is there a secular trend in the public opinion of unions in canada over time

64
UNIVERSITY OF WATERLOO Department of Economics Have Canadians Become More Anti- Union? An Analysis of Public Opinion Poll Data Spanning Three Decades. Mark Vanspall

Upload: mark-vanspall

Post on 12-Apr-2017

45 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Is there a secular trend in the public opinion of unions in Canada over time

UNIVERSITY OF WATERLOODepartment of Economics

Have Canadians Become More Anti-Union? An Analysis of Public Opinion Poll Data Spanning

Three Decades.

Mark Vanspall

20278874

Econ 472-Senior Honours Essay

Mikal Skuterud

Dec 14, 2012

Page 2: Is there a secular trend in the public opinion of unions in Canada over time

Abstract

This paper attempts to identify whether or not there is a secular trend in the public

opinion of unions in Canada. Determining whether or not there is a secular trend in the public

opinion of unions has important implications for both unions and governments as it could

facilitate better decision making. In the case of governments, public opinions could be assessed

to estimate the popularity and support of strike-breaking or anti-union legislation, while unions

could assess public opinion to try and determine the probability of success of strikes and other

strategic negotiation tactics.

Looking at public opinion poll data from the years 1976 – 2008 there is a clear increasing

trend in the public opinions of unions. In this paper OLS regression methods are used to attempt

to identify the determinants of the public’s decision making process of whether or not to support

unions. Determinants such as: age, geography, education, inflation rates, media coverage, union

density, strike activity and the gini coefficient are examined.

No identifiable secular trend emerges in the public opinion of unions over time; however,

several key variables have predictive power that reinforces a secular time trend in opinions;

which indicates that there could, in reality, be an increasing secular trend in public opinion from

1976-2008. Further, inflation is identified as the biggest determinant of trends seen in this area.

Page 3: Is there a secular trend in the public opinion of unions in Canada over time

Have Canadians Become More Anti-Union? An Analysis of

Public Opinion Poll Data Spanning Three Decades.

1 Introduction

There are many reasons one may be interested in the public opinion of unions. For a

perspective union, formation and effectiveness can both be related to public opinion. Unions

require mandatory dues and as such must have a positive opinion within their membership to

form and continue operations. Public opinion of unions can be related to the success of general

bargaining and labour negotiations as a hostile climate can hurt solidarity and hamper the ability

of a union to achieve its goals (Bok and Dunlop 1970). Union density is falling over time and

union leaders are looking to maintain power and knowing public opinion trends could enable

more effective policy and growth plans (Poole 1981). Unions may need to develop new ways of

operating to be successful if the climate in which they operate is changing (Lewenza 2012).

Collective bargaining can be a drawn out process and involve the threat or use of strikes, and

public opinion could clearly be a deciding factor. The battle for public opinion can be easily seen

in the current NHL lockout as it is clearly important for both sides. In the 2004 – 2005 lockout

public opinion was identifiably skewed to the owner’s position and this helped pressure the

players into massive concessions (CBC 2012).

On the flip side, legislators and executives may take cues from public perceptions in how

they approach general bargaining and strikes. In Canadian legislation, certain industries have

been declared essential services, and can be denied the right to strike and are instead offered

binding interest arbitration (Treasury Board of Canada Secretariat). Given the political import of

1

Page 4: Is there a secular trend in the public opinion of unions in Canada over time

these decisions it is not unreasonable to expect that public opinions may dictate which services

are declared essential. The TTC strike of 2011 is a prime example of this. In many communities

public transportation while important for many people would not be considered essential. In

Toronto, where a much larger percentage of the population uses public transportation, Mayor

Ford had an easy time pushing for the essential designation and the Toronto Transit Commission

Labour Dispute Resolution Act was passed doing just that. From a policy standpoint many

government and public workers are unionized and public opinion is absolutely necessary for

leaders deciding on wages, in addition to decisions restricting the right to strike of certain

workers. Recent laws passed in Michigan limiting unions, in particular mandatory membership

and dues, have sparked protests and anger that also illustrates the clear link between legislative

decisions about unions and public perception (San Francisco Chronicle). Understanding

Canadian opinions about these issues and unions as a whole can help Canadian legislators if they

decide to enact similar measures, as Ontario Premier hopeful Tim Hudak promises to do if

elected (Globe and Mail).

2 Literature Review

Despite the obvious benefits of assessing the public opinion trends, very few studies

addressing unions are longitudinal in nature (Klandermans 1986). Previous research has

examined the trend of British public opinion over time, specifically between 1971 and 1981

using Gallup poll responses, and found they were generally stable (Martin and Little 1986).

Further research into a positive growth in public opinion in the Britain beyond 1979 found that

the tendency was due to the reduction of inflation and the volume of strikes (Bain and Edwards

1988). Strikes can be disruptive, and affect the lives of non-union members and possibly could

2

Page 5: Is there a secular trend in the public opinion of unions in Canada over time

be violent, providing a militant view of unions to independent observers. A study on declining

union density within the United States found that declining union density was highly correlated

with the decline of public opinion of unions (Lipset 1986). This could be argued either way

though as it could be that as union density falls there is less awareness of unions, or exposure to

union advertising, that leads to a decrease in public opinion, or that decreasing public opinion

causes a fall in union density.

With respect to the Canadian studies Riddell conducted a cross sectional comparison of

the years 1950-58 and 1976-82, again using the Gallup poll responses, and found that Canadian

public opinions are becoming less favourable (Riddell 1986). This poses the question of whether

this trend continues and whether it is time trend or reflective of different economic conditions

between the 50’s and late 70’s. Public opinions of unions do appear to be sensitive to economic

and socio-political changes (Johnston 1986). Lowe and Krahn conducted a cross-secional study

on Winnipeg and Edmonton for the years 1987 and 1981 and found that a negative view of

unions, specifically that they are too powerful, was not increasing over time. However, they did

suggest that the key causal agent for increasing support in Calgary as opposed to Winnipeg was

the result of poor economic performance and high unemployment. They also posited that

community studies could be of greater benefit than poll data for interpolating trends (Lowe and

Krahn 1989). While this paper is interested in Canadian opinions, it will be beneficial to look

into provincial data when possible to try and capture this sentiment.

Another purported determinant of the public opinion of unions is media coverage of

unions. This could be related to the level of strike activity as media coverage of strikes is biased

towards strike reporting (Schmidt 1993). It could be that this bias tends to negatively affect

public attitudes towards unions over time (Schmidt 1993). It may also be that media is more of a

3

Page 6: Is there a secular trend in the public opinion of unions in Canada over time

moderating influence and does not create this negative view (Duffy 1981). It is clear from

previous research examining the factors of public opinion in general and over time that there are

many contradicting conclusions depending on the country that is studied, the sources used, as

well as the timeframe observed. This may be because there are valid reasons to form an opinion

either way depending on the specific outcome being looked at. In the case of conflicting effects

of inflation it could be argued unions are a bad thing as they drive up the price level and cause

inflation, or that unions are good thing for workers as they enable quicker raise adjustments in

the case of rising price levels (Lowe and Krahn 1989). Another area with well-established

conflicting opinions is the social outcomes that could result from unions. People may view

unions as a tool to raise the standards of living of the lower ends of society, but at the end of the

day it may be that governments are a more effective means for addressing job concerns

(Johnston 1986).

3 Methodology

3.1 Initial Data

This paper will use survey responses to the question: “How much confidence do you have

in unions?” from the Gallup polls and the Canadian Election Survey to attempt to assess whether

there is a secular trend in the opinion of unions over time. When looking at an individual’s

support of unions, it may be important to note that an individual’s opinion of unions may vary

based on different aspects of the union activity (Furnham 1984). To account for this, several

macroeconomic variables, that may be viewed as outcomes of unions can also be examined. For

the timeframe of 1976 – 1989 the Gallup poll asked respondents “Do you think that labour

unions are a good thing or bad thing for Canada?” For the purpose of analysis only respondents

4

Page 7: Is there a secular trend in the public opinion of unions in Canada over time

who gave a definitive response have been included. For the period of 1991-2000 the Gallup poll

asked the question “Would you tell me how much confidence you, yourself, have in labour

unions – a great deal, quite a lot, some or very little. For the purposes of analysis a binary

variable, labeled supunion, was created with a value of 1, indicating support of unions, if the

respondent answered “a great deal” or “quite a lot”, and 0 otherwise. The trend in public opinion

based off this poll is therefore just identifying those who are decisive in their support for unions,

which may also indicate they believe strongly in most or all aspects of union behaviour. For the

time frame of 1993-2008 the mail back survey of the Canadian Election Study asked respondents

“How much confidence do you have in unions? – A great deal, quite a lot, not very much, or

none at all.” Again the dummy variable, supunion, was coded 1, indicating support, for

respondents who replied “a great deal” or “quite a lot”, and 0 otherwise. Because of the

difference in questions a variable, sample, is also included in all analysis to account for question

bias.

When regressing the dummy variable, supunion, on the intercept we can see an

interesting result. Sample1 is a dummy variable for the Gallup poll responses from 1976- 89

asking whether unions are good or bad for Canada. Sample2 is a dummy variable for the

responses to the Gallup poll from 91-2000, while Sample3 is a dummy variable for the responses

to the Canadian Election Study’s mail back survey. When discounting the level effects that are

caused by the form and coding of the question a clear trend can be seen of increasing support of

unions over time. This is surprising given that falling union density rate over time would be

expected to either be a cause or consequence of less favourable attitudes towards unions. This

leads to the question of whether a decrease in union density may in fact have the opposite effect.

It is also important to note that the observations are distributed roughly around .5. This supports

5

Page 8: Is there a secular trend in the public opinion of unions in Canada over time

the use of ordinary least squares linear regression as opposed to the maximum likelihood

regression model. All analysis in this paper is done using the OLS regressor.

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 20100

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

f(x) = 0.0033808826530613 x − 6.44886225918383R² = 0.537312193333234

f(x) = 0.00806481151515152 x − 15.8747103684848R² = 0.508908390945886

f(x) = NaN x + NaNR² = 0

Sample 1Linear (Sample 1)sample 2

Supun-ion

Support of Unions Over Time

Year

By controlling for the variable “sample” in the regression model:

supunion = β 0 + β 1t + β 2sample1 + β3 sample2 + ε

where sample1 and sample2 are dummy variables and t is a time index in which 1976 is 1.

we can identify the effect of time independent of the level effects caused by the question

differences. We can interpret β 2 and β3 as these level effects. The results from this regression are:

Table 1

Regression Results of Supunion on Time and Sample

6

Page 9: Is there a secular trend in the public opinion of unions in Canada over time

Variable b St. Err. t

t

sample1

sample2

constant

N= 23292

R2= 0.1082

.0075855

.428482

-.0607189

.1241728

.0007236

.0153119

.0074989

.0185234

10.48

27.98

-8.10

6.70

*all R2 reported= adjusted R2

From this regression we can see a large time trend of .76 percentage point increase in the

probability of support for unions per year. This trend is also significant, meaning we can reject

the null hypothesis that the effect of time is zero at the 5% significance level, which is observed

in the high t statistic. However it could be argued that the time trend is in fact not linear. To test

this hypothesis we can estimate the models:

supunion = β 0 + β 1t + β 2t2 + β3 sample1 + β4sample2 + ε (quadratic)

supunion = β 0 + β 1t + β 2t2 + β3 t3+ β4sample1 + β5sample2 + ε (cubic)

supunion = β 0 + β 1t + β 2t2 + β3 t3+ β4t4+ β5sample1 + β6 + ε (quartic)

Looking at the graph comparing the resulting time trends from these models:

7

Page 10: Is there a secular trend in the public opinion of unions in Canada over time

1970 1975 1980 1985 1990 1995 2000 2005 20100

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

linearquadraticcubicquartic

Regression of Supunion on Time

Year

For regression results refer to Table 1 in appendix.

We can see that over the sample period 1976-2008 there is some variation between the

linear model as compared to the quadratic, cubic and quartic models. The associated R2 for these

models are: Quadratic = .1095, Cubic = .1099, Quartic = .1102. This variation is not too drastic,

and doesn’t account for much of the variation seen in the data and for the analytical purposes of

this paper, the linear model will be used.

3.2 Personal Attributes

Now that a significant time trend has been identified, the question becomes whether it is

well defined or if there is a significant omitted variable bias. Given the variable of interest is

time; it definitely could expected that elements, not yet introduced into the model, may be

correlated with the support of unions as well as with t. To truly identify if there is a unique time

8

Page 11: Is there a secular trend in the public opinion of unions in Canada over time

trend in the support of unions, the elements that affect people’s opinions of unions must also be

introduced into the model.

Age may be a determinant of whether a person supports a union or not. Young people,

under the age of 25 are much less likely to be members of unions (Bryson et al. 2005). As people

age they gain life experiences and knowledge that will affect their opinions on certain issues.

They may: become members of unions, join a non-unionized employer, become upper

management or know people who have had interactions with unions. These life experiences may

make an individual more supportive of unions or more cynical of unions as they age. Again it is

not unreasonable to expect that the relationship between age and the support of unions is non-

linear. People have very different needs as they age, a person working in their prime could have

a very different opinion than a retired person. To exam this relationship the following models

were regressed:

supunion = β 0 + β 1age + β 2sample1 + β3 sample2 + ε

supunion = β 0 + β 1age + β 2 age2 + β3 sample1 + β4sample2 + ε

supunion = β 0 + β 1age + β 2age2 + β3 age3+ β4sample1 + β5sample2 + ε

supunion = β 0 + β 1age + β 2age2 + β3 age3+ β4age4+ β5sample1 + β6 Sample2 + ε

Looking at the graph of the resulting age trends:

9

Page 12: Is there a secular trend in the public opinion of unions in Canada over time

10 20 30 40 50 60 70 80 90 100 1100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

LinearQuadraticCubicQuartic

Different Models for Regressing supunion on age

Age

For the regression results see Table 2 in appendix.

It is important to note that the age range of interest in this graph is age <85 as there are

very few instances of respondents over this age. Looking at the graph there is some variation

between the models but they are fairly close, especially over the prime age of individuals. The

R2 for the linear, quadratic, cubic, and quartic models are: .1095, .1109, .1118 and .1134

respectively. Therefore, in the final model only the linear age term will be included. The linear

age regression coefficient is -.0022 with a t statistic of -12.01 which allows us to reject the null

hypothesis that the effect of age on supunion is zero. This indicates that as people age they tend

to grow more cynical of unions due to life experience or changing values.

It may be that women and men have systematically different opinions about unions. This

may be a result of differing levels of union participation. It may be that women are less career-

10

Page 13: Is there a secular trend in the public opinion of unions in Canada over time

oriented as they have children. To test this, a dummy variable, male, equaling 1 for men is

included in the regression.

It is reasonable to expect that respondents living in different locations have significantly

different opinions about unions. This could be a result of different economic conditions in the

provinces or varying levels of union density or activity. Different industries are more or less

prevalent in different geographical locations. It also could be a result of different taxation in

between provinces, and different levels of welfare in different geographical locations. It could

also be a result of different political tendencies across provinces, for example Alberta is

historically very conservative, and different political affiliations tend to have varying views on

unions. To test for differing opinions across provinces a dummy variable is included in the model

for each province. To avoid the dummy variable trap Ontario is chosen as the baseline and

excluded from the model, meaning all province coefficients will be a reflection of the difference

in opinion between that province and Ontario.

It is reasonable to expect that as one gains more education their opinion of unions will

change. This could be a result of a different likelihood of joining a union. It could also be that as

one gains more education they gain a better or simply new understanding of how unions affect

the economy. It could be that teachers and professors pass their views on to students. As students

enter university and college they specialize and these specializations could lead to unique

opinions about unions. To test for the effects of education, several dummy variables are included

in the model. Respondents are classified as having no high school education, high school

education, some post-secondary education, having completed college, or having a university

degree. Again to avoid the dummy variable trap those with no high-school are excluded from the

11

Page 14: Is there a secular trend in the public opinion of unions in Canada over time

model so that the coefficients on the other education variables can be interpreted as the

difference in opinions about unions than those with no highs school diploma.

It could be that being born in different periods will also lead to differing views on unions.

It may be that the differences in economic conditions and union activity during formative years

could lead to drastically different opinions of unions. It could be that living through different

historical events like war or political unrest may lead to differing values that may affect one’s

opinions of unions. To test for these effects dummy variables are created for respondents

depending on which decade they are born: before1920, 20s, 30s, 40s, 50s, 60s, 70s, or after 1980.

Those born before 1920 are omitted from the model so coefficients can be interpreted as the

difference in opinions about unions between the decade of birth and being born before 1920.

These periods were chosen so that there is a roughly equal amount of respondents born in each

period. These periods are somewhat arbitrary, for example there is not much difference in being

born between 1929 and 1930, but they will still provide insight as to whether cohort effects exist.

3.3 Macroeconomic Attributes

There are a variety of reasons we would expect the current unemployment rate to affect

public opinion of unions. A well known facet of unions is that they raise the wages above

equilibrium rates. Given this, it is reasonable to expect high unemployment would cause a

pervasive negative view of unions. However it has also been argued that during periods of high

unemployment workers may be more susceptible to bullying by management, so unions can

protect the interests of workers (Lowe and Krahn 1989). Provincial unemployment rates are used

in the regression analysis as one would expect that they would have a bigger effect on opinions,

as the majority of people are most concerned about the local job climate. Unemployment rates

12

Page 15: Is there a secular trend in the public opinion of unions in Canada over time

were obtained from the Labour Force Survey (LFS). Previous studies have posited that the union

density is a predictor of union rates (Lipset 1986). Union density rates were obtained from

stats.oecd.org. Union density is declining over time and this may be expected to lead to a more

negative view of unions. One would probably expect a union member to be more favourable of

unions as they have directly or indirectly chosen to join a union. Having a union member in

one’s household could also significantly affect ones opinion of unions. If one’s spouse is a union

member, they may get a higher level of pay as a result. It could be that union members with

negative experiences are affected the other way but there is little doubt that there will be an

identifiable effect. To test for this, respondents are given a dummy variable equal to 1 if they or

a member of their household is a member of a union.

It may be that current strike activity affects people’s support of unions. More labour

disputes and strikes could lead to views that unions are militant or ineffective, or they may

disrupt everyday activities of people leading to negative views (Bain and Edwards1988). To test

this, the number of strike hours lost per month per year in Canada, as obtained from the Labour

Force Survey (LFS) estimates, is included in the model.

It may be that people view the primary objective of unions should be to lift the bottom

members of society to higher earnings. One way we could model this is by looking at the Gini

coefficient. This is one measure of the dispersion of wealth with a higher score indicating a

greater disparity between the rich and poor. This is increasing over time as the coefficient

obtained when gini is regressed on time is .0018.

It could be that most people are not aware of the actual level of strikes. Therefore a better

measure of the effect of strike activity may be the effect of media coverage of strikes. To attempt

13

Page 16: Is there a secular trend in the public opinion of unions in Canada over time

to capture this effect, a measure of the number of articles in which strikes and unions appear in

the first paragraph of articles in the Globe and Mail was obtained through the Factiva periodicals

database. This measure labeled globeart is analyzed in the regression in addition to the number

of labour hours lost to strikes. Note this could not be obtained for 1976 so those observations are

dropped in subsequent analysis.

There are conflicting views on the effect of inflation on the support of unions. It could be

that unions cause inflation and therefore a higher inflation could lead to negative views (Bain

and Edwards 1988), or that unions help ensure livings standards in the case of inflation(Lowe

and Krahn 1989). It also may be that inflation concerns have decreased over time given that

Canada has maintained relatively stable inflation in recent history. To test this, provincial

inflation calculated from Consumer Price Indices (CPI) from Statistics Canada is included in the

model. Note inflation for the years 1978 and 1976 was not obtained so observations from that

year have been dropped.

It is human nature to be envious or jealous of those who are doing well. This may arise in

the case of unions if given two equivalent workers; the union member receives better pay or

benefits. This effect may be bigger if a less qualified union member gets paid more than a non-

union member. To test for this effect a measure of the union wage premium was constructed

using data from the Labour Force Survey (LFS) by subtracting the logarithm of the hourly wage

of non-union members from those covered by collective bargaining agreement. We could also

expect interaction effects between union and non-union members. Note that this was only

available starting in 1997.

4 Results

14

Page 17: Is there a secular trend in the public opinion of unions in Canada over time

Including the personal variables of age, gender, province, education and cohort into the model

leads to:

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5male + β6NFLD + β7PEI +

β8NS + β9NB + β10QUE + β11MAN + β12SASK + β13ALB + β14BC + β15 HS + β16SomePS

+ β17college + β18Uni + β1920s + β2030s + β2140s + β2250s + β2360s + β2470s + β25after1980

+ ε

The results of this regression are:

Table 2

Results of supunion regressed on personal variables

Variable b St. Err. t

Sample1

Sample2

Time

Age

Male

NFLD

PEI

NS

NB

QUE

.4199226

-.0667914

.0097208

-.0039056

.0019385

.1144033

.0365501

.0247479

.0362482

.0702841

.0152863

.0075171

.0012035

.0009799

.0058605

.0192484

.0255508

.0166407

.0177458

.0076703

27.47

-8.89

8.08

-3.99

0.33

5.94

1.43

1.19

2.04

9.16

15

Page 18: Is there a secular trend in the public opinion of unions in Canada over time

MAN

SASK

ALB

BC

HS

Some Post-Secondary

College

University

20s

30s

40s

50s

60s

70s

After 1980

Constant

Obs= 23292

R2 = 0.1246

-.0062977

.0030781

-.0643011

.0223685

-.0244156

-.0416294

-.0399268

-.0421845

.0118766

-.0158691

-.0172754

-.0457343

-.0807569

-.0429485

.0025056

.3041292

.0149275

.0147711

.0107833

.0098577

.008603

.0099761

.009847

.0092722

.0189516

.0256307

.0340324

.0427245

.0514785

.0611867

.0735187

.0653449

-0.42

0.21

-5.96

2.27

-2.84

-4.17

-4.05

-4.55

0.63

-0.62

-0.51

-1.07

-1.57

-0.70

-0.03

4.65

Looking at the results of this regression there are a few interesting implications. First the

time coefficient .0097 has increased from .0076 in the regression of supunion on time alone after

the inclusion of these personal variables. The variation increased but the coefficient is still

significant at the 5% level. The R2 also increased from R2= 0.1082 to R2 = 0.1246. The negative

16

Page 19: Is there a secular trend in the public opinion of unions in Canada over time

effect of age on the support of unions leads to an increase in the effect of time, because of

Canada’s aging population. This aging population is reflected in the sample, if age is regressed

on time the coefficient is .2789. The results suggest that the effect of gender on the support of

unions is very small, and we cannot reject the null hypothesis that it is in fact zero at the 5%

significance level. The effects of the province of respondents are very straight forward and fall in

line with what might be expected. Newfoundland and Quebec are significantly more supportive

of unions than Ontario, while Alberta is significantly less supportive. While other provinces tend

to be more supportive of unions than Ontario, their opinions are fairly close and only in BC and

New Brunswick are we able to reject a null hypothesis of no effect at a 5% significance level.

The effect of education also seems to push the time trend higher as it is clear that

opinions of unions seem to drop a bit as one completes high-school and definitely drops if the

respondent completed at least some post-secondary education. Over time people are becoming

more educated, supported by a positive coefficient if education is regressed on time, and one

would expect this to drop the opinions of unions over time.

Finally the cohort effects are very small and the null hypothesis of zero effect cannot be

rejected for any decade. As previously stated the chosen cohorts are somewhat arbitrary but we

can still infer that there are no or very small cohort effects. Because of the insignificance of

gender and cohort they are excluded in further analysis. If the same regression is run without

these variables the time trend is very close although slightly smaller at .0089 with a t statistic of

12.15. See Table 3 appendix for full results.

Inclusion of unemployment leads to the model:

17

Page 20: Is there a secular trend in the public opinion of unions in Canada over time

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment

See Table 4 in Appendix for full results.

When this regression is run the coefficient on unemployment rates is -.0040 with a t

statistic of -2.09. This result can be interpreted as people are .40 percentage points less likely to

support unions given a one point increase in the provincial unemployment rate. We can reject the

null hypothesis that provincial unemployment rates, in fact, have no effect on people’s support of

unions. The inclusion of unemployment rates does reduce the time trend from .0088 to .0079

though this is still a rather significant time trend independent of the variables included in the

model thus far.

When union density is included in the regression model it leads to:

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment + β19uniondensity

See Table 5 in appendix for full regression results

When this regression is run the coefficient is surprisingly negative at -.0016 however the

t statistic is -0.79. As union density is falling over time in Canada, this reduced the effect of time

on union support to .0075. However we cannot reject the null hypothesis that the effect of union

density has no effect on the support of unions. Therefore because of this ambiguity a better

representation of the effect of the effect of falling union density may be the personal level

18

Page 21: Is there a secular trend in the public opinion of unions in Canada over time

variable of whether or not a member of the respondents household is a union member. The

falling union density over time is reflected in the respondents as household union membership

regressed on time produces a coefficient of -.0022. This leads to the model:

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment + β19HHUnionMembership

See Table 6 in appendix for full regression results

Not surprisingly, the estimates suggest that respondents with union members in their

household are strongly predisposed to be supportive of unions compared to non-union members.

The coefficient of household union membership is .1782 with a t statistic of 27.94. Here we can

easily reject the null hypothesis that this effect is in fact zero at the 5% significance level. This

positive trend reinforces the idea that there is a secular time trend in increasing support of

unions. The coefficient of time in this regression is .0091 with a t statistic of 10.88. It is also

worth noting the difference in the adjusted R2 statistics of these two models, as the union density

model was 0.1236 and the household union membership was 0.1520. This means this second

model fits the data significantly better. Therefore for further analysis on household union

membership and not union density will be included.

Inclusion of the number of labour hours lost to strikes leads to the model:

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment + β19HHUnionMembership + β20Strikehours

19

Page 22: Is there a secular trend in the public opinion of unions in Canada over time

For full regression results refer to Table 7 appendix.

Looking at the results the expected negative effect of more strikes on the opinions of

unions can be seen as the coefficient for strike hours is -.0000446 with a t statistic of -2.37.

Although this is an extremely small effect, it is per hour of labour so given high variation it could

have an impact on public opinions. This can be seen in the fact that we cannot reject the null

hypothesis that this effect is in fact zero at the 5% significance level. However the inclusion of

strike hours lowers the R2 slightly, meaning it does not make the model fit the data any better.

Including the gini measure, as recorded by Statistics Canada, leads to the model:

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment + β19HHUnionMembership + β20Strikehours + β21Gini

See Table 8 in appendix for complete regression results.

There is a large negative coefficient on Gini in this regression of -2.9928 with a t statistic

of -5.80, which can be viewed as an increasingly negative public view of unions as the dispersion

between the rich and poor increases. This increases the time trend significantly to .0135 with a t

statistic of 9.31. It is hard to identify the causal relationship here in the minds of the public. The

inclusion of Gini does increase the R2 a little, from .1522 to .1533. It may be that this is not a

consideration that the majority of people have when deciding on their support for unions and this

result is picking up spurious correlation. What is evident is that it is not reducing the secular

trend in public opinions over time.

20

Page 23: Is there a secular trend in the public opinion of unions in Canada over time

The poor predictive value of strike hours may be a reflection of a lack of knowledge in

the general public. To better capture this effect in the decision making process of the public,

strike hours is replaced with a measure of media coverage of strikes. This leads to the model:

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment + β19HHUnionMembership + β20globeart + β21Gini

For complete regression result see Table 9 in appendix.

Looking at the results, the coefficient of globeart is similar to that of strike hours, b=

-.0017 but with a higher t statistic of -4.35. It also has a much larger effect on the time coefficient

reducing it to .0072 with a t statistic of 3.69. This indicates that media does in fact have an effect

on the public opinion of unions. It must be noted that while a general result can be drawn from

this there is bound to be a high uncertainty with using one measure of media coverage. The

variations in articles may not necessarily reflect how much awareness the public has of strikes, as

the form of the paper changes over time. For example there are more sections and different

authors with different opinions. It is also interesting that the coefficient for Gini drastically

drops, to -1.1124 with a t statistic of -1.58 lending credence to the idea that this is not a

consideration of an average individual in their decision about whether or not to support unions.

Including provincial inflation leads to the model:

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment + β19HHUnionMembership + β20globeart + β21Gini +

β22provincialinflation

21

Page 24: Is there a secular trend in the public opinion of unions in Canada over time

For full regression results see Table 10 in appendix

The results of this regression are very interesting, the inflation coefficient is -.0075 with a

t statistic of -3.42. This implies that as inflation increases people become less supportive of

unions. The effect of provincial inflation on time is very suggestive. In this regression the

coefficient on time is .0017 which is significantly smaller than previous results. Perhaps more

telling is the t statistic for time, which is now 0.76, meaning one fails to reject the null hypothesis

that there is no secular trend in the support of unions over time. If globeart and gini are removed,

given their subjective nature, the time coefficient is still much smaller (see Table 11 in appendix

for complete results) at .0027 with a t statistic of 2.09. While the null hypothesis would not be

rejected at a 5% significance level in this regression, if one takes the idea of some negative

correlation between media coverage of strikes and public opinion of unions, which is reasonable

given the previous research into this relationship in the United States (Schmidt 1993), then the

null hypothesis of zero effect at the 5% significance level would likely be rejected. Also note that

the effect of the dropped variables is also negligible to these key results. If the initial regression

of supunion on sample and time is run without observations from 1976 and 1978 the time

coefficient increases from .0076 to .0077, implying that the effect of provincial inflation is

bigger than the effect of lost observations.

Including the measure of wage premium and accounting for interaction effects leads to the

model:

supunion = β0 + β1sample2 + β2t + β3age + β4NFLD + β5PEI + β6NS + β7NB + β8QUE +

β9MAN + β10SASK+ β11ALB + β12BC + β13HS + β14SomePS + β15College +

22

Page 25: Is there a secular trend in the public opinion of unions in Canada over time

β16University+ β17Unemployment + β18HHUnionMembership + β19globeart + β20Gini +

β21provincialinflation + β22wagepremium + β23hhunion*wagepremium

For full regression results refer to Table 12 in appendix

Looking at the results of this regression one can see a very large coefficient on the union

premium of -.2615 which suggests that as the wage premium for union member grows the public

opinion of unions falls, and a large positive coefficient on the interaction effects of .0989

suggesting as the wage premium grows union members become more supportive of their unions.

However the t statistics for these coefficients are -1.09 and 1.39 respectively so the null

hypothesis that these effects are zero cannot be rejected at the 5% significance level. Because of

this fact combined with the severely reduced R2 of .0507 these variables will be left out of the

regression.

5 Conclusions

There is a clear trend of increasing support for unions in Canada from 1976 – 2008. After

analyzing components that could be incorporated in the decision process of how the public form

their opinions of unions the following model best represents the support of unions from 1976 to

2008.

supunion = β0 + β1sample1 + β2sample2 + β3t + β4age + β5NFLD + β6PEI + β7NS + β8NB

+ β9QUE + β10MAN + β11SASK + β12ALB + β13BC + β14HS + β15SomePS + β16college +

β17Uni + β18Unemployment + β19HHUnionMembership + β20globeart + β21Gini +

β22provincialinflation

Table 3

23

Page 26: Is there a secular trend in the public opinion of unions in Canada over time

The Results of the Ordinary Least Squares Regression

Variables b St. Err. t

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

HHunion

globeart

Gini

.4229384

-.0621281

.0016906

-.0017991

.2806546

.1983724

.0873261

.1087373

.1256822

-.0297746

-.0216571

-.0706928

.0247258

-.0313082

-.0374687

-.0602123

-.0606356

-.0224847

.1732808

-.0001787

-.2943564

.0206513

.0076367

.0022291

.000197

.0322291

.0316005

.0191287

.020947

.011332

.0156541

.0156258

.0114349

.0102153

.009017

.0102973

.010006

.0094446

.0029625

.006695

.0000432

.7499253

20.48

-8.14

0.76

-9.13

8.71

6.28

4.57

5.19

11.09

-1.90

-1.39

-6.18

2.42

-3.47

-3.64

-6.02

-6.42

-7.59

25.88

-4.13

-0.39

24

Page 27: Is there a secular trend in the public opinion of unions in Canada over time

Provincial Inflation

Constant

R2 = 0.1451

Obs= 20851

-.0074529

.6124605

.0021779

.2230462

-3.42

2.75

From these results there are three main conclusions that can be reached:

1. One cannot definitively state that the increasing public opinion is a secular trend over time, as

one cannot reject the null hypothesis that the time coefficient is not in fact zero.

2. While we cannot definitively state that there is an increasing secular public opinion of unions,

the majority of covariates analyzed had a clear impact on the support of unions that tended to

increase the secular time trend. This suggests that in reality there could be a secular increase in

the support of unions over time.

3. The biggest factor leading to the increasing support of unions appears to be the reduction and

stabilization of provincial inflation.

Further analysis of these regression results does provide some other useful insights. There

appears to be significant differences given geographical locations within Canada. People tend to

become less supportive of unions as they age, and as they receive more education. However the

type of education doesn’t appear to have a huge effect, suggesting this is a difference in life

experience and goals as opposed to a fundamental aspect of any particular education system. In

Canada, both the amount of education and the age of the population are increasing which

supports the idea of a secular trend of increasing support of unions from 1976 – 2008. People

tend to become more cynical of unions as unemployment increases; this may be a result of the

25

Page 28: Is there a secular trend in the public opinion of unions in Canada over time

fact unions raise the wage above equilibrium values. There does not appear to be a pervasive

opinion that unions are a tool for social equalization. This may be a result of well-defined legal

and political systems that are seen as better tools to reach these goals.

With respect to media coverage of unions it appears that the level of coverage of strikes over

time has affected people’s opinions of unions. This could be reflective of the decline of the

union’s social status. On the surface, looking at the big effect that union membership has on the

support of unions, the falling union density over time reinforces a secular time trend. However, it

may be that as union density has declined, people hear less about unions, there are fewer strikes

and people may believe unions have become less powerful and, therefore, worry less about

negative externalities that may arise, like corruption.

26

Page 29: Is there a secular trend in the public opinion of unions in Canada over time

Appendix

Table 1

Results of Supunion Regressed on Time

Variable b St. Err. t

t

t2

sample1

sample2

constant

R2= 0.1095

Obs= 23292

.0188369

-.0002834

.4708031

-.0722534

.0265434

.0020378

.000048

.0168957

.0077438

.0248172

9.24

-5.91

27.87

-9.33

1.07

t

t2

t3

sample1

sample2

constant

R2= 0.1099

Obs= 23292

.007967

.0006006

-.0000176

.4947763

-.0735191

.0323508

.0038819

.000273

5.35e-06

.0183971

.0077517

.0248747

2.05

2.20

-3.29

26.89

-9.48

1.30

t

t2

t3

-.0173958

.0038739

-.0001582

.0089008

.0010692

.0000447

-1.95

3.62

-3.54

27

Page 30: Is there a secular trend in the public opinion of unions in Canada over time

t4

sample1

sample2

constant

R2= 0.1102

Obs= 23292

1.96e-06

.5275729

-.0749639

.04691

6.19e-07

.0211094

.0077637

.0252913

3.17

24.99

-9.66

1.85

Table 2

Results of Supunion Regressed on Age

Variable b St. Err. t

Age

sample1

sample2

constant

R2 = 0.1095

Obs= 23292

-.0022033

.2773917

-.1006012

.4160115

.0001834

.0077489

.0069052

.0100911

-12.01

35.80

-14.57

41.23

Age

Age2

sample1

sample2

constant

R2 = 0.1109

-.008333

.0000636

.2741687

-.10253

.5477623

.0010175

.0000104

.0077607

.006907

.0237588

-8.19

6.12

35.33

-14.84

23.06

28

Page 31: Is there a secular trend in the public opinion of unions in Canada over time

Obs= 23292

Age

Age2

Age3

sample1

sample2

constant

R2 = 0.1118

Obs= 23292

-.0274764

.0004843

-2.84e-06

.2714935

-.1027765

.811458

.0039058

.0000835

5.59e-07

.0077744

.0069035

.0571158

-7.03

5.80

-5.08

34.92

-14.89

14.21

Age

Age2

Age3

Age4

sample1

sample2

constant

R2 = 0.1134

Obs= 23292

-.1085889

.0032212

-.000041

1.87e-07

.2731515

-.1022642

1.639322

.0129484

.0004249

5.83e-06

2.85e-08

.0077715

.0068977

.1383301

-8.39

7.58

-7.03

6.57

35.15

-14.83

11.58

Table 3

Regression Results of Personal Variables Omitting Gender and Cohort

Variables b St. Err. t

29

Page 32: Is there a secular trend in the public opinion of unions in Canada over time

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

Constant

R2 = 0.1235

Obs= 23292

.4244988

-.0671639

.0088378

-.002602

.1134528

.0372681

.0247839

.0353604

.0711148

-.0066681

.0024748

-.0641586

.0225853

-.0263522

-.0410781

-.0445653

-.0454725

.2297021

.0152368

.0075063

.0007274

.00019

.0192512

0255611

.0166477

.017755

.007671

.0149351

.0147788

.0107886

.0098619

.0085644

.0099696

.009782

.0092064

.0214251

27.86

-8.95

12.15

-13.69

5.89

1.46

1.49

1.99

9.27

-.045

0.17

-5.95

2.29

-3.08

-4.12

-4.56

-4.95

10.72

Table 4

Regression Results From Inclusion of Provincial Unemployment Rates

30

Page 33: Is there a secular trend in the public opinion of unions in Canada over time

Variables b St. Err. t

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

Constant

R2 = 0.1236

Obs= 23292

.4064867

-.0668922

.0079165

-.0026077

.1471443

.0613827

.0357001

.0485352

.0820449

-.0107693

-.0032475

-.0683742

.0252959

-.0259398

-.0407324

-.0440578

-.0452419

-.0039857

.2778308

.0175047

.0075069

.0008505

.00019

.0251091

.0280433

.0174468

.0188398

.0092839

.0150624

.0150293

.0109748

.0099462

.0085661

.0085661

.0097843

.0092064

.0092064

.031454

23.22

-8.91

9.31

-13.72

5.86

2.19

2.05

2.58

8.84

-0.71

-0.22

-6.23

2.54

-3.03

-4.09

-4.50

-4.91

-2.09

8.83

31

Page 34: Is there a secular trend in the public opinion of unions in Canada over time

Table 5

Regression results when union density is included

Variables b St. Err. t

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

Union Density

.4059635

-.0666246

.0074682

-.0026066

.1337012

.0519019

.0310564

.0429064

.0774938

-.0090346

-.0007628

-.066701

.0242724

-.0259001

-.0407124

-.0439957

-.0450448

-.0023766

-.0016398

.0175172

.0075145

.0010214

.00019

.0302994

.0304873

.0184042

.0201336

.0109157

.0152207

.0153527

.0111761

.0100297

.0085663

.0099704

.0097847

.0097847

.0097847

.0020685

23.18

-8.87

7.31

-13.72

4.41

1.70

1.69

2.13

7.10

-0.59

-0.05

-5.97

2.42

-3.02

-4.08

-4.50

-4.89

-0.85

-0.79

32

Page 35: Is there a secular trend in the public opinion of unions in Canada over time

Constant

R2 = 0.1236

.3309864 .0740651 4.47

Table 6

Regression Results of Including Household Union Membership

Variables b St. Err. t

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

.4303811

-.052317

.0091174

-.0018927

.1172761

.0662348

.0364901

.0425793

.0692943

-.0146353

-.0073081

-.0596871

.0145939

-.0315564

-.036612

-.0566837

.01724

.0074027

.0008377

.0001887

.0247221

.0275857

.0171619

.0185333

.0091436

.014817

.0147845

.0108001

.0097912

.0084285

.0098085

.0096351

24.96

-7.07

10.88

-10.03

4.74

2.40

2.13

2.30

7.58

-0.99

-0.49

-5.53

1.49

-3.74

-3.73

-5.88

33

Page 36: Is there a secular trend in the public opinion of unions in Canada over time

University

Unemployment rates

Household union

membership

Constant

R2 = 0.1527

Obs= 23292

-.0549966

-.0032368

.1782362

.1617485

.0090627

.0018762

.0063804

.031218

-6.07

-1.73

27.94

5.18

Table 7

Regression Results When Labour Hours Lost to Strikes is Included in the Model

Variables b St. Err. t

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

.4349104

-.0531343

.0076466

-.0018819

.1468663

.0883411

.0466236

.0549597

.0791499

-.0184293

-.0125618

.017344

.00741

.0010427

.0001887

.0276969

.0291191

.0176854

.0192545

.010045

.0149019

.0149485

25.08

-7.17

7.33

-9.97

5.30

3.03

2.64

2.85

7.88

-1.24

-0.84

34

Page 37: Is there a secular trend in the public opinion of unions in Canada over time

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

HHunion

Strike hours

Constant

R2 = 0.1522

Obs= 23292

-.0636783

.0171395

-.0320167

-.0371473

-.0571846

-.0552182

-.0068496

.1780333

-.0000446

.234278

.0109296

.009849

.009849

.0098101

.0096364

.0090623

.0024178

.0063803

.0000188

.043726

-5.83

1.74

-3.80

-3.79

-5.93

-6.09

-2.83

27.90

-2.37

5.36

Table 8

Regression Results From Including Gini Coefficient

Variable b St.err. t

Sample1

Sample2

T

Age

NFLD

PEI

.4070361

-.0512435

.0134541

-.0018936

.1864314

.1151885

.0179857

.007412

.001445

.0001886

.0285056

.0294645

22.63

-6.91

9.31

-10.04

6.54

3.91

35

Page 38: Is there a secular trend in the public opinion of unions in Canada over time

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

HHunion

Strike hours

Gini

Constant

R2 = 0.1533

Obs= 23292

.0601283

.0722122

.0928176

-.0239618

-.0205839

-.0687433

.0199677

-.0328054

-.037782

-.0578585

-.0568523

-.0115263

.1783318

7.40e-06

-2.992812

1.15954

.0178257

.0194695

.0103108

.0149219

.0150019

.0109568

.0098542

.0084251

.0098039

.0096304

.0090604

.0025471

.0063761

.0000208

.5159567

.1653902

3.37

3.71

9.00

-1.61

-1.37

-6.27

2.03

-3.89

-3.85

-6.01

-6.27

-4.53

27.97

.36

-5.80

7.01

Table 9

Regression Results Replacing Strike Hours with Globeart

Variables b St. Err. t

36

Page 39: Is there a secular trend in the public opinion of unions in Canada over time

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

HHunion

globeart

Gini

Constant

R2 = 0.1538

.4287592

-.0535827

.0071888

-.001924

.2311022

.1554228

.0733432

.0842406

.1116534

-.0307813

-.0268736

-.0689193

.0244902

-.0335549

-.0400736

-.0590665

-.0599356

-.0159473

.1777231

-.0001698

-1.112357

.7114185

.0187532

.0074554

.0019467

.0001907

.0279036

.0293045

.0179585

.0195923

.0102594

.0150875

.0151201

.0110028

.0099168

.0085954

.0099183

.0097141

.0091679

.0024278

.0064667

.000039

.7059709

.210207

22.86

-7.19

3.69

-10.09

8.28

5.30

4.08

4.30

10.88

-2.04

-1.78

-6.26

2.47

-3.90

-4.04

-6.08

-6.54

-6.57

27.48

-4.35

-1.58

3.38

37

Page 40: Is there a secular trend in the public opinion of unions in Canada over time

Obs= 22525

Table 10

Regression with Provincial Inflation Included

Variables b St. Err. t

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

.4229384

-.0621281

.0016906

-.0017991

.2806546

.1983724

.0873261

.1087373

.1256822

-.0297746

-.0216571

-.0706928

.0247258

-.0313082

-.0374687

-.0602123

-.0606356

.0206513

.0076367

.0022291

.000197

.0322291

.0316005

.0191287

.020947

.011332

.0156541

.0156258

.0114349

.0102153

.009017

.0102973

.010006

.0094446

20.48

-8.14

0.76

-9.13

8.71

6.28

4.57

5.19

11.09

-1.90

-1.39

-6.18

2.42

-3.47

-3.64

-6.02

-6.42

38

Page 41: Is there a secular trend in the public opinion of unions in Canada over time

Unemployment rates

HHunion

globeart

Gini

Provincial Inflation

Constant

R2 = 0.1451

Obs= 20851

-.0224847

.1732808

-.0001787

-.2943564

-.0074529

.6124605

.0029625

.006695

.0000432

.7499253

.0021779

.2230462

-7.59

25.88

-4.13

-0.39

-3.42

2.75

Table 11

Regression with provincial Inflation and Without Globeart and Gini

Variables b St. Err. t

Sample1

Sample2

T

Age

NFLD

PEI

NS

NB

QUE

MAN

.3921882

-.0642707

.0027159

-.0018018

.2338274

.1629527

.070827

.0904373

.1097074

-.0232088

.0187286

.0075776

.0013021

.0001971

.0312513

.0310935

.0189375

.0207295

.0110057

.015625

20.94

-8.48

2.09

-9.14

7.48

5.24

3.74

4.36

9.97

-1.49

39

Page 42: Is there a secular trend in the public opinion of unions in Canada over time

SASK

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

HHunion

Provincial Inflation

Constant

R2 = 0.1443

Obs= 20851

-.0132702

-.0651165

.0215151

-.0304537

-.0367433

-.059738

-.0594498

-.0169039

.1736293

-.0072063

.4172811

.0155707

.0114102

.0102107

.0090239

.010306

.0100146

.0094487

.0028117

.0067008

.0021436

.051631

-0.85

-5.71

2.11

-3.37

-3.57

-5.97

-6.29

-6.01

25.91

-3.36

8.08

Table 12

Regression Including Wage Premium and Wage Premuim*HHunion

Variables b St. Err t

Sample2

T

Age

NFLD

PEI

-.057269

.0072155

-.0010436

.1692896

.0874713

.0120779

.0079509

.0002978

.1094656

.1068876

-4.74

.91

-3.5

1.55

0.82

40

Page 43: Is there a secular trend in the public opinion of unions in Canada over time

NS

NB

QUE

MAN

SASK

ALB

BC

HS

Some post-secondary

College

University

Unemployment rates

HHunion

Globeart

gini

Provincial Inflation

Union premium

Union

premium*HHunion

Constant

R2 = 0.0507

Obs= 9714

-.0356566

.0531451

.0802219

-.0873228

-.0703859

-.0864481

-.0068095

-.051431

-.0546847

-.0787903

-.0703196

-.0138767

-.0960244

-.0001144

.1826176

-.0039215

-.2615254

.0988516

.955606

.0720631

.0782314

.036093

.055346

.05526

.0260757

.0178725

.0146563

.0160569

.0155808

.0145162

.0085385

. 1952469

.0001679

2.001193

.0099611

.2400441

. 0709162

.9822524

-0.49

0.68

2.22

-1.58

-1.27

-3.32

-0.38

-3.51

-3.41

-5.06

-4.84

-1.63

-0.49

-0.68

0.09

-.039

-1.09

1.39

0.97

41

Page 44: Is there a secular trend in the public opinion of unions in Canada over time

References

Bain, S, Bain and P.K. Edwards . “Why Are Trade Unions Becoming More Popular? Unions and Public Opinion in Britain” British Journal of Industrial Relations (November 1988): 311-26

Bok, Derek C. and john T. Dunlop, Labour and the American Community (New York: Simon and Schuster, 1970)

Bryson, A et al. “Youth-Adult Differences in the Demand for Unionization: Are American, British, and Canadian Workers All That Different?” Journal of Labor Research 26 (January 2005): 155-67

Canadian Elections Study; mail back survey

Duffy, N.F. “Influences on Public Opinion od Unions and Industrial Relations.” Journal of Industrial Relations 23 (December 1981): 417-29

Furnham, Adrian. “The Protestant Work Ethic, Voting Behaviour and Attitudes to the Trade Unions.” Political Studies 32 (September 1984): 420-36

Gallup poll

http://www.cbc.ca/sports/hockey/nhl/story/2012/12/12/sp-nhl-hockey-survey-lockout.html

http://www.sfgate.com/news/article/More-protests-after-Michigan-right-to-work-action-4111896.php

http://www.tbs-sct.gc.ca/pubs_pol/hrpubs/TBM_11B/esao-esea-eng.asp

http://www.theglobeandmail.com/report-on-business/economy/jobs/will-right-to-work-michigan-puncture-ontarios-factories/article6281563/

Johnston, Richard. Public Opinion and Public Policy in Canada: Questions of Confidence. Toronto: University of Toronto Press, 1986.

Klandermans, P.G. “Psychology and Trade Union Participation: Joining, Acting, Quitting.” Journal of Occupational Psychology 59 (September 1986): 189-204.

Lewenza, Ken. "Birth of a Super-Union." Canadian Business 85, no. 16 (October 15, 2012): 24

Lipset, S.M. “Labour Unions in the Public Mind.” In Unions in Transition: Entering the Second Century. S.M. Lipset. San Fransisco: Institute for Contemparary Studies, 1986.

Lowe, S, Grahm and Harvey Krahn. “Recent Trends in Public Support for Unions in Canada” Journal of Labour Research (Fall 1989): 391-410

OECD “Http://stats.oecd.org/Index.aspx?QueryId=20167#20167”

Poole, Michael, Theories of Trade Unionism (London: Routledge and Kegan Paul,1981)

42

Page 45: Is there a secular trend in the public opinion of unions in Canada over time

Riddell, W, Craig. “Canadian Labour Relations: An Overview.” Canadian Labour Relations. Edited by W. C. Riddell Toronto: University of Toronto Press, 1986.

Rosier, Martin and Tim Little. “Public Opinion, Trade Unions and Industrial Relations.” Journal of Occupational Psychology 59 (September 1986): 259-72

Schmidt, E, Diane. “Public Opinion and Media Coverage of Labour Unions.” Journal of Labour Research 14 (Spring 1993): 151-64

Statistics Canada

43