The Determinants of Trade Union Density in CrossCountry Comparisons: Theortical Opulance and Empiric

1
The Determinants of Trade Union Density in
Cross-Country ComparisonsTheortical Opulance
and Empirical Destitution

• Bernd Brandl
• University of Vienna
• Department of Industrial Sociology

1st TURI network Conference The future of trade
union structures and strategies
2
Introductory remarks and motivation

• Researcher offer a number of determinants which
  possibly explain differences in the level of
  trade union density across countries and shifts
  over time
• From a macro-perspective (socio-)economic,
  political and institutional factors are used to
  explain differences in trade union density across
  countries and over time
• Empirical studies (usually) use aggregate pooled
  time series to investigate the empirical
  relevance of determinants on basis of different
  models
• Results of empirical studies are mixed
  regarding the relevance of specific determinants

3
Introductory remarks and motivation

• The heterogeneity in empirical results is
  unsatisfactory for researchers and policy
  makers as there is uncertainty about what
  determinants are really important
• The aim of this paper is to investigate
  systematically the empirical relevance of
  determinants in explaining variations of trade
  union density across countries and over time
  using a Bayesian Model Averaging approach
• The work allows the identification of variables
  that are robust in explaining trade union
  density, i.e. provide explanatory power
  independent of what (specific) theory or model is
  used!

4
Contents

• Determinants of trade union density
• The empirical relevance of determinants and model
  uncertainty
• Bayesian Model Averaging and robust determinants
• Summary and conclusions

5

• Determinants of trade union density

6
I. Determinants of trade union density

• Trade union density A country comparison
• Averages from 1970 to 2000.
• Source OECD

7
I. Determinants of trade union density

• Trade union density A country comparison
• Shifts over time from 1970 to 2000.
• Source OECD

8
I. Determinants of trade union density

• Literature is rich in offering theories and
  models that explain trade union density (across
  countries and in time)
• Literature is also reach in specifying
  determinants (factors, variables) that explain
  differences in trade union membership across
  countries and changes over time
• Institutional determinants (number of union trade
  unions, organization of trade unions ...)
• Economic determinants (unemployment rate,
  business cycle )
• Socio-economic determinants (demographic
  characteristics, structure of the economy,
  education of population, religion )
• Political determinants (Corporatism, political
  orientation of government )

9
I. Determinants of trade union density

• Analyzed variables in cross-country studies

10
I. Determinants of trade union density

• Analyzed variables in cross-country studies

11
I. Determinants of trade union density

• Analyzed variables in cross-country studies

12
I. Determinants of trade union density

• Analyzed variables in cross-country studies

13
I. Determinants of trade union density

• Analyzed variables in cross-country studies

14
I. Determinants of trade union density

• Analyzed variables in cross-country studies

15
I. Determinants of trade union density

• Analyzed variables in cross-country studies

16
I. Determinants of trade union density

• Analyzed variables in cross-country studies

17
I. Determinants of trade union density

• Analyzed variables in cross-country studies

18
I. Determinants of trade union density

• Analyzed variables in cross-country studies

19
I. Determinants of trade union density

• Analyzed variables in cross-country studies

20
I. Determinants of trade union density

• Analyzed variables in cross-country studies

21
I. Determinants of trade union density

• Analyzed variables in cross-country studies

22
I. Determinants of trade union density

• Analyzed variables in cross-country studies

23
I. Determinants of trade union density

• Analyzed variables in cross-country studies

24
I. Determinants of trade union density

• Analyzed variables in cross-country studies

25
I. Determinants of trade union density

• Analyzed variables in cross-country studies

26
I. Determinants of trade union density

• Analyzed variables in cross-country studies
• 76 Variables
• 67 Interaction terms (variables)
• 143 Regressors
• There are much more other variables which are
  reasonable!
• Further lags
• Further interactions

27
II. The empirical relevance of determinants and
model uncertainty
28
II. The empirical relevance of determinants and
model uncertainty

• (Usually) theoretical literature offers testable
  predictions and can used as a basis for empirical
  studies
• Unfortunately, not only predictions of theories
  are differing, but also the results of empirical
  studies are heterogeneous.
• One reason for the heterogeneity Estimation of
  different models!
• Depending on what combination of regressors the
  investigator chooses to put into his regression
  different significant determinants of union
  density are achieved!

29
II. The empirical relevance of determinants and
model uncertainty

• Examples

30
II. The empirical relevance of determinants and
model uncertainty

• A phenomenon that can be observed in
  empirical studies
• One study (based on a specific theoretical model)
  concludes that a specific variable has no
  significant influence
• Another study (based on a slightly different
  specification) concludes that the same variable
  has a significant positive influence
• Another study (based again on a slightly
  different specification) concludes that the same
  variable has a significant negative influence
• Reasons for estimating different models or for
  the existence of model uncertainty
• Theories are different (and competing)
• Theory is not precise enough in offering the
  true model so that empirical researchers have
  to check alternative specifications
• Small sample size (number of observations
  limited) so that a selection has to be made

31
II. The empirical relevance of determinants and
model uncertainty

• Comments on model uncertainty
• Model uncertainty is pervasive in social science
• There is (almost) no theory in social science
  that is strong enough to dictate a single model
  specification
• All models are wrong, some are useful Box
  (1979)
• The typical case is one in which a number of
  variables are plausible predictors
• The problem is how to decide which specification
  to use
• Bayesian model averaging is (at least) an
  interesting approach in this context

32
III. Bayesian Model Averaging and robust
determinants
33
III. Bayesian Model Averaging and robust
determinants

• Bayesian Model Averaging (BMA)
• Bayesian statisticians reject idea of a single,
  true estimate
• Instead each parameter has a distribution!
• It is not believed that any of the models is
  actually correct all models are used as proxies
  for some unknown underlying model
• BMA provides a coherent mechanism for accounting
  for model uncertainty as probabilities to
  different possible models are attached
• The idea of BMA is to average across several
  models instead of selecting one model
• It takes the K variables and runs a regression on
  all 2K subsets of the K variables, before
  averaging over all these models

34
III. Bayesian Model Averaging and robust
determinants

• Bayesian Averaging of Classical Estimates (BACE)
• Sala-i-Martin, Doppelhofer and Miller, AER, 2004
• The BACE approach constructs estimates by
  averaging weighted OLS coefficients across models
• The weights given to individual regressions have
  a Bayesian justification similar to the BIC
• Advantages of the BACE approach
• In contrast to a standard Bayesian approach that
  requires the specification of a prior
  distribution for all parameters, BACE requires
  the specification (assumption) of only one prior
  hyper-parameter the expected model size ( 7).
• The interpretation of estimates is
  straightforward
• the weights applied to different models are
  proportional to the logarithm of the likelihood
  function corrected for degrees of freedom.

35
III. Bayesian Model Averaging and robust
determinants

• Determinants of trade union density BACE
  results
• Dependent variable(s)
• Share of trade union members in relation to the
  total number of employees (i.e. trade union
  density)
• Yearly percentage change
• (Source and definition OECD)
• Cross-country panel data set (balanced)
• OECD countries (Australia, Austria, Belgium,
  Canada, Denmark, Finland, France, Germany,
  Ireland, Italy, Japan, Netherlands, New Zealand,
  Norway, Portugal, Spain, Sweden, Switzerland, UK,
  USA)
• Time 1970 to 2000
• (missing values!)
• 55 (theoretically grounded) variables
• Number of trade unions strike activity
  Ghent-System participation of unions in
  socio-economic policy making centralization of
  wage bargaining, unemployment rate, trade
  openness,

36
III. Bayesian Model Averaging and robust
determinants

• Determinants of differences in the level across
  countries
• Robust determinants - Level ( 15)
• Number of union confederations (-)
• Ghent-System ()
• Inflation ()
• Union activities in policy-making ()
• Business activities in policy-making (-)
• Centralisation of bargaining level (-)
• Yearly change of wage and salary earner
  Extension practice Closed/Union shop practice,
  Bargaining governability, Change of compensation
  per employee, change of wage rates, change of
  unit labour costs, change in productivity

37
III. Bayesian Model Averaging and robust
determinants

• Determinants of trade union density
• Robust Determinants - Change ( 10)
• Ghent-System ()
• Left parties in government ()
• GDP growth (-)
• Trade openness (-)
• Change in total labour force (-)
• Change in unemployment rate ()
• Change in productivity () Population growth
  (-), Change employment share (-), one period
  lagged change in union density ()

38

• Summary and conclusions

39
IV. Summary and conclusions

• What we already knew!
• Trade union density is different in different
  countries
• Trade union density is declining in many
  countries
• There are a lot of theories and studies available
  which aimed to explain differences in the level
  of trade union density and the decline over time
• We also knew that some theories and studies are
  saying this and some are saying that
• I knew that that this is not satisfactory
• at least I have the impression that the current
  state of research is unsatisfactory because we do
  not know what to do after reading all these
  studies

40
IV. Summary and conclusions

• What do we know now?
• We still do not know which theory is the true
  one
• But we know that some theories are explaining
  trade union density better than other theories
• And there are variables which are robustly and
  highly correlated with trade union density and
  there are variables that are not correlated with
  trade union density
• These variables are of GENERAL relevance (not
  only specifical)!
• For the problem at hand the paper showed that
• Only 15 variables (out of 55) are robustly
  correlated with the level of union density
• Only 10 variables (out of 55) are robustly
  correlated with the change

41
IV. Summary and conclusions

• What do we know now?
• We still do not know which theory is the true
  one
• But we know that some theories are explaining
  trade union density better than other theories
• And there are variables which are robustly and
  highly correlated with trade union density and
  there are variables that are not correlated with
  trade union density
• These variables are of GENERAL relevance (not
  only specifical)!
• For the problem at hand the paper showed that
• Only 15 variables (out of 55) are robustly
  correlated with the level of union density
• Only 10 variables (out of 55) are robustly
  correlated with the change
• In fact Only few determinants are able to
  explain trade union density
• BUT These few determinants are able to provide
  us an instrument for changing the situation,
  i.e. policy makers (for example trade unions)
  may use these few variables to increase the
  number of members in trade unions!
• The general relevance allows a very high degree
  of certainty that something can be changed!

42
IV. Summary and conclusions

• The relevance of these (few) variables for trade
  unions
• Trade union density (memberships) varies with the
  business cycle
• Variables in particular unemployment, inflation
  and economic growth
• hard to use as a policy instrument for trade
  unions
• Trade union density depends on country specific
  traditions (i.e. Ghent system) and on global
  economic trends
• traditions and global trends are also hard
  to use a
• policy tool (instrument)

43
IV. Summary and conclusions

• The relevance of these (few) variables for trade
  unions
• There are two determinants that might be
  considered by trade unions to increase their
  memberships
• Centralisation of collective bargaining
• Trade unions should bargain collective agreements
  on the right level, i.e. not too central and
  not too decentralized
• There is an optimal level of bargaining in
  between
• Number of union confederations
• The more united and integrated unions are the
  more members they have!
• It is not easy to unify different trade unions
  because of different traditions
• but it is possible because trade unionists are
  the ones who are able to change the situation

44
IV. Summary and conclusions

• FINAL REMARKS
• THERE ARE DEFENITELY NO EASY WAYS (NO EASY
  INSTRUMENTS) FOR TRADE UNIONS TO INCREASE THEIR
  MEMBER SHARES
• BUT THERE ARE WAYS!
• The analysis identified ways and instruments
  that can be used and which are (extremely)
  empirically relevant.
• There might be other ways (country specific
  ways) but the two ways identified by this work
  will work with a high probability!