Networking in the Penumbra: Public Management, Cooptative Links, and Distributional Consequences

ABSTRACT

Recently, some researchers have begun to emphasize that governance networks can have a “dark” side as well as a relatively benign one. This work thus far has mostly emphasized illegal and violent forms of action. But even networks that operate in accord with the law often include covert, or at minimum difficult-to-observe, ties and pathways of influence. Likewise, managers involved in networking engage in behavior that is only partially observable. They can use their discretion to shape network action in ways that are hard to discern—and therefore analyze. Research on such subjects is limited, and the practical results of such behavior can include lack of accountability as well as inequitable distribution of outputs and outcomes. This article shows that individual network nodes can work to bias the actions of the organization in ways that are likely to benefit the organization's more advantaged clientele. This study reports on tests for differential impacts of managerial interactions of and with various network partners. The argument is supported with an analysis of performance data from 500 organizations over an eight-year period. Networks and managerial networking can generate considerable benefits but can also have a less palatable impact that managers and scholars need to consider seriously.

ACKNOWLEDGEMENTS

An earlier version of this article was presented at the 21st EGOS Colloquium, Freie Universität Berlin, Sub-theme on Control of Legal and Illegal Networks, 30 June–2 July 2005.

This paper is part of an ongoing research agenda on the role of public management in complex policy settings. We have benefitted from the helpful comments of George Boyne, Stuart Bretschneider, Gene Brewer, Amy Kneedler Donahue, Sergio Fernandez, H. George Frederickson, Carolyn Heinrich, Patricia Ingraham, J. Edward Kellough, Laurence E. Lynn, Jr., H. Brinton Milward, Sean Nicholson-Crotty, David Peterson, Hal G. Rainey, Bob Stein, and Richard Walker on various aspects of this research program. Needless to say, this article is the responsibility of the authors only.

Notes

R-squared, .63.

Standard error, 7.18.

F, 368.94.

N of cases, 4114.

Dummy variables for individual years not reported.

All equations control for teachers' salaries, percent state aid, class size, teacher experience, percent of teachers not certified, percentage of black, Latino and low income students, managerial quality, managerial stability, employee stability and yearly dummy variables.

^#Significant p < .10, two-tailed test.

All equations control for teachers' salaries, percent state aid, class size, teacher experience, percent of teachers not certified, percentage of black, Latino and low income students, managerial quality, managerial stability, employee stability and yearly dummy variables.

*Significant p < .05, two tailed test.

All equations control for teachers' salaries, percent state aid, class size, teacher experience, percent of teachers not certified, percentage of black, Latino and low income students, managerial quality, managerial stability, employee stability, and yearly dummy variables.

*Significant p < .05.

^#Significant p < .10.

All equations control for teachers' salaries, percent state aid, class size, teacher experience, percent of teachers not certified, percentage of black, Latino and low income students, managerial quality, managerial stability, employee stability and yearly dummy variables.

*Significant p < .05, two tailed test.

^#Significant p < .10, two tailed test.

For example, one might use nepotism, patronage, or corruption to overcome the fragmental structures of networks.

Modeling the performance of overall networks of organizations, including efforts to manage the network, is considerably more complex than modeling the performance of an organization operating within an interdependent setting. Work is proceeding on aspects of the first-mentioned subject as well, although this effort goes beyond the reach of the present article. For an early analysis incorporating both managerial networking and also structural networks, see O'Toole and Meier (2004b). For initial modeling at the network level, see Meier and O'Toole (2005a).

Since we are exploring a set of systemic “biases,” controlling for past performance would wipe out a portion of such regularized impact, which should be expected to be relatively stable, especially over fairly limited periods of time. When an autoregressive set of estimations is used, the results are consistent with the findings in this paper but are weaker and less likely to be statistically significant.

Independent means that the school district is not subordinate to another unit such as a city. Independent districts have their own elected board, have the ability to tax and set budgets, and acquire bonding authority by a vote of the residents.

Districts responding to the survey, conducted during 2000, were no different from nonrespondents in enrollment, enrollment growth, students' race, ethnicity and poverty, or test scores. There were slight differences in a few other factors. Respondents had 0.48 more students per class, paid their teachers $200 more per year, but had annual operating budgets of about $100 per student less.

Whether or not school boards should be included in this measure is open to question. School boards also form a principal-agent relationship with the superintendent (see O'Toole, Meier and Nicholson-Crotty 2005). This study examines managerial networking, rather than the impact of the full network structure. Both can be important. As explained in detail in Meier and O'Toole (2005b), certain research questions can best be studied by large-N investigation of a portion of the interactions in networked settings. For details of the validity and reliability of this measure, see Meier and O'Toole (2005b).

While this measure includes both self-initiated and also reactive interactions by superintendents, other analysis shows that the majority of top-managerial interactions are of the former type (Goerdel 2006). Also, as shown in Meier and O'Toole (2005b), the correlation of the measure over a two-year period is approximately 0.5 when the top manager has not been replaced—a fairly impressive result for a behavioral measure. When the superintendent has been changed, however, the correlation over time is effectively zero. Consequently, this measure is an appropriate one for tapping managerial networking.

In 2002 the TAAS was replaced by the Texas Assessment of Knowledge and Skills (TAKS). At times during this period, the exit exam was given in the tenth grade rather than the eleventh.

The various pass rates do not correlate as highly as one might imagine. The intercorrelations between the Anglo, black and Latino pass rates are all in the neighborhood of .6, thus suggesting the overlap is only a bit more than one-third.

The relationship between the percentage of students taking the tests and the test scores in Texas is actually positive but explains less than two percent of the variance.

Consideration was given to using dropouts as a performance measure, but this idea was rejected for two reasons. First, the state of Texas changed its method of calculating dropouts during this time period. Second, the dropout measures are widely conceded to contain substantial errors and misreporting.

Nonetheless, the same controls listed in Table 2 were included in all eight other regression analyses.

Additional information

Notes on contributors

Laurence J. O'Toole

Laurence J. O'Toole, Jr. ([email protected]) is the Margaret Hughes and Robert T. Golembiewski Professor of Public Administration and head of the Department of Public Administration and Policy in the School of Public and International Affairs at the University of Georgia. Much of his research focuses on public management in network settings.

Kenneth J. Meier

Kenneth J. Meier ([email protected]) is a Distinguished Professor of Political Science and the Charles Gregory Chair in Liberal Arts at Texas A&M University. His two major research agendas are quantitative tests of public management theory and the investigation of the role of race and ethnicity in public policy.