unconventional narrow-channel studies. There are many more studies documenting these unconventional channels in contexts ranging from purchases of government bonds, mortgage bonds, and corporate bonds.  Asset pricing evidence strongly supports sizable unconventional channels.  Additionally evidence on quantities, such as loan originations and firm/household behavior, is consistent with unconventional narrow channel mechanisms.

• [49:58] Asset pricing theory with narrow channels. Reconciling the evidence with theory requires going from a complete markets model towards a segmented market model in which a financial intermediary’s pricing condition is paramount in setting asset prices. The forefront of modeling work in this area is in papers such as Vayanos and Vila (2021). Their research develops a model of the treasury market yield curve delivering risk premia which depend on asset supplies.  The model can explain quantitatively the impact of QE on the Treasury yield, in line with empirical estimates. The model can also speak to conditional impacts of QE: when yield curve arbitrageur (hedge funds, bond dealers) risk bearing capacity is low, risk premia are higher, and so QE has a larger impact.

• [1:00:19] Macroeconomic effects. Conventional monetary policy shocks affect real interest rates, which shifts both the user cost of capital and the household borrowing/saving rate. These, in turn, affect investment and consumption, which leads to shifts in employment and output.  When considering the macroeconomic impact of QE, we need to account for the evidence that QE affects the purchased asset price significantly more than other asset prices.  This factor adds considerable nuance to the transmission mechanism of unconventional policy. For example, purchases of corporate bonds of an investment grade company can be expected to affect the yields on the purchased bonds, but may not impact the firm’s investment behavior if the marginal source of capital is not the purchased bond.  There is evidence for this sort of null-effect in the research assessing the firm impact of both the Fed and ECB’s corporate bond purchase programs. Furthermore, there is evidence for macroeconomic transmission mechanisms via affecting an intermediary’s financial conditions; this is the macroeconomic counterpart of Vayanos and Vila’s asset pricing model.      

• [1:11:52] Further modeling, policy implications, and summary. The evidence and mechanisms at play in the macro transmission indicate that much more research is needed to understand macroeconomic effects.  Research needs to mix corporate financing and financial intermediation factors. Thus from the standpoint of understanding the impact of QE on the macroeconomy and thus guiding policy, research is still in the insights stage. We know that the asset market targeted matters for transmission and thus should affect the design of optimal policy. We know that crisis interventions are more powerful than non-crisis interventions.  We know that communication about QE policies matters. We have a good idea of the relevant channels of the impact of asset purchases on asset prices, but we are not at the stage of having normative models to guide optimal policy and policy communications.