Discussion by Valerie A. Ramey Brookings March 23, 2012 1 1. Summary of DeLong and Summers view. 2. Notable elements of their model. 3. 4. 5. Test of ability of government spending to reverse hysteresis effects. Test of differential multipliers at times of slack and low interest rates A word of caution on extrapolating from past interest rates. 2 Traditional Elements -Most of the time output is supply-determined, so output is equal to potential. - During times such as the Great Recession and aftermath, output is below potential and is demand-determined. - Government spending can raise output when it is below potential. 3 Less Traditional Elements - Hysteresis effects of current output gaps on future potential output. Operate through the effect of reduced investment on future capital and the effect of unemployment on worker skills and labor force attachment. - Interest rates at the ZLB 4 For a variety of parameter values short-run increases in government spending during a slump can pay for themselves in the long-run. Thus, we shouldn’t let fear of government budget deficits prevent us from enacting another stimulus package now. 5 It’s the Keynesian Version of “Supply-Side Economics” Or Say’s Law: “Government spending creates its own financing” 6 The DS model is missing the “GE” of the DSGE model - No general equilibrium. No specification of the assumptions about fundamentals, such as preferences, technology, and resource constraints. Policies are evaluated solely by effect on output, not on household welfare. Overall, the model lacks the rigor of modern macro models. 7 The reason that we can criticize modern macroeconomic models is that they are explicit about their assumptions about fundamentals such as preferences, technologies and resource constraints. Of course, being precise means making simplifying assumptions that are often questionable – there are many elements of modern macro models that I question. However, replacing them with models based on vague intuitive ideas is not an improvement. 8 Hysteresis effect on future potential output Multiplier 𝜂µ𝑟 𝑟<𝑔+ 1 − µ𝑟 growth rate of potential output 9 DS assume not only that an output gap today leads to lower future potential output but also that raising output with government spending will reverse this effect. It is not obvious to me that an increase in government spending would create the private investment and skill-building jobs required to reverse the effect. We can test this hypothesis on U.S. data: if G raises Y in the short-run, it should have a persistent effect on output. 10 EVAR from Ramey (2011): 1939q1-2008q4 gdp -.1 -.5 0 0 .1 .5 .2 1 .3 1.5 .4 government spending 4 8 12 16 20 0 4 8 12 16 quarter quarter total hours nonresidential investment 20 -1.5 -.1 -1 0 -.5 .1 0 .2 .5 .3 0 0 4 8 12 quarter 16 20 0 4 8 12 16 20 quarter Blanchard-Perotti SVAR shows the same patterns 11 Standard VARs suggest that the effect of government spending on output lasts only as long as the government spending. 12 My recent paper (Ramey (2012)) uses a more precise way to estimate the multiplier in a VAR and using IV. It looks at the effect of government spending on private spending (Y – G). This method indicates that multipliers are significantly below unity – about 0.5 when tax effects are accounted for. 13 DS and numerous others have argued that the multiplier may be higher when there is slack in the economy and when interest rates are at the Zero Lower Bound. In principle, it is possible to test these hypotheses on historical data. 14 Monthly Data from 1931 to 1951 Civilian Unemployment Rate 15 10 percent 1931 1935 1939 1943 1947 1951 0 0 5 1 percent 2 20 3 25 3-Month Treasury Bill Rate 1931 1935 1939 1943 1947 1951 15 ∆𝑌𝑡 ∆𝐺𝑡 ∆𝑌𝑡−1 = 𝛽0 + 𝛽1 + 𝛽2 + 𝜀𝑡 𝑌𝑡−1 𝑌𝑡−1 𝑌𝑡−2 • Monthly data, 1933m1 – 1951m12 • Gordon-Krenn GDP and government spending data • My unemployment series to create slack variable • IV using lagged government spending (my news variable had low first-stage f-statistics) • Allow β’s to differ if unemployment>7% ∆𝑌𝑡 ∆𝐺𝑡 ∆𝑌𝑡−1 = 𝛽0 + 𝛽1 + 𝛽2 + 𝜀𝑡 𝑌𝑡−1 𝑌𝑡−1 𝑌𝑡−2 • Estimate of β1 = 0.593, s.e. 0.116 during nonslack times. Increment to β1 during slack times is -0.031, s.e. 0.530. • The estimates also show that the multiplier during this period of accommodative monetary policy doesn’t seem to be any bigger than for the 1939-2008 period studied in my previous work. DeLong and Summers look at historical data on long-term government interest rates to argue that it is unlikely that interest rates will rise significantly. It is always wise to bear in mind the Lucas Critique – historical data was not generated in a regime when entitlements led to projections of every-rising deficits. Consider the following example: 18 A Tale of Two Countries 3 4 5 6 7 Long-Term Interest Rates on Government Debt 2000 2002 2004 month Country A 2006 2008 Country B 19 0 5 10 15 20 25 Long-Term Interest Rates on Government Debt 2000 2004 2008 2012 month Country A Country B 20 20 25 Long-Term Interest Rates on Government Debt 10 15 Greece 0 5 U.S. 2000 2004 2008 2012 month Country A Country B 21 DeLong and Summers offer the intriguing and thought-provoking idea that temporary stimulus packages might be self-financing. I have qualms about the lack of rigor of their model. It would be great to see these ideas incorporated in a modern macro model. I can’t find any empirical evidence to support their key mechanisms. 22