The International Wage Flexibility Project

1
The International Wage Flexibility Project

• Presentation at Workshop on Downward Nominal Wage
  Rigidity
• Norges Bank, 12 June 2003

2
What?

• 13 Country study of wage inflation
• Using micro data on individual and occupational
  wages
• Intent is to study the nature, extent, and
  implications of wage rigidity in the presence and
  absence of inflation

3
Country Teams

• Austria
• Belgium
• Denmark
• Finland
• France
• Germany
• Italy

• Norway
• Portugal
• Sweden
• Switzerland
• United Kingdom
• United States

4
Who Is On Norwegian Team?

• Jan Morten Drystad Norwegian Institute of Science
  and Technology
• Steinar Holden University of Oslo
• Kjell Salvanes Norwegian School of Economics and
  Business

5
Starting Point for Some History GS

• Groshen and Schweitzer / Grease and Sand
• As some have suggested, inflation may grease
  economy and facilitate relative price adjustment
• On the other hand, inflation may distort
  usefulness of price signals thus fouling the
  works (sand)

6
What Groshen and Schweitzer Did

• Analyze firm level data on wages paid in
  different occupations.
• Estimate an analysis of variance model of within
  firm occupational wage changes with firm and
  occupation effects (ie. regression with firm and
  occupation dummy variables).

7
What they Were Looking For

• If firms make more errors in setting wages when
  inflation is high, the variance of the
  coefficients on the firm dummies should grow with
  the rate of inflation.
• If inflation facilitates relative wage
  adjustments then the variance of the coefficients
  on the occupation dummies should increase with
  inflation.

8
What They Found

• Both the variance of the coefficients of
  occupations and firms increased with inflation
• Welfare effects of sand and grease comparable and
  roughly offsetting
• Several tests suggest the identification strategy
  is appropriate

9
History of Project(continued)

• Erica and Mark begin recruiting project teams
  (Fall 2000).
• Erica and Mark share plans for project with me
  and I suggest broadening the range of questions
  to be addressed (Winter 2001)

10
History of Project(continued)

• Im invited to help with project planning
• In line with new concept of project we recruit
  additional country teams
• Organizational meeting held at ECB 22-23 October
  2001

11
At The Meeting

• Country teams presented
• descriptions of data available
• descriptions of wage setting institutions
• preliminary attempts to replicate Groshen-
  Schweitzer
• wage change histograms
• Some Very Interesting Patterns Emerged

12
Observations

• Preliminary analysis suggested sand more
  important than grease, but data werent uniformly
  good for replication of G-S (few and inconsistent
  occupational categories)
• Nearly all countries had micro data on individual
  wage histories
• Several country teams expressed concern that G-S
  identification strategy wouldnt work given their
  wage setting institutions
• It became clear that focusing on Sand and Grease
  alone would miss much of what is important about
  wage rigidity in Europe.

13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
Five Types of Wage Rigidity

• Nominal Rigidity
• Menu costs
• Downward nominal rigidity
• Real Rigidity
• Downward real rigidity
• Insensitivity to fundamentals (unemployment and
  productivity)
• Bargained or legislated wage floors

18
(No Transcript)
19
First Attempt To Estimate Rigidity Had Several
Problems

• We tried to adapt Altonji and Deveraux ML method
  to include real rigidity
• Couldnt tell real rigidity or measurement error
  from non-normality in notional wage distribution
  (this became obvious at authors meeting in Bonn
  last October).
• Also, discussions at meeting last fall suggested
  that in many countries real rigidity didnt take
  form assumed in model.

20
New ApproachUse information in correlation
between years

• Abowd and Card (1989) suggest that wage changes
  have two components
• permanent changes
• transient (one period) changes (which result in
  negative serial correlation of wage changes)
• New method identifies transient changes as errors
  and uses covariance and frequency of sign
  switching in changes to identify error rate and
  error variance.
• This information is used to identify
  non-parametric estimate of true wage
  distribution.

21
Validating Primary Assumption

• Results for US largely fit with those of other
  studies (nearly complete downward nominal
  rigidity).
• German data has virtually no errors and estimated
  covariances are tiny and sometimes positive
  (implied sd of error when cov lt 0 is about 1.5)
• Will shortly check time series properties of
  Gottschalk true wage changes.

22
We assume that process generating
observed wage changes is
where uU-c,1-c This implies
We can then estimate c by matching theoretical
number of switchers and actual number of
switchers (method of moments)
23
Now we assume that underlying true distribution
is a sequence of point masses at specific
intervals and that the error is normal when
present. From that we can compute the expected
fraction of observations in any interval of the
observed wage distribution given the fraction of
observations at each mass point. We can then write
Where mo is the vector of observed fractions of
observations in each range, m is the fraction of
observations in the true distribution at each
mass point and T is the matrix which has columns
containing the fraction of observations from each
mass point that are expected to be observed in
each interval of the true distribution. We keep
T square for simplicity so...
is an estimate of the true distribution.
24
Example Using PSID Datat

• Only wage earners
• Use reported wages and not PSID computed wages

25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
Nominal Rigidity Estimates

• Spike vs tail (with and without correction for
  menu costs)
• Lower tail vs. upper tail (symmetry)
• Below zero vs. same category above zero in
  another year (constant distribution as in Kahn
  (1997) -- not yet implemented)

30
Menu Cost Rigidity

• Categories around zero vs. reflection to other
  side of median (symmetry)
• Categories around zero vs. same categories in
  other years when they arent around zero
  (constant notional distribution Kahn)

31
Real Rigidity

• Downward real rigidity by symmetry method (use
  different estimates of expected rate of inflation
  and compare mass below to mass above)
• Downward nominal by constant distribution
  (estimate reduction in mass by contrasting values
  in different periods -- not yet implemented).
• Asymmetry After removing spike
• measure skew
• measure difference between mean and median
• excess mass in spikes other than zero

32
Real RigidityInsensitivity to Fundamentals

• Estimate Phillips Curve with median instead of
  mean
• Estimate mean of notional wage distribution using
  modified Kahn method

33
(No Transcript)
34
(No Transcript)
35
Real Rigidity Measures
36
Whats Next?

• Country Teams
• finish descriptions of wage setting institutions
• replications of G-S (where possible)
• histograms of wage change distributions for each
  year data are available
• discussion of data reliability
• run program to do rigidity estimates
• unique country projects

37
Whats Next (continued)

• Project leaders, in cooperation with country
  teams,
• develop hypothesis about institutional sources of
  rigidity
• develop data set on institutions
• panel meta-analysis of causes and consequences of
  rigidity and sand
• Culminating in meeting in Winter 2003