Comments on

1
Comments on Labor Force and Wage Dynamics

• by French, Mazumder and Taber

2
Discussant dilemma

• Two approaches
• Interpret results
• Focus on methods
• I follow the latter
• Results are preliminary
• Conference to guide revisions

3
Question

• How do wage and employment dynamics
• differ over the business cycle?
• differ across workers?
• Policy relevance
• Do recessions affect wage growth?
• Does welfare to work have potentially large long
  term effects?
• Do the least advantaged have different dynamics?

4
Overview

• Methodological issues
• Data
• Presentation

5
Methodology

• General comments
• Quality is what I expect from these authors
• Face difficult issues that cant be avoided
• Technique not for techniques sake

6
Methodology

• Issue 1-- Returns to Tenure
• Standard framework--wages grow
• Within job due to
• returns to job specific skills (tenure)
• returns to general skills (experience)
• Between job due to
• better job match
• offset due to lost returns to tenure

7
Methodology

• Authors dont try to separate returns to tenure
  from returns to experience
• Im sympathetic since
• Requires additional strong assumptions
• Previous studies have shown low returns to tenure
  for less educated
• But it would be useful to policy community
• Do welfare recipients lose wage gains when they
  job hop?

8
Methodology

• Suggestion
• Change exposition to focus on
• wage changes within jobs and between jobs
• rather than on returns to experience and job
  match
• This is all that is identified in this model
• Study doesnt provide information on
• Returns to specific and general human capital
• Whether policy should encourage or discourage job
  hoping

9
Methodology

• Issue 2 -separating returns to experience from
  shifts in wages that affect everyone
• Suppose average wages of workers increase by 4
• Does this reflect
• returns to experience (and tenure)?
• aggregate shifts that affect everyone?
• Answer important given focus on cycle

10
Methodology

• Authors
• Reject using non-employed to get time effects
• unemployed not proper control group
• Propose using new entrants to get time effects
• Assumes
• Quality of entrants is independent of cycle
• Quality of match is independent of cycle
• Suggestion
• Parameterize time effects by trend and cyclical
  var
• This is identified
• Dont need to assume parameters change annually
• Can test rather than eyeball cyclical patterns

11
Methodology

• Issue 3 Selection
• Wages increase if
• workers earn more
• low wage workers drop out
• Solution requires identifying assumption
• Authors assumption work is independent of
• Cyclical variation in wages
• make hay while the sun shines
• Job match
• make hay when farm next door offers job
• No suggestion for this problem

12
Data

• Use 1984 to 2001 SIPP panels
• Good information on
• Wages
• employer
• Problem with monthly data
• Seam bias
• 96 and 01 panels collect employer id only once a
  wave
• Suggestion treat wave as unit of observation

13
Presentation

• Provide better evidence on cyclicality of series
• Currently cycle is in the eye of the beholder
• Indicate recessions
• Plot confidence intervals
• Parameterize trend and cycle
• Provide links to trend and cycle in wage
  inequality

14
Presentation

• Provide information on diversity within education
  groups
• Does mean experience of education group apply to
  most members?
• Are most (all) welfare recipients likely to have
  same experiences as average experience of
  dropouts?
• Previous evidence suggests not
• Non-managerial job in food industry
• In poor family

15
(No Transcript)
16
(No Transcript)
17
Presentation

• Suggestion
• Show percentiles of distributions
• Quantile regressions

18
Conclusion

• Study addresses important question
• Difficult issues
• Well addressed by authors
• Technique is necessary
• Cross--tabs wont do
• Early draft that should lead to good paper