Asymmetry and probit models: the case of labour unions

Abstract

Binary choice models fail to capture the strength of one's preferences. An attempt is made to measure the context and environment which influence choice. Union elections provide some insight into the process because the electorate may vote either for or against a particular union. The roles that general sentiments as well as specific preferences play in an election can be examined. Negative sentiment dominates the positive and apathy in affecting the election outcome. In the case of union elections the administrative process does not appear to have an impact.

Notes

¹In fiscal year 1963 there were a total of 6871 elections, excluding decertification elections. The union won 59%. Of the total valid votes cast, 441969, the union received 59.9%. In 2001, there were 3076 elections, of the total valid votes cast, the union received 195845. The union won 52% of the elections. Twenty-Eighth Annual Report of the National Labor Relations Board, 1963 and 2001.

²See for example NLRB (1997).

³The board classifies cases as RC, RM or RD. Most elections are RC cases ‘filed by an employee or group of employees, or any individual or labour organization acting their behalf, for an election to certify a representative for collective bargaining in an appropriate unit.’ RD cases occur when ‘[A] petition filed by employees alleging that the union previously certified or currently recognized by the employer as their collective-bargaining representative no longer represents a majority of the employees in the appropriate unit and seeking an election to determine this.’ RM refers to petitions filed by an employer.

⁴The value of the SE did not change very much when OLS was used without correcting for heteroscedasticity.

⁵A detailed discussion of the estimated coefficients is available upon request.

⁶The election unit need not be the bargaining unit, though it may be. Workers in a particular department may vote to join a union that already represents some employees. The bargaining unit would then represent both the existing members and the new members.

⁷The excluded category includes: professional – clerk, clerical, guards and other.

⁸Seeber and Cooke (1983) found that consent elections were more likely to lead to an election loss. In this data set, however, there were not a sufficient number of consent elections to test their significance.

⁹ Elections Conducted by the NLRB (July 2000), p. 4.

¹⁰Farber concludes the decline may be due to different behaviour in small and large firms or the decline over time in organizations likely to organize. Unfortunately, Farber did not look at the kinds of election units.

¹¹Seeber and Cooke (1983) found a similar result.