Job Mobility and Wage Dynamics

1
Job Mobility and Wage Dynamics

• Valerie Smeets
• Very preliminary comments welcome
• Prepared for the Internal Seminar UC3M

2
Motivation

• Job transitions (patterns) and wages
• Common framework is search and matching models
• Able workers should match with better jobs
  (Becker, 1973)
• Workers change jobs to improve their match
• Switching decision may be influenced by firm
  specific HC and switching costs
• Firm specific HC destroyed for switchers
• Workers with more firm specific human capital and
  better matches are less likely to leave

3
Motivation (contd)

• Number of past jobs
• Ambiguous predictions about cross sectional
  correlation between the number of jobs a worker
  has held and his match quality
• Workers who remain at their job for a long time
  have high matches while those who switch firms
  have low matches ? past jobs negatively related
  to wages
• Workers take better matches at other firms as a
  substitute for acquiring specific capital ?
  past jobs positively related to wages

4
Literature

• Job seniority and wages
• Large returns to firm tenure (2 to 3 per year)
• Unobserved heterogeneity lead to biased returns
  estimates (Abraham and Farber (1987), Altonji and
  Shakotko (1987), Topel (1991), Abowd, Kramarz and
  Margolis (1999))
• The large firm tenure coefficients go away when
  industry, occupational or career tenure are
  included (Neal (1995), Kambourov and Manovskii
  (2002) and Pavan (2005))

5

• Job mobility and wages
• Mincer (1986)
• Short and long run wage gains from separations
• PSID, 1970-1981 (all ages)
• Positive wage gains in the short run
• Movers never catch up higher wage of stayers
  (flatter life time trajectories)
• Abowd, Kramartz and Roux (2006), Buchinsky,
  Fougere, Kramarz and Tchernis (2003)
• Wage and mobility (firm policies, firm
  performance)
• French matched employer-employee data
• Includes firm changes (more) to control for
  bias in worker heterogeneity wrt previous
  employment

6

• Job changes patterns and wages
• Mincer-Jovanovic (1981)
• National Longitudinal Survey 1966-1976
• No effect of past firm changes for young workers
• Negative effect for old workers
• Light and McGarry (1998)
• NLSY (1979-1993), young white men only (8 first
  years of career)
• Positive or no effect of firm changes in first
  2 years, negative effect of total changes
  overall
• Extremely small effects

7
Outline

• Estimation strategy
• Data description, sum stats and results with
  Danish matched employer-employee data
• Verification using US data (NSLY) brief data
  description, sum stats and results with US data
• Conclusions

8
Estimation model

• Correlation of variables with unobserved job or
  worker characteristics will produce biased OLS
  estimators of ß

9
Two empirical possibilities (I)

• Consistent with costly search model of Jovanovic
  (1979)
• No HC gain to accumulate experience in multiple
  firms
• Workers who remain at their job for a long time
  have high matches while those who switch firms
  have low matches
• Costly SHC only for workers planning to stay
• ?
• Firm tenure biased upward and firm changes biased
  downward (ßft gt 0 and ßc lt 0)

10
Two empirical possibilities (II)

• Some workers take better matches at other firms
  as a substitute for acquiring specific human
  capital
• These workers may switch because of low search or
  switching costs, scare skills or plain luck at
  securing offers
• ?
• Firm tenure biased downward and firm changes
  biased upward (ßft lt 0 and ßc gt 0)

11
Data - Denmark

• Danish Integrated Database for Labor Market
  Research (IDA) Statistics DK
• Matched employer-employee dataset of entire
  Danish population (1980-2001)
• Combination of 3 datasets
• Common keys to match them back
• Individual personnel characteristics (age, sex,
  education, family status, location, occupation,
  etc.)
• Job plant, industry, location, jobs, wage,
  etc., November data
• Firms nbr plants, old ID number, etc.

12
Data - Denmark

• Sample
• Private sector, primary job only
• Workers aged 25 to 44
• 25 - Avoid students
• 45 - Data limitations in computation of labor
  history variables
• 2001 cross section and 1996-2001 panel
• 731,358 individuals or 4,314,813 data year
• Controls
• Education (4 groups), part time, female, firm
  size, industry dummies, firm and worker fixed
  effect

13
Data - Denmark

• Human capital variables
• Experience computed back to 1969 by Stat DK
• Firm tenure using 1980-2001 history
• Industry tenure - using 1980-2001 history
• Firm changes variables
• past firm changes after last graduation
• Decompose firm changes into exits and non exits
• Intervals for firm changes 1-2, 3-5, 6-9, 10 up
  to check for non linearity
• Disaggregate exits and non exits by two age
  intervals 25-34 and 35-44

14
(No Transcript)
15
(No Transcript)
16
Histogram of firm changes (after last graduation)
17
Histogram of exits and non exits
18
Estimations

• Wage (log) on HC variables, firm changes
  controls
• 2001 sample, men full time, firm worker fixed
  effects (1996-2001 sample)
• Wage (log) on HC variables, exits and non exits
  controls
• 2001 sample, men full time, college, college men,
  manufacture, non manufacture
• Wage growth (log) on HC variables, firm changes
  and controls
• 1996-2001 sample, men full time, college, 2001
  and 1996 firm fixed effects

19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
Data - US

• National Longitudinal Survey of Youth (NLSY79)
• Data prepared by Pavan (2005)
• Computed workers labor history, including firm
  tenure, industry tenure and career tenure using
  NLSY work history files (1978-1994)
• Sample
• 1979-1993
• Full time male workers, primary job
• Men only, age 25 and older 25,37
• 1935 individuals or 12,776 data year
• Older sample than Light and McGarry (1998)

23
Data - US

• Firm changes variables
• past firm changes after age 25
• exits and non exits after age 25
• Intervals for firm changes and non exits
• Controls
• Experience, firm tenure, industry tenure, career
  tenure, black, education
• Estimations log wage
• Panel, without and with individual fixed effect
• past firm changes, exits and non exits
  (intervals)

24
Summary statistics US (NLSY)
25
Histogram of firm changes (after 25)
26
(No Transcript)
27
US Data (NLSY) - Wages and Exits vs. Non Exits
28
Conclusions

• Denmark
• Cross section panel - the effect of past firm
  changes increases with the number of changes
• 2 to 7 in cross section
• 17 to 23 in panel
• US
• Cross section - Only moving up to 3 times affect
  your wage (7-6,3-6)
• Panel - the effect of past firm changes increases
  with the number of changes (13 to 23)

29
Conclusions

• New empirical fact - workers who switch firms
  repeatedly in the past earn higher wages
• Workers who move end up with better match than
  workers who do not move
• Contredict Mincer- Jovanovic (1981) and Light
  McGarry (1999)
• Different groups of workers have different effect
  of job transitions patterns on wages
• Returns to firm changes dominate returns to SHC
  effect

30
Conclusions

• Correlation btw firm changes and wage become very
  large once workers fixed effects are introduced
• Some negative selection in the time invariant
  ability of workers who switch firms
• Robust to all specifications
• Number of job changes, exits or non exits, age
  groups, fixed effect
• Similar results for Denmark and the US
• Policy implications for labor market flexibility

31
(No Transcript)
32
(No Transcript)