ESTADO ACTUAL DE LA POLÍTICA INFORMÁTICA EN MÉXICO
2.4 ACUERDO DE LOS LINEAMIENTOS PARA LA OPERACIÓN DEL REGISTRO PÚBLICO DE COMERCIO
place size effect may be biased downward by the exclusion of these establishments.
"One of the peculiar features of the CSA survey is that workplace size is only
recorded for respondents who are self-employed. Thus, workplace size is not
included in the CSA membership equation.
Preliminary Analysis
magnitude of the estimated coefficients is gained by calculating the marginal
effect of a one-unit change in the variable of interest on the probability of
union membership with all independent variables set at their mean values.
Thus, we see that the estimated probability of membership for an individual
who is the sole employee in an establishment is approximately 34.5 percentage
points lower than that of a similar individual employed in a workplace with
more than 500 employees.
TABLE 3.3
Union Membership and Workplace Size: Estimated Coefficients and Marginal Effects
Workplace Size Category Estimated Coefficient
Standard Error Margmal Effect on Probability of Membership ( % ) ' Employee works alone -1.12"" 0.31 -34.5
2-20 employees -0.62™ 0.10 -22.2 21-50 employees - 0 . 2 6 " 0.11 -10.0 51-100 employees -0.12 0.12 -4.6 101-500 employees 0.06 0.11 2.5 More than 500 employees - - - LR Test of Joint Significance (x^) 8 0 . 0 4 ~ Notes:
'significant at 10% (two-tailed f-test for coefficients) "significant at 5 % (two-tailed /-test for coefficients) " s i g n i f i c a n t at 1 % (two-tailed /-test for coefficients)
'Effect of a one-unit change in the independent variable on the probability of union membership with all independent variables set at their mean values.
Chapter 3
It is interesting to note that while workers in smaller establishments have a significantly lower probability of union membership, the workplace size effect diminishes rapidly. Indeed, a worker employed in an establishment with 51-100 workers does not have a significantly lower predicted probability of membership than a worker with similar attributes employed in an establishment with more than 500 employees.
The estimated workplace size coefficients are consistent with the proposition that there are economies of scale associated with larger establishments and that, as a consequence, unions are more likely to organise larger workplaces, and to offer workers in these workplaces a more extensive range of union services. Ultimately, this translates into a higher probability of union membership for workers in larger establishments. However, it would seem that any organising economies are largely exhausted once a workplace size threshold of about 50 employees is reached.
The social custom model offers a possible explanation for the observed drop-off in the magnitudes of the estimated workplace size coefficients. To start, recall that a critical mass of union members is required to sustain a non- zero equilibrium level of union density. Now, it can be argued that unions will be more inclined to devote the organising resources necessary to reach the critical mass when the potential number of members to be gained is greater. However, we might expect the social bonds between workers in larger, more impersonal workplaces to be weaker than in smaller workplaces. Consequently, workers in large establishments may place less weight on their workplace reputations than workers in smaller establishments.
Preliminary Analysis
Thus, it can be argued that there are countervailing forces operating.
On the one hand, unions are more likely to organise larger plants because they
offer the prospect of more members to be gained. But, on the other hand,
larger workplaces offer a less fertile environment for union membership to be
sustained by social custom because the bonds between workers in larger plants
are weaker.
Industry of Employment
Industry variables are routinely included in models of individual union status.
They are usually interpreted as proxying several factors affecting the monopoly
gains to be made by unions. In particular, they are assumed to proxy variables
affecting union power, and the rents available to be captured from firms.^^
If unions are more successful in extracting rents in a certain industry,
does this mean that an individual employed in the industry will have a higher
probability of union membership than one employed in an industry where
unions are less successful? Many researchers assume that the answer is Yes,
arguing that when unions capture greater rents, the benefits of union
^''In the context of this scenario, the IMA estimates suggest that the organising effect dominates intially, but that once an establishment size of around 5 0 employees is reached the organising effect is offset by a diminution of the social bonds between workers. It should be noted, however, that estimates to be presented in Chapter 5 suggest that, in open shops, the economies of scale/organising effect only dominates until a workplace size of about 20 employees is reached.
^^For example, different industries have different concentration and capital-labour ratios, and face different product market conditions. These variables are expected to influence both the costs and benefits of unionisation (see, for example, Hirsch and Berger, 1984).
Chapter 3
membership are greater and that, consequently, there is a greater incentive for
workers to join a union. In Chapter 2, however, we argued that the way in
which the rents are distributed is of crucial importance. If, for instance, the
rents are distributed as private goods, then the greater monopoly union gains
give workers an added incentive to join the union. On the other hand, if the
rents are distributed as collective goods, individuals have an incentive to free-
ride, and the nexus between the magnitude of the monopoly gains and the
probability of union membership is more tenuous.^^
An alternative reason for including industry dummies is that they
capture variations in the intensity of union activity across industries.
Assuming that trade union activity is a function of organising costs, unions are
expected to be less active in those industries which are more costly to
organise." Thus, workers in industries characterised by small, widely
dispersed workplaces with high rates of labour turn-over are expected to be
less likely to become union members because there is less union activity in
^^Recall, however, that it can be argued that the prospect of greater gains f r o m collective action provides additional impetus for a social custom of union membership to become established and sustained among a group of workers.
" H i r s c h and Addison (1986: p.31) argue, in the context of unionism in the United States, that greater above-competitive rents may not only be associated with higher potential benefits of unionisation (through higher wages, for example), but may also be associated with higher unionisation costs as firms will resist union organising activity more vigorously when there is more at stake. Thus, the existence of greater rents in a given industry may not necessarily translate into a higher rate of unionisation in that industry.
While Australian firms may not be able to resist unionisation in the same way as American firms, it may nevertheless still be true that firms which feel they have more to lose f r o m union activity will be prepared to increase the costs of unionisation (for example, by not providing a facility for payroll deductions of union fees, or by limiting union access to the workplace strictly to the letter of award requirements).
Preliminary Analysis
these industries than in other industries.
Several of the earlier Australian studies have established a statistically significant link between industry of employment and probability of union membership (Christie, 1992; Miller and Rummery, 1989; and Crockett and Hall, 1987). However, we find only tentative support for the proposition that the probability of union membership depends on industry of employment (see Table 3.4).
Of all the industry variables included in our estimated equations only the transport industry coefficient is significant in both.^^ The wholesale and retail industry coefficient is significant only in the CSA equation, while the finance and business services coefficient is significant in the IMA equation but not the CSA equation. In the IMA equation a worker in the communications industry has a significantly higher estimated probability of membership. (However, none of the respondents to the CSA survey are employed in the communications industry so a communications coefficient is not estimated).
^^It should be noted that, in this section, any comparisons between the IMA and CSA estimates focus on the significance of the coefficients rather than their signs or magnitudes. This is because the estimated equations have different specifications and the coefficients are not directly comparable. For example, because workplace size is not controlled In the CSA equation, differences In the workplace size may be a source of differences in the coefficients If industry of employment Is correlated with workplace size.
Chapter 3
TABLE 3.4
Industry of Employment: Estimated Coefficients
Issues in Multicultural Australia Class Structure of Australia
s.e Marginal Effect ( % ) ' 0 s.e. Marginal Effect (%)'• Industry of Employment: Agriculture Mining -0.37 -0.35 0.28 0.38 -13.9 -13.3 0.09 0.31 3.7 Manufacturing 0.03 0.12 1.2 -0.01 0.19 -0.3 Electricity, gas and
water
0.20 0.27 8.1 0.19 0.46 7.3
Construction 0.28" 0.17 11.0 -0.05 0.32 -2.1 Wholesale and retail
trade -0.12 0.14 -4.8 -0.48"" -0.22 -18.6 Transport and storage 0 . 4 4 " ° 0.17 17.4 0.52"" 0.22 19.4 Communication 0.52"" 0.23 20.3 - - -
Finance, property and business services 0.30"" 0.14 12.0 -0.09 0.18 -3.5 Public administration - - - - Community services - - - - Recreation and personal services -0.05 0.16 -2.0 -0.25 0.25 -9.8 LR test of joint significance (x") 2 8 . 6 2 ~ 1 5 . 0 6 " Notes:
"significant at 10% (two-tailed f-test for coefficients) "significant at 5% (two-tailed f-test for coefficients) •""significant at 1 % (two-tailed f-test for coefficients)
'Effect of a one-unit change in the independent variable on the probability of union membership with all independent variables set at their mean values.
While only a small number of industry coefficients are statistically significant at the 5 percent level or better, this should not be seen as evidence that the influences they are assumed to proxy are not significant determinants of union membership. Indeed, this point serves to highlight a shortcoming of
Preliminary Analysis
the previous studies of union membership which, it should be noted, is also shared by this thesis. In particular, theory suggests that a range of firm and industrial characteristics will be related to union membership. However, typically the only proxies for these variables available in the data are industry dummies. Moreover, the industry dummies are conventionally entered in the membership equation at such a high degree of aggregation that they are very imprecise instruments for the influences they are thought to proxy.^^
Public Sector Employment
A key feature of unionism (not only in Australia but in virtually all OECD countries)^'' is that public sector employees have a higher rate of union membership than otherwise similar workers in the private sector (Visser, 1991). In the case of Australia, the Australian Bureau of Statistics reports that 67 percent of Australian workers in the public sector are union members while only 29 percent of private sector workers are union members (ABS, 1993).
The influence of public sector employment on the probability of union membership has previously been investigated by Crockett and Hall (1987) and Deery and DeCieri (1991). Both studies find a highly significant and positive
^^The industry dummies in our membership equations (and in the previous Australian studies) are constructed at the 1 digit or major classification level. More disaggregated dummy variables could be constructed (at the t w o or, possibly, 3 digit level). H o w e v e r , many of the industry categories would then have too f e w cases for the coefficients to be determined with a reasonable degree of accuracy.
^'^n S w e d e n and Demark union density rates in the private and public sectors are approximately equal (Visser, 1991). However, by international standards both of these countries have very high overall rates of union membership.