Technology Shocks and Aggregate Fluctuations: How Well Does the RBC Model Fit Postwar U.S. Data?

1
Technology Shocks and Aggregate Fluctuations How
Well Does the RBC Model Fit Postwar U.S. Data?

• Jordi Galí (CREI and UPF)
• and
• Pau Rabanal (IMF)

2
Two Questions on The Role Of Technology in the
Business Cycle

• Question 1 Can a calibrated DSGE model driven
  by technology shocks generate economic
  fluctuations that resemble those observed in the
  postwar U.S. economy?
• ANSWERYes.
• It makes sense to think of fluctuations as
  caused by shocks to productivity, Cooley and
  Prescott (1995)
• ...the main criticisms levied against
  first-generation real business cycle models have
  been largely overcome, King and
  Rebelo (1999)

3
Two Questions on The Role Of Technology in the
Business Cycle

• Question 2 Have technology shocks played a
  central role as a source of the postwar U.S.
  business cycle?
• Present paper
• - overview of recent attempts to answer that
  question.
• - focus on a key feature of the U.S. business
  cycle the positive comovement of output and
  labor input measures (Figure 1).
• TENTATIVE ANSWER Most likely not.

4
(No Transcript)
5
Plan of the Presentation

• Present evidence from VAR model on the behavior
  of employment, productivity and technology.
• Discuss extensively possible pitfalls in
  estimation.
• Possible theoretical explanations.
• Present further evidence from an estimated DSGE
  model with nominal and real rigidities.
• Conclusions and directions for future work.

6
Estimating the Effects of Technology Shocks
(Galí, AER 99)

• A benchmark model
• subject to the restriction
• where

7
Estimating the Effects of Technology Shocks
(Galí, AER 99)

• Data
• y nonfarm business sector output.
• n nonfarm business sector hours.
• normalized by working age population.
• (consistent with ADF and KPSS) tests.
• Sample period 19481-20024.
• Evidence Figures 2 and 3.

8
(No Transcript)
9
(No Transcript)
10
The VAR-based Evidence

• Variance of output and hours explained by
  technology shocks 5 and 7 percent respectively.
• Correlation of output and hours conditional on
  technology shocks -0.08.
• Variance of output and hours explained by
  non-technology shocks 95 and 93 percent
  respectively.
• Correlation of output and hours conditional on
  non-technology shocks 0.96.

11
Implications

• Rejection of a key prediction of the RBC model
  the positive comovement of output, labor and
  productivity in response to technology shocks.
• Technology shocks cannot be a dominant source of
  economic fluctuations (independently of ones
  favorite model).

12
Empirical Work with Similar Findings

• Using constructed technology measures
• Blanchard, Solow and Wilson (1995).
• Basu, Fernald, and Kimball (1999).
• Using indicators of technological innovation
• Shea (1998).
• Using alternative VAR identifying assumptions
• Francis and Ramey (2003a).
• Francis, Owyang and Thedorou (2003).
• Using industry-level data
• Kiley (1996).
• Francis (2001).
• Using international data
• Galí (1999) G7. Galí (2004) Euro Area.
• Francis and Ramey (2003b) U.K.
• Carlsson (2000) Swedish manufacturing.

13
Possible Pitfalls in the Estimation of the
Effects of Technology Shocks

• 1. Do capital income tax shocks play a role?
• Equilibrium condition for the capital/labor
  ratio
• where
• Capital income tax rates could be a source of
  unit root in labor productivity.
• Time series constructed by McGrattan (1994) and
  Jones (2002).

14
(No Transcript)
15
Possible Pitfalls in the Estimation of the
Effects of Technology Shocks

• Result 1 no significant correlation with
  identified technology shocks.
• Result 2 no significant response of tax rates
  to identified technology shocks. Francis and
  Ramey (2003a).

16
Possible Pitfalls in the Estimation of the
Effects of Technology Shocks

• 2. Relationship with Basu, Fernald, and Kimball
  (1999).
• BFK and VAR shocks
• Simple regression

17
(No Transcript)
18
Possible Pitfalls in the Estimation of the
Effects of Technology Shocks

• 3. Alternative VAR Specifications
• Christiano, Eichenbaum and Vigfusson (2003)
  level specification findings and encompassing
  analysis.
• Our assessment a battery of alternative
  specifications
• Per capita hours, total hours, employment rate.
• Difference, level, quadratic detrending.
• Others
• Francis and Ramey (2003b) difference and
  detrended specifications most plausible.
• Fernald (2004) the discrepancy associated with
  the level specification vanishes when we allow
  for a plausible break in trend productivity.

19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
Possible Pitfalls in the Estimation of the
Effects of Technology Shocks

• 4. Investment Specific Technology Shocks (Fisher,
  2003)
• Two types of technology shocks
• aggregate, sector neutral (standard RBC).
• investment specific (as in Greenwood, Hercowitz
  and Krusell).
• Identification only I-shocks can have a
  permanent effect on the relative price of
  investment.
• Evidence
• N-shocks low correlation between output and
  hours. Small contribution to their variance at
  business cycle frequencies.
• I-shocks high positive correlation between
  output and hours. Contribution to variance highly
  sensitive to labor input measure
• levels above 50 percent of var(n)
• difference or detrended below 20 percent.

23
(No Transcript)
24
Explanations The Role of Nominal Frictions

• Sticky prices by their own do not generate a
  negative comovement of hours and output to a
  technology shock.
• One simple case would be when
• and the central bank does not alter the money
  base.
• But this result is not general once we move away
  from money-based rules.

25
Explanations The Role of Nominal Frictions

• We show in the paper that the dynamics of output
  and employment in a very simple model can be
  written as
• Under plausible assumptions and

26
Explanations The Role of Nominal Frictions

• Under sticky prices and a monetary policy rule of
  the Taylor-type

27
Explanations The Role of Nominal Frictions

• Under what conditions will
• Strong nominal rigidities.
• A weak monetary policy response to inflation.
• A strong concern for output stability.
• Low sensitivity of aggregate demand to interest
  rate changes.
• Calibration hours elasticity is -0.87.
• Evidence Gali, Lopez-Salido and Valles (2003),
  Francis et al. (2004), Marchetti and Nucci
  (2004), Chang and Hong (2003).

28
.... But Real Frictions Are Also A Possibility.

• Habit formation in consumption and capital
  adjustment costs (Francis and Ramey, Lettau and
  Uhlig).
• Extremely slow technology adoption (Rotemberg).
• Labor-augmenting technical progress with low
  capital-labor substitutability (Francis and
  Ramey).
• Low substitutability between domestic and foreign
  goods (Collard and Dellas).

29
Technology Shocks in an Estimated DSGE Model

• We have analyzed the comovement of output and
  hours using a bivariate VAR. In this section, we
  use a 5-variable DSGE model to examine
• The nature of the shocks that have played a
  determinant role as a source of postwar U.S.
  cycles
• Real versus nominal rigidities what are their
  relative merits?
• Ingredients
• Habit formation in consumption.
• Staggered price and wage setting a la Calvo.
• Flexible indexation of wages and prices.
• A Taylor rule with interest rate smoothing.
• Five exogenous shocks technology (permanent),
  preference, price markup, wage markup and
  monetary policy.
• The model is similar to Smets and Wouters
  (2003a), Ireland (2004), Rabanal (2003), and is
  estimated using Bayesian methods.

30
Technology Shocks in an Estimated DSGE Model

• Preferences and Technology

31
Technology Shocks in an Estimated DSGE Model

• Log-linear dynamics

32
Technology Shocks in an Estimated DSGE Model

• Log-linear dynamics

33
Technology Shocks in an Estimated DSGE Model

• Shocks

34
Technology Shocks in an Estimated DSGE Model

• Bayesian Estimation
• Data (xt)
• The likelihood function is evaluated with the
  Kalman filter, using the state-space form
  solution.
• The posterior is obtained using the
  Metropolis-Hastings algorithm with 500,000 draws.
  As in Rabanal (2003), Fernandez-Villaverde and
  Rubio-Ramirez (2004), Lubik and Schorfheide
  (2004), Smets and Wouters (2003).

35
Main Findings

• Parameter estimates (Table 5)
• Significant habit formation.
• Significant price stickiness.
• Wage stickiness negligible.
• Relative stability over time.
• Second moments (Table 6)
• Reasonably good fit.
• Technology shocks generate a negative comovement
  between hours and output.
• Technology shocks have a negligible role in
  explaning the variance of output or hours at
  business cycle frequencies.
• Dominant driving force preference/demand shocks.

36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
Main Findings

• Impulse-responses Negative response of hours to
  a positive technology shock.
• Nominal versus real frictions both sticky prices
  and habit formation are sufficient to generate,
  on their own, the negative comovement between
  output and hours.
• Corollary the effects of technology shocks will
  depend on the monetary regime in place. But even
  under a regime that minimizes nominal distortions
  they are likely to generate non-RBC-like
  fluctuations.

40
(No Transcript)
41
solid line technology component
(BP-filtered) dashed line U.S. data (BP-filtered)
42
Conclusions

• Have technology shocks played a central role as a
  source of postwar U.S. business cycles?
• ANSWER most likely not.
• If they have, or if they do in the future, the
  channels are likely to be quite different from
  those emphasized in the literature.

43
Directions for Future Work

• According to McGrattan, by leaving out investment
  in our model we are favoring the importance of
  the preference shock. But Smets and Wouters
  (2003) show similar results to ours in an
  estimated DSGE model with investment.
• Also, in work in progress, in a model with
  nominal rigidities, capital accumulation and 5
  shocks (neutral and investment-specific
  technology shocks, labor supply, government
  spending and monetary), I find that demand shocks
  and labor supply shocks still play a predominant
  role.