The Hangover: The Lasting Impact of Early Conflict on the Duration of Bargaining Relationships

Abstract

Although studies have shown that employer opposition negatively affects the likelihood of success at the certification and first-contract stages, there is little to no indication of what the long-term impact of such opposition might be for relationships that successfully conclude a first agreement. Using survival analysis on nearly three decades of data from Ontario, this article investigates whether relationships that experience early conflict encounter a “hangover”—a heightened likelihood of dissolution—that extends into the representation phase of the process, meaning beyond the settlement of the first agreement. Using unfair labor practice charges and first contract arbitration applications as proxies for conflict, the author finds evidence of a hangover for relationships that exhibit a turbulent start. Further, findings suggest that relationships that experience this early conflict also have a higher likelihood of dissolution throughout the entire relationship.

Keywords

duration analysis employer behavior empirical analysis industrial relations theory labor–management relations private sector union decline union organizing

Given the downward trajectory of the unionization rate in many countries over previous decades, much of industrial relations scholarship has been spent on trying to uncover the determinants of the decline and the effect of legislative action meant to target it. With much of the focus of this scholarship and legislation on union organizing, it is unsurprising that these studies have concentrated on the initial phases of the process, namely the certification and first-contract stages (see, for example, Cooke 1983, 1985; Weiler 1983, 1984; Johnson 2002; Riddell 2013). Although the importance of such studies cannot be overlooked—they identified bottlenecks in the unionization process at which the pursuit of union representation may be extinguished—equally important to recognize is that these stages are only a portion of a larger unionization process. Obtaining certificates of representation and first contracts do represent successes; however, employees do not undertake the sometimes long, arduous procedure of obtaining union representation for solely the achievement of either of those early steps. Rather, employees presumably undertake it to obtain continued representation by a union that allows for continued bargaining over their terms and conditions of work. Thus, it is imperative to ask: Are bargaining relationships sustainable?

The main goal of this article is to investigate the sustainability of bargaining relationships, particularly when the relationship is born of conflict. Many of the preceding studies on the earlier stages of the process focus on the deleterious effect of employer opposition. A number of studies have shown that employer opposition that takes place during the certification stage has an impact not only on reducing union success in organizing but also on reducing the likelihood that workers are able to secure a first contract (Ferguson 2008; Riddell 2013). Even if workers are successful in organizing and securing a first contract, however, the bargaining relationship they establish is still more likely to dissolve than will the bargaining relationships established without such conflictual beginnings. This early conflict in the pre-relationship period, prior to the settlement of the first contract, makes the ensuing bargaining relationship more vulnerable to dissolution than bargaining relationships that begin peacefully, suggesting that those relationships that experience early conflict are marred by it.

This article undertakes this investigation by merging two governmental data sets from the Canadian province of Ontario for a nearly three-decade period. The data set from the Ontario Labour Relations Board (OLRB) focuses on the organizing phase. Through unfair labor practice (ULP) charge filings and first contract arbitration (FCA) applications, these data allow for the identification of bargaining relationships that experience early conflict. First contract arbitration is a provision of labor law common in the Canadian provinces in which either of the parties to a bargaining relationship may apply to have their first agreement settled by the labor board or by a private arbitrator when the parties are unable to settle it themselves. Generally, unions apply for FCA when they feel that the employer is frustrating the negotiation of a first agreement. Thus, FCA applications provide another measure for early conflict, and employer resistance in particular, that may take place at the beginning of bargaining relationships. The other data set has longitudinal contract data from the Ontario Ministry of Labour (MOL), which allows for the observation of the overall duration of relationships. The combination of these two data sets allows me to examine whether the incidence of early conflict at the beginning of the relationship has an impact on the duration and likelihood of dissolution of bargaining relationships after the relationship is established.

My study finds a greater likelihood of relationship dissolution for bargaining relationships that experience conflict prior to the settlement of a first contract. This outcome alone is not a novel finding. Bentham (2002) found that the presence of employer actions generally considered to be ULPs prior to the settlement of the first agreement was associated with a statistically significant increase in “early decertification.” Bentham examined only the effect of ULP charges through the term of the first collective agreement and only for dissolution through decertification, which leaves us with only a partial understanding of the total impact that early conflict has on the sustainability of bargaining relationships. This current study provides a more comprehensive depiction of the impact of early conflict by examining its influence beyond the first agreement and for the variety of ways in which a relationship may end. Using the period 1985 to 2012, I estimate that the presence of ULP charges prior to a first contract correlates with an approximately 33 to 37% higher likelihood of relationship dissolution throughout the entire relationship as compared to those relationships that are absent such charges. Likewise, FCA applications associate with an approximately 56 to 65% increase in the likelihood of relationship dissolution as compared to those relationships that do not involve an application. These findings have a number of important implications for our understanding of the sustainability of bargaining relationships. They show that the circumstances in which relationships begin are extremely important for their long-term viability, as the relationships that experience early conflict may be irrevocably marred by it. Further, the likelihood of dissolution remains higher throughout the entire relationship, as compared to their counterparts that do not experience discord. Last, when accounting for the different manners in which relationships may dissolve, the relationships that experience early conflict are more likely to end through decertification than through business closure, signifying a deliberate attempt to exit the relationship.

Background

Research on certification elections and the negotiation of first contracts has long established that these stages represent significant hurdles in the unionization process. Further, much of it has documented that the success of unions at these stages has worsened over the past half century (see, for example, Weiler 1983, 1984; Ferguson 2008 for the United States; Johnson 2002; Riddell 2013 for Canada). In attempting to explain these trends, many studies highlight the role that employer opposition has played, typically using unfair labor practice charges as a proxy for such opposition. Clearly, employer opposition affects these early stages of the process, whereas there is little to no indication of what effect, if any, employer opposition might have on outcomes beyond the settlement of the first contract.

To my knowledge, only one empirical study attempts to investigate the impact of early employer opposition beyond the settlement of the first agreement. Using survey evidence of approximately 400 bargaining relationships that achieved certification during the period 1991 to 1993 across eight Canadian provinces, Bentham (2002: 179) found that the occurrence of “actions commonly recognized as unfair labour practices” during the organizing phase correlates with a 46 to 57% increase in the likelihood of “early decertification.”¹ Somewhat curiously though, she did not find any statistically significant correlation between actual filed ULP charges and the likelihood of early decertification. Although this discrepancy is perplexing, the former finding provides the only indication of a possible lasting effect for relationships that experience a turbulent beginning.² Even with this finding though, the influence of employer opposition at the beginning of the process has been investigated only from the certification election through to the conclusion of the first agreement, and it has been investigated only for relationships that have ended through decertification. As shown in Figure 1, this may provide only a partial glimpse into the impact of early conflict as many bargaining relationships end in subsequent rounds of bargaining and do so in a number of ways. Bentham (2002) provided us with an indication of the damage that early conflict may have on the viability of bargaining relationships, yet her focus on decertifications taking place through only the term of the first agreement results in an incomplete appreciation of the role that early conflict plays in bargaining relationship duration and dissolution.

Figure 1.

Bargaining Relationship Dissolutions by Round in Ontario, 1985–2014

Thus, what might we expect the impact of employer opposition to be for newly established bargaining relationships beyond the settlement of the first agreement? It is possible that the institution of collective bargaining may not be (or rather, may no longer be) transformative enough to salvage a relationship that has been marred by conflict. As mentioned above, it has already been shown that employer opposition can negatively affect the likelihood of success at the certification and first-contract stages of the process (for the certification stage, see Cooke 1983; Bronfenbrenner 1997; Riddell 2001; Ferguson 2008; for the first-contract stage, see Cooke 1985; Bronfenbrenner 1994; Bentham 2002; Ferguson 2008; Riddell 2013). Perhaps, the negative influence of employer opposition will continue into or have a similar effect on the representation phase of the process. Absent Bentham’s (2002) article though, there has been very little empirical investigation into the representation phase of the unionization process and thus, little to no empirical evidence that this may be the case. One might expect that the decertification literature could provide some indication as to how employer opposition influences relationships beyond the settlement of the first agreement; however, many of the decertification studies focus only on the outcome of the election and, therefore, disregard when in the process the decertification takes place (see, for example, Anderson, Busman, and O’Reilly 1982;³ Dickens, Wholey, and Robinson 1987; Meyer and Bain 1994). Thus, what may have precipitated the decertification could be the conflict that was experienced at the beginning of the relationship or it could be something much later in the process.

By contrast, early theories on bargaining relationship development by seminal scholars in the field of industrial relations may provide us with a different hypothesis. During the post–WWII boom in unionization, when many bargaining relationships were developing and perhaps maturing, these scholars theorized about the conditions that were required for the development and maturation of relationships. Although perhaps not their main focus, their theorizing pointed toward the hypothesis that relationships may be endangered in their infancy because of the inexperience of the parties in dealing with each other. Each of them seemingly alludes to a trial period during which the parties attempted to come to an acceptable agreement over the role and function that each would play in the relationship, and that the absence of such an agreement could bring about conflict and would perhaps be a harbinger for the death of the relationship.

Ross (1948: 107) theorized that an initial “basic power settlement” needed to take place for a “political compatibility” to develop between labor and management, the presence of which could contribute to the stability and maturity of the relationship. This political compatibility represents a situation in which both sides are potentially able to realize the benefits of a relationship. For the employer, examples of these benefits include “the union as an agency for supplying qualified workers, imposing industrial discipline, and contributing generally to the management of the enterprise” (ibid.: 107). The absence of this power settlement, however, could result in conflict that prevents the establishment of a new bargaining relationship or the continuation of an existing one. Likewise, Dunlop’s (1958) systems theory of industrial relations posited that the parties in a system create a “web of rules” to govern the system and the interactions of those within it. Further, each system is bound together by an ideology, which is composed of the individual ideologies of the parties. When the ideologies of the parties are in congruence, or rather compatible, then this brings stability to the system. If the ideologies of the parties are incongruent, then this could bring instability and conflict to the system.

Lester (1958) further theorized that both the union and the relationship were able to mature through the stable predictability that is produced by the continued interactions of the union and management through several rounds of bargaining. It is through such interactions that the parties bridge, at least partially, some of the divide that separates them initially. Furthermore, these subsequent interactions and negotiations may help the parties come to a shared understanding and mutual respect of each other, which may move them past any preconceptions that the other is a threat to their own security. Additionally, the parties are likely to develop shared mechanisms through which they can overcome the conflict inherent in the bargaining process and, as the agreements become more intricate, come to depend on each other for the contract’s administration. These developments subsequently result in bargaining that becomes more routine and less conflictual, which according to both Ross (1948) and Lester (1958) is indicative of a mature bargaining relationship. Thus, the postwar theorists hypothesized that collective bargaining would become entrenched for those relationships that were able to survive this initial period as these trends could transform relationships into ones of mutual respect and cooperation, allowing them to persist over the medium to long-term.

Paul Weiler later adopted an assumption similar to that of the postwar theorists, and in marrying it with the “rogue employer hypothesis”—that the misfortunes of unions were in some part caused by the opposition of employers at the start of relationships—developed first contract arbitration as a policy to mitigate the harmful impact of employer opposition and to facilitate the development of lasting bargaining relationships. Indeed, in designing the initial first contract arbitration statute in British Columbia, Weiler assumed that the opposition from employers resulted from their inexperience in dealing with a union, an anxiety that might be overcome if the parties were forced (either directly or through the threat of arbitration) into a first contract. The goals of FCA legislation include the prevention of first-contract work stoppages (see Johnson 2010) and the fulfillment of the expectation that workers who choose union representation will be met with a collective agreement (see Riddell 2013); it was also hoped that the impact of the legislation would extend beyond the settlement of the first agreement.

This first agreement was seen as absolutely vital in that it allowed the union to become established in the workplace and to demonstrate the value of collective bargaining and working under a collective agreement to the members of the bargaining unit. Ultimately, Weiler hoped that such a “trial marriage” would lay the foundation for lasting, mature bargaining relationships by familiarizing the parties and normalizing collective bargaining (Weiler 1980). Indeed, in describing this goal, he wrote that “first contract arbitration attempts to do more than simply settle a past dispute: it also seeks to install the union firmly within the plant and to foster the kind of relationship that most likely would have arisen had the employer not acted in bad faith” (Weiler 1984: 409). Under Weiler’s assumption, one might expect that once the union has a foothold in the workplace and the parties become familiarized, the employer will learn that the union and the collective agreement do not represent threats to the organization. This might lead to the employer’s acceptance of the decision of his or her workforce to obtain union representation such that the bargaining relationship would be able to endure. Implicit in the above quote, however, is that this assumption hinges on a change in the employer’s attitude and behavior. The initial first contract arbitration statute was meant as a remedy for illegal acts committed by the employer, and it was hoped that righting those wrongs would produce the hypothetical relationship “that most likely would have arisen had the employer not act acted in bad faith” (ibid). The imposition of the agreement, or rather the threat of one, does not change the employer though and so a bargaining relationship may persist only when an opposed employer changes his or her pattern of behavior.

Weiler subsequently abandoned his hope that FCA in British Columbia might foster lasting bargaining relationships because nearly all of the cases in which the labor board imposed an agreement in the first five years of its operation resulted in the bargaining unit’s decertification (Weiler 1980). Weiler’s skepticism may have been unwarranted: first, Weiler was basing his observation on only 12 cases. Subsequent studies on other provinces that had more accessible forms of FCA or that imposed agreements with more favorable terms and conditions were shown to potentially produce better outcomes pertaining to this goal (McDonald 1987; Sexton 1987). Second, and perhaps of more importance, only a small percentage of first-contract negotiations go through the first contract arbitration process and even fewer result in an imposed agreement (Johnson 2010). Cases in which the parties are unable to come to an agreement, such that it requires the issuance of an arbitration award, are very likely to be the worst of the worst. For the vast majority of nascent bargaining relationships outside of that process, or even those within it that are able to come to a settlement, the question remains as to whether an appreciable difference occurs in the likelihood of the relationship coming to an end for those relationships that experience a turbulent beginning compared to those that do not. Thus, perhaps Weiler’s initial assumption may hold more widely for those relationships that do not require the imposition of a first agreement by a third party.

This article, therefore, seeks to answer the question of whether a lasting effect from a conflictual start to a bargaining relationship persists beyond the settlement of the first agreement. One hypothesis to this question may be supported: perhaps the institution of collective bargaining is not transformative enough to salvage a relationship that has been marred by conflict. Or rather, perhaps in line with the rogue employer hypothesis, opposed employers will always be opposed employers and a collective agreement will not change their desire to operate union-free. Thus, we might expect that conflict prior to the representation phase will increase the likelihood of the relationship ending. Alternatively, perhaps as was assumed by the postwar theorists and was hoped for by Weiler, the institution of collective bargaining is sufficiently transformative to move the parties past their initial misgivings about each other such that the relationship may be salvaged, despite early conflict, when they conclude an agreement and enter the representation phase. In seeking to answer this question, this article is the first to investigate the influence of employer opposition at the start of relationships beyond the first agreement, when numerous relationships continue to end, and it is the first to account for the variety of ways in which they might end. Furthermore, it investigates not only if there is an effect but also how persistent it may be, the findings of which have both theoretical and policy implications.

Data and Methodology

Given the large size and long time frame of the decline in union density in the United States, much of the deunionization literature has focused on that country. This trend may make the United States the ideal setting to investigate a subject that is hypothesized to be one of the main determinants of the decline, but the type of longitudinal relationship data that is needed to perform this investigation does not exist in the United States. However, declining union density is by no means exclusive to the United States. Since 1980, many advanced industrial nations have experienced decreases in their levels of union membership, especially the Anglo-American liberal market economies (Schmitt and Mitukiewicz 2011; Colvin and Darbishire 2013). The decline in the union densities of the United States and Canada from 1981 to 2004, a time frame that substantially overlaps with the period of analysis in this study, was remarkably similar (Warner 2012). In addition to this similarity, the United States and Canada are often viewed as natural comparators in the field of industrial relations since the provincial legislations governing industrial relations in Canada are substantially based on the US National Labor Relations Act (NLRA). As mentioned in the introduction, this has led to simultaneous research in both countries that has focused on the difficulty of bargaining units in obtaining certification and first contracts and that has investigated the influence of employer opposition at various stages of the unionization process. Some of this research aimed to inform the policymakers of both countries, particularly that which focused on policies proposed in the US Employee Free Choice Act (Johnson 2010; Riddell 2013; Weinberg 2015). Further, because of the commonalities in their economies, institutions, cultures, and actors,⁴ the United States and Canada have also been used as comparators for other labor market interventions, including employment law (Colvin 2006) and unemployment insurance (Kuhn and Riddell 2010). Finally, the province of Ontario is arguably the most suited for generalizations to the United States for a number of reasons (Riddell 2013). It is viewed as one of, if not the most, business-friendly provinces, it has a unionization rate that is comparable to the more heavily unionized states, and it has enacted labor law changes in recent decades that have moved it toward those that are found in the United States. The Ontario Labour Relations Act requires that the parties to a collective agreement submit the agreement to the Ontario MOL, which means that agency has a database that is a complete census of settled collective agreements in the province. These longitudinal contract data permit the investigation of whether early conflict prior to the representation phase of the unionization process has lasting effects and may allow for cautious generalization of this study’s findings to the other Canadian provinces and the United States.

The complete data used in this study are the result of a merge between the Ontario MOL’s Collective Bargaining Information Services (CBIS) Database and the OLRB’s Reports. The former data source is used to track bargaining relationships through the number of completed rounds of bargaining. This data set also records the date and manner in which the relationship ended if it took place prior to 2015. The reasons for termination listed by the MOL that are used to define dissolution in the analysis below include decertifications, business closures, and a category for “others.”⁵ The latter data source is used to obtain information on the experience of relationships prior to the settlement of the first agreement, for example, certifications, unfair labor practice charge filings,⁶ and first contract arbitration applications.

Since my interest is focused on the population of newly formed bargaining relationships in Ontario, the CBIS Database ultimately determined the relationships included in the study. Given constraints in the CBIS Database, the period of analysis used in the study is 1985 to 2012 for newly formed relationships.⁷ With this population of interest, all relationships in the database that are not observed from their inception are omitted from the analysis. Further, the CBIS Database contains information on all relationships operating in the province, warranting a number of additional omissions before the data can be analyzed. The factors and processes that influence the duration of bargaining relationships, including employer resistance, across the provincial and federal jurisdictions and the public and private sectors are likely to vary greatly, calling their comparability into question. Therefore, this study focuses solely on private-sector bargaining relationships and omits any relationships under the federal jurisdiction or in the public sector. Additionally, all bargaining relationships in the construction industry are omitted from the analysis since many of those relationships are not comparable to those within other industries as they tend to arise and disappear with the acceptance and completion of construction projects. Following these omissions, and a few others,⁸ I had 4,076 private-sector non-construction relationships remaining from which I attempted to find matching information in the OLRB Reports.⁹

The main criterion used for merging the two data sets was the dyad of the employer and union names. In the CBIS Database, the MOL tried to link many of the first agreements to the certifications by including the case file number given to the certification by the OLRB. Where available, this case file number was used in the first round of the merge to identify certificates that matched to first agreements. When there was no file number or an incorrect one,¹⁰ further information, such as the location and the disposition month and year, was used to identify matches in subsequent rounds of the merge. As a result, I was able to successfully merge 3,386 certificates to a relationship, which represents approximately 83% of the private-sector non-construction relationships that remained following the above omissions. This number, combined with the inclusion of relationships that involved voluntary recognition by the employer, provided a sample of 3,860 relationships to use in the analysis, which includes slightly more than 14,000 collective agreements. ULP case filings were matched onto any relationships that listed the same parties and for which the disposition date of the ULP was found to take place during the period between a year prior to the certification/disposition date and the ratification date of the first agreement. This process resulted in matching 834 relationships with ULP charge filings, which represents approximately 22% of the relationships included in the analysis.¹¹ The merging of FCA applications resulted in the matching of at least one first contract arbitration application onto 113 relationships in the CBIS Database, representing slightly less than 3% of the relationships.

With the focus of this study being the effect of early conflict on the length of bargaining relationships and the probability of the dissolution of those relationships, the appropriate method for such an investigation is survival analysis, also known as event history or duration analysis. This method is preferable because it easily handles relationships that are right-censored or, rather, those that do not experience a dissolution by the end of the period of analysis (Box-Steffensmeier and Jones 1997). Specifically, this study employs the semi-parametric Cox proportional hazards regression model (Cox 1972).¹² This model assumes that the dissolution of bargaining relationships occurs as part of a continuous process, meaning that it can take place at any point in time, and uses days as the unit of analysis time.¹³ The dependent variable is the hazard rate, which is the estimate of the probability of experiencing a dissolution during a unit of analysis time, given that it has not experienced dissolution in any of the previous units of analysis time.¹⁴ The Cox model allows for 1) the investigation of what characteristics of the bargaining unit and negotiations influence the duration and dissolution of bargaining relationships and 2) the exploration of how the probability of dissolution faced by bargaining relationships evolves over time. For a list of summary statistics for the independent variables used in the estimation, see Table 1.

Table 1.

Summary Statistics

Variables	Mean	Standard deviation	Minimum	Maximum
ULP charges (general)	0.215	0.411	0	1
FCA application	0.025	0.156	0	1
Bargaining unit size	62.582	172.263	1	4,996
Log bargaining unit size	3.211	1.235	0	7.550
Log bargaining unit size(t)	3.225	1.306	0	8.516
Work stoppage	0.042	0.200	0	1
Conservative government	0.344	0.475	0	1
NDP government	0.130	0.336	0	1
Liberal government (base)	0.527	0.499	0	1
Pre-OLRA 1986 (base)	0.084	0.278	0	1
OLRA 1986	0.359	0.750	0	1
OLRA 1993	0.211	0.408	0	1
OLRA 1995	0.182	0.385	0	1
OLRA 2000	0.165	0.371	0	1
Primary	0.007	0.083	0	1
Manufacturing	0.386	0.487	0	1
Transportation	0.062	0.241	0	1
Trade	0.165	0.371	0	1
Education	0.009	0.097	0	1
Health	0.030	0.171	0	1
Other services	0.340	0.474	0	1

Notes: ULP, unfair labor practice; FCA, first contract arbitration; NDP, New Democratic Party; OLRA, Ontario Labour Relations Act.

Negative regression coefficients produced by the Cox model correspond with a decrease in the likelihood of relationship dissolution, and positive regression coefficients correspond with an increase in the likelihood of relationship dissolution, but they are difficult to interpret beyond this. Exponentiating the regression coefficients transforms them into hazard ratios, which allows for an easier interpretation of the magnitude of the effect. These hazard ratios can be interpreted as the percentage difference in the likelihood of dissolution caused by a one-unit change in the value of the covariate. A coefficient above 1 corresponds with an increase in the likelihood of dissolution, and therefore an expected decrease in the length of the relationship; a coefficient below 1 corresponds with a decrease in the likelihood of dissolution, and therefore an expected increase in the length of the relationship. For example, if an independent variable has an estimated coefficient of 1.25 then this would indicate a 25% increase in the likelihood of relationship dissolution for a one-unit change in the value of the variable. A coefficient of 1 indicates that the corresponding covariate has no influence on the likelihood of relationship dissolution.

Finally, the Cox model assumes that the effect of independent variables is proportional and constant with respect to time. Known as the proportional hazards (PH) assumption, it is of particular theoretical interest to this article. If the explanatory variables used to model conflict at the beginning of relationships are shown to violate this assumption, then this result indicates that the effect of the variables changes throughout the relationship. On the one hand, this outcome may provide evidence toward the transformative nature of collective bargaining in that parties are able to overcome a turbulent start to their relationship (assuming that the effect of the variable is waning). On the other hand, if the ULP charge and FCA application variables are shown not to violate this assumption, this would indicate that the effects of the variable do not change throughout the relationship. This model may provide evidence toward the hypothesis that relationships that experience conflict at their outset are marred by it, such that the effect of a turbulent start may always predispose them to a higher likelihood of dissolution.¹⁵ Using a test of whether the main conflict variables of interest violate the proportional hazards assumption, this study investigates how persistent the effect of early conflict, if there is one, might be.

Results

Preliminary analysis using Kaplan-Meier estimates of the survival function for private-sector non-construction bargaining relationships in Ontario indicates that relationships that are affected by either ULP charge filings or FCA applications prior to the first agreement both exhibit statistically significantly worse survival experiences than do relationships that lack either of those occurrences.¹⁶ To illustrate, the median estimated duration for a relationship with ULP charge filings is 4,684 days (or 12.82 years), which compares to an estimated median duration of 7,305 days (or 20.00 years) for relationships without such filings—a difference of 2,621 days (or 7.18 years). Similarly, the median estimated duration for a relationship that files an application for access to FCA is 2,414 days (or 6.61 years), which compares to an estimated median duration of 6,729 days (or 18.42 years) for relationships that never seek access to the FCA system—a difference of 4,315 days (or 11.81 years).

Although the preliminary analysis provides us with some initial indication of the harmful impact of early conflict, it does so without controlling for other factors. Table 2 presents the Cox regression estimates for the influence of the incidence of ULP charge filings prior to the settlement of the first agreement and applications for first contract arbitration on the likelihood of bargaining relationship dissolution. Each specification controls for the log of the initial size of the bargaining unit. The columns in the table further differ by the inclusion of various fixed effects that control for time-invariant characteristics of industries, unions, and regions pertaining to the relationships. Column (4) includes independent variables for the calendar year in which the relationship began, which control for the macroeconomic environment during that year and ensure that the results are not influenced by when relationships began during the period of analysis (Wooldridge 2001). As one can see from the table, slight variations in the coefficients exist depending on the inclusion of the various controls, but the estimates are quite robust. The presence of ULP charge and FCA application filings statistically significantly associate with an increased likelihood of relationship dissolution as compared to relationships absent either of those filings. The filing of any ULP charges prior to the settlement of the first agreement associates with a 33 to 37% increase in the likelihood of dissolution. The filing of an application for access to FCA associates with a 56 to 65% increase in the likelihood of dissolution. Thus, it appears that at least an initial adverse effect occurs for relationships that begin amid conflict as measured by the presence of ULP and FCA filings.

Table 2.

Cox Regression Estimates for Private-Sector Bargaining Relationships in Ontario, 1985–2012

Variables	(1)	(2)	(3)	(4)	(5)	(6)
ULP charge(s)	0.285***	0.317***	0.314***	0.299***	0.337***	0.295***
	(0.058)	(0.061)	(0.061)	(0.061)	(0.096)	(0.061)
	[1.329]	[1.373]	[1.369]	[1.348]	[1.401]	[1.343]
FCA application	0.498***	0.455***	0.455***	0.482***	0.477***	0.785***
	(0.135)	(0.138)	(0.137)	(0.140)	(0.140)	(0.215)
	[1.645]	[1.576]	[1.575]	[1.619]	[1.612]	[2.193]
Log bargaining unit size	−0.108***	−0.155***	−0.161***	−0.163***	−0.163***	−0.162***
	(0.021)	(0.023)	(0.023)	(0.024)	(0.024)	(0.024)
	[0.898]	[0.857]	[0.851]	[0.850]	[0.850]	[0.850]
ULP charge(s)*time					−0.006
					(0.011)
					[0.994]
FCA application*time						−0.054*
						(0.031)
						[0.947]
Industry effects	Yes	Yes	Yes	Yes	Yes	Yes
Union effects	No	Yes	Yes	Yes	Yes	Yes
Region effects	No	No	Yes	Yes	Yes	Yes
Cohort effects	No	No	No	Yes	Yes	Yes

Observations	14,034	14,034	14,034	14,034	14,034	4,225,269
Relationships	3,860	3,860	3,860	3,860	3,860	3,860
Dissolutions	1,612	1,612	1,612	1,612	1,612	1,612

Notes: Robust standard errors clustered at the relationship level are presented in the parentheses below the regression coefficient. Hazard rate is presented in the brackets. Log bargaining unit size is the log of the size of the bargaining unit listed for the first agreement in the relationship. Industry dummies representing primary, manufacturing, transportation, trade, education, health, and other services are included in all of the specifications. Regional effects are based on the Ministry of Labour’s economic region codes. Hazard ratio is the exponentiation of the corresponding coefficient. ULP, unfair labor practice; FCA, first contract arbitration.

Statistically significant at the .10 level; ** at the .05 level; *** at the .01 level.

The control variable for the log of the initial size of the bargaining unit is shown to statistically significantly correlate with a decrease in the likelihood of dissolution. Depending on the specification, each 1% increase in the initial size of the bargaining unit associates with an approximate 10 to 15% decrease in the likelihood of dissolution. The size of the bargaining unit has been found to negatively correlate with the likelihood of winning a certification election (see Heneman and Sandver 1983 for a review; Cooke 1983; Ferguson 2008), but positively correlate with the likelihood of achieving a first agreement (see Bain 1981 for a review; Cooke 1985) and defeating a decertification attempt (Krislov 1956; Anderson, Busman and O’Reilly 1979; Chafetz and Fraser 1979; Elliott and Hawkins 1982; Ahlburg and Dworkin 1984; Dickens et al. 1987; Meyer and Bain 1994). Thus, it appears that this finding is largely in line with the findings of other studies that larger bargaining units are more successful than smaller ones in outcomes beyond certification. This result suggests that unions should perhaps attempt to organize larger bargaining units, as those relationships may prove more durable.

As stated in the previous section, the proportional hazards (PH) assumption that the effect of independent variables does not vary over time is of particular interest for this study as it suggests how persistent the effect of early conflict might be. The simplest way to check for a violation of the PH assumption is by interacting independent variables with analysis time. If the coefficient on the interaction term is shown to be statistically significant then this shows that the effect of the independent variable varies over time. This check is performed in the last two columns of Table 2 for the ULP and FCA variables, respectively. The interaction term in column (5) involving the ULP charge variable is statistically insignificant, indicating that it does not violate the PH assumption. Thus, it appears that the effect of the ULP covariate is time-invariant, suggesting that those relationships that involve a ULP charge experience a statistically significantly higher likelihood of dissolution throughout the entire relationship as compared to relationships that do not involve ULP filings at their outset. The interaction term in column (6) involving the FCA application variable, however, is marginally statistically significant, indicating that it does violate the PH assumption.¹⁷ Thus, although it seems quite clear from the Cox regression estimates in Table 2 that a turbulent start to a bargaining relationship has a lasting impact, how lasting the association of that bad start might be may depend on the measure that is used as a proxy for conflict.

Following the estimation performed in Table 2, I inspected the baseline hazard functions produced after controlling for the covariates included in the regressions.¹⁸ These functions show how the likelihood of relationship dissolution changes over time for a hypothetical relationship with all of the independent variables set to the mean based on either unobserved or omitted variables (except for the main variables of interest on which they are stratified). The baseline hazard functions reveal that the risk of dissolution is increasing through the first 1,500 days (or nearly four years) of relationships, before beginning to decline. The shape of the hazard indicates that the presence of ULP charge and/or FCA application filings do not explain all of the risk of dissolution that relationships face initially. Therefore, it appears that all relationships, regardless of the level of conflict, or rather those used here to measure conflict, may go through somewhat of a trial period. Given that first contracts tend to be two to three years in duration, this trial period generally extends beyond the first agreement. This observation may provide some evidence toward the hypothesizing of the post–WWII theorists that all relationships may go through an initial period of a heightened risk of dissolution, perhaps because of the inexperience of the parties in dealing with each other and the need to build trust and cooperation.¹⁹

Table 3 presents a number of robustness checks to investigate the sensitivity of the estimates in Table 2 and to assess alternate explanations for the effect of early conflict on the likelihood of dissolution during the representation phase of the unionization process. Column (1) investigates whether events that take place after the settlement of the first agreement account for the influence of the ULP and FCA covariates. Indeed, one might assume that if an employer is willing to commit acts that necessitate the filing of ULP charges or FCA applications during the organizing phase, then that same employer may be willing to take actions against the union during the representation phase. Thus, this specification includes a binary variable for the incidence of a work stoppage during any round of bargaining. It also allows the log size of the bargaining unit to vary throughout the duration of the relationship to control for the shifting of work into or out of the bargaining unit. The next three columns investigate whether the external environment may influence the estimates of conflict by controlling for the political, legal, and economic environments, respectively. Column (2) examines the political environment through the inclusion of dummy variables for when the historically business-friendly Progressive Conservative (PC) party or the historically labor-friendly New Democratic Party (NDP) form the government in power (leaving the more centrist Liberal Party as the base case). Column (3) controls for the four major revisions to the Ontario Labour Relations Act during the period of analysis. Martinello (2000) showed that both the political party in power and the legislation that it enacts affect the certification, decertification, and ULP filing activity within Ontario.²⁰ Column (4) attempts to control for the economic environment by including the provincial unemployment rate and finally, column (5) includes all of the controls of the previous four specifications. As one can see from the estimates presented in the table, the inclusion of these variables have little to no impact on the estimates of the main variables of interest as compared to Table 2. The coefficients for a ULP filing prior to the first agreement remain essentially unchanged, whereas the estimates of the impact of the FCA application variable result in a slight decline in magnitude only when the model accounts for subsequent work stoppages and/or changes to the size of the bargaining unit. Thus, the impact of early conflict, as proxied by ULP charge and FCA applications filings, on the duration of bargaining relationships appears to be quite robust, regardless of the environment external to the bargaining relationship or subsequent events for which the model is able to control.

Table 3.

Robustness Tests

Variables	(1)	(2)	(3)	(4)	(5)
ULP charge(s)	0.294***	0.294***	0.299***	0.299***	0.290***
	(0.061)	(0.061)	(0.061)	(0.061)	(0.061)
	[1.342]	[1.342]	[1.349]	[1.348]	[1.337]
FCA application	0.432***	0.492***	0.484***	0.482***	0.452***
	(0.140)	(0.141)	(0.140)	(0.140)	(0.140)
	[1.540]	[1.635]	[1.622]	[1.620]	[1.571]
Log bargaining unit size	−0.210***	−0.162***	−0.163***	−0.163***	−0.210***
	(0.022)	(0.024)	(0.024)	(0.024)	(0.022)
	[0.811]	[0.851]	[0.849]	[0.850]	[0.810]
Work stoppage	0.306***				0.316***
	(0.096)				(0.095)
	[1.358]				[1.371]
Conservative government		0.372***			0.412***
		(0.085)			(0.085)
		[1.451]			[1.509]
NDP government		0.401***			0.428***
		(0.131)			(0.130)
		[1.493]			[1.533]
OLRA 1986 regime			0.103		0.097
			(0.171)		(0.178)
			[1.108]		[1.101]
OLRA 1993 regime			−0.060		−0.049
			(0.432)		(0.413)
			[0.942]		[0.952]
OLRA 1995 regime			−0.107		−0.064
			(0.472)		(0.453)
			[0.898]		[0.938]
OLRA 2000 regime			0.384		0.531
			(0.585)		(0.567)
			[1.467]		[1.701]
Unemployment rate				−0.004	−0.031*
				(0.016)	(0.016)
				[0.996]	[0.969]

Observations	14,032	14,034	14,034	14,030	14,028
Relationships	3,860	3,860	3,860	3,860	3,860
Dissolutions	1,612	1,612	1,612	1,612	1,612

Notes: Robust standard errors clustered at the relationship level are presented in the parentheses below the regression coefficient. Hazard rate is presented in the brackets. Hazard ratio is the exponentiation of the corresponding coefficient. Log bargaining unit size is the log of the size of the bargaining unit listed for the first agreement in the relationship in columns (2)–(4), but it is allowed to vary with each collective agreement in columns (1) and (5). All specifications include the industry, union, region, and cohort effects included in Table 2. ULP, unfair labor practice; FCA, first contract arbitration; NDP, New Democratic Party; OLRA, Ontario Labour Relations Act.

Statistically significant at the .10 level; ** at the .05 level; *** at the .01 level.

Finally, the previous estimates pooled together all of the circumstances under which relationships might end. As was shown in Figure 1, however, relationships can dissolve in a variety of ways and the effect of conflict may differ across them. Table 4 performs an analysis that looks at how the covariate estimates differ based on whether the relationship ended through decertification or plant closure. As one can see in column (1), the presence of ULP charges and FCA applications significantly correlates with the likelihood of decertification, and the magnitude of the estimates is vastly higher than the pooled estimates. The presence of a ULP charge filing correlates with a nearly 80% increase in the likelihood of decertification, and an FCA application associates with an almost 180% increase in the likelihood of decertification. Column (2) shows that only a ULP charge filing associates with a marginally statistically significant increase in the likelihood of a plant closure, with it estimated to correlate with a nearly 16% increase. In the case of a decertification, it is clearly an attempt to get out of the bargaining relationship, but a plant closure may occur for a variety of reasons, some of which may have little or nothing to do with the bargaining relationship itself. This difference may explain the decreased magnitude and significance when looking at the association of early conflict with plant closures as compared to decertifications.

Table 4.

Cox Regression Estimates by Type of Relationship Dissolution

Variables	Decertifications	Plant closures
ULP charge(s)	0.583***	0.145*
	(0.101)	(0.078)
	[1.792]	[1.156]
FCA application	1.025***	−0.093
	(0.180)	(0.222)
	[2.788]	[0.911]
Log bargaining unit size	−0.230***	−0.134***
	(0.039)	(0.030)
	[0.795]	[0.875]

Observations	14,034	14,034
Relationships	3,860	3,860
Dissolutions	565	1,011

Statistically significant at the .10 level; ** at the .05 level; *** at the .01 level.

Conclusion

Using data that allow for both the identification of relationships that experience early conflict and the observation of the total duration of relationships, this study finds that a turbulent start to a bargaining relationship has a lasting deleterious effect that may persist beyond the first agreement and well into the representation phase of the unionization process. Conflict is a difficult concept to operationalize, however, and so how persistent the effect might be may depend on how conflict is measured, or rather proxied. Results of the proportional hazards tests indicate that the effect of ULP filings do not vary with time, which is suggestive of a rogue employer hypothesis: either opposed employers remain opposed or relationships are marred by the conflict that occurs at their outset. Thus, it appears that relationships that experience ULP filings prior to the settlement of the first agreement face a higher risk of dissolution throughout the entire relationship as compared to those relationships that are absent any ULP filings during that portion of the process. Although it does appear that the influence of FCA applications is time-varying, perhaps suggestive of a trial marriage hypothesis in which those relationships may be salvageable and will endure, it seems more likely that the relationships that have to resort to the filing of an FCA application may be so conflictual that the association appears to diminish with time simply due to the dissolution of those relationships. This possibility is certainly something that should be assessed in future research. Also, a large disparity occurs in the association of conflict across the different types of dissolution. The incidence of a troubled start to a bargaining relationship correlates with a greatly increased risk of the relationship ending through decertification, whereas the association may be minor to nonexistent with regard to plant closures.

The focus of this study is the influence of early conflict on the duration of bargaining relationships, but it may also lend some empirical evidence to the hypothesizing about bargaining relationship development espoused by the aforementioned post–WWII theorists. As the post-estimation baseline hazard curves show, the beginning of relationships is a hazardous period, regardless of the variables used to proxy for conflict in this study. This finding could be caused by omitted variables that also represent conflict at the beginning of relationships or it could be because of the duration dependence of relationships. If it is the latter, this heightened hazard may indicate that all relationships go through somewhat of a trial period during which the relationship is more precarious early on in the life cycle rather than later, perhaps because of the inexperience of the parties and the need to settle the role and the function that each will play in the relationship.

This investigation into whether a hangover occurs for bargaining relationships that begin amid conflict is vital for a number of reasons. From a theoretical perspective, collective bargaining has historically been viewed as a potentially transformative process in which the parties come to an understanding of the role and function that each holds in the relationship. Further, bargaining may result in the development of joint machinery and a shared responsibility for the maintenance of both the relationship itself and the collective agreement. A finding that bargaining relationships that experience conflict from their outset due to the opposition of one of the parties are subject to a much higher propensity for dissolution may temper these notions. Indeed, given the decreasing relevance of unions—as exhibited by declines in the unionization rate in many countries and the findings of the research investigating the growth of management opposition to unionization—the capacity for collective bargaining to produce the mature, lasting relationships that Paul Weiler and the postwar theorists predicted may have diminished. From a policy perspective, much of the recent legislative reform, both actual and proposed, has focused on improving the outcomes of unions through the organizing phase. Employees presumably do not organize to solely achieve certification or a first agreement, however, and instead do so for the continued ability to influence the terms and conditions of their employment. Legislative reforms targeted to improve the achievement of certification and first agreements may help unions get their foot in the door, yet the impact and focus of such legislation must surely be reconsidered if that door is subsequently slammed shut.

Footnotes

Acknowledgements

I thank Alex Colvin, Adam Seth Litwin, Chris Riddell, and Samuel Kleiner for their helpful comments. I also thank the staff at the Ontario Ministry of Labour, especially Katherine Waterhouse, for providing me with access to much of the data that was used in this article and for answering all accompanying questions about it. Finally, I am very grateful to Mark Gough for assisting with the data.

The data in this paper may be obtained from the respective governmental agencies. The statistical programs are available from the author upon request at bw81@queensu.ca.

1

Bentham defined “early decertification” as a decertification that takes place within the first two open periods during which an application for decertification or a union raid may take place; the first period is the two to three months preceding the one-year anniversary of the certification and the second period is the two to three months preceding the expiration of the first agreement.

2

It may provide an indication, but it is not definitive. Of the 29 relationships in her sample that ended in “early decertification,” she never divulged how many of them occurred in each of the two open periods. Thus, we cannot know how many of the decertifications involved relationships that settled a first agreement, nor if the influence of employer opposition varies for the relationships that decertified in the first period, presumably without an agreement, versus those that decertified in the second period at the expiration of the first agreement. This approach did not enable Bentham to assess whether the settlement of a first agreement and the representation provided to the workers through that agreement during the representation phase may have mitigated the impact of the early employer opposition.

3

did investigate the opposition of employers, although only during the decertification campaign, and did control for the number of years in which the bargaining unit was represented. They did not relate the two or attempt to investigate the influence of early conflict on the likelihood of the unit decertifying, however.

4

Many companies and unions operate on both sides of the US–Canada border.

5

The “other” category includes voluntary terminations of bargaining rights by the union, employer service contract terminations, and the termination of bargaining rights by the OLRB.

6

ULP filings in which either party could be the applicant are included in the data set. Unfortunately, the indices do not list the type of ULP that is filed, so it is not possible to investigate whether certain ULPs are more harmful than others. However, the analysis includes only ULP filings with an employer and a union listed as the parties. This should filter out ULP filings pertaining to duty-of-fair-representation claims made by a bargaining unit worker against the union since that type of ULP should have little to no impact on the relationship between the union and the employer.

7

Since the period 1982 to 1984 accounts for a vast number of the relationships in the data set that begin with an agreement labeled as a “renewal agreement,” indicating the addition of existing relationships, it is likely that this is approximately when the database was created. This realization led to the choice of 1985 as the year to begin my analysis since the database should account for all, or at least the vast majority of, relationships beginning from that point moving forward. The year 2012 is used as the final year from which to analyze nascent bargaining relationships because of a change in the system used by the Ministry of Labour to maintain the database. During this period, an average of 325 bargaining relationships started each year.

8

There were 19 bargaining relationships in which the public-sector dummy variable varied across the bargaining relationship. The switch in sectors is likely attributable to data entry error or privatization. These relationships were also omitted from the estimation found below. There were also four bargaining relationships in which the number of employees in the bargaining unit was listed as 0 at some point during the relationship. These relationships were also dropped due to the missing data point.

9

This tally does not include relationships in which the union was voluntarily recognized by the employer as there would be no certification in the OLRB records to match those relationships. However, those are valid relationships and are included in the analysis performed below.

10

With the change in their system, only the file number for the most recent certification in a relationship is recorded. This means that any relationship in which a union displacement took place, the initial certification file number was overwritten.

11

It is not possible in some cases to identify which relationships the ULP matches onto when there are multiple relationships with the same parties, but by noting when this may be the case, I can investigate how the matching of ULPs influences the results of the estimation. The merging of ULPs for which there was no possible conflict (i.e., the match was unique) resulted in the matching of 718 relationships with one or more ULP charges, which represents approximately 19% of the relationships in the sample. All of the analyses performed below were also carried out with this restricted matching of ULPs, the results of which produce the same conclusions. These estimates are available from the author upon request.

12

The Cox model is given by $h (t | X_{j}) = h_{0} (t) \exp (X_{j} β_{x})$ where the hazard rate for subject j is determined by the baseline hazard, h₀(t), multiplied by the exponentiation of a vector of independent variables, X_j, and a vector of estimated regression coefficients, β_x . In this model, the baseline hazard is left unspecified and is actually conditioned out of the estimation. See Cleves, Gould, and Gutierrez (2004) for an in-depth discussion of this model.

13

This approach is as opposed to a discrete-time approach in which either 1) it is assumed that the failures take place only during discrete points in time or 2) the process is assumed to be continuous, but it is interval censored. The latter means that the event is known to have occurred within some interval of time, but the exact timing is unknown. At first glance, one might think that an argument can be made for using the discrete-time approach with bargaining rounds used as the units of analysis time, especially since decertification applications may be made only in the “open periods” of a contract. This method applies only to the application for decertification, however, whereas the actual decertification may take place at any point in time after such an application. Thus, the continuous-time approach may be preferable to the discrete-time approach, especially since the manners in which relationships may fail, especially business closures, fit a continuous process better than a discrete one. Note that the estimates produced in the continuous-time approach may be biased if numerous units of analysis experience the event at the same time. Within the data used here, the number of these “ties” never approach the rule of thumb—greater than 5% of units at risk of failing at the same time—that is likely to lead to severe bias (: 44).

14

Mathematically, the dependent variable is $\lim_{Δ t \to 0} \frac{\Pr (t + Δ t > T > t | T > t)}{Δ t} .$

15

This hypothesis is evocative of concept of imprinting from organizational theory in which the founding conditions of an organization are theorized to potentially have a lasting impact on its survivability. In this study, the bargaining relationship is the analogue to the organization and it tests whether what occurs in the initial founding stages of a relationship has lifelong consequences.

16

These estimates are available from the author upon request.

17

According to Allison (2010), when a covariate violates the PH assumption, the estimate represents an “average” effect for that covariate throughout analysis time. Thus, the estimate on the first contract arbitration covariate can be thought of as the average association of this covariate with relationship duration for private-sector non-construction bargaining relationships in Ontario from . The PH assumption for the main variables of interest was also checked using two other common tests: 1) based on the variation in the scaled Schoenfeld residuals and 2) log-log survival plots. These tests largely corroborated the findings of the check discussed in the text. The ULP variable is found not to violate the assumption whereas the FCA variable does. These results are also available from the author upon request.

18

These figures are available from the author upon request.

19

This finding also dovetails well with the application of the concept of imprinting mentioned in footnote 15. also wrote about a “liability of newness” in which younger organizations were more fragile than older ones because they face a number of problems that older organizations do not, such as a dependence on the cooperation of strangers and a lack of legitimacy. It appears that bargaining relationships may also face something similar to a liability of newness.

20

This study finds that the political party in power, and not the legislation it enacts, seems to influence the duration and dissolution of bargaining relationships during the representation phase.

References

Ahlburg

Dennis A.

Dworkin

James B.

1984. The influence of macroeconomic variables on the probability of union decertification. Journal of Labor Research 5(1): 14–28.

Allison

Paul D.

2010. Survival analysis. In Hancock

Gregory R.

Mueller

Ralph O.

(Eds.), The Reviewer’s Guide to Quantitative Methods in the Social Sciences, pp. 413–24. New York: Routledge.

Anderson

John C.

Busman

Gloria

O’Reilly

Charles A.

III.

1979. What factors influence the outcome of decertification elections? Monthly Labor Review (November): 32–36.

Anderson

John C.

Busman

Gloria

O’Reilly

Charles A.

III.

1982. The decertification process: Evidence from California. Industrial Relations 21(2): 178–95.

Bain

George Sayers

. 1981. Certifications, first agreements, and decertifications: An analytical framework. Ottawa: Ministry of Labour, Government of Canada.

Bentham

Karen

. 2002. Employer resistance to union certification: A study of eight Canadian jurisdictions. Relations Industrielles/Industrial Relations 57(1): 159–87.

Box-Steffensmeier

Janet M.

Jones

Bradford S.

1997. Time is of the essence: Event history models in political science. American Journal of Political Science 41(4): 1414–61.

Bronfenbrenner

Kate

. 1994. Employer behavior in certification elections and first contract campaigns: Implications for labor law reform. In Friedman

Sheldon

Hurd

Richard

Oswald

Rudolph

(Eds.), Restoring the Promise of American Labor Law, pp. 75–89. Ithaca, NY: ILR Press.

Bronfenbrenner

Kate

. 1997. The role of union strategies in NLRB certification elections. Industrial and Labor Relations Review 50(2): 195–212.

10.

Chafetz

Israel

Fraser

Chris R. P.

1979. Union decertification: An exploratory analysis. Industrial Relations 18(1): 59–69.

11.

Cleves

Mario A.

Gould

William W.

Gutierrez

Roberto G.

2004. An Introduction to Survival Analysis Using Stata. College Station, TX: Stata Press.

12.

Colvin

Alexander J. S.

2006. Flexibility and fairness in liberal market economies: The comparative impact of the legal environment and high performance work systems. British Journal of Industrial Relations 44(1): 73–97.

13.

Colvin

Alexander J. S.

Darbishire

Owen

. 2013. Convergence in industrial relations institutions: The emerging Anglo-American model? ILR Review 66(5): 1047–77.

14.

Cooke

William N.

1983. Determinants of the outcomes of union certification elections. Industrial and Labor Relations Review 36(3): 402–14.

15.

Cooke

William N.

1985. The failure to negotiate first contracts: Determinants and policy implications. Industrial and Labor Relations Review 38(2): 163–78.

16.

Cox

David R.

1972. Regression models and life tables. Journal of the Royal Statistical Society, Series B, No. 34: 187–220.

17.

Dickens

William T.

Wholey

Douglas R.

Robinson

James C.

1987. Correlates of union support in NLRB elections. Industrial Relations 26(3): 240–52.

18.

Dunlop

John T.

1958. Industrial Relations Systems. Carbondale, IL: Southern Illinois University Press.

19.

Elliott

Ralph D.

Hawkins

Benjamin M.

1982. Do union organizing activities affect decertification? Journal of Labor Research 3(2): 154–61.

20.

Ferguson

John-Paul

. 2008. The eyes of the needles: A sequential model of union organizing drives, 1999–2004. Industrial and Labor Relations Review 62(1): 3–21.

21.

Heneman

Herbert G.

III Sandver

Marcus H.

1983. Predicting the outcome of union certification elections: A review of the literature. Industrial and Labor Relations Review 36(4): 537–59.

22.

Johnson

Susan J. T.

2002. Card check or mandatory representation vote? How the type of union recognition procedure affects union certification success. Economic Journal 112: 344–61.

23.

Johnson

Susan J. T.

2010. First contract arbitration: Effects on bargaining and work stoppages. Industrial and Labor Relations Review 63(4): 585–605.

24.

Krislov

Joseph

. 1956. Union decertification. Industrial and Labor Relations Review 9(4): 589–94.

25.

Kuhn

Peter

Riddell

Chris

. 2010. The long-term effects of unemployment insurance: Evidence from New Brunswick and Maine, 1940–1991. Industrial and Labor Relations Review 63(2): 183–204.

26.

Lester

Richard A.

1958. As Unions Mature: An Analysis of the Evolution of American Unionism. Princeton, NJ: Princeton University Press.

27.

Martinello

Felice

. 2000. Mr. Harris, Mr. Rae and union activity in Ontario. Canadian Public Policy/Analyse de Politiques 26(1): 17–33.

28.

McDonald

Alistair Peter

. 1987. First contract arbitration in Canada: An analysis of the legislation in five labour jurisdictions in Canada. School of Industrial Relations Research Essay Series, No. 17. Kingston, Ontario: Industrial Relations Centre Press, Queen’s University.

29.

Meyer

David

Bain

Trevor

. 1994. Union decertification election outcomes: Bargaining unit characteristics and union resources. Journal of Labor Research 15(2): 117–36.

30.

Ontario Labour Relations Board. January 1985–December 2012. Ontario Labour Relations Board Reports. Toronto: Government of Ontario.

31.

Ontario Ministry of Labour. 2012. Collective Bargaining Information Services Database. Toronto: Government of Ontario.

32.

Prentice

Ross L.

Farewell

Vern T.

1986. Relative risk and odds ratio regression. Annual Review of Public Health 7: 35–58.

33.

Riddell

Chris

. 2001. Union suppression and certification success. Canadian Journal of Economics/Revue canadienne d’économique 34(2): 396–410.

34.

Riddell

Chris

. 2013. Labor law and reaching a first collective agreement: Evidence from a quasi-experimental set of reforms in Ontario. Industrial Relations 52(3): 702–36.

35.

Ross

Arthur

. 1948. Trade Union Wage Policy. Berkeley: University of California Press.

36.

Schmitt

John

Mitukiewicz

Alexandra

. 2011. Politics matter: Changes in unionization rates in rich countries, 1960–2010. Washington, DC: Center for Economic and Policy Research.

37.

Sexton

Jean

. 1987. First contract arbitration in Canada. Labor Law Journal 38(8): 508–14.

38.

Stinchcombe

Arthur L.

1965. Social structure and organizations. In March

James G.

(Ed.), Handbook of Organizations, pp. 142–93. New York: Routledge.

39.

Warner

Kris

. 2012. Protecting fundamental labor rights: Lessons from Canada for the United States. Washington, DC: Center for Economic and Policy Research.

40.

Weiler

Paul

. 1980. Reconcilable Differences: New Directions in Canadian Labour Law. Toronto: Carswell Company.

41.

Weiler

Paul

. 1983. Promises to keep: Securing workers’ rights to self-organization under the NLRA. Harvard Law Review 96(8): 1769–827.

42.

Weiler

Paul

. 1984. Striking a new balance: Freedom of contract and the prospects for union representation. Harvard Law Review 98(2): 351–420.

43.

Weinberg

Bradley

. 2015. A quantitative assessment of the effect of first contract arbitration on bargaining relationships. Industrial Relations 54(3): 449–77.

44.

Wooldridge

Jeffrey M.

2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press.