ABSTRACT
Representative bureaucracy theory assumes that representative organizations are better able to represent the needs of particular citizens and groups of service users and that they may perform especially well on their core tasks. In particular, recent work suggests that street-level bureaucrats are more likely to become active representatives than upper-level bureaucrats for whom a critical mass of representation is required to influence policy outcomes. The positive contribution of street-level bureaucrats may also be enhanced where they are accorded greater discretion or are better able to draw upon their connections with clients. In this study, we test these hypotheses by exploring the relationship between gender and minority ethnic representation and the performance of fire authorities in England. Our study reveals that more representative fire authorities tend to be more effective organizations. We find little evidence of a critical mass effect at upper levels, but street-level representation does affect organizational performance. Performance impacts are greater in policy areas with more discretion and opportunity for co-production.
Notes
Note: N=138.
Notes: Number of observations=138. Robust standard errors shown in parentheses. Formal hypotheses were evaluated with a one-tailed test, other hypotheses with a two-tailed test. Dummy variables for first two years of analysis not reported.
+ p≤0.10;
* p≤0.05;
** p≤0.01.
Notes: Number of observations=138. Standard errors shown in parentheses. Formal hypotheses were evaluated with a one-tailed test, other hypotheses with a two-tailed test. Dummy variables for first two years of analysis not reported.
+ p≤0.10;
* p≤0.05;
** p≤0.01.
Notes: Number of observations=138. Standard errors shown in parentheses. Formal hypotheses were evaluated with a one-tailed test, other hypotheses with a two-tailed test. Dummy variables for first two years of analysis not reported.
+ p≤0.10;
* p≤0.05;
** p≤0.01.
NOTES
For example, the operational rules for English fire authorities have been amended so chief female firefighters are allowed to be the first to enter burning buildings, having previously been excluded on the grounds they may be pregnant.
More representative bureaucracies could improve organizational performance for two other reasons. Greater representation of the different groups within society will mean that more and possibly better ideas will inform agency actions. In addition, by not discriminating in the job market, organizations are likely to get higher-skilled employees and that will increase productivity (Becker Citation1993). Unfortunately, we cannot distinguish among these three theoretical paths for improving performance with the current data.
Near identical results to those presented in the article were obtained when using the indicators simply measuring the level of representation of women and minority ethnic people amongst firefighters.
In theory, the nonlinear relationship between representativeness at the senior level and performance should follow a U-shaped path with the coefficient for the base term being negative and the coefficient for the quadratic term being positive.
Ideally, it would be informative to include both gender and racial representation in the same equation; however, organizations with substantial gender representation tend to be those that also have substantial racial representation, thus generating collinearity. This overlap means that it is difficult to sort out the independent effects of each specific type of representation. Our additional analysis showed that racial representation had an independent effect on organizational effectiveness over and above the impact of gender representation. Gender representation had an independent effect on special services over and above the impact of racial representation. These findings must remain tentative given the need for more cases where gender and racial representation are not tightly linked.
Similar results to the OLS estimates presented in Tables were derived when using Zellner's seemingly unrelated regression estimator to control for correlations between the residuals from the separate models for each measure of fire service performance. We also obtained near-identical results for the fire service effectiveness model using an ordered probit estimation.
The numbers of special services incidents, population density, and the proportion of the working-age population of minority ethnic origin were not normally distributed so logged versions of these variables were included in the statistical models.
We are grateful to IPMJ Editor Steve Kelman for suggesting this additional test of the effects of representativeness.