Union competition in collective bargaining and effects on the wage increase settled in agreements. The Spanish case.

Manuel Pérez TrujilloUniversidad Autónoma de Madrid

mp.trujillo@uam.es

Santos M. RuesgaUniversidad Autónoma de Madrid

ruesga@uam.es

Julimar da Silva BicharaUniversidad Autónoma de Madrid

julimar.dasilva@uam.es

Abstract

The transition to Spanish democracy simultaneously resulted in the democratization of labour relations, consolidating Union elections as the main mechanism of labour representation, determining, based on the number of candidates elected, ability to represent and, therefore, the Union’s institutional power. This situation introduces incentives for the competition among Trade Unions, in order to maximize the number of elected representatives. In turn, conditioning their behaviour in negotiation, weakening the labour Union’s cohesion as well as the power of labour representation against business organization. All in all resulting in worse outcomes in the agreement and, therefore, lower wage increases.

1. Introduction

In Spain, after the fall of dictatorship and the transition process to democracy, democratization of labour relations begins to take place, the role of social partners to propose agreements that could govern labour conditions in the workplace are recognized, breaking with the previous model where high state intervention prevailed (Card & de la Rica.2006). During this process, representation of workers in the productive sphere is determined by a Trade Union model based on the results achieved in Union elections held at the workplace. 1

1 Originally elections were every four years, however, since 1994 there is a new electoral system, allowing Union elections to be held at any time, provided there is prior agreement between interested parties (Jódar and Jordana, 1999). Workers' Statute determines that the enterprise council should be formed in work places that have 50 or more workers. The number of committee members increases in proportion to the number of employees of the enterprise. The representation also extends, in proportion to the number of employees, in establishments employing 11 to 49 workers by the staff representative model. Aditionally and optionally work places with 6 to 10 employees can elect a staff representative.

1

Every worker —18 and older and with at least 1 month in the enterprise—, participate in Union elections with no restrictions on Union representation, extending this to all employees regardless of Union membership. The agreements reached in collective bargaining would affect all workers of the area within which the agreement is settled. Non-discrimination between members and non-members in implementing the settled conditions in collective agreements discourages workers to be part of the Union, therefore impacting membership. Spain thus stands out as one of the countries with the lowest Union affiliation rates among European economies. In spite of that, low membership does not affect the bargaining power of the Union, which is legally guaranteed by the election results, which determine the institution’s representation (Maluquer & Llonch, 2005).

Depending on the number of representatives elected, measured amongst all enterprises or workplaces of a sector or concrete territorial level, the degree of representativity that the Union has is established2 allowing it to negotiate in higher levels of bargaining3 (Kholer & Calleja, 2012).

Representativeness allows Unions to extend their intermediation role not only at the workplace, but also to the economic, political and social areas (Jodar, Martin & de Alós-Moner, 2004), giving the institution capacity for dialogue with other social agents (Jodar & Jordana,1999), organizational strength and ability to negotiating through conflict situations (Baylos, 2006), as well as increased financial resources (Hamann & Martinez, 2003). Since representation is guaranteed on the basis of electoral results, the Trade Union has incentives to compete with the other organizations in order to increase the number of representatives elected through polls.

At the same time, the negotiating committee that sets terms for agreements may include representatives from various Trade Unions. By making an impact on the behaviour of the Unions during the negotiation process on the subsequent outcome that will occur at the polls (Strom, 1990), incentives for competition in a negotiating committee where Union representation is heterogeneous will be higher, leading each Union to defend its own electoral program during the development of the negotiation. In that respect, Union competition tends to impair the power of the Union as an instance for labour representation, limiting their ability to achieve better results in agreements with the employers organization (Akkerman, 2008) and (Craypo, 1986).

Although the influence of Union behaviour on macroeconomic performance has been widely analyzed, as demonstrated by extensive existing literature, there is a limited number of studies about Trade Union competition and, in particular, on the results achieved in agreements. Thus, there is a need to study this issue in greater depth. This analysis focuses on contrasting the negative effect that Trade Union competition can exert on the bargaining power of the Union and the results achieved in agreements, essentially bargaining on wage increases. The peculiarity of

2 Union representativeness criteria is set out in the Estatuto de los Trabajadores of 1980 and is regulated by Articles 6 and 7 of the Ley Orgánica de Libertad Sindical de 1985. For a Union to be considered as representative it shall obtain in Union elections at least 10 percent of total staff representatives of the members of the enterprise councils and relevant government bodies. Regionally of 15 percent is the minimum level.3 Agreements signed at higher negotiation levels affect automatically, through the principle of eficacia general -erga omnes ex ante clause- all workers, both union members and non-members, and companies included in the scope of bargaining (Card & the Rica, 2006). This leads to an inclusive model of industrial relations (Jodar Vidal & Alos-Moner, 2011). Said principle rules a centralized character to collective bargaining and strengthens the institutional role of the Union, especially those majority Unions, giving it the ability to act as a negotiator on the basis of industry or territory –for example at regional level- (Baylos, 1991).

2

the Spanish model, where the incentives for inter-Union competition are high in seeking to increase the organization’s power (Kholer, 1995), turns the Spanish collective bargaining model as ideal to contrast this hypothesis.

The study is divided into three sections. The first will discuss objectives of Trade Unions from a theoretical approach and of how from this perspective, considering the particular characteristics of the Spanish Trade Union model, Unions tend to compete. The following section develops the empirical analysis, detailing the database used, the indicator chosen to measure the degree of Union competition at the negotiating committee and the econometric model used depending on the characteristics of the variables that compose it. Lastly, the conclusions drawn in from the analysis are presented.

The results attained provide evidence of the negative effect that Union competition generates on the negotiation outcomes, weakening Trade Union’s cohesiveness, especially in decentralized areas of negotiation, where the Union interacts directly with constituents and the pressure to achieve better election results is greater.

2. Trade Union elections and Competition between Unions

The knowledge about Union behaviour and its objectives is important for understanding different economic processes affecting enterprises contracting decisions, wages, labour disputes, inflation and the social and political movement (Gahan, 2002). In industrial relations there exists a relationship of asymmetric power between workers and employers (Kaufman, 2010) and (Traxler, 2003), being the role of the Trade Union that of representing workers, through collective organization, to counteract inequality of power with the entrepreneur and to influence on income distribution of production gains.

During the 20th century analysis of Union behaviour and its objectives have been at the core of economic and social sciences, with two prevailing different approaches. On the one hand, the economic dimension, where Unions are presented as monopoly agents whose main interest lies in wage increases. Under this approach the Union exerts greater pressure on the labour supply by setting wages above the level of equilibrium.

The main defender of this theory is John Dunlop, who, in 1944, took the first step in theorizing Union behaviour. He argued that, to analyze the Union from an economic perspective, it should behave as a maximising agent, and its main objective would be to maximize total wage earned by the workers it represents. Under this approach, Union acts as a monopoly on the labour supply, setting wage levels unilaterally —restricted by the demand for employment from the enterprise—, maximizing a utility function in addition to individual preferences of their clients (Cahuc and Zylberberg, 2004) and (Kaufman, 2002).

On the other hand, a second vision on the analysis of the Union stands out: the institutionalist approach, where the Union is analyzed as a mechanism for representation of the collective voice of workers, defending their interests and proposals in the workplace (Freeman & Medoff, 1979). The theoretical reasoning tends to interpret the Union only as an economic agent and not as a social actor, resulting in an incomplete analysis. Under the institutionalist approach, the political nature of the Union is essential in the effort to improve understanding of its organization and objectives (Rees, 1989).

3

The institutionalist approach comes to prominence with the work of Ross (1948) Trade Union Wage Policy, written as a critique of the theoretical approach of Trade Union behaviour as presented by Dunlop in 1944, resulting in the famous debate between them. Ross, identifies that the Union as an organization does not possess identical interests to those of their constituents, as Dunlop states, thus marking a clear distinction between the two approaches. The Trade Union under Ross approach is not an "enterprise" whose sole purpose is to "sell jobs", Trade Union organization is a political institution that operates in an economic setting and represents "labour vendors". On this premise, Ross reaches the conclusion that the negotiation is aimed not only at improving wages and conditions of employment, which he defines as "worker-oriented provisions", but it also aims improve the institutional status of the Union, taking into consideration the "Trade Union-oriented provisions” –also named “institutional provisions”- (Ross, 1948). In this regard, the Union is much more than the sum of preferences of their clients –members- and as an institution its interests are not solely market-oriented, European Trade Unions being a clear example of it —and in particular the Spanish case —, where Union’s performance acquires a political dimension that transcends the social sphere (Turnbull, 2003).

However, this fact does not rule out the concept that the Union has monopoly power and that it uses it to increase wages, as the theoretical approach posits, but it involves an enlargement of the Union’s dimension, beyond the economic perspective (Freeman & Medoff 1979). With the addition of the political approach, the analysis of the Trade Union acquires more complexity, and one must take into account aspects such as the relationship between leaders and their constituents, the way they are organized and make decisions, thus preferences conditioning all, behaviour and objectives in the negotiation process (Kaufman, 2002).

The context surrounding the Union thus becomes a key element at the strategic level, being conditioned by the regulatory framework, entrepreneurial strategies and the economic, political, social and institutional conditions prevailing in each country, the setting and the way of adaptating to those conditions are key in the success of the Union as an institution (Tzannatos & Aidt, 2006) and (Hall et al.; 2011). In a context of labour relations as the Spanish one, where representation is the key aspect to define the power of the Trade Union as an organization, the institutional dimension plays a fundamental role in the study of their behaviour and objectives.

In Europe, the representation of workers can rest, according to different criteria of representativeness, on two or more Unions, giving rise to multi-Unionism (Akkerman, 2008). Focusing the approach on the Spanish case, elections are the means through which the organization achieves the ability to dialogue with other social agents (Jodar & Jordana, 1999), affecting on its organizational strength and negotiating conflict capacity in the economic, political and social system (Baylos, 2006).

In elections voters elect their representatives from the different candidates to join as staff delegates or members of the enterprise council. As well as the individual decision to join or not to join a Trade Union is conditioned by aspects related to the work setting, the personal situation in employment, the perception of the individual of the Union institution and its representatives, the expected success of the Union to achieve its objectives —on behalf of workers—, the proximity of the organisation with the individual or ideology (Jodar Vidal & de Alós- Moner, 2011), (Martínez & Fiorito, 2009), (Juravich & Shergold, 1988) y (Summers, Betton & Decotis, 1986), the decision to vote or not for a Trade Union representative will also be affected by these reasons. The worker then assesses costs and benefits he gets when choosing between the different candidates.

4

Elected representatives are the collective voice of workers in labour relations, defending their interests. If the worker wishes to maximize their well-being, behaving rationally —like conventional economic theory posits—, interests will focus on improving their income levels, working conditions and stability in employment (Summers, Betton & Decotis,1986). Also, Unions, in order to achieve better results in the elections, must offer the workers greater benefits in the negotiation process than those presented by the rival organisations (Ross & Hartman, 1960) and (Akkerman 2008), therefore conditioning negotiation objectives of the Union.

However, the main objective of the Union is not to maximize the number of votes itself. As Strom (1990: 573) says: "Votes have no intrinsic value to party leaders. They are simply a means, and in democratic societies, an important one, toward office or policy benefit, (i.e. representativeness)". Returning to the concept of Ross on "provisions aimed at the Trade Union", if the election results give the ability to participate at higher levels of negotiation, conferring power of dialogue with the State and employers (Kholer & Calleja, 2012), as well as increased financial resources (Hamann & Martinez, 2003) 4, the Union will be motivated to compete with the other organizations to increase the number of votes. In this regard, the way the Union acts during the bargaining process will bring future electoral gains. Thus each Union has an incentive to defend its electoral program before the rest5. This can lead to an increase in competition and therefore weaken Trade Union’s cohesion in bargaining process against employers (Strom, 1990).

Since elections are developed at the enterprise level, the core of the competition focuses on this level (Baylos, 1991). The results achieved in the agreements are conditioned by two factors:

1. Proximity between the negotiators –union representatives- and their voters –employees-, which means the latter exert greater pressure on the first to promote their demands and interests (Canal & Rodríguez, 2004)

2. The degree of rivalry between Trade Union organizations involved in the agreement.

At the enterprise level, the Union interacts directly with voters, and must respond to the demands made by workers as to improve their chances of winning an election. However, elections can generate incentives that emphasize the rivalry between Unions, thus they compete to get the most votes possible. Consequently this leads to a weakening of Trade Union cohesion in the negotiation process that can affect the outcomes in agreements.

However, what has been described does not imply that rivalry does not occur in higher levels of enterprise negotiations; but, it can be mitigated by:

1. The extension mechanisms and the free rider effect, where the Union represents a large number of workers who are mostly not affiliated, affect negatively on the

4 In the Spanish case, membership accounts for about 70 percent of the financial resources of the major Union confederations (Hamann & Martinez, 2003), but these resources are not able to finance the Union's activity. This makes public funding key for maintaining an organizational structure (Richards, 2008).5 The parties are encouraged to maintain a stable political position over time, as voters seek to minimize uncertainty about future policies that they can develop (Strom, 1990), which implies a certain rigidity in modifying their electoral program.

5

pressure that Union has to achieve better negotiation outcomes (Canal & Rodriguez, 2004).

2. Also, extending the agreement to a large number of companies and workers and the perspective of its impact on the economy may influence in moderating Trade Union demands, in order to internalize externalities associated with the development of the negotiation that may affect macroeconomic performance —mainly in terms of inflation and unemployment— (Flanagan, 1999)

Both factors lessen the effect of Trade Union rivalry on the results achieved under the agreement.

Focusing the argument on Trade Union competition, election results and, in particular, the composition of the negotiating committee are key factors. The negotiating committee in Spain can include Trade Union representatives from various organizations, depending on election results. If all the representatives belong to a single organization, competition is unlikely to occur 6, preventing fragmentation of labour representation in the bargaining process against the enterprise organization, and vice versa (Mishell, 1986). However, when two or more compiting Trade Unions are paticipating in the bargaining process7, their bargaining power decreases (Craypo, 1986), involving at least “two (or more) potential attack sources”; a) rival Unions and b) employers organization (Summers, Betton & Decotis,1986). This fact impairs the power of the Trade Union as the unit of labour representation and thus weakens their ability to achieve the desired outcomes in bargaining (Akkerman, 2008).

The effect of Union competition in the development of collective bargaining has been tested on different aspects of the labour market, however, the number of studies on the subject is lower in comparison with the existing literature about the effects exerted by Trade Unions on macroeconomic performance. Despite this, and in accordance with this analysis, the empirical evidence reached by Mishell (1986)8, for the United States, about the effect of Union concentration -stands out as a proxy of the degree of competition- in the salary —defined as the hourly wage—. Mishell gets positive and significant evidence about the effect exerted by Union concentration on the compensation between different industries —in terms of union membership; when this is concentrated in a small number of Unions— concluding that the negotiated wage is affected positively when Union concentration is high.

6 Even so, the power of the Trade Union may be negatively affected by potential conflicts at the intra-Union level between leaders and their opponents within the organization itself (Rees, 1989). This in turn may affect the results achieved in the negotiation.7 There is great difficulty in being able to measure the degree of "competitive Unionism", because competition depends not only on the number of Trade Unions, but also on the relationships between them. Thus, a negotiating committee where there is greater heterogeneity in Union representation of workers does not directly imply a greater or lesser degree of Union competition. In this personal aspects such as the closeness of the individuals involved in the negotiation or existence of common strategies are ignored, which would result in a smaller effect of competition on the outcomes settled in the agreement. However, there is a greater likelihood of rivalry and competition incentives to decrease when a Union is dominant in the negotiation process and vice versa when the composition of the negotiating committee has a higher degree of Union heterogeneity (Mishell, 1986).8 The analysis carried out by Mishell uses business data for USA, adding each enterprise according to their sector or industry area, from the survey of Expenditure for Employee Compensation for 1968, 1970 and 1972, taking the years grouped together in the analysis.

6

3. Empirical Analysis

This study focuses on the Spanish case, using information, for its empirical development from the Base de Datos de Convenios Colectivos (BDCC) drawn up by the Ministerio de Empleo y Seguridad Social (MEySSS). This Data Base contains micro-data from all the agreements registered during each period in the Ministry, providing relevant information of the settlements reached at negotiations and the characteristics of agreements9 (Ruesga et al; 2007).

The period of analysis includes data from 2001 until 200710, the target analysis unit are agreements registered during this time frame. The adhesion to other agreements, arbitration procedures or extension of agreements are excluded. The sample provides an unbalanced panel of 38.998 agreements. Among the sample of agreements there is a total 9.083 individual agreements with an average of 4.3 observations per agreement across the seven years records.

The use of collective agreements as a unit of analysis allows to capture more accurately the effect exerted by Trade Union competition on the results of the bargaining, reflecting this in settlements of agreements —with special emphasis on the salary increase—, obtaining personalised information of each bargaining process.

The sample of agreements will be divided into three levels, all agreements, firm agreements and settlements at higher levels –upper than firm level-, in order to capture more accurately the effect of the concentration degree of Unions on the nominal wage increase settled. The division between enterprise and higher corporate level agreements arises under the logic that the expected effects of the Union concentration on the results and, in particular, on the negotiated wage, may not be similar at both levels. In this respect, Trade Union elections and the greater proximity between negotiators and their voters at the firm level are a decisive factor that affect the incentives that each Union has to compete with other organizations, and this fact may affect the Trade Union cohesion and negotiation outcomes. On the contrary, at higher bargaining levels, the effect of settlements on a great number of companies and workers and the perspective of its impact on the economy can influence moderation of Union demands. This would lead to a potential greater coordination among actors who negotiated the agreement and, thus, lower incentives to the rivalry at this level for organizations.

On the other hand, assuming that collective agreements at firm level are the 80 per cent of the overall, it is necessary to apply a differentiated analysis in order to obtain greater homogeneity of information that improves the analysis.

In order to adequately contrast the hypothesis, trying to prove the negative effect exerted by a lower Union concentration at the negotiation committee —as proxy for the degree of competition between Trade Unions— on the settled wage increase, it is essential to obtain a measure of "Union competition" that approximates to this concept. Numerous studies use Trade Union membership rate (Scheuer, 2011) (Bryson, Ebbinghaus & Visser, 2011), (Bowdler & Nunziata, 2007) and (Koeniger, Leonardi & Nunziata, 2007), both at macro level —all workers affiliated nationwide—, as well as micro level —an specific firm or sector—, or the coverage rate

9 The BDCC provides a total of 516 variables related to the salary, working hours and holidays agreed, as well as clauses on employment and contracting, bargaining scope, among others (Ruesga et al., 2007).10 This period is used due to increased availability of data in the database and homogenization in building them, 2007 being the latest available at the time of analysis.

7

of the Agreement11 (Posen & Popov, 2006) to measure the power and degree of influence of the Trade Union in the labour market.

However, both measures do not fully account for the Union’s influence capacity in the Spanish model, as well as the possibility of identifying the extent of existing Union competition. Membership in Spain is irrelevant —this because of the erga omnes clause and the disincentive that it brings about for affiliation—, while the coverage rate is not a "high-quality measure" to accurately calculate influence and power of Trade Unions12 (Bover, Bentolilla & Arellano, 2001). According to the Spanish representation model, results of Trade Union elections may be useful to measure the influence degree of the Trade Unions (European Commission; 2011), determining the degree of strength or weakness of the Union cohesion at the negotiating committee and their effects on competition and the settlements. Consequently, Union elections are a key element in identifying the strength of the Trade Union in Spain. Thus an indicator that takes into account election results must be used so as to quantitatively proxy Trade Union’s power.

The BDCC provides information about the composition of the negotiating committee and the number of representatives belonging to each Union. This information enables for the calculation of an indicator that measures the degree of association of Unions at the negotiating committee. This analysis uses the Herfindahl-Hirschman Index (IdHH) as a measure of the degree of concentration and degree of Trade Union cohesion at the negotiating committee. This

is calculated as the ∑i=1

N

pij2 being pij the proportion of the total number of representatives

belonging to a Trade Union "i" of the total amount of Union representatives at the negotiating committee of an Agreement "j". Its value is restricted between a minimum close to 0 and a maximum equal to 1. Value 1 expresses that all Trade Union representatives in the negotiating committee belong to the same Union, considering that as this value approaches 0, Union heterogeneity is greater. According to the hypothesis, when there exists a larger number of Unions represented at the negotiating committee it is likely for Trade Union competition to be higher —under the assumption of competition— that otherwise, where a Union dominates the Union representation at the committee. The IdHH thus serves as a proxy of Union competition degree.

3.1. Econometric Model.

Having different units —collective agreements— over different periods of time, panel data is the proposed econometric analysis. This technique allows to obtain consistent estimates in presence of omitted variables. Analysis of panel data specifies a single effect α iwhich includes

11 The coverage rate is the appropriate indicator in countries where there are extension mechanisms that expand the effect of the agreements to all workers and companies in a given area, regardless of affiliation. However, their use is less widespread compared to the Unionization rate for its lower variability, with the coverage rate more stable over time and introducing greater difficulty when analyzed in models which take into account the temporal variability (Koeniger, Leonardi & Nunziata, 2007).12 For measuring the Spanish coverage rate one has to resort to estimates, since no official statistical agency is responsible for its evaluation. There are also additional problems, because in the calculation of the number of workers covered by collective agreements there may be overlaps in different level agreements, in some cases counting more than once workers affected by two or more agreements simultaneously, pluriemployees. This leads to a risk of overestimating the number of workers covered which would in turn affect the calculation of coverage ratio (Ruesga et al., 2007).

8

non observed effect levels of each individual not identified in our analysis that affects the model. The models commonly used in the analysis of panel data are the models of fixed effects (FE) and random effects (RE). The differentiating factor between both FE and RE estimators, is the assumption that is set on the individual effect α i and the degree of relationship between it and the rest of regressors (Wooldridge, 2002), (Greeen, 2008) and (Baltagi, 2008).

The model of this analysis is expressed as:

W ¿=β1+αi+β2∗IdHH ¿+ β3∗δ j+β4∗φl+β4∗μt+βk−5∗f (control variables)¿+u¿ (1)

Being "i" the unit of analysis, "t" time period, "j" region where the agreement applies and "l" the industry of the enterprise or sector in which it is settled.

In equation (1) the nominal wage increase in basis points settled in the Agreement (Wit) is function of the degree of Union concentration (IdHHit), a distinct individual effect for each agreement (αi), two dummy variables that identify fixed effects that, on the one hand affect equally all Agreements belonging to the same region, different with respect to other regions 13 (δ j) and, on the other, those that affect the industry to which the enterprise belongs or the sector in which the agreement is settled14 (φ l), different from the rest. It also includes a temporary dummy variable (μt). This variable allows to capture all the unobservable effects that affect all Agreements equally each year. Finally, it includes a set of control variables that improve the explanatory nature of the model. These variables are associated to factors that influence on the Union’s bargaining power and the results achieved in the Agreement15, including:

1) Variables associated with the setting in which negotiation takes place: Strikes or protests took place during the bargaining process –Yes = 1/No = 0– (Conf ¿), Agreement is settled in a high-technology and/or high knowledge setting. The Agreements

are grouped by the degree of technological intensity that is associated to the industry to which they belong according to the classification performed by EUROSTAT16 –high technology = 1/no high technology = 0– (HighT i),

The agreement is negotiated for the first time –Yes = 1/No = 0– (FirstAg it), Temporary scope of the Agreement –Anual = 1/Plurianual = 0– (TimeExt¿), Scope of Agreement negotiation, –firm level = 0/upper than firm level = 1– (BargCont i ¿, For sector agreements, territorial scope of the agreement – local = 1, interprovincial = 0,

provincial = 2, regional = 3, inter-regional = 4 and national = 5– (TerrCont i) For enterprise Agreements, the company in which the Agreement is settled is public or

private –Private = 1/Public = 0– (Private¿)

13 Regions are classified between 0 to 19, with 1 to 17 values for each Autonomous Community plus Ceuta (18) and Melilla (19). Also the value 0 corresponds to those agreements that are negotiated above provincial levels (see Annex I).14 The agreements are divided into 20 branches depending on the activity in which the enterprise or sector in which the agreement is signed (see Annex II).15 In Annex III presents the descriptive statistics of the variables.16 EUROSTAT Classification: “High-technology and knowledge based services aggregations based on NACE Rev.2”: http://epp.eurostat.ec.europa.eu/cache/ITY_SDDS/Annexes/htec_esms_an3.pdf

9

2) Variables that identify clauses related to the salary included in agreements: Wage Guarantee clauses –cláusulas de salvaguarda- are settled17,–Yes = 1/No = 0– (ClasGar¿),

Allowances linked to productivity are established –Yes = 1/No = 0– (ProdIncent¿)

3) Variables that identify the inclusion of Agreement clauses related to employment or working day:

Clauses related to employment and hiring are established in the Agreement, for example, net employment creation, employment continuity, employment creation to replace early retirees, etc –Yes = 1/No = 0– (Hire¿),

Implementation in the Agreement of working day reduction clauses –Yes = 1/No = 0- (JournRed¿),

The use of variables such asHighT i, BargCont i, TerrCont i or Private¿ make reference to the external economic setting and the pressure that it exerts on the results achieved in the agreement. On the one hand, the higher the degree intensity of technology or use of knowledge, the more negotiators tend to limit their objectives. This because enterprise or companies that adhere to the agreement are subject to an increasing competitive pressure that forces them to establish greater control over labour costs, driving to a certain wage moderation (Garret & Way, 1999) and (Traxler & Bradl, 2010). Also, the inclusion of Spain in the Economic and Monetary Union implies a greater control of public expenditure at the national level, resulting in the moderation of salary increases in the Administration. This is order to meet budget control policies agreed upon in the Pact of Stability and Growth (Glassner, 2010).

Furthermore, the necessary differentiation of agreements by levels where negotiation took place allows to differentiate the effects that the setting has on the negotiation. The effect of Union rivalry on results achieved in settlements may be less when the agreement is settled at higher corporate levels, where there is a higher chance for better coordination.

In turn, to include in the model variables such asConf ¿or FirstAg¿ make reference to specific internal setting aspects. The existence of strikes or protests during the development of the agreement exerts pressure on the enterprise to give in to labour demands (Rees, 1989). This can possitively affect the settled wage increase. On the contrary, when the agreement is negotiated for the first time, the degree of success in the settlement could be affected by potential inexperience at negotiating by the parties that participate in the agreement.

Another aspect taken into account in the equation is the duration of the Agreement (TimeExt¿). To negotiate an agreement for a greater time period —higher than a year— negotiators may tend to set higher wage increases in order to avoid possible future income losses due to price level increases.

Lastly, the model includes variables related to the criteria set to establish salary increase and structure (ClasGar¿ and ProdIncent¿), as well as criteria of hiring and working day (Hire¿

and JournRed¿) complementing the analysis.

17 Wage guarantee clauses –cláusulas de salvaguarda- act when the price index provided in the agreement agreed to update the wage index is lower than actual prices, offsetting the loss of purchasing power on the part of employees to incorporate the difference on salary.

10

Inclusion of no time-variant variables in the model like, for example, HighT i,BargCont io TerrCont i, must be considered when developing the econometric model. They introduce a potential inconsistency risk in the coefficients of the model FE conditioning use of econometric techniques. The specific conditions of the FE model, that establish as an assumption that the individual effect is constant in time and correlated with the independent variables 18, implies that the existence of any constant variable included in the model, can mimic the behaviour of such effect (Green, 2008), by introducing bias in the estimators calculated (Wooldridge, 2002), so the explanatory capacity of the FE model becomes limited.

An alternative is to apply the transformation WITHIN, subtracting from each individual observation the group’s mean –each agreement being a different group from others–, expressing variables as:

X̌ ¿=X ¿−X i

Thus, transformation eliminates variables constant over time and, hence, the individual effects (α i) from the analysis, correcting inconsistency problems. However, the transformation shows two significant defects. On the one hand, all constant variables are suppressed. Their coefficient is not calculated, and, on the other, the WITHIN estimator is not completely efficient, ignoring the variation between agreements in the analysis (Hausman & Taylor, 1981).

A second alternative is proposed by Hausman and Taylor (1981), using an estimate obtained using the technique of instrumental variables, in order to solve the inconsistency problem and bias in model estimates, including as a needed restriction the existence of regressors, both time-variant as well as time-invariant, uncorrelated with individual effect (α i).

This new estimator applies random-effects Generalized Least Square (GLS) transformation on the variables of the model –appliying GLS estimation-, being:

X̂ ¿=X ¿−θ̂ iX i

Where:

θ̂i=1−( σ̂u2

σ̂ u2+T i σ̂α

2 )12

This transformation does not eliminate time-invariant regressors, by calculating their coefficients. However, it does neither eliminate individual effects and, therefore, existing correlation between them and the regressors. This results in a problem of endogeneity. To eliminate endogeneity the instrumental variables technique is applied, using exogenous regressor instruments, both time-variants and time-invariants, included in the equation and various transformations thereof, for example, those expressed in deviations from the group mean observations of the same unit – applying this transformation to those time-variant variables- or simply the group´s mean. Thus, a number of tools that enable a consistent calculation of the

18 This will relax the orthogonality condition between the explanatory variables and the individual effect that must be fulfilled in RE estimation.

11

coefficients are obtained, so it is applied in the estimate of equation (1) the method proposed by Hausman and Taylor (HT).

Additionally, the pooled OLS and the RE models will be included as alternatives to FE model. In the first, pooled OLS, the model is performed without specifying the existence of individual heterogeneity among analysis subjects, considering each analysis unit independent of the others, being able to calculate the estimators consistently. To test the assumption of no individual heterogeneity in the model, corresponding to the pooled OLS specification, the Lagrange Multiplier test for random effects posed by Breusch & Pagan (1980) is conducted, at doing the null hypothesis is that variance of individual effects is zero: σ αi=0:, thus accepting lack of heterogeneity, and therefore confirming the assumption made to take as valid the pooled OLS model.

In the second alternative proposed, RE, determines that the individual effects (α i) are uncorrelated with the regressors used in the equation. This assumption is necessary to obtain a consistent estimator calculated. To test this assumption we apply the Hausman test (1978), comparing regressors coefficients of HT and RE models, with the purpose of obtaining evidence about the possible existence of correlation between the regressors and the individual effect. This results in the null hypothesis raised in the contrast:

Hº: E [α i/X ¿= 0]Accepting Hº indicates the lack of correlation between the individual effect and

regressors, being correct the HT estimator for calculating the coefficients in equation (1), yielding an optimal linear unbiased estimator (Baltagi, Egger & Kesina, 2012) (Hausman & Taylor, 1981) and (FitzRoy & Kraft, 2005).

3.2. Analysis of results.

Table 1.1 presents the results of pooled OLS, RE and HT estimates applied as well as the Lagrange Multiplier test and the Hausman test, dividing the analysis into three levels: all agreements, firm agreements and agreements made at upper levels than firm level.

The specificity of the HT model implies the existence of endogenous variables correlated with the individual effect to obtain a consistent estimate of the coefficients to be calculated. The individual effects that apply to the equation are those unobserved characteristics inherent to the agreement that impact the settlement and that, in our case, have been considered as constant over time.These characteristics are those associated with the setting in which the agreement is developed, determining the bargaining power and behaviour of agents and thus the settlement. In this respect, the variables included in the model define the negotiation’s setting: HighT i, BargConti, TerrCont i, Private¿, Conf ¿and FirstAg¿ are considered to be endogenous and thus, potentially correlated with individual heterogeneity. Thus instrumental variables to obtain consistent estimates of the coefficients have to be used due to the potential endogeneity caused by non-varying over time variables (HighT i, BargCont i y TerrCont i).

The Hausman test applied shows the consistency of HT estimator only when analysis is applied to higher corporate level agreements, accepting the null hypothesis that indicates the absence of correlation between the regressors and individual heterogeneity, there being no

12

evidence to accept this hypothesis in the analysis performed for all agreements and firm agreements, providing greater consistency in the estimated coefficients calculated with RE. Also, the pooled OLS estimation is rejected in all models used, not rejecting the existence of variability in the individual effects (σ αi≠0).

The results of the analysis indicate a disparate effect of IdHH on the wage increase. This is differentiated according to the level of bargaining. At analyzing the agreemtns as a whole, it appears that the effect of Union’s concentration on the negotiated wage increase is positive and significant, taking that for every increase of one tenth of IdHH the salary increases about 0.12 percentage points (RE-GLS). This positive relationship is maintained when analysising enterprise agreements only. The estimated coefficient for IdHH at this level indicates that, as Union’s concentration at the negotiating committee increases by a tenth, wage increase experiences 0.078 percentage points increase (RE-GLS).

Moreover, and contrary to the results obtained at the firm level and all agreements, analyzing the IdHH effect on wage increase at higher bargaining levels than firm level we observe a negative and not significant relationship (-0.081), indicating the possible absence of effect between the two variables.

The results attained concur with the hypothesis. Expecting a positive relationship between the degree of Union’s concentration-lesser rivalry at the negotiating committee-and wage increases settled in agreements at the firm level, which could be related to the greater proximity between negotiators and their constituents, implying more pressure on the negotiation outcomes (Canal & Rodriguez, 2004), added to that the effect of Union elections and greater incentives to rivalry. Meanwhile on the other hand, at higher corporate levels of negotiation, Union competition and Union objectives are moderated in order to internalize the externalities of the agreement (Flanagan, 1999), reducing the impact of competition on the settled wage increase, there being no significant evidence of association between concentration effect at the negotiation committee on the results of the agreement at this level.

Analyzing in detail the control variables, results indicate that the clauses relating to recruitment and hiring (Hire¿) have no significant impact on any of the models analyzed, on the settled wage increase. Regarding the existence of clauses relating to the reduction in working hours (JournRed¿), there is a positive and significant effect on all levels of negotiation on the salary increase, and this is greater on enterprise agreements compared to higher level agreements.

In reference to the establishment in the agreement of clauses related to the existence of allowances linked to productivity (ProdIncent¿) or fixing wage guarantee clauses -cláusulas de salvaguarda- (ClasGar¿), we observe that, for the first, the effect of wage allowances is significant only in the analysis of all agreements and agreements at firm level. The estimate shows a negative effect on the settled wage increase equivalent approximately to -0.05 percentage points (RE-GLS). Although this effect is not high, the significance of the coefficient indicates that productivity incentives replace the fixed wage increase, introducing greater flexibility that allows the organization to improve matching labour costs to the economic situation (Ruesga et al, 2012). Moreover, the estimated coefficient for fixing wage guarantee clauses - cláusulas de salvaguarda- which offset the loss of purchasing power in sight of negative price index forecast in the

13

agreement, shows a negative and significant relation in all models of about -0.7 percentage points moderating the settled nominal wage increase.

Regarding the context in which the agreement is developed, results show that at the firm level the setting has a greater effect on the negotiated wage increase compared to higher corporate negotiation levels. Temporary scope of the agreement (TimeExt¿) has a positive and significant effect on all levels —with a higher effect at the firm level— with respect to the nominal wage increase, indicating that as the agreement’s time scope is larger, the wage demands tend to be higher and may avoid possible future losses of purchasing power on the part of employees by a changing price index.

With respect to conflict (Conf ¿), results show that it acts as a mechanism of pressure from workers, affecting positively and significantly on the negotiated wage increase at all negotiation levels. Moreover, when the agreement is negotiated for the first time (FirstAg¿) the estimated effect on nominal wage increase is slim, and this is not significant. This gets significant only for high corporate level agreements.

Similarly, when all agreements are analyzed, the variable that identifies bargaining context (BargCont¿) indicates that wage increase is higher and significant —0.35 percentage points (RE-GLS)— at higher corporate level agreements compared to the enterprise level, there being more moderation at the firm level.

Moreover, in the enterprise agreements, we observe that estimated wage increase for private firms (Private¿) is higher —showing a significant effect equivalent to 0.52 percentage points (RE-GLS)— compared to the public level, with a greater wage moderation in the latter. This could be motivated by the expenditure level control and adjustment to the criteria set forth in The Stability and Growth Pact.19

Finally, high levels of technology and knowledge associated with the industry or business sector in which the agreement is settled (HighT i), have a negative effect on the wage increase agreed —on all bargaining levels—. It is possible to explain this fact by the competitive pressure to which companies operating in a context driven by innovation, tend to be subject to, which promotes wage moderation (Garret & Way, 1999) and (Traxler & Bradl, 2010). Salaries tend to be determined by allowances linked to productivity or business profits (Ruesga et al, 2012).

In short, in relation to the hypothesis of this analysis, the results indicate that Union competition has a negative effect on nominal wage increases on settled results, said significant effect is negative at firm level, consistent with the hypothesis proposed in this piece. However, it is necessary to study in more detail the features of this result. It is important to analyze the effect of the competition according to the number of workers affected by the collective agreement. This will be done through the same model posed for enterprise agreements, disaggregating them by the number of workers affected by the agreement. In this regard we present four different models, grouping agreements according to the dissagregation mechanism raised by Izquierdo, Moral & Urtasun, (2003) 20:

19 Similar results can be seen in the estimate by Izquierdo, Moral & Urtasun, (2003), where lower wage increase in public enterprise agreements are explained by budget constraint measures in the years 1993, 1994, motivated by the economic crisis, and 1997, a result of compliance with the convergence criteria to join the European Monetary Union.

14

Agreements that affect fewer than 30 workers, Agreements affecting between 30 and 100 workers, Agreements affecting between 101 and 500 workers, Agreements that affect more than 500 workers.

3.3. Analysis gathering enterprise agreements by number of employees affected by the agreement.

The results for this second analysis are presented in Table 1.2 —APPENDIX IV— with the frequency table corresponding to model variables disaggregated by each level of grouping according to the number of workers affected. In all models the estimate shows that HT estimation is appropriate, accepting the null hypothesis raised in the Hausman test (Hº: E [α i/X ¿= 0]), allowing the calculation of consistent estimators

A breakdown of enterprise agreements by the four groups described, shows that the effect of Union wage concentration has a positive and significant effect only on agreements that affect more than 500 workers, this effect being greater -0,5 percentage points- compared to the previous estimate raised for collective agreements at firm level (Table 1.1). This estimated effect has a comprehensive explanation in the economic literature. In the negotiation processes where the number of workers affected by agreement is high, it can be noted that:

trade Unions have a greater ability to exert pressure (Nunziata, 2001), and the increased capacity for collective action and therefore a warning on

entrepreneurs with the existing possibility of a strike (Wallesteirn and Western, 2000),

increasing the bargaining power of the Trade Union and therefore deriving all on more favorable salary increases settled in Agreements (Ruesga et al, 2007). Also, at this level, Trade Union rivalry is larger derived from the largest number of representatives to be elected in Union elections21, where incentives exist between Unions to compete with each other to increase their degree of representativeness. Consequently, the variable IdHH will have a significant effect in the model, interacting positively with nominal wage increase settled on the collective agreement.

As for the results of this second analysis for the control variables, we observe that the setting has a greater effect on those companies where the number of workers affected is at an intermediate level -between 30 and 500 employees-. For those cases, the setting is significant in determining the wage increase. In turn, also at this level -between 30 and 500 workers-, it should

20 We decided to disaggregate the sample into four levels. Also it was decided not to include the variable corresponding to the number of workers affected by agreement in the analysis as an explanatory variable. This was done because the variable would be highly correlated with the degree of Union concentration (IdHH), introducing an inconsistency risk in the calculation of estimates. This occurs because the number of elected representatives in the elections depends on the number of workers in the workplace where the agreement is implemented (Article 62 and 66 of the Law of the Workers' Statute of 1985). In this respect, there is a greater likelihood that those workplaces with a high number of workers affected by the agreement, have a lower degree of Union concentration, since the number of candidates to choose from at elections is higher when compared to those centers where is chosen only one representative, where representation is logically unique and IdHH presents a value of 1. There may be a high degree of correlation between the two variables that could lead to the existence of multicollinearity in the model.21 From 501 employees in the workplace or the number of representatives or enterprise council members to choose is between 17 and 75 candidates, to be chosen, depending on the size of the workplace (Article 66 of the Statute Law Workers ).

15

be noted that productivity incentives tend to have a significant negative effect on the settled nominal wage increase (ProdIncent¿), as well as in the estimate for all enterprise agreements. This variable is not significant in the other levels.

Lastly, we again observe a negative and significant effect of using wage guarantee clauses –cláusualas de salvaguarda- in the Agreement (ClasGar¿) similar to the estimate presented for all firm agreements, without discriminating on the number of workers affected by agreement, being also similar the effect of reduced working hour clauses ( JournRed¿) on the settled wage increase.

4. Final Remarks

Trade Union elections and representation criteria in Spanish collective bargaining are fundamental to define the institutional power of the Union, enabling it to act as spokesperson of workers with the State and enterprise organizations in different areas at the economic, political and social levels.

As noted by Downs (1957: 34-35): "[Party] members are motivated by their personal desire for the income, prestige, and power… Since none of the appurtenances of office can be obtained without being elected, the main goal of every party is the winning of elections. Thus, all its actions are aimed at maximizing votes". This fact conditions the behaviour and objectives of the Trade Union in negotiations. It is affected by the interests of their constituents, but also for its own interests as an institution (Ross, 1948), in order to increase its institutional representation.

In elections workers show their level of confidence about the different candidates running through their vote. If the worker seeks to maximize his welfare, his interests will focus on improving income levels, working conditions and job security (Summers, Betton & Decotis, 1986). Also in the interests of workers, Trade Unions, to achieve better results in the elections, have to offer greater benefits in the negotiation process than those promised by rival organizations (Ross & Hartman, 1960) and (Akkerman , 2008), in order to get more votes, thus fostering competition. Under this premise, when Union representation at the negotiating committee is composed of representatives from various organizations, greater heterogeneity, the level of competition is higher, and it may negatively affect the bargaining power of the Union as a collective against the employers. This is the hypothesis at the core of this analysis, which aims to obtain evidence for the Spanish case, on the effect exerted by Union competition in the outcome of negotiations and, in particular, on the nominal wage increase.

The results obtained in the empirical analysis show a distinct effect on the degree of Union concentration (IdHH) on the salary increase as the field of negotiation. At enterprise level, and especially in those where the number of workers represented in the agreement is high, more than 500 workers, there is a positive and significant effect of the degree of concentration on the wage increase bargained. Thus, at the negotiating committee when there is less diversity of association, Trade Union cohesion is greater and therefore the effect on wage increase improves. Given that, otherwise, Union competition may affect the representation unity of workers at the negotiating committee, weakening their power against the business organization, thus getting worse results in the Agreement.

16

The results also highlight that the greater the number of workers represented, the greater the Union’s ability to put pressure (Nunziata, 2001) and, therefore, the greater the threat of collective action (Wallesteirn and Western, 2000) increasing the bargaining power of the Trade Union with the employer in the development of the agreement (Ruesga et al, 2007).

Moreover, the estimated relationship between the degree of Union concentration and wage increases settled in bargaining levels upper than the enterprise levels tends to be less noticeable and not significant, indicating the absence of an effect of the Union concentration on increasing wage in agreements. This result can be explained by the existing incentives to moderate wage demands in areas of centralized bargaining in order to internalize the externalities of the agreement which may affect macroeconomic performance (Flanagan, 1999), reducing the impact of competition on the settled wage increase. Likewise, the existence of greater incentives for cooperation, directing Trade Union action towards serving business interests in the political arena (Ross & Hartman, 1960), implies that “Unions are able to “exchange" wage restraint for tax or welfare benefits that advance social justice” (Turnbull, 2003), resulting in an exchange of objectives with political favours that moderates their wage claims. Also, the extension mechanisms and the free rider effect imply that negotiation at higher levels where the Union represents a large number of workers who are mostly unaffiliated having that this pressure to achieve better results in negotiation is smaller (Canal & Rodriguez, 2004), so that it would mitigate the effect of competition.

In conclusion, these results confirm our hypothesis, providing evidence of the negative effect that Union concentration, as an estimate of the degree of competition, generates on negotiation, especially in decentralized levels with a high number of employees affected by the agreement. In this setting, the Union interacts directly with voters and puts pressure to achieve better electoral results, having a higher colletive bargaining capacity at enterprise level, where elections take place.

17

5. ANEX

ANEX I. Number of collective agreements by Spanish region.

Number of collective agreements

Region Firm Upper level Region Firm Upper

level

Madrid 2.292 378 Aragón 1.091 357

Extremadura 395 194 Asturias 812 219

Andalucía 4.110 1.216 Navarra 758 220

Castilla-La Mancha 1.063 521 La Rioja 323 120

Galicia 1.916 588 País Vasco 2.486 521

Murcia 405 299 Ceuta 140 66

Islas Canarias 1.479 236 Melilla 71 18

Comunidad Valenciana 1.782 780

Cantabria 714 199

Cataluña 4.020 674

Islas Baleares 440 142

Castilla y León 1.911 1.139

18

ANEX II. Number of collective agreements by economic sector.

Number of collective agreements

Region Firm Upper level Region Firm Upper

level

Agriculture, forestry and fishing 1.300 385 Real State 77 4

Mining and quarrying 372 63 Professional, scientific and technical activities 445 108

Manufacturing 8.327 2.726 Administrative and support service activities 2.518 670

Supply of electricity, gas, steam and air conditioning 275 7 Public Adm. and Military Defense 3.881 1

Water supply, sewerage, waste management and decontamination

2.195 129 Education 406 120

Construction 408 385 Health and Social services 1.384 407

Wholesale and retail trade, repair of motor vehicles and motorcycles

1.783 1.996 Arts, entertainment and recreation activities 1.406 190

Transport and storage 2.530 1.019 Other activities 512 266

Hotels and Restaurants 576 360 Activities of households as employers of domestic staff or as producers of goods and services for own use

10 35

Information service and communication 1.297 135

Financial and insurance activities 241 49

19

20

ANEX III. Descriptive statistics of the variables used in the analysis.

All levels Firm level Centralized level

Variables Average St. Dev. Min. Max. Average St. Dev. Min. Max. Average St. Dev. Min. Max.

W ¿ 309,8151 124,3701 0 2773 302,2458 126,1052 0 2773 334,8449 116,6300 0 2455

IdHH ¿ 0,7014 0,2680 0,01 1 0,7485 0,2719 0,01 1 0,5454 0,1820 0,01 1

Hire¿ 0,4484 0,4973 0 1 0,4533 0,4978 0 1 0,4319 0,4953 0 1

JournRed¿ 0,3386 0,4732 0 1 0,3307 0,4704 0 1 0,3648 0,4814 0 1

ClasGar¿ 0,4977 0,5000 0 1 0,4574 0,4981 0 1 0,6313 0,4824 0 1

ProdIncent¿ 0,2704 0,4441 0 1 0,3151 0,4645 0 1 0,1225 0,3279 0 1

Conf ¿ 0,0688 0,2532 0 1 0,0685 0,2526 0 1 0,0700 0,2551 0 1

FirstAg¿ 0,1694 0,3751 0 1 0,2076 0,4056 0 1 0,0430 0,2030 0 1

TimeExt¿ 0,0951 0,2934 0 1 0,0914 0,2882 0 1 0,1075 0,3098 0 1

HighT i 0,2739 0,4460 0 1 0,3204 0,4666 0 1 0,1202 0,3252 0 1

BargCont i 0,2321 0,4222 0 1 - - - - - - - -

Private¿ - - - - 0,7842 0,4113 0 1 - - - -

ANEX IV. Descriptive statistics of the variables used in the model disaggregating to agreements based on number of workers affected by collective agreement.

Number of workers <30 Number of workers ≥30 & ≤100 Number of workers >100 & ≤500 Number of workers >500

21

Average St. Dev. Min. Max. Average St. Dev. Min. Max. Average St. Dev. Min. Max. Average St. Dev. Min. Max.

W ¿ 311,923 145,368 0 2773 306,219 124,392 0 2210 296,717 116,539 0 2150 277,263 98.472 0 1322

IdHH ¿ 0,9637 0,1372 0,01 1 0,7852 0,2513 0,01 1 0,6052 0,2481 0,01 1 0,4942 0,2123 0,01 1

Hire¿ 0,3158 0,4646 0 1 0,4143 0,4926 0 1 0,5394 0,4984 0 1 0,7062 0,4555 0 1

JournRed¿ 0,2860 0,4519 0 1 0,3304 0,4703 0 1 0,3667 0,4819 0 1 0,3284 0,4697 0 1

ClasGar¿ 0,3878 0,4872 0 1 0,4197 0,4935 0 1 0,5166 0,4997 0 1 0,6020 0,4895 0 1

ProdIncent ¿ 0,2056 0,4041 0 1 0,2964 0,4567 0 1 0,3885 0,4874 0 1 0,4428 0,4968 0 1

Conf ¿ 0,0297 0,1697 0 1 0,0467 0,2111 0 1 0,0983 0,2977 0 1 0,1648 0,3711 0 1

FirstAg¿ 0,3387 0,4733 0 1 0,2137 0,4099 0 1 0,1230 0,3284 0 1 0,1127 0,3163 0 1

Private¿ 0,0217 0,1457 0 1 0,0334 0,1799 0 1 0,0453 0,2081 0 1 0,0917 0,2886 0 1

TimeExt¿ 0,1110 0,3142 0 1 0,1007 0,3010 0 1 0,0788 0,2695 0 1 0,0404 0,1969 0 1

HighT i 0,3316 0,4708 0 1 0,3311 0,4706 0 1 0,2856 0,4517 0 1 0,3733 0,4837 0 1

22

Table 1.1 Econometric analysis for all collective agreements, agreements at firm level and agreements settled at upper levels than firm level.All agreements Firm level Level upper than firm level

Variables Pool. data^ RE-GLS^ HT-GLS Pool. data^ RE-GLS^ HT-GLS Pool. data^ RE-GLS^ HT-GLS

IdHH ¿ 14,21 (2,48) *** 10,83(3,54)

*** 12,05(4,78)

** 10,22(2,66)

*** 7,79(3,67)

*** 7,83(4,21)

* -7,98(7,77)

-2,27(11,62)

-8,16(9,18)

Bargaining EnviromentConf ¿ 16,82

(2,86)*** 15,18

(3,69)*** 11,36

(3,28)*** 14,07

(3,14)*** 13,71

(4,03)*** 11,02

(3,95)*** 23,35

(6,57)*** 17,70

(8,31)** 12,79

(5,90)**

HighT i -10,96(3,22)

*** -9,30(4,04)

** 177,98(30,05)

*** -13,24(3,33)

*** -11,18(4,13)

*** 149,16(27,09)

*** -11,31(11,60)

-10,50(14,73)

-138,20(70,01)

*

FirstAg¿ -3,67(1,85)

*** -4,09(2,33)

* -7,94(2,67)

*** -5,05(1,87)

*** -4,75(2,35)

** -7,87(2,85)

*** 4,08(8,25)

-2,86(8,45)

-1,49(7,92)

TimeExt¿ 16,92(1,89)

*** 13,21(2,13)

*** 9,26(2,58)

*** 19,42(2,25)

*** 15,15(2,53)

*** 10,78(3,17)

*** 5,60(3,47)

6.89(3,85)

* 5,98(4,42)

BargConti 32,56(1,61)

*** 35,04(2,55)

*** 53,42(10,75)

*** - - - - - -

TerrCont i - - - - - - Yes Yes Yes

Private¿ - - - 52,56(2,78)

*** 52,38(3,93)

*** 31,62(9,89)

*** - - -

Wage and JourneyClasGar¿ -71,43

(1,25)*** -70,41

(1,73)*** -70,63

(1,58)*** -72.09

(1,42)*** -70,72

(1,98)*** -70,08

(1,81)*** -75,99

(2,66)*** -72,55

(3,49)*** -73,38

(2,86)***

ProdIncent¿ -8,66(1,35)

*** -5,96(1,79)

*** -4,86(2,16)

** 19,42(2,25)

*** -5,04(1,91)

*** -6,55(2,15)

*** 2,24(3,69)

1,67(4,73)

1,21(4,68)

Hire¿ 0,47(1,21)

0,61(1,65)

0,58(1,60)

1,53(1,39)

1,55(1,84)

0,58(1,85)

1,15(2,47)

1,09(3,60)

1,45(2,98)

JournRed¿ 9,89(1,27)

*** 7,82(1,53)

*** 7,56(1,43)

*** 11,44(1,50)

*** 9,31(1,80)

*** 9,18(1,67)

*** 7,36(2,30)

*** 4,62(2,89)

4,95(2,64)

Constant 331,82(5,06)

*** 332,19(7,91)

*** 337,97(13,05)

*** 284,19(7,72)

*** 287,30(11,55)

*** 317,07(21,40)

*** 340,79(19,14)

*** 356,83(29,74)

*** 505,56(71,89)

***

Number of Observations 38.998 38.998 38.998 29.943 29.943 29.943 9.055 9.055 9.055Number Goups - 9.083 9.083 - 7.515 7.515 - 1.568 1.568LM test / Hausman test - 0,0000 a 0,0198 b - 0,0000 a 0,0330 b - 0,0000 a 1,0000 b

23

Tabla1. 2. Econometric analysis for enterprise agreements by the number of workers affected by collective agreement.< than 30 wrokers ≥ than 30 workers and ≤ than 100 workers > than 100 workers y ≤ than 500 workers

Variables Pool. data^ RE-GLS^ HT-GLS Pool. data^ RE-GLS^ HT-GLS Pool. data^ RE-GLS^ HT-GLS

IdHH ¿ 7,30(13,23)

10,05(15,98)

2,27(17,17)

8,26(4,53)

* 6,23(6,02)

4,60(6,65)

0,96(4,85)

-1,44(6,63)

-2,45(7,10)

Bargaining EnviromentConf ¿ 6,86

(9,17)11,61

(12,39)18,92

(17,57)28,71(8,21)

*** 28,32(10,31)

*** 24,10(7,72)

*** 13,44(4,13)

*** 8,09(5,41)

-4,12(5,63)

HighT i -19,13(8,39)

** -18,70(9,46)

** 165,58(67,34)

** -12,13(4,40)

*** -7,62(5,71)

147,45(36,64)

*** -6,55(5,29)

-4,06(6,68)

131,53(65,15)

**

FirstAg¿ -4,48(3,59)

-3,51(4,01)

-5,77(6,49)

-7,82(3,00)

*** -5,43(4,01)

-2,79(4,96)

-1,07(3,99)

-2,57(5,29)

-2,71(6,19)

TimeExt¿ 10,98(4,74)

** 7,37(5,19)

-2,61(8,64)

24,04(3,54)

*** 19,00(4,43)

*** 15,35(5,02)

*** 21,29(3,90)

*** 16,38(4,03)

*** 11,68(5,14)

**

Private¿ 51,80(6,31)

*** 55,62(7,98)

*** 23,70(29,20)

49,84(5,35)

*** 47,94(8,10)

*** 29,23(14,50)

** 63,09(4,76)

*** 61,74(6,68)

*** 53,05(18,36)

***

Wage and JourneyClasGar ¿ -72,80

(3,07)*** -73,54

(3,92)*** -71,87

(4,38)*** -71,00

(2,38)*** -72,69

(3,40)*** -71,91

(2,94)*** -72,47

(2,57)*** -67,95

(3,50)*** -68,26

(2,96)***

ProdIncent¿ 1,98(4,07)

4,06(4,85)

2,86(5,39)

-11,53(2,38)

*** -8,76(3,21)

*** -9,30(3,51)

*** -8,06(2,30)

*** -7,83(3,10)

** -9,54(3,43)

***

Hire¿ -0,37(3,72)

-0,87(4,39)

-0,22(4,55)

2,25(2,28)

6,93(3,13)

** 6,74(3,01)

** 3,39(2,16)

0,62(2,78)

0,09(2,90)

JournRed¿ 8,66(3,74)

** 9,78(4,30)

** 9,85(4,23)

** 13,04(2,40)

*** 6,73(2,92)

** 6,34(2,76)

** 12,15(2,48)

*** 10,28(2,92)

*** 11,02(2,66)

***

Constant 296,14(21,29)

*** 285,53(27,77)

*** 333,47(57,43)

*** 272,44(11,38)

*** 282,87(16,53)

*** 309,42(33,20)

*** 279,82(16,92)

*** 294,82(22,61)

*** 313,78(41,66)

***

Number of Observations 7.372 7.372 7.372 10.690 10.690 10.690 9.406 9.406 9.406Number Goups - 2.316 2.316 - 3.028 3.028 - 2.320 2.320LM test / Hausman test - 0,0000 a 0,9999 b - 0,0000 a 0,9038 b - 0,0000 a 0,9968 b

24

> than 500 workersVariables Pool. data^ RE-GLS^ HT-GLS

IdHH ¿ 45,38(11,56)

*** 50,47(15,76)

*** 50,07(18,82)

***

Bargaining SettingConf ¿ 3,98

(5,56)4,03

(5,95)9,81

(7,47)HighT i -19,85

(10,95)* -24,78

(13,60)* 50,68

(103,49)FirstAg¿ 1,08

(6,90)0,86

(9,06)4,25

(10,26)TimeExt¿ 16,06

(9,80)21,12

(10,73)** 25,77

(10,51)Private¿ 31,25

(7,37)*** 28,79

(11,27)** 8,59

(21,99)Wage and JourneyClasGar¿ -69,65

(4,66)*** -67,39

(6,62)*** -64,15

(5,65)***

ProdIncent¿ 1,08(6,90)

-4,64(5,56)

-7,00(5,83)

Hire¿ -2,31(4,34)

-3,29(6,27)

-5,23(5,57)

JournRed¿ 11,61(4,38)

*** 9,46(5,16)

* 8,14(4,61)

*

Constant 295,76(25,85)

*** 293,25(36,02)

*** 324,16(77,08)

***

Number of Observations 2.475 2.475 2.475Number Goups - 523 523LM test / Hausman test - 0,0000 a 1,0000 b

Fort he two estimates (Table 1.1 y Table 1.2) it is shown: “^”: applying a robust covariance matrix. All models include time, regional and sectoral dummies variables in the analysis. *: significance level to 90%, **; significance level to 95% y ***: significance level to 99%. “a”) It is referred to the Lagrange Multiplier test, “b”) It is referred to the Hausman teste. Values inside () referred to Standard Desviation.

6. References

AKKERMAN, A. (2008), “Union Competition and Strikes: the Need for Analysis at Sector Level”, Industrial and Labor Relations Review, Vol. 61, Number 4, p. 445 – 459.

BALTAGI, B. (2008), Econometric Analysis of Panel Data, 4th Edition, John Wiley and Sons¸ United Kingdom.

BALTAGI, B.; EGGER, P.H. & KESINA, M. (2012), “Firm-level Productivity Spillovers in China's Chemical Industry: A Spatial Hausman-Taylor Approach”, in: CEPII - GEP - Ifo Conference: China and the World Economy, Munich 6 - 7th September 2012, last view on November 27th, 2012: http://www.cesifo-group.de/portal/page/portal/CFP_CONF/CFP_CONF_2012/Conf-gep12-Felbermayr/Paper/gep12_Egger.pdf

BAYLOS GRAU, A. (1991), “La intervención normativa del Estado en materia de relaciones laborales”, en Miguélez Lobo, F. & Prieto, C. (1991): Las relaciones laborales en España, Ed. Siglo Veintiuno de España Editores, S.A, p. 289 – 306, Madrid (Spain).

________________ (2006), “Representación en la empresa y audiencia electoral. Notas sobre el mecanismo de representación español”, en Romero Rodenas, M., J. (2006): Aspectos conflictivos de las Elecciones Sindicales, Editorial Bomarzo, p. 11 - 24. Madrid (Spain).

25

BOVER, O.; BENTOLILLA, S. & ARELLANO, M. (2001), The Distribution of Earnings in Spain during 1980s: the Effects of Skill, Unemployment, and Union Power , Center for Economic Policy Research, Discussion Paper Number 2770, London (United Kingdom).

BOWDLER, C. & NUNZIATA, L. (2007), “Trade Union Density and Inflation Performance: Evidence from OECD Panel Data”, Economica, Volume 74, p. 135 – 159.

BREUSCH, T.S. & PAGAN, A.R. (1980), “The lagrange multiplier test and its application to model specification in econometrics”, Review of Economic Studies, Volume 47, p. 239 – 253.

BRYSON, A.; EBBINGHAUS, B. & VISSER, J. (2011), “Introduction: Causes, consequences and cures of union decline”, European Journal of Industrial Relations, Volume 17, Number 2, p. 97 – 105.

CAHÚC, P. & ZYBELBERG, A. (2004), Labor Economics, MIT Press, Boston (USA).CANAL, J.F. & RODRÍGUEZ, C. (2004), “Collective Bargaining and Within-firm Wage Dispersion

in Spain”, British Journal of Industrial Relations, Volume 42, Number 3, p. 481 - 506.CARD, D. & DE LA RICA, S. (2006), “Firm-Level Contracting and the Structure of Wages in

Spain”, Industrial and Labour Relations Review, Volume 59, Number 4, p. 573 – 592.CRAYPO, C. (1986), “The Sources of Union Bargaining Strength”, in Bureau of National Affairs

(1986): The Economics of Collective Bargaining: Case Studies in the Private Sector , Washington, D.C. (USA).

DOWNS, A. (1957), An Economic Theory of Democracy, Ed. Harper and Row, New York (USA).DUNLOP, J. (1944), Wage determination under trade unions, Macmillan, New York (USA).EUROPEAN COMMISSION (2011), Industrial Relations in Europe 2010, Publications Office of the

European Union, Luxemburg.FITZROY, F. & KRAFT, K. (2005), “Co-odetermination, Efficiency and Productivity”, British

Journal of Industrial Relations, Volume 43, p. 233 - 247.FLANAGAN, R.J. (1999), “Macroeconomic Performance and Collective Bargaining: An

International Perspective”, Journal of Economic Literature, Volume 37 (September), p. 1.150 – 1.175.

FREEMAN, R.B. & MEDOFF, J.L. (1979), The Two Faces of Unionism, National Bureau of Economics Research, Working Papers, Number 364, Cambridge (USA).

GAHAN, P.G. (2002), “(What) do unions maximise? Evidence from a survey data”, Cambridga Journal of Economics, Number 26, p. 279 – 286.

GARRET, G. & WAY, C. (1999), “Public Sector Unions, Corporatism, and Macroeconomic Performance”, Comparative Political Studies, Volume 32, Number 4, p. 411 – 434.

GLASSNER, V. (2010), The Public Sector in the Crisis, European Trade Union Institute (ETUI) Working Paper series 2010-07, Brussels (Belgium).

GREEN, W.H. (2008), Econometric Analysis, 6th Edition, Pearson Prentice Hall, New Jersey (United States).

HALL, A.T.; FIORITO, J.; HORN, M. & LANDFORD, C. R. (2003), “Unions as Organizations: Strategy vs. Enviroment”, The Journal of Labor and Society, Volume 14, p. 571 – 587.

HAMANN, K. & MARTÍNEZ, M. (2003), “Strategies of Union Revitalization in Spain: Negotiating Change and Fragmentation”, European Journal of Industrial Relations, Volume 9, Number 1, p. 61 - 78.

HAUSMAN, J.A. (1978), “Specification Test in Econometrics”, Econometrica, Volume 46, Number 6, p. 1.251 – 1.271.

HAUSMAN, J.A. & TAYLOR, W. E. (1981), “Panel Data and Unobservable Individual Effects”, Econometrica, Volume 49, Number 6, p. 1.377 – 1.398.

IZQUIERDO, M.; MORAL, E. & URTASUN, A. (2003), El sistema de negociación colectiva en España : un análisis con datos individuales de convenios, Documento Ocasional nº 0302 del Banco de España (Servicio de Estudios), Madrid (España).

JÓDAR, P., & JORDANA, J. (1999), “Current development in trade unions elections”, European Industrial Relations Observatory, last view on September 21st, 2012: http://www.eurofound.europa.eu/eiro/1999/02/feature/es9902298f.htm

26

JÓDAR, P.; MARTÍN, A. & DE ALÓS-MONER, R. (2004), “El sindicato hacia dentro. La relación entre la organización y los trabajadores desde el análisis de la afiliación”, Papers, Number 72, p. 112 – 144.

JÓDAR, P.; VIDAL, S. & DE ALÓS-MONER, R. (2011), “Union Activism in an Inclusive System of Industrial Relations: Evidence from the Spanish Case Study”, British Journal of Industrial Relations, Volume 49, p. 158 - 180.

JURAVICH, T. & SHERGOLD, P.R. (1988), “The Impact of Unions on the Voting Behavior of their Members”, Industrial and Labor Relations Review, Volume 41, Number 3, p. 374 – 385.

KAUFMAN, B.E. (2002), “Models of Union Wage Determination: What Have we Learned Since Dunlop and Ross?”, Industrial Relations Journal, Volume 41, Number 1, p. 110 – 158.

______________ (2010), “Institutional Economics and the Minimum Wage: Broadening the Theoretical and Policy Debate”, Industrial and Labor Relations Review, Volume 63, Number 3, p. 427 – 453.

KHÖLER, H-D. (1995), El movimiento sindical en España. Transición democrática. Regionalismo. Modernización económica, Ed. Fundamentos, Madrid (España).

KHÖLER, H-D. & CALLEJA, J.P. (2012), “Transformations in Spanish trade union membership”, Industrial Relations Journal, Volume 43, Number 3, p. 281 – 292.

KOENIGER, W.; LEONARDI, M. & NUNZIATA, L. (2007), “Labor Market Institutions and Wage Inequality”, Industrial and Labor Relations Review, Vol. 60, Number 3, p. 340 – 356.

MALUQUER, J. & LLONCH, M. (2005), “Trabajo y relaciones laborales”, in Carreras, A. & Tafunell, X.: Estadísticas históricas de España (2º Ed), Fundación BBVA, p. 1.156 – 1.214, Madrid (Spain)

MARTÍNEZ, A.D. & FIORITO, J. (2009), “General Feelings Toward Unions and Employers as Predictors of Union Voting Intent”, Journal of Labor Research, Volume 30, Issue 2, p. 120 – 134.

MISHELL, L. (1986), “The Structural Determinants of Union Bargaining Power”, Industrial and Labor Relations Review, Volume 40, Number 1, p. 90 – 104.

NUNZIATA, L. (2001), Institutions and wage determination: a multi-country approach , Working Economic Papers, Number 29, University of Oxford, Oxford (UK).

POSEN, A. & POPOV, D. (2006), Has EMU Had Any Impact on the Degree of Wage Restraint? , Working Paper Series, Institute for International Economics, Number 06, Massachussetts (USA).

REES, A. (1989), The Economics of Trade Unions (3rd Ed.), The University of Chicago Press, Chicago (USA).

RICHARDS, J.A. (2008), El sindicalismo en España, Documento de trabajo 135/2008 de la Fundación Alternativas, Madrid (Spain).

ROSS, A.M. (1948), Trade Union Wage Policy, University of California Press, Los Angeles (USA).

ROSS, A.M. & HARTMAN, P. (1960), Changing Patterns of Industrial Conflict, Ed. John Wiley and Sons, New York (USA).

RUESGA, S.M. ET AL. (2007), Análisis Económico de la Negociación Colectiva en España. Una Propuesta Metodológica, Comisión Consultiva Nacional de Convenios Colectivos, Ministerio de Empleo y Seguridad Social de España, Madrid (Spain).

_______________ (2012), Instrumentos de Flexibilidad Interna en la Negociación Colectiva, Comisión Consultiva Nacional de Convenios Colectivos, Ministerio de Empleo y Seguridad Social, Madrid (Spain).

SCHEUER, S. (2011), “Union membership variation in Europe: A ten-country comparative analysis”, European Journal of Industrial Relations, Volume 17, Number 1, p. 57 – 73.

STROM, K. (1990), “A Behavioral Theory of Competitive Political Parties”, American Journal of Political Science, Volume 34, Number 2, p. 565 – 598.

SUMMERS, T.P.; BETTON, J.H. & DECOTIS, T.A. (1986), “Voting For and Against Unions: A Decision Model”, Academy of Management Review, Volume 11, Number 5, p. 643 – 655.

27

TRAXLER, F. (2003), “Bargaining (De)centralization, Macroeconomic Performance and Control over the Employment Relationship”, British Journal of Industrial Relations, Volume 41, Issue 1, p. 1 – 27.

TRAXLER, F. & BRADL, B. (2010), “Collective Bargaining Inter-sectoral Heterogeneity and Competitiveness: A Cross-National Comparison of Macroeconomic Performance”, British Journal of Industrial Relations, Volume 50, Issue 1, p. 73 – 98.

TURNBULL, P. (2003), “What Do Unions Do Now?”, Journal of Labor Research, Volume 24, Number 3, p. 491 – 527.

TZANNATOS, Z. & AIDT, S. (2006), “Los sindicatos, los resultados microeconómicos y lo que cuenta para economistas y empresas”, Revista Internacional del Trabajo, Volume 125, Number 4, p. 287 – 312.

WALLERSTEIN, M. & WESTERN, B. (2000), “Unions in decline? What has changed and why”, Annual Review of Political Science, Number 3, p. 355 – 377.

WOOLDRIDGE, J. M. (2002), Econometric Analysis of Cross Section and Panel Data, The MIT Press, Cambridge (United States).

28