Employment and Unions

1
Employment and Unions
Andrew E. Clark (Paris School of Economics -
CNRS) http//www.parisschoolofeconomics.com/clark-
andrew/
APE Masters Course
2

• The study of institutions, how we came to have
  them, and what effect they have on economic
  outcomes is a growing topic in Economics. And
  arguably is a move towards a more unified social
  science (maybe).
• Common Law (past legal precedents or judicial
  rulings are used to decide cases at hand) or
  Civil law (codified statutes and ordinances)
• Central Bank Independence
• Divorce Law
• Minimum Wages/EPL/the Standard Workweek
• Trades Unions
• I consider the last of these here, and especially
  wonder about their effect on employment.

3
Online OECD Employment database
http//www.oecd.org/document/34/0,3343,en_2649_33
927_40917154_1_1_1_1,00.htmlunion
4
Unions are in decline
5

(No Transcript)
6

(No Transcript)
7

(No Transcript)
8

(No Transcript)
9

• What effect would we expect unions to have in the
  labour market?
• And so what might we expect to see as union
  presence diminishes?

10

• Lets meet an old friend The labour demand
  curve.

This shows how employment is determined in a
competitive market.
11

• The labour demand curve is defined by the choice
  of employment to maximise profit at a given wage.

12

• Derive this formally.
• pF(L) wL
• Along an iso-profit curve, d?0. Totally
  differentiate
• d? pF(L)dL wdL Ldw 0
• So dw/dL (pF(L)-w)/L
• dw/dL 0 where pF(L) w, which is the equation
  of the labour demand curve.

13

• A key difference between workers (labour) and
  other inputs (land, materials, machinery)
  workers can express themselves, and bargain. In
  particular, they cannot be forced to work.
• They can form unions.
• What are the implications of trades unions for
  the level of employment? Both at the firm (micro)
  and the economy (macro) level.
• Macro is not ? micro representative agents may
  lead us astray

14

• Empirical regularities regarding unions.
• 1) Productivity
• Higher union wages might shock the firm into
  being more productive
• I used to think of this is being a bof
  argument.
• But I have now changed my mind. This is very
  simply set out in Mayneris, Poncet and Zhang
  (2015), Improving or Disappearing Firm-Level
  Adjustments to Minimum Wages in China, IRES
  Discussion Paper No. 2016-27.

15

• Higher union wages may yield firm-level
  efficiency gains.
• Assume that firms have to choose between two
  production processes
• A high-tech process with low constant marginal
  labor requirements but a high fixed adoption cost
• A low-tech process with high marginal labor
  requirements but no adoption cost.
• Higher wages widen the marginal-cost gap between
  the high- and low-tech technologies.

16

• Keeping quantities constant, the opportunity cost
  of adopting the high-tech process falls with
  wages.
• Firms which may have previously preferred the
  low-tech process may switch to paying the fixed
  cost required for the high-tech process.
• Low wages here act as a disincentive for the
  adoption of more efficient production techniques.
• Unions, and indeed minimum wages, overtime rates
  (les 35 heures) and labour taxes can increase
  productivity

17

• b) Improvements in worker morale and cooperation
  union provides a collective voice (rather than
  exit).
• Implication of b) labour turnover should be
  lower amongst union members, ceteris paribus.
• c) Protect membership and block innovation
  (Luddites). Which reduces worker effort (and so
  productivity)

18

• Empirical evidence from the US has produced only
  mixed evidence some positive, some negative.
• This is true in both levels and growth
• UK evidence suggests a lower productivity level
  results in terms of productivity growth are again
  mixed.
• Bear in mind that productivity is famously
  difficult to measure, and is most often
  calculated as a residual.

19

• Measure output as Q f(K, L, M)
• Log-linearised as
• Q alnK ßlnL ?lnM
• Productivity measured as Q QQ-hat
• Very noisy

20

• 2) Capital
• The other factor of production
• Unions tend to be associated with lower levels of
  physical investment and RD.
• But unions are associated with greater levels of
  worker training (which is unsurprising if they
  are also associated with lower worker quit rates
  as workers stay on longer, it is more likely that
  the firm will recoup its investment).

21

• 3) Working Conditions
• A positive effect (unsurprising)
• Dont forget causality. Unions may improve
  conditions, but poor work conditions predict
  unionisation too.
• Our conditional correlation coefficient here
  measures the net effect of positive (direct) and
  negative (selection) phenomena
• So that it represents a lower bound of the causal
  effect.
• The opposite selection effect is found for
  police and crime
• The same causal argument can be made with respect
  to the productivity findings. If low-productivity
  firms are more likely to unionise then the
  negative productivity-union relationship in
  cross-section is biased downwards.

22

• 4) Wages
• Unions are associated with higher worker wages
  they are also associated with lower wage
  inequality (the wage effects of unions are felt
  most strongly for the low-paid).
• There is a direct union effect via their
  bargaining wages for covered employees.
• There are also indirect effects
• Longer tenure ? firm-specific human capital
  investments
• Threat of unionization raises wages in the
  non-union sector.
• Job losses in union sector, resulting in excess
  labor supply to non-union sector.
• These will affect the wages of both members and
  non-members

23

• Its actually not that obvious to come to a
  conclusion about the union wage effect.
• The union wage differential is given by
• di(wiU-wiN)/wiN
• In practice, we dont observe the same individual
  i as both a union member and a non-union
  member. We thus compare across individuals
• d(wU-wN)/wN

24

• The problem here is that union members (who give
  us the wU) might have very different observed and
  unobserved characteristics to non-union members
  (who supply the wN figure).
• i.e. There is omitted variable bias.
• Typically estimate
• Ln wage ?U ßX e

25

• Ln wage ?U ßX e
• We then then have (total differentiation)
• dw/w ?dU
• So that the elasticity of wages w.r.t. unions is
• dw/dU x U/w (?w/w) x 1 ?

26

• This is a partial-equilibrium estimation. In
  practice the wage of non-union members will be
  determined by some regression as well.
• And the probability of union coverage will be
  some function of the difference between union and
  non-union wages.
• Difficult to do this well need an exogenous
  change in union membership (that does not
  directly affect wages).

27

• Estimates of the mark-up in the UK are around
  7-12.

28

• The wage premium in the UK and the US has been
  pretty stable over time.
• Unions have persisted in high mark-up sectors
  (composition effect)
• Weaker unions reduce non-union wages too
• Rising competition will not affect industries
  with market power or the State sector.
• The union wage effect is higher for manual and
  low-skilled workers (wage compression, as noted
  above).

29

• The union wage premium has been suggested as one
  reason why union density has been falling (wage
  effect larger than the productivity effect) but
  this doesnt explain why the mark-up hasnt faded
  away above.
• Another is the change in industrial composition.
• A third is institutional change in the legal
  environment, worker preferences, government
  regulation of the labour market.

30

• 5) Job Satisfaction
• Typically found to be negative.
• Careful of causality (desire for union
  representation is negatively correlated with
  satisfaction)
• Are union jobs really worse, or do unions just
  allow people to moan more (give them a voice)?
  Test this via an objective correlate of
  satisfaction quits.
• Model
• quit q(X, satisfaction, union,
  unionsatisfaction).

31

• What happens when workers unionise?
• Union utility function
• U Lu(w) (M-L)u(b)
• M is union membership, L is employment of union
  members. Totally differentiate for the
  indifference curve
• dLu(w) Lu(w)dw u(b)dL 0
• So dw/dL -u(w)-u(b)/Lu(w)

32

• Utility increases North-Westwards.
  Negatively-sloped as long as u(w) gt u(b)
  indifference curves become horizontal at wb.

At wb utility from employment and unemployment
is identical the union does not care about
employment
33
Models of Union Behaviour

• MONOPOLY UNION (Dunlop)
• The union makes all of the decisions here, by
  choosing both the level of wages and the level of
  employment, subject only to the constraint that
  the final point be on the labour demand curve.

34

• 2) RIGHT TO MANAGE (Nickell and Andrews)
• More realistic. Employment is chosen by the firm,
  wages are bargained over.
• Nash bargaining solution
• Max B Lu(w)(M-L)u(b)?pF(L)-wL 1-?
• w
• s.t. the labour demand curve (pF(L)-wL)
• The parameter ? reflects the unions relative
  bargaining power.
• If ?0 then the firm has all the bargaining
  power
• If ?1 then the union has all the bargaining
  power

35
RTM solution is somewhere along the LD curve,
between wMU and b.
36
Both the Monopoly Union and the RTM models have
the implication that higher wages are associated
with lower employment because we are moving up
and down the labour demand curve. All of the
MU/RTM solutions are on the labour demand curve.
All of these solutions are (in general)
inefficient. We therefore turn to the.
37

• 3) EFFICIENT BARGAIN (McDonald and Solow)
• Here both wages and employment are negotiated
  between the firm and the union. We have the same
  Nash maximand as above, but now maximised over
  both w and L.
• The EB reflects the general principle that for
  efficiency, we should always bargain over all of
  the elements upon which our utility depends.
• Only let others decide when we do not care about
  these elements (when our utility does not depend
  on them).

38
An EB point is not
The MU point is inefficient
Efficient points are given by the tangent between
the IC and the iso-profit curve. At any other
point, we can make one of the firm and the union
better off without making the other one worse
off. Tracing these out produces an upward-sloping
curve in (w, L) space. Note that when
indifference curves are flat, the tangent to the
iso-profit curve will be on the labour demand
curve. This is true at wb.
39
CC the Contract Curve
40
In the EB model, there is a positive relationship
between wages and employment. Moving up the CC
curve implies greater union bargaining power. The
union takes part of the payoff of more power in
wages, part in employment. The problem with the
EB is that it doesnt seem to describe the real
world very well. While unions do negotiate over
wages, they most often probably dont negotiate
over employment. Therefore in most of the
bargains that we observe, higher wages are
associated with lower employment. As we now that
unions are associated with higher wages, surely
its a no-brainer to say that they reduce
employment?
41
The Macroeconomics of Trade Unions

• The microeconomic analysis above was partial
  equilibrium. We didnt consider the effect of one
  firms wage and employment decisions on other
  firms.
• In the macro analysis, we will take these effects
  on board. This changes the results.
• Consider an economy with trade unions. There are
  a number of different dimensions along which we
  can describe the degree of unionisation of that
  economy.
• The percentage of workers who are union members
  (Union Density)
• The percentage of workers who have their pay and
  working conditions decided by union bargaining
  (Collective Bargaining Coverage)
• How many unions there are
• At which level do unions bargain firm, industry,
  economy? (Degree of centralisation)
• Do different unions coordinate their bargaining
  activities? (Degree of coordination)

42
Inspiration for this literature
The wildly different macro performances of OECD
countries in the 1970s and 1980s following the
first two oil-price shocks. Some, such as Japan,
Austria and the Nordic countries, seemed to have
levels of unemployment and inflation that were
persistently lower than those in other
countries. The bad pupils the rest of Western
Europe, and North America. Different macro
performance could come from macro variables
(interest rates, exchange rates) or from
institutions. We here particularly think of
trades unions.
43
The initial focus of this work was on bargaining
centralisation. This can range from the
establishment-level right up to the country level.
Branch/ Industry
Establishment
Firm
Country
OECD countries operate at wildly-different points
on this scale see my attempt at a ranking in
Table 3.3. of the Employment Outlook chapter.
Nordics to the right N. America to the left
much of Western Europe in the middle There is
movement between countries, and within countries
over time.
44
So what?

• We are interested in this because of its
  potential impact on economic performance.
• A distinction (vastly oversimplified)
• Eurosclerosis school centralised bargaining is
  a rigidity which prevents labour markets from
  clearing
• Corporatists believe that centralised
  institutions may help us to overcome market
  failure.
• Both imagine a linear relationship between
  bargaining centralisation and performance this
  assumption was challenged in a well-known paper
  by Calmfors and Driffill in Economic Policy in
  1988.

45
CD do think that there is a ranking in terms of
centralisation and economic performance they
just suggest that the middle is worse than the
two ends. Initial analysis carried out in terms
of the unemployment rate.
Unemployment
Centralisation
Based on two key elements the first is
externalities.
46
Externalities

• Are others hurt by the higher wage a union
  bargains for its members?
• My members receive higher wages, and are thus
  happy.
• What about others?
• Calmfors (1993) identifies six potential routes
  via which higher wages for a union might reduce
  the utility of those not covered by the bargain.
• Consumption prices (lower real wages)
• Input prices
• Fiscal effect (have to pay for others
  unemployment)
• Unemployment (less chance of getting another job
  if you are yourself unemployed).

47
5) Relative utility uu(y/y) 6) Efficiency
wages. If ee(y/y), then effort by uncovered
workers will fall. Suppose that individuals are
not altruistic. To what extent will these
externalities be taken into account? Decentralised
bargaining will take none of them into account
those who profit from higher wages represent only
a small percentage of those who are hurt by
them. As we move from left to right on the
centralisation scale, negations cover a greater
percentage of people we take more notice of the
costs of higher wages relative to their
benefits. At the logical extreme, if we bargain
for everyone in the country, then those who
profit and those who are hurt are the same
people externalities are internalised.
48
Internalising externalities/centralisation leads
to wage moderation. The relationship between
centralisation and wages is negative, therefore
so is that between centralisation and
unemployment.
Unemployment, wages
Centralisation
The second key element is the price elasticity of
demand, which determines the derived elasticity
of labour demand.
49
By what percentage will employment fall if wages
rise by 10? Monopolists pass all of the wage
increase onto prices, if demand is totally
inelastic, so there is no employment effect. As
the number of rival products or substitutes
rises, the price elasticity of demand rises, and
so does the elasticity of labour demand. Under
perfect competition, only a slight increase in
wages and thus price leads the firm to lose all
of its market. Price elasticity imposes market
discipline the lower is the level of
centralisation, the more market substitutes there
are, the greater is the elasticity.
Decentralisation brings wage moderation
50
Unemployment, wages
Centralisation
Put these two elements (internalisation and
elasticity) together and youve got what? CD
argue that this produces a hump-shaped
relationship between unemployment and
centralisation. I reckon they might be right but
I also think it could produce all kinds of other
shapes too.
51

• Lets take it to the data to decide.
• CD used four performance indicators.
• Unemployment rate
• Employment rate
• Okun Index (unemployment plus inflation rates)
• API (unemployment plus current account deficit as
  of GDP)

52
CD analysed 17 OECD countries, 1974-1985. Classifi
ed according to centralisation of the collective
bargaining system split up into three
groups. The first two columns in each panel of
Table 3.2 show CDs results. The next two
columns in each panel show my updating to
1986-1996 using the same countries and method
53
Note jacknife-style worries about the
classification of Switzerland
54
Unemployment rate Centralised best ?unemployment
is hump-shaped. Less obvious when
updated. Employment rate U-shaped in 1974-85.
Less obvious when updated. Okun Index Not
U-shaped. API Sort of hump-shaped, perhaps.
55

• Drawbacks of CD
• No statistical testing. Some of these means
  (many?) are not significantly different from each
  other.
• Bivariate doesnt take into account density, CB
  coverage or coordination.
• Countries are assumed to have the same
  classification over a 20-year period, even though
  there is substantial movement.
• Update CD to mid-90s. Add Spain and Portugal.
  And run regressions.
• Classifications of countries in Table 3.3.

56
(No Transcript)
57
A) Centralisation. Higher in Nordic countries
lower in Canada, Japan, NZ and the US. A
decentralisation tendency in the UK, Australia,
Denmark and NZ. A centralisation tendency in
Italy and Portugal. B) Centralisation is no use
if centralised agreements are renegotiated at
more local levels. Take union coordination into
account. This is higher in Austria, Germany and
Japan lower in Canada, the UK, NZ and the US. A
tendency towards less coordination.
58
(No Transcript)
59
A) TU Density. France is strange. TU density has
been drifting downwards. OECD Trade Union
Density 1980 1990 1994 46 40 40 It is
currently 17. Because TU density is falling in
countries, and newer members (Turkey, Mexico)
have low TU density rates. B) CB coverage is
usually greater than membership. See Figure
3.1. OECD Collective Bargaining
Coverage 1980 1990 1994 72 70 68
60
One of the big issues in this literature has been
the wild proliferation of indices. There are as
many indices of country bargaining regimes as
there are authors, I reckon. Happily, they do
seem to be pretty well correlated (see Table
3.4).
61
(No Transcript)
62

• Simple correlations
• I have five macro performance indicators
  (five-year averages)
• Unemployment rate
• Employment rate
• Inflation
• Real wage growth
• Wage inequality

63
(No Transcript)
64
Table 3.5 shows Spearman rank correlation
coefficients. Thus looking for a linear
relationship. Not much comes out of that
(although unions associated with less wage
inequality). To identify non-linearities, recode
rank. Instead of 1, 2, .., 18, 19 use an
ascending-descending rank (middle worse or better
than ends?) 1, 2, , 8, 9, 10, 9, 8, .2,
1. Any positive correlation implies
middle-ranked countries have more of whatever it
is were looking at than the ends do. Only
result a hump-shape with inflation, but even
this goes away by 1994.
65

• Pooled Regressions.
• Make up dummies for centralised/coordinated and
  intermediate.
• Countries can change classification over time.
• The omitted category is decentralised.
• If there is indeed a U-shaped relationship, then
  intermediate should do worse than decentralised
  (the omitted category), and so attract a negative
  estimated coefficient.
• We dont have a strong prediction about whether
  centralised/coordinated should do better or worse
  than decentralised (the two ends of the U-shape).

66
should have worse performance than the OC
(decentralised) and centralised
67
OLS regressions in Table 3.6. N 57
observations. TU membership E, -
inequality CB coverage unemployment,
inflation, -E Centralised - unemployment,
inflation, inequality Intermediate -
inflation, inequality. What a mixed picture. One
thing we can say is that there is no evidence of
a multivariate hump-shape. Strong inequality
result. Is centralised best after all?
68
Fixed-Effect Regressions. Look at changes within
one country over time. See Figure 3.2. What
happened to countries that decentralised/became
less coordinated? Note that we are back to
looking at a monotonic relationship
here Unemployment rose, employment fell, and
inequality rose.
69
(No Transcript)
70
(No Transcript)
71

• Conclusion
• Micro union theory suggests that higher wages
  increase unemployment.
• Macro theory suggests that there is a hump-shape
  with respect to centralisation.
• Difficult to find clear evidence.
• Unions certainly raise wages. But they might
  raise productivity too (human capital, discourage
  quits, voice leading to higher effort). So were
  still on the labour demand curve, but the whole
  thing has shifted to the right.
• We have considered only price-taking firms. When
  firms have market power, then the effect of wages
  on employment is not so clear (see Alan Mannings
  Monopsony work).
• All the empirical literature is probably flawed
  anyway unionism is not exogenous. Find a good
  instrument (recognition ballots?)

72

• Am I out of date?
• Is anyone really interested in unions anymore
  anyway?
• To put it another way, will North American
  journal editors publish me if I work on this?

Your Journal Editor Lives Here
73
(No Transcript)
74

• Union density is certainly in free fall in the
  US.
• 2016 BLS figures. 14.6M union members in the US
  (10.7) 16.3M covered workers (12.0).
• Private union density is now 6.4 (public sector
  density is 34.4).
• But thats not so true in Europe.
• One phenomenon which is on the rise in the US is
  licensed occupations.

75
(No Transcript)
76

• Occupations are becoming increasingly restricted.
  
• Have to meet certain standards to practice the
  profession
• Or have to be certified to use the job title
  (others can carry out same duties, but not use
  the job title).
• As such, there are restrictions on entry and the
  supply of labour
• Which increases the price of services and wages.
• Applies to Dentists and Doctors (thank
  goodness)... and Lawyers.
• But also to hairdressers and florists

77

• Licensed Occupations now account for over 20 of
  jobs. Great deal of inter-State variation from
  5 in Kansas to 30 in California.
• There is huge State variation in union density in
  the US as well, from New York (25) to North
  Carolina (2) figures for Kansas and California
  are 7.4 and 16
• Argument for licensing is that it increases the
  quality of service.

78

• Yet Kleiner finds no cross-State difference in
  customer complaints in occupations that are
  licensed in Wisconsin, but not in Minnesota.
• Equally insurance premia for malpractice by
  medical practitioners are the same in licensed
  and unlicensed States.
• Licensing is associated with restrictions on
  labour supply, however (reduced the growth rate
  of employment by about 20).
• Consequently led to higher wages (10-17) sound
  familiar?

79

• Postcript
• Two questions about institutions
• What effect do institutions have on outcomes?
• Why do we have institutions in the first place?
• There are likely factors, Z, which help bring
  about institutions.
• We have an empirical problem if the same Z have a
  direct effect on outcomes too.
• Instrumenting institutions is a growth area in
  empirical economics