1. Brookings Papers of Economic Activity
The disappointing recovery of output after 2009 Fernald, Hall, Stock and Watson Discussion
Lucrezia Reichlin
Brookings Papers of Economic Activity
March 23-24, 2017

2. What kind of recovery? Not the “classical” recession and recovery

3. Key question for FHSW
Why has the recovery from last recession been so slow?
Establish features once variables are “cleaned” from cyclical component
Do growth accounting to compare with previous recessions and establish role of TFP, labor force participation etc.

4. Cyclical adjustment obtained by using Okun’s law

5. This recovery …. QUESTIONS: Slow because change in trends or cycle?
If trends, has the change occurred prior to the Great Recession?
What are the drivers?

6. Main findings of the paper
Slower output recovery mostly explained by slow TFP growth (71% of shortfall), LFPR and K/POP
When adjusted by the cycle the shortfall is larger. Interpretation: the shortfall due to trend and noise (about 50-50)
TFP started slowing before the crisis therefore main factors precede the Great Recession
Weak investment and capital growth are not key factors for slow recovery
Rule out demand driven alternative explanations of slow recovery but finds some role for fiscal policy

7. Alternative ways to analyze the data ….
Estimate VAR and compute counterfactual paths conditioning on historical correlations and observed unemployment
RESULTS: counter-factual paths higher than actual for output

8. Conditionally on unemployment GDP is higher than realized

9. Alternative ways to analyze the data ….
➔ This is in line with FHSW: indicates that unemployment rate recovered earlier than other variables
If we define the cycle via Okun’s law, the cyclically adjusted series simply reflect this fact
➔ POTENTIAL PROBLEM: Leans heavily on a stable Okun’s Law
During last recovery unemployment recovered relatively fast and this lowers cyclically adjusted output
FHSW may under-estimate (slow) cyclical contribution to recovery

10. Results on GDP change when sample starts in 1990

11. Other issues
TFP has large irregular component not clearly interpretable
Role of capital and investment not clear
➔ define alternative decomposition to address these problems

12. Alternative time series model
Variables in level
T is a random walk with drift
C is a stationary process

13. Transformations for the decomposition in ln-levels
Variable
Transformations for the decomposition in ln-levels
Civilian Unemployment Rate: 16 yr + (SA, %)
Levels
Real Gross Domestic Product (SAAR, Bil.Chn.2009$)
Natural logarithms
Real Private Nonresidential Fixed Investment (SAAR, Bil.Chn.2009$)
Output, business sector (2012=100)
Hours, business sector (2012=100)
Labour productivity, business sector (2012=100)
Capital input (2012=100)
TFP, business sector (2012=100)

14. Results Output gap shows unusually slow cyclical recovery
GDP trend slowed down earlier than TFP trend
Slow recovery is both about the trend and the cycle
The trend in TFP has large idiosyncratic component (which is what the FHSW decomposition attributes to the irregular).
TFP has also a cyclical component correlated with non-residential investment

15. Output gap: it took a long time to go back to a possibly slowing trend

16. TFP and GDP trends: correlation weak

17. But the cyclical component of TFP is strongly correlated with and leading the cycle of non residential investment

18. The role of investment
VAR model suggests investment weaker than path conditional on unemployment contrary to FHSW results
Factor model suggests role for TFP and investment in explaining weak cyclical output

19. Alternative decompositions – different stories
Alternative decompositions may attribute some medium term fluctuations in TFP to the cycle rather than the trend
These maybe related to demand factors
Alternative is structural decomposition based on VAR Blanchard-Quah identification
Demand shock: neutral on TFP in the long-run
Supply shock: unrestricted

20. TFP has a sizeable cyclical component: ¼ of variance between 2 and 10 quarters

21. Conclusions GDP trend slowed down before the crisis
But slow-recovery also explained by cyclical component including TFP and investment
TFP trend also slowed-down before the crisis but later than GDP trend – not easily interpretable
TFP has persistent fluctuations possibly related to demand

22. My own comment
There is idea that recession had a scarring effect that lowered LFPR but did not affect U3 unemployment rate in the long-run
Erceg and Levin (2014) present evidence from state level data
But Okun’s law method will treat whole decline as in the trend/noise components once the U3 rate has normalized