Labour Turnover

1
Labour Turnover

• What do we know?
• CAED, Cardiff, Sept 2005

2
Motivation

• Important
• a big cost to firms/management issue
• subject to state regulations many places
• Large differences in turnover rates between
  different economies, subject to a policy
  discussion in EU about the impact of regulations
• impact on
• the functionning of the labour market
• productivity, and profit of firms
• Wage growth of individuals
• The development of employer-employee data
• Own work in the field

3
Overview

• Models/theories for worker turnover
• The amount of turnover
• International Comparisons
• The role of regulation
• Own estimates
• Where do workers go after separations?
• What do workers get out of separations?
• What are the overall costs when people loose
  their jobs and cant find new employment?
• What is the impact on firms?
• Do non-pecuniary factors play a role?

4
Theories for turnover

• Parsons, 1977. A search model building on Stigler
  ( 1961, 1962), McCall (1970), Mortensen (1970b),
  Gronau (1971), Salop (1973), Nelson(1970),
  Parsons(1973). Learning Models Rothschild(1974)
  and De Groot(1968).
• Factors of importance to the job searching
  worker
• The distribution of wage offers
• The cost of search
• the explicit or implicit job duration
• the workers state of knowledge of market
  conditions
• the contracting terms
• And for the employed worker
• the work hours rigidity of his current job

5
Quit behaviour

• The search model suggest that the wage rate
  becomes a policy instrument of the firm higher
  wages lower quits.
• Salop, 1973,
• Wages should raise in firms which find workers
  suddenly more valuable in order to reduce quits
• A sudden increase in hiring costs higher wages

6
Layoff behaviour (Parsons)

• Closely related to the theory on specific human
  capital, Oi, 1962, Becker, 1962, Rosen, 1968,
  Parsons, 1968 and Salop, 1973
• Training costs are born by the firm and must be
  repaid by productivity gt w
• Implications
• Minor fluctuations in product demand are less
  likely to drive productivity below wages the
  higher level of spec HC
• Firms should more readily lay off workers when
  unemployment is high because workers will be
  available for rehire temporary layoffs as an
  inventory function where the training investment
  is maintained state subsidy in many countries

7
Temporary layoffs

• As a long run contract between the firm and the
  employee
• the firm takes the main risk, but sometimes
  subsidized, Italy in one extreme and DK as
  another
• Feldstein, 1975, Hamermesh, 198? and Jensen and
  Westergaard-Nielsen, 1990.
• The premium to workers is a wage premium
• But how to prevent workers from taking a better
  offer when times are good and jobs are plenty?
• HRM methods comes into the picture bonding, job
  satisfaction, challenges on the job, training
  possibilities, career etc (to be discussed later)

8
Turnover over the business cycle according to
Parsons

• Mild fluctuations in the separation rates cover
  larger fluctuations in the quit and the layoff
  rates, which are neg. correlated
• Layoffs rise in years of recession
• Quit rates fall in years of recession and
  increase in the upswing
• Residuals do not vary

9
Summing up on Parsons

• Search based models to explain quits,
• quits are related neg. to wages
• Specific HC theory to explain layoffs, room for
  HRM methods to explain retention
• Layoffs neg related to spec HC, HC and tenure
• Temporary layoffs are a consequence but also
  noise for the observer

10
Jovanovic, 1979 and 1984

• Jovanovich combines a search model with a
  matching model to yield movements of workers from
  job to job and into and out of employment.
• He uses the assumption that a workers contact
  with employers occur as a Poisson Process
• In Jovanovic, 1984 he introduces three different
  reservation wages job-to-job, job-to-unemployment
  , and unemployment-to-job transitions.

11
J 2

• Job to Job transitions
• Main points are
• After a worker has worked for a while, the worker
  will be willing to accept a job with a lower
  current wage, because now he knows the future
  wage trajectory
• The job-to-job hazard rate decreases with w and
  increases with t. It is assumed that general
  skills, transferable across firms, are
  accumulated on the job.

12
J 3

• Unemployment to Job transitions
• The rate of movement from unemployment to
  employment is a constant .
• This rate exceeds the job-to-job rate. This means
  that a worker has a bigger chance of getting a
  job while unemployed than when he is employed.
• The constancy of the hazard rate reflects the
  assumed performance of unemployment insurance.
• Job to Unemployment transitions
• unemployment is something the worker can always
  have.
• The job to unemployment hazard rate must
  initially increase. In his 1979 article he makes
  the same prediction and Flinn and Heckmann (1983)
  find that it does initially increase, but that it
  eventually declines.

13
Theories for quits and lay offs

• Becker, Landes and Michael (1977)
• Labour turnover is always efficient or joint
  wealth maximizing. The voluntary/involuntary
  distinction is inappropriate.
• The firm and worker dissolve their match if and
  only if their total value as separated exceeds
  the combined value of the match. In this world,
  potentially inefficient separations results in
  wage adjustments or side payments to preserve the
  optimal assignments.
• At some wage the firm would want to keep the
  worker, but the worker would choose to leave at
  a higher wage the firm would want the worker to
  leave but the worker would want to stay
• In praxis If a worker asks for a wage increase,
  he does not get, he quits
• If the worker refuses to accept a wage cut he is
  laid off.

14
Hall/Lazear

• The idea Excessive layoffs in bad times and
  excessive quits in good times stem from the same
  weakness in worker-firm relations
• The claim is that there are too many inefficient
  lay-offs and quits.
• Too many (inefficient) layoffs occur when
  conditions are poor in the firm but even worse in
  the market
• Too many (inefficient) quits take place when
  conditions in the market are good but even better
  in the firm

15
The difference between lay offs and quits

• McLaughlins model (McLaughlins, 1991)
• Quits are voluntary separations
• Lay offs are involuntary separations
• The classification depends on whether the demand
  for a wage revision is initiated by the worker or
  the firm
• Separations are censored wage revisions
• The worker knows his outside wage offer R but
  does not know his productivity with his incumbent
  employer.
• Wages are governed by a counteroffer-matching
  contract
• If productivity is exceeding the outside offer
  new pay is equal to the value of productivity, if
  not the worker quits, due to censored wage
  increase
• If the firm observes that the worker productivity
  falls, initiates a wage cut, if the wage is below
  the value of productivity and the worker rejects
  it, there is a censored wage cut
• This means that quits separate to higher paying
  jobs and layoffs to lower

16
McLaughlin 2

• Wages are governed by a counteroffer-matching
  contract
• If productivity is exceeding the outside offer
  new pay is equal to the value of productivity, if
  not the worker quits, due to censored wage
  increase
• If the firm observes that the worker productivity
  falls, it initiates a wage cut, if the wage is
  below the value of productivity and the worker
  rejects it, there is a censored wage cut
• This means that quits separate to higher paying
  jobs and layoffs to lower
• Estimates a model with these features
• Conclusion
• Attempt to argue that the distinction between q
  and l is not meaningful
• The estimates generally supports this model
• The reason why we still distinguish between q and
  l it is useful in describing the conditions of
  the employment transition.

17
Monopsony

• Manning, 2003, claims that it is meaningful to
  discriminate between quits and layoffs
• The reason is that wages as in the Lazear/Hall
  model are fixed ex ante and not altered in
  response to shocks, there is rigidity, and it
  becomes meaningful to distinguish thus
  disagreeing with McLaughlin
• Manning finds 15 points of evidence of monopsony
  on the labour market wage dispersion prevails,
  employer size effects, separation rates vary with
  wages, gender gap,

18
Definitions

• Which measures Those, who were employed at
  establishment q in year t who were not employed
  at that establishment in year t1
• Separations
• Two sides of separations a third
• if employers take the initiative lay offs
• If employees take the initiative quits
• Life turnover death, sickness, retirement,
  military service
• In our case, we cannot distinguish!
• Empirical question if we can trust answers
  employers and employees

19
International Comparisons, average tenure ordered
and median tenure
Source OECD, 1996
20
Skewness (as an indicator of the feasibility of
short employment spells?)
21
Regulation

• Employment protection
• No lay off possible
• Layoff through legal process
• Firing for wrong doing through labor court
• Firing for wrong doing
• Layoffs have to be negotiated
• Mass layoffs have to be negotiated
• Layoffs with long notice
• Layoffs with short notice
• Layoffs with no conditions

• Compensation
• No UI, bribing
• Redundancy pay
• Experience related UI
• UI for layoffs
• UI for all spells of unemployment

22
International Comparisons

• In countries with regulation
• Difference between mean and median is larger
  because there will be a number of people with
  short tenure and some with long tenure
• more tend not to get a job at all. They will not
  even get a short tenure.
• Low turnover means that firms are reluctant to
  hire
• Employees are reluctant to move
• Matches will tend to get a lower quality

23
Empirical studies of worker turnover
1Anderson Meyer, 2MarshallZarkin, 3Kidd,
4Parsons, 5 Pencavel
24
Other studies

• A number of studies have shown
• high worker and job turnover
• (Davis Haltiwanger and Schuh, 1996), 19
• (Albæk and Sørensen, 1998) and (Werwatz et al,
  1999) for DK, 30, /- for other industries.
• job creation and destruction are obtained through
  a simultaneous hiring and firing process

25
Hypotheses on turnover over the cycles

• Separations are changing between 27 and 33 in
  DK. Separations are lowest in low growth years,
  indicating that the low number of quits are
  dominating the increase in layoffs (also AM,
  1994)
• Hires are highly procyclical (like AM, 1994)
• The result is that youth and entrants are more
  vulnerable to business cycles

26
Hires and Separations (DK)
27
Separations over timethe coefficients in DK
28
Emperical prob that Person stays one add.
year, 5 more, or 10 more years Given tenure in
2000
Completed spells
Prob staying increaseswith tenure
29
The probability that a person stays 20years
given 5 years of tenureCompleted spells
30
Conclusions with respect to turnover

• High
• for low tenure
• For young people
• Some are still staying for long periods

31
What happens to people being separated from their
job?

• (Where did they go? By Anders Frederiksen and
  NW-N, 2003) and Ibsen and NW-N, 2005)
• The ideaFollow all job leavers into different
  destinations (Blanchard and Diamond, 1990)
• Cyclicality, Personal characteristics (Anderson
  and Meyer, 1994)
• Plant closures/conditions of previous employment
  (Gibbons and Katz, 1991)

32
Worker turnover where did they go?
33
Observation 1Where did people go?

• Of 30 leaving a job between two years,
• 20 get a new
• 10 become unemployed or leave the labour force

34
Strategy

• (in Ibsen and W-N, 2005)
• 1. estimation of logit for getting immediate
  employment after separation
• 2. estimate competing risk model for getting
  employment orleaving the labour force
• 3. Data spell data, where we have followed each
  person week for week.

35
Probability that the person finds a job right
away, 1
36
2
37
Observation 2

• Moving from declining industry is more difficult
  than moving from other industries because of loss
  of industry specific HC
• A lemon effect? (Gibbons and Katz) a neg coeff.
  should be seen if person is leaving a growing
  firm
• Person specific factors seem highly important

38
Hazard function for finding employmentgiven a
spell of unemployment
39
Hazards continued
40
What did we find?

• There is a substantial job creation and
  destruction.
• Newly created jobs require more education than
  destroyed jobs
• The relation between job destruction and turnover
  12, but skew distribution
• Of 30 leaving a job between two years,
• 20 get a new
• Industry/firm effects
• Moving from declining industry is more difficult
  than moving from other industries because of loss
  of industry specific HC
• A lemon effect? (Gibbons and Katz) to be seen if
  you are leaving a growing firm workers from
  closing plants are doing better
• It looks like person effects are more important
  for the hazards moving out of unemployment and
  getting a job immediately after a separation than
  growth etc

41
Conclusions

• Flows out of jobs and destinations are cyclical
• The conditions of the work place matters for the
  probability getting the next job and matters for
  unemployment evidence for Lemons
• Workers involved in job closures are more likely
  to get a job than others
• Growing firms tend also to re-allocate more

42
A case The Danish Textile and clothing industry

• Descriptive work
• The decline of the textile industry in DK Late
  70s About 80,000 people employed in the textile
  and apparel industry
• In 2000 there were 15,000 and the GDP
  contribution was larger
• Similar but not as strong development in other
  European countries
• Bjerring Olsen, Ibsen and W-N, 2003

43
Tentative conclusions concerning the closure of
the textile and clothing industry

• Most of the adjustment has happened via lower
  accessions
• Exits are not excessive
• A large part of those made redundant in one
  establishment has found employment in other parts
  of the same firm
• Exit destinations are of the same composition as
  other parts of the economy
• Educational compostion of the work force has
  shifted
• Young people with little education do not find
  their first job in textiles to the extent they
  used to!

44
The costs of turnover, where people do not find a
new job immediately
45
What do workers get out of moving job?
46
empirical studies on outcome of job mobility

• Borjas NLSMM
• labour turnover may lead to a significantly
  higher wage in the new job than in the old job,
  but
• when they reach midlife the short run
  advantages of labour mobility are less important,
  leaving non-mobile men with significantly higher
  wage rates
• The general result is that mobility increases the
  wage level -
• Light and Garry estimate wage functions on NLSY,
  where mobility is an argument. Find that workers
  who undergo persistent mobility have lower
  log-wage paths than less mobile workers.

47
Jump between wage trajectories
C
B
A
48
Empirical strategy

• Estimate wage functions for all employers
• Wage function including firm specific element
• Observing alternative wage offers?
• Use the past pattern of mobility to limit the
  distribution of possible wage offers to employees
  of particular firm
• Characterizing all wage trajectories by
  individual level, firm wage level, alternative
  wage, and the growth element of these concepts
• Run probit on changing job or not with value of
  future job trajectories as explanatories.

49
Specific Theories for turnover

• Human Capital approach (Becker, 1962 and Oi,
  1962)
• highlights the inverse relationship between job
  mobility and the investments in job specific
  skills.
• you could expect that job changes are associated
  with substantial wage gains

50
Personnel Policy

• This model suggests that firms will use wages and
  employment conditions as policy instruments to
  retain people
• What we find in Bingley and W-N, 2003.
• The firm matches outside wage offers to key
  persons
• Conditions for matching offers
• highly productive/profitable workers
• What keeps society on the edge of efficiency
• Constraints from general concerns about internal
  wage distribution (Hall/Lazear model)
• Too much mobility

51
Data

• Privately employed, white collar, 20-59 years
  old, 1980-2001
• CCP sample of all IDA (Danish Population)
• Continuously employed deleted people with
  longer spells of unemployment
• We observe 1827 big firms and numerous small
  firms (collapsed to 112 industries). No closures
• Link between firms and employees Nov

52
Elapsed tenure since 1964
53
Probability moving
54
Immediate earnings gain
55
Wage gain distributions
56
Distribution of wage gains
57
Percentiles distribution, wage gain for stayers
58
Wage growth for movers
59
Wage concepts

• Calculate
• Person residuals Xb-w
• Xb fe
• Xb fe (tenure0) and fe is weighted
• Change in person residuals
• Change in Xb fe
• Change in Xb fe

• Hypotheses
• A large person specific residual gives
  lower/higher prob moving
• A firm, that pays more will have fewer moving
• If competing firms are paying more, higher
  mobility
• Fast track move less/more?
• Fast wage growing firms have less movements
• Fast wage growth in other firms will increase
  mobility

60
Results, firm random effects, probability of move
61
Conclusions

• Shown that a credible but simple Human Capital
  model can actually contribute to explaining
  mobility.
• Wage differentials play a key role in mobility
• Shows some of the tools firms can use economic
  tools matter
• Interesting to see if
• Less wage compression in another country or over
  time makes a difference?
• regulatory settings make a difference is there a
  higher threshold when regulations?

62
What are the costs to firms of turnover

• Including firm and HRM characteristics

63
Firms perspective

• The real cost of personnel changes are with
  respect to hiring
• Bingley and W-N, 2003 show that cost of expanding
  with one employee can cost as much as 1/3 of a
  man year
• Reducing the work force costs nothing when there
  is no redundancy pay
• So why raid other firms?
• If value of future productivity (incl costs of
  hiriring and recruiting gt future wage costs
• General HC industry clusters

64
Arguments on worker tenure and turnover - demand
side

• Turnover is good for a firm because it brings new
  ideas and new general human capital to the firm
• High wage firms may be more selective in whom
  they retain.
• Firms are more reluctant to lay off employees if
  they possess larger amounts of firm specific HC
  or OC
• Firms may pay a high wage as a compensating wage
  differential for temporary employment or they may
  other benefits

65
Turnover and wage policy

• large differences between firms regarding pay,
  25, Abowd, Kramarz and Margolis, 1999, BW-N,
  1998, Nikolaisen, 1998 on employer data.
• Changes in remuneration systems create quits and
  different selection of new hires
• Safelite Glass, Lazear, 2001
• No relationship between wage growth and tenure at
  firm level, Barth, 1997
• Higher wage level and higher wage growth reduce
  worker flows from mid size and large firms, Barth
  and Dale-Olsen, 1999
• These and other studies are partial - either
  describing one firm or dealing with one
  relationship at a time.
• Leonard and van Audenrode (1997) worker flows neg
  corr with profit
• Leonard, Mulkay and van Audenrode (1999) higher
  productivity pos corr with profit and wage growth

66
Research strategies

• Single firm or multiple firm studies?
• Data availability is often the decisive factor
• Gibbons and Waldman (1999)
• "several of the existing findings come from a
  small number of studies ... and these studies
  typically focus on a very small number of firms.
  Before much more effort is expended on
  developing theories.. it is crucial to know which
  findings are specific to the particular firms
  studied and which are representative of careers
  in a range of occupations, industries, countries
  and time periods"

67
Personnel Policy and Profit by Bingley and W-N,
2003?

• The novelty of this paper
• relationship between wage level, wage growth,
  separations, hires and tenure and economic
  performance of the firm
• we observe 7118 firms (all gt20 employees) in a
  country
• Purely a statistical analysis
• Factors, to be controlled for
• characteristics of the work force, e.g. AGE
• industry effects,
• specific firm effects
• Hard to control for exogenous shocks to the firm,
  even harder for single firms
• endogeneity workplaces setting a high wage for
  other reasons may also retain workers and for the
  same reasons have a good performance - IV methods
  are needed

68
Equations

• Profit function, where X is a matrix of exogenous
  firm characteristics, Y is endog. firm
  characteristics, epsilon ..
• Instrumenting Y's by same X's and Z's (local
  labour market characteristcs)

69
Endogeneity

• we need an instrument that determines (correlated
  with) the personnel policy and which does not
  determine firm performance
• neighbouring labour markets are candidates
• defined as the most popular home municipality
  outside the work municipality
• the 2nd most popular.... the 3rd most......
  the 4th.........
• the means of all firm characteristics in the
  municipality
• unobserved firm specific factors - fixed effects
  estimation takes care of those - everything is in
  deviations from the mean
• Or we have to call for experiments

70
Raw correlations
71
Estimation results, IV, fixed effects
Profit per worker in 1980 DKK
72
Results

• Hires reduce profit - are costly - recruitment
  and hiring costs
• Separations increase profit due to selection
• this indicates that most separations are actually
  layoffs, where the employer has weeded out the
  least productive workers
• Or separations make the remaining workers to work
  harder remember this estimate is only valid in
  a aproximity to the mean

73
Conclusion

• Personnel policy has a strong relation to firm
  economic performance
• More hiring is associated with lower profit
• More separations are associated with higher
  profit
• Shows that separations are not random but rather
  used as a management tool
• Partly justification of the high turnover found
  in Denmark

74
Is it all pecuniary?

• Studies on HRM and job mobility
• Study on job satisfaction and turnover

75
HRM practices

• HR-practices may have different impact on quits,
  examples
• HR high-involvement practices selective hiring
  of people with high general skillsdesign of work
  to provide opportunities for individual
  discretion and ongoing learningongoing
  investment in training, high wages and
  perdormance management systems that build trust
• High inverse relationship between
  High-involvement practices and quits
• Firms less selective in hiring experience more
  quits (Arthur 1994, Shaw, Delery, Jenkins and
  Gupta, 1998)
• The design of work through motivation (Hackman
  and Oldham, 1980)
• Training, employment security, high relative pay
  are likely to induce employee attachment and
  commitment and is negatively correlated to quits
  (CottonTuttle, 1986, Shaw et al, 1998)
• In customer-contact settings, high quit rates
  increase the costs of recruitment, but negatively
  affect sales

76
HRM hypotheses

• High involvement HR-practices contributes to
  create a capable work force, long term employees
  are likely to be more productive and will be
  negatively correlated with quit rates
• Employee quit rates will mediate the relationship
  between high-involvement human resource practices
  and sales growth

77
HRM-methods

• Study of 326 similar call centers (R. Batt, 2002)
• General managers surveyed (less optimistic than
  HRMs
• Measuring high-HRM involvement
• Survey, 3 indexes, additive
• Skill level years of formal schoolingyears
  school ojt needed to become proficient
• Work design employee collaboration in
  teamsindividual discretion
• Involvement enhancing HR-incentives ongoing
  training, employment security, electronic
  performance monitoring
• Quits excluded discharges, retirements,
  transfers and promotions
• Results Coefficient in quit rate function skill
  index () work design-HR-incentives -.
• Sales Growth function coeff to quits neg coeff.
• Causality is an issue!

78
Job satisfaction and turnover

• N. Kristensen and N.W-N, 2004, using ECHP-data in
  the EU-project Epicurus
• Less satisfied workers are more likely to move
• But even the satisfied do move
• Side benefit the value of job security is found
  to be important in UK but job content is more
  important in DK.

79
Attempts to conclude

• Turnover at national level varies a lot
• Regulatory policies are different but outcome is
  not completely predictable firms and workers
  find other roads to optimize
• We do not have a full understanding of the real
  costs to society but have a suspicion
• Turnover varies a lot at enterprise level
• Part is due to regulatory policies
• Probably important for the efficiency of
  companies
• Different levels of turnover calls for different
  HRM practices
• Still a lot of uncharted land