1
Featherbedding, dismissal protection and labor relations:
The effects of trade unions on workers’ effort on the job
Florian Sniekers
This paper investigates the effect of trade unions on manual workers’ effort on the job. It replicates the research of Green and McIntosh (1998) to look at efficiency wages, dismissal protection and cooperative labor relations, but adds, like Freeman and Medoff (1984), featherbedding to the analysis. A factor analysis and three ordered probits were performed on the 1990 Workplace Industrial Relations Survey of the UK, and led to results supporting Freeman and Medoff, but very different from Green and McIntosh. If the labor union is strong, featherbedding has a negative effect on effort, while if the union is weak, downward communication has a negative, and upward communication has a positive effect. If these variables are controlled for, the power of the trade union has no direct effect on effort. If a primary sector with more union membership and all accompanying variables is constructed by factor analysis, trade unions still have no net effect, refuting Bowles(2004). Efficiency wages and, therefore, union dismissal protection are not supported to have an effect on effort. These results suggest that featherbedding and communication improvement by a union cancel each other out, resulting in a zero net effect of trade unions on effort.
1. Introduction
In mainstream literature, trade unions are primarily associated with their impact on wages and employment. If unions bargain about wages only, wage increases will result in less employment within that firm or industry. For that reason, trade unions are said to be one of the causes of higher unemployment (Gärtner, 2006, p. 419).
However,this situation is not Pareto-efficient for the trade union and the employer. If union and employer negotiate about employment as well, combinations of wages and employment benefiting both parties are possible. For that reason, this is called efficient bargaining(McDonald & Solow, 1981). However, a negotiation like this does not lead to efficiency in an allocative sense, because more employment is created at the given wage than a profit-maximizing firm would choose itself. Labor is not put to use where it is most productive, and there might simply be not enough work for the extra people employed. By means of rules about how the work should be carried out, more employment must be created. This is called featherbeddingand can reduce thelabor productivity.
Freeman and Medoff (1984, p. 163-165) argue thatfeatherbeddingis one of the reasons that trade unions can reduce the effort on the job. However, according to them, unions not only have a negative impact on effort. Trade unions can also increase the engagement of the workers and improve the possibilities for communication between employer and employees. The effort on the jobwill be increased when the trade union contributes to cooperative labor relations. Consequently, the question is then whether the first or the second effect on effort dominates.Freeman and Medoff argue that both effects cancel each other out.
However, effort on the job might be influenced by the level of pay as well. If one assumes that effort is onlyimperfectly observable and not verifiable, efficiency wages come into consideration (see e.g. Shapiro & Stiglitz, 1984). Effort cannot be enforced, but the employer can dismiss a worker thought to be shirking. By offering wages higher than the best alternative, the worker gets a cost of losing his job. To avoid these costs,he or sheincreases his efforts at higher wages.
Bulkley and Myles (1996) show that if trade unions negotiate wages, and efficiency wages apply, the effort is higher than in the situation of complete competition without a trade union. Even if the union negotiates about the effort level for dismissals as well, effort can still increase. In such a bargain, easily contractible efforts will indeed decrease, but by increasing wages, imperfect observable efforts have to rise as part of the deal to compensate for the increased cost of job loss. In these cases, the employer still determines employment and is responsible for the monitoring of the workers.
Apparently, workers would like to increase the wage even if they have to provide more effort for it. However, they cannot increase their wage more, even if their higher effort would generate enough revenues (pay for itself), because for the employer effort is non-verifiable, and he or she cannot be sure that a higher wage indeed will be compensated by enough effort. Consequently, as Bowles points out (2004, p. 277, 278), this situation is Pareto-inefficient.[1] However, if in some way a higher effort and wage can be contracted, circumventing the enforceability problem, a Pareto-improvement is possible. As was the case for efficient bargaining, according to Bowles (2004, p. 283, 284) again a trade union can make this improvement. He argues that a union can bargain with the employer for a higher wage and a higher effort, and enforce the efforts of the workers by peer monitoring. In this way, a union can contribute to cooperative labor relations and reduce the number of dismissals for cause. For Bowles (2004, p. 286), this explains the existence of primary labor markets.
Sampson (1993) even assumes that trade unions negotiate wages, employment, and effort at the same time, but arrives at similarconclusions. If trade unions can be held responsible for effort levels, reducing the opportunities for free-riders, a situation of profit-sharing and high levels of efforts can be reached.In addition, Bandyopadhyay (1997) shows within an infinitegame theoretic model, that a cooperative outcome with higher wages, more employment and higher effort levels is possible.
However, Green and McIntosh (1998) argue that next to effects on wages, employment, and effortin the ways described above, trade unions have an impact on dismissal protection as well. In sectors with a strong representation of trade unions, it is harder to dismiss a worker. This allows free-riders to enter the game again. Similarly, one can imagine that a trade union without peer monitoring might want to reduce the employer’s monitoring, to alleviate the pressure of the cost of job loss. Bowles (2004, p. 294) too acknowledges thatthe secondary labor market is characterized by stronger labor discipline effects. Green and McIntoshconclude that trade unions have no net impact on the efforts of manual workers. On the one hand union members do not provide much effort because they have protection against dismissal, on the other hand cooperative labor relations (as with Bowles) and human resource management contribute to more efforts.
This research replicates the research of Green and McIntosh (1998) to examine the impact of efficiency wages, dismissal protection and cooperative labor relations in the presence of trade unions, thereby including the possibility of Pareto-improving cooperation. On top of that, this study addsfeatherbedding to the analysis, in imitation of Freeman and Medoff (1984). With all these factors together, this research can offer an answer to the question what the effectis of trade unions on effort on the job.
Paragraph 3 will discuss the data for this study, data that Green and McIntosh (1998) already used. Paragraph 4 pays attention to the methodology. The results will be discussed in paragraph 5, whereupon in paragraph 6 can be concluded what the effect is of trade unions on workers’ effort. First, in the next paragraph the theoretical model is discussed.
2. Theoretical model
Maybe the most common assumption in labor economicsis that workers prefer to provide less effort e, at least in the economically relevant ranges (Bowles, 2004, p. 271), and prefer to earn a higher wagew:
.
It is in the employer’s interest to induce as much effort as possible for lower spending on wages, monitoring or other techniques. However, since effort is hard to verify, in many cases it is impossible to incorporate this in a contract.
As said, an employer can increase effort by paying efficiency wages, in this way generatinga cost of job loss. Workers will provide more effort to prevent their dismissal, if their current job pays more than the weighted average of the chances of the best alternative or unemployment. Unemployment results in a benefit, a situation which occurs with the chance of the unemployment rate, and the best alternative is the wage of another employer, a situation occurring with the complementary chance. Then, the cost of job loss cjlresults in:
in whichwis the current wage, puis the unemployment rate, b the unemployment benefit and the wage of the best alternative.
In negotiating wages, a trade union will either as unintended part of the bargain, or as part of a negotiated Pareto-improvement, increase effort. The latter is the Bowles’ (2004, p. 284) causal chain of cooperative labor relations. The former is the pure effect of an increased cost of job loss.According toGreen and McIntosh (1998), at the same time, trade unions limit the possibilities for dismissal, and therefore reduce effort. Moreover, a trade union that is not only interested in a higher wage, but also in the level of employment, will try to introduce featherbedding practices. Because of this, the effort is reducedfurther. However, Freeman and Medoff (1984, p.163) argue that trade unions might improve the communication between employer and workers and increase involvement, contributing to cooperative labor relations. In figure 1, all these effects are presented as follows:
Figure 1 Causal chainof trade unions on effort
Next to the impact of trade unions on effort on the job, other factors have an effect on effort as well. For instance, an employer can improve the human resource management hrmto increase the involvement,and as a result of that, the effort of theworkers. Human resource management can contain several components, such as communication and consultation, but always consists of methods associated with the ‘carrot’ rather than the ‘stick’ (Green & McIntosh, 1998, p. 367). Omitting technical and demographical control variables for the moment, this results in the following model of effort e:
in whichF stands forfeatherbedding and clr for cooperative labor relations. The hypotheses of this study concern the signs of the firstthree coefficients, and the impact of trade unions on them, as presented in figure 1 above. The next paragraph discusses the datafor this study.
3. Data
This study can be conducted with micro data at the company level. The most important variable is the effort on the job. It is very hard to measure this variable objectively, and for that reason this research is limited to subjective data. However,even then datasets seldom contain questions concerning effort, as a result of which this study must be conducted with relatively old data, the Workplace Industrial Relations Survey of 1990. This survey is the largest and most detailed survey concerning labor relations in Great Britain, covering 2061 companies with more than 25 employees. The surveycontains interviews with managers, representatives of manual workers, and workers themselves. The range of this survey permits a study of all mentioned effects on effort.
Effort is measured on a five-point scale by a workers’ representative,comparing the intensity of their work with the intensity in other companies. The representative was in all but five cases a union official, who is assumed to be able to make such a comparison, since he or she has to negotiate about it. Although a subjective variable is concerned, Green and McIntosh (1998, p. 368, 369) showfor this dataset that the problems of social-desirable answers, an unclear reference point, and an unclear meaning of `intensity’are not large. Moreover, in a consistency-check the intensity correlates with the productivity. For these reasons, this effort variable seems appropriate for this study.
Because the representative is almost always a union official, no comparison can be made between companies without and with a trade union. Instead, in imitation of Green and McIntosh (1998, p. 372), this study makes a distinction between companies which are only accessible for trade union members (a closed shop) and companies with recommend membership on the one hand, and all other companies on the other hand. This reflects the strength of the union, and the possibilities for cooperation.
This dataset also contains data concerning the incomes and the hours worked, whereupon wages can be calculated as the weighted average of all manual workers. With the regional unemployment rate of 1991 (unfortunately 1990 cannot be found) of the Office for National Statistics in the United Kingdom, the costs of losing a job can be constructed. The regional distinction is based on the eleven government office regions.[2]
The survey of the managers investigates whether a number of aspects of human resource management are applied in the company concerned. Consequently,they are asked whetherthey undertook attempts to increase the involvement of the employees in the past three years, and whether a Joint Consultative Committee (JCC) of managers and workers exists. Two-way communication (meetings), downward communication (by themanagement chain or company newsletters) and upward communication (surveys and suggestion schemes) are examined as well. These variables result in five dummies.
In addition, managers are asked whether they negotiate about both wages and employment, likeefficient bargaining, and whether they feel themselves limited in control by given rules about the implementation of the work, which can be calledfeatherbedding. Besides, it is asked howthe labor relations can be described and whether people are dismissed for cause. Therefore, this leads to an additional set of four dummies. Finally, the proportion of supervisors with respect to manual workers is calculated. The remaining variables are control variables and are discussed in the appendix.
Table 1 presents the means and standard errors of the variables considered, both for the entire sample, and for those 232 observations which are eventually used since they have values for all variables. Generally, the used sample tends towards a stronger trade union influence, more human resource management, and more innovation in the management, but the differences remain relatively small. However, the changes in the existence of a JCC, and the union density are more doubtful. On top of that, on average only larger companies stay in the sample. Finally, the proportion supervisors changesconsiderably as well, but that is because earlier measurement errorsof companies with only supervisors now turn out to be excluded. Unfortunately, the number of observations is reduced considerably.
The means and the number of observations do not correspond with those of Green and McIntosh (1998), although the same dataset has been used. Moreover, this study tries to replicate the variables which are used in both studiesas precise as possible. For that reason, the differences that arise are hard to explain. Of course this study continues with the variables as constructed here, so that, interestingly, other results can be expected as well. The next paragraph will address how the variables can be applied.
Table 1 Averages and standard errors variables, for the entire and reduced sample
variable / entire sampleobserv. mean s.e. / 232 observ.
mean s.e.
effort
union req./recomm.
average wage
unemployment rate
cost of job loss
worker involvement
JCC exists
2-way communication
commmunication down
communication up
result pay
union density
new machinery
new work practices
new microtechnology
incr. flexibility
good lab. relations
proportion male
prop. unskilled
prop. full time
proportion ethnic
prop. supervisors
size
featherbedding
eff. bargaining
dismissals / 655 3.731 .873
1100 .486 .500
1594 4.188 1.225
2061 .0861 .015
1594 -1.085 .360
2054 .498 .500
2057 .454 .498
2057 .760 .427
2057 .817 .387
2057 .456 .498
2049 .823 .381
1668 .542 .421
1387 .549 .498
1387 .458 .498
1368 .397 .489
2055 .482 .500
2053 .890 .313
1693 .635 .355
1687 .469 .390
1802 .755 .328
1222 .349 .477
1654 .916 7.421
2061 3.462 1.723
2027 .199 .400
1104 .562 .496
2030 .631 .483 / 3.759 .859
.569 .496
4.421 1.104
.0852 .014
-1.025 .322
.595 .492
.75 .434
.797 .403
.944 .230
.586 .494
.647 .479
.846 .203
.690 .464
.672 .470
.578 .495
.694 .462
.879 .327
.684 .327
.432 .367
.827 .263
.237 .426
.189 .446
5.134 1.017
.388 .488
.668 .472
.772 .421
4. Methodology
Because the effort variable is ordinal, it can only be estimated by anordered probit, an estimation method with coefficients very hard to interpret. Even the sign of the coefficients cannot unambiguously be understood as the effects’ direction, but has a clear meaning at the lowest and highest value of the variable only. At least, the ordered probit allows for some interpretation.
In our theoretical framework, effort is a function of the variables described above. The first of these was the cost of job loss. When, as is often assumed, the alternative wageis equal to the currentwage, and the unemployment benefit is a fractionρof that last wage, the equation is . This can be reduced to what is equal to the sum of the logarithms of wages, the regional unemployment rate and a constant which for this reason can be ignored. This variable allows us to test the effect of efficiency wages.
Second, featherbedding and efficient bargainingare included in the probit to estimate their impact on effort. Third, few dismissals are an example of cooperative labor relations, as are an explicit positive statement of these relations. Fourth, the five examples of human resource managementare included: the three varieties of communication, the existence of a JCC, and attempts to increase workers’ involvement. Fifth, attention is paid to the proportion of supervisors as acomplementaryway to generate more effort. Sixth, the strength of the trade union is measuredby the union density and whether the trade union is required or recommended, or not. Finally a series of control variables is included, concerning organizational and technological changes, the composition of the work force, and the size, region and industry of the establishment.