Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Technological innovation in public organizations
(USC Thesis Other)
Technological innovation in public organizations
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Technological Innovation
In Public Organizations
By Matthew Michael Young
Committee:
Juliet Musso, chair
William Resh
Lisa Schweitzer
A Dissertation Presented to the
Faculty of the Graduate School
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
Doctor of Philosophy
Public Policy and Management
Degree Conferral: August 2017
ii
For Ma.
Who’da thunk it?
Copyright 2017 by Matthew Michael Young
All rights reserved.
iii
Abstract
This dissertation considers how technological innovations are implemented in local
governments. Public organizations are under increasing pressure to implement new, innovative
technologies in order to be more transparent, efficient, and improve service delivery. But both the
challenges public managers face in implementing such innovations as well as the question of
whether they improve outcomes in practice remain under-examined. This work uses two recent
innovations – open data and the introduction of mobile applications and social media in municipal
311 systems – to directly address these issues.
The first essay critically examines open data implementation by local governments. It draws
upon the theoretical framework of Digital Era Governance (DEG) to argue that open data is a case
study in the type of “radical disintermediation” expected by DEG, and that this framework is useful
for understanding how open data can be both beneficial and detrimental. Open data’s potential
benefits include service digitization; increased transparency and accountability; and enhanced
economic development. Its risks include the use of data to shame organizations and individuals; loss
of control over service delivery; goal displacement; and replacing public employees with technocratic
elites. The essay then identifies how public managers can mitigate the risks posed by open data.
The second essay examines implementation of open data policies by large municipalities
across the United States. Drawing from research on public sector innovation, it analyzes the factors
that promote or inhibit open data implementation. The findings indicate that service-oriented
departments make more files available on average than administrative departments, and that of all
service-oriented departments, those whose missions include economic development make the most
files available. Department size and administrative capacity are associated with more files per
department, but city-level institutional characteristics had no appreciable impact. Demand-side
pressures at the city level, however, were all associated with increased data availability. The essay
concludes with a discussion of the implications for public managers looking to promote open data
implementation in their organizations.
The third and final essay analyzes whether technology adoption in a preexisting service
delivery system can improve efficiency, and considers whether such efficiency improvements come
at the expense of worsened inequality in service provision. It uses longitudinal analysis to causally
assess how the introduction of a smartphone application called Open311 and the integration of
Twitter affect the time required to resolve reports in San Francisco’s 311 system. Results show that
effects vary by technology; reports entered through Open311 get resolved faster than others, while
Twitter reports take longer. Both technologies offer improvements over traditional reporting
methods for resolution of service issues in historically disadvantaged communities, though unlike
Open311, Twitter’s effect remains consistent over time.
This research makes several contributions to the field of public management. First, it
identifies and details the potential benefits and risks that public managers face when implementing
open data. It also creates a novel, nationwide database of open data implementation in large US
cities, and uses these data to identify how institutional and organizational characteristics promote or
inhibit implementation in practice. This dissertation also provides empirical and methodological
contributions to the field through its causal estimates of how implementing new technology affects
organizational performance in responding to requests for government services.
Keywords
Public management, innovation, technology, digital era governance, coproduction, open data, 311
iv
Table of Contents
Acknowledgements .................................................................................................................. vi
Introduction ............................................................................................................................... 1
Theoretical and Empirical Context: Tension Between Bureaucracy and Innovation .................... 3
Overview and Contributions ................................................................................................................... 5
The Promise and the Peril: Open Data In Local Government ............................................... 10
INTRODUCTION ..................................................................................................................................... 11
OPEN DATA AND DIGITAL ERA GOVERNANCE ................................................................... 13
Administrative and Policy Context ....................................................................................................... 14
Open Data as Digital-Era Governance ................................................................................................ 16
Motivation for Adopting Open Data ................................................................................................... 19
THE PROMISE AND THE PERIL OF OPEN DATA .................................................................... 19
Open Data’s Promise .............................................................................................................................. 20
The Perils of Open Data ........................................................................................................................ 25
IMPLEMENTING OPEN DATA ......................................................................................................... 34
Addressing Goal Displacement ............................................................................................................. 35
Addressing Shame, Loss of Service Delivery Control, and Privileging Technocratic Elites ....... 37
CONCLUSION ........................................................................................................................................... 39
Open Data Implementation in US Cities ................................................................................ 42
INTRODUCTION ..................................................................................................................................... 43
OPEN DATA IN CONTEXT ................................................................................................................. 45
Open Data as Innovation ....................................................................................................................... 47
Factors Affecting Implementation ....................................................................................................... 50
DATA AND METHODS ......................................................................................................................... 59
Data ........................................................................................................................................................... 60
Model Specification ................................................................................................................................. 64
v
RESULTS ..................................................................................................................................................... 65
DISCUSSION AND IMPLICATIONS ................................................................................................. 70
The Medium is the Message: Innovation in Municipal Coproduction Systems .................... 76
INTRODUCTION ..................................................................................................................................... 77
DISINTERMEDIATION IN SERVICE DELIVERY: SAN FRANCISCO’S 311 SYSTEM ... 79
Coproduction as Disintermediation ..................................................................................................... 80
E-government, Digital Era Governance, and Coproduction ........................................................... 82
Disintermediation and Innovation in 311: The Case of San Francisco .......................................... 85
DATA AND METHODS ......................................................................................................................... 89
RESULTS ..................................................................................................................................................... 94
311 Usage Patterns By Technology ...................................................................................................... 95
Model Results ........................................................................................................................................... 99
DISCUSSION AND IMPLICATIONS ............................................................................................... 103
Conclusion ............................................................................................................................. 108
EXTENSIONS AND FUTURE RESEARCH AGENDA .............................................................. 111
Understanding How Open Data Are Used In Practice ................................................................... 111
Comparing Open Data to Other Service Digitization Arrangements .......................................... 112
Measuring The Impact of Fiscal Shocks on Open Data and Digital Service Provision ............ 113
Identifying the Effect of Managerial Changes in Open Data Implementation ........................... 113
Open Data and Performance Management: Complements or Complications? .......................... 114
Conclusion .............................................................................................................................................. 114
References ................................................................................................................................................... 116
vi
Acknowledgements
I have been truly fortunate to enjoy the guidance and support of a great network of mentors,
colleagues, family, and friends. First and foremost I want to thank my committee, beginning with my
advisor Juliet Musso. Juliet, thank you so much for years of phenominal support. You helped me
stay focused on what mattered, believed in my work, and always showed me where and how I could
do better with honest, direct, and constructive criticism. I’ve never had nor could I have hoped for a
better boss. Thank you, Lisa Schweitzer, for all of your amazing feedback and suggestions, not only
on my dissertation, but also on everything else, from life in the academy to Kant and the NSA and
everything in-between. I already miss stopping by your office on a lark; next time the drinks are on
me. You’re an incredible, brilliant person and I’m much better for knowing you. Bill Resh, thank you
so much for jumping right in and helping me grow as a scholar from the minute you landed at Price.
It’s been fun as hell to learn from you and I’m grateful for all of your guidance, wisdom, and help –
especially with managing the job market. But for everything all three of you have given me over
these years, I am most grateful for your friendship. Thank you.
There is no way that I could have navigated graduate school without the amazing Price School staff.
Chris Wilson, you are an absolute treasure to whom I am deeply indebted for all of your work
keeping the whole system pumping. Another special thank you goes to Suzanne Alexander. Suzanne,
you were an unbelievable advocate and resource for all of the years I was lucky enough to have you
as my program coordinator. Thank you for your friendship and for all of the candy corn. Aubrey
Hicks, thank you for your friendship and all of the awesome Bedrosian Center programming.
Thanks as well for giving me a reliable place to work in RGL. And thank you to everyone else who
keeps the whole edifice upright and running. I will miss you all.
Another round of thanks goes to the other Price faculty that I had the pleasure to know, learn from,
and work with over the years. Dan Mazmanian and Gary Painter, thank you so much for your
support and effort in helping me as a student in your classes and later in our time running the MPP
practicum. Dan, thank you as well for all of your help with my job search. Thanks to Nicole
Esparza, Yan Tang, Raphael Bostic, Tony Bertelli, and everyone else who helped me learn the tools
of the trade and develop as a scholar. This also includes colleagues in both PhD programs: Jovanna
Rosen, Vincent Reina, Ryan Merrill, Robert Jackman, Anthony Orlando, Haram Lee, Yusun Cho,
Brettany Shannon, Seva Rodnyansky, and everyone else.
I owe an eternal debt to my friends who have supported me and made life fun, especially when
graduate school wasn’t. Bryan and Victor, he’s really proud and lucky to know that he’ll always have
you and me in his corner. Thanks to Nat “That Bastard Nat Baldwin” Baldwin, Jon, Chris, and Ben
for all of the amazing memories on and off the mountain. Anna and Evan, thank you for being such
good friends and making my time in LA infinitely more fun. Nicole and Vincent, the same goes for
you, but thanks as well for letting us use your apartment during my dissertation defense and for
having the cojones to glitterbomb your own place in my honor. John Vilandre, you’re an amazing
guy and thank you in particular for all of your help on my application letters for graduate programs
all those years ago. Nate, I will always remember all the good times we shared and how we learned
from each other and grew over those halcyon days. And thanks to each and every one of you that
ever covered me or cut me a steal of deal on account of being a poor graduate student. I’ll never
forget it, and rest assured it’ll get paid forward. To all of my board, tabletop, and online gaming
friends across the world, thanks for the fun distractions and for helping me nail quick arithmetic and
vii
basic probability. A special thank you goes out to Max Barry for all of your help on my dissertation
in spite of all the other demands on your time and attention as I flailed about trying to use Python.
To my parents: thank you for everything that you sacrificed, put up with, and turned a blind eye to
as I made my way to where I am now. Ma, can you friggin’ believe this? We beat the odds together,
and I’m so very proud of and grateful for you. Pop, we went from Cheetos and fans on a hot day at
Garwood Way to dueling doctorates (mine’s shinier). You’re a great guy and I’m lucky to have you
as my father. I’m also incredibly grateful to Martha and Vin for treating me like one of their own,
right down to the inherited living room furnishings. You’ve always been there for me, Ta. Thanks
and love as well to the entire Massachusetts contingent: Loretta, Mark, Barbara, Paul, and the whole
host of cousins that always have my back. Thanks are also due to my new family. Jennifer and Rich,
thank you so much for welcoming me into your family, offering your love and wisdom, and
especially for raising such an amazing daughter. Sarah Mawhorter, you are without question the best
thing that’s ever happened to me, let alone come out of my time in graduate school. I am humbled
by your brilliance, warmth, love, and support. Thank you so much for everything you’ve done for
me along the way, especially for helping me believe in my work and myself. You make everything
fun. I couldn’t ask for a better partner.
1
Introduction
This dissertation contributes to research on public sector innovation by analyzing how local-
level public organizations implement new technologies following the rise of “big data” and the
concept of “Smart Cities.” Data-driven decision-making practices are nothing new under the sun in
local government; city police departments have used such systems since the 1990s (D. C. Smith &
Bratton, 2001). But the drastic increase in information and communications technology (ICT)
capacity – with correspondingly drastic decreases in cost – has made it possible for organizations of
all sizes and budgets to generate and analyze data on scales hardly imaginable ten years ago. At the
same time, the introduction and diffusion of smartphones has led to a sea change in how end users
receive and engage with the services they use on a daily basis – including, increasingly, those
provided by local governments.
These changes have not gone unnoticed in the fields of public administration and
management. In 2006 Dunleavy et al. proposed a new framework they called Digital Era
Governance to replace New Public Management based explicitly on the potential for new
technologies to reshape public organizations and how they interact with citizens. They argued then
and in subsequent work that advances in ICT could be harnessed to help reconsolidate public
organizations that had been atomized under New Public Management; reincorporate the provision
of services that had been outsourced to the private and nonprofit sectors; and do so while also
restructuring the way in which citizens request and receive services. This last element of Digital Era
Governance – what the authors called “radical disintermediation” – proposed to replace the street-
level bureaucrats traditionally responsible for fielding public requests for services with direct,
technologically-mediated access to the government’s ICT infrastructure (Dunleavy, Margetts,
Bastow, & Tinkler, 2006; Margetts & Dunleavy, 2013).
2
Other scholars building on the Digital Era Governance framework have argued that public
organizations’ use of open source standards will free them of contracts with technology vendors
whose proprietary systems carry a high degree of asset specificity (Fishenden & Thompson, 2013).
Others argue that open source standards will inevitably lead to an isomorphic process by which the
public services provided directly by governments will consist of generic and inexpensive core, or
“utility” services, while a free and open market – again thanks to open software standards – will lead
to a hypercompetitive market where more dynamic services are provided by developers as user-
focused products (Fishenden & Thompson, 2013). In this way, the “systems bureaucrat” is obviated;
technology now serves as the sole intermediary between core government systems and the public
(Fishenden & Thompson, 2013; Fountain, 2001). At the same time, others have produced case
studies of entrepreneurial leaders in US cities who have begun introducing the kinds of innovations
called for under Digital Era Governance, arguing that there are now two types of municipal public
managers: those who innovate, and those who fall behind (Goldsmith & Crawford, 2014).
A separate but related line of research seeks to understand whether and why organizations,
and public organizations in particular, adopt innovative practices (Borins, 2001; Damanpour &
Schneider, 2006; Fernandez & Wise, 2010; Hansen, 2011; Jun & Weare, 2010; Walker, 2008, 2014;
Walker, Avellaneda, & Berry, 2011). Private technology firms like IBM and Google, as well as non-
profit organizations like Bloomberg Philanthropies and Code for America, have also deliberately
built expectations for these types of services and for “Smart City” style solutions to urban problems.
In the end, whether the result of an intrinsic belief in the value of technology, a convincing sales
pitch, the promise of grant funding and/or volunteer labor, or simply mimetic isomorphism, cities
of all sizes across the US are implementing Digital Era Governance-style reforms, two examples of
which are open data and mobile application-based services.
3
What is missing from the current work on Digital Era Governance-style reforms and
innovations, however, is explicit attention to the challenges posed by implementation. These
challenges have been examined with respect to prior ICT-based public sector reforms (Coursey &
Norris, 2008; Fountain, 2001; Jun & Weare, 2010; D. F. Norris & Moon, 2005; Reddick, 2004; West,
2005). The research presented here seeks to build on and extend this body of critical analysis to
identify and understand the organizational and institutional factors that affect the kind of
disintermediating innovations that are central to the Digital Era Governance framework. The three
essays that follow consider the public sector’s challenges in implementing technological innovations,
analyze organizational factors that influence success, and assess technology’s potential to promote
more effective and efficient public goods provision. These essays draw from the research in public
sector innovation and e-government, coproduction, institutional analysis, and organization theory.
Theoretical and Empirical Context: Tension Between Bureaucracy and Innovation
By design, bureaucracies are insulated from outside influences in order to ensure the
consistent, impartial, and equitable provision of public goods and services (Merton, 1968; Weber,
2009). The organizational structures that promote stability also become sources of resistance to
mandates from elected officials that conflict with the bureau’s mission and operations. For example,
the New Public Management (NPM) movement of the 1980s sought fundamental change in
bureaucratic prerogatives, promoting competition through contracts for services as well as
coproductive strategies to improve customer service (Bovaird, 2007; Gunn, 1988; Hood, 1991).
Rapidly developing information and communication technologies (ICT) and, later, the
commercialization of the internet via the world wide web were considered one mean to these ends
(Ae Chun et al., 2012; Bimber, 2003; J. Musso, Weare, & Hale, 2000). The literature suggests,
however, that staff can resist or coopt innovations during the implementation process (Damanpour
& Schneider, 2009; Walker, 2008). The first essay explores how technological innovation in public
4
organizations aligns with and differs from existing bureaucratic and institutional norms, as well as
how management resolves these tensions in practice through the implementation process.
Benefits associated with e-government generally include efficiency gains derived from
lowered transaction costs and, notably, new modes of organization that allow for increased focus on
citizen service (Fountain, 2001; Heeks, 2002). While e-government has enjoyed rapid, broad
diffusion (Jun & Weare, 2010), there is little evidence that its implementation has had a substantive
impact on core municipal functions (Coursey & Norris, 2008; D. F. Norris & Reddick, 2013; West,
2004). This gap may occur in part due to organizational resistance and to the tendency for many e-
government systems to merely replicate traditional analog service provision with digital technologies
rather than use technology to radically change provision. For example, first-generation e-
government systems typically sought cost reduction in information provision and service delivery by
implementing transactional services such as payments of fines and license and permit requests and
renewals (J. Musso et al., 2000; Weare, Musso, & Hale, 1999). The ability to renew a driver’s license
online certainly reduced transactions costs without posing a threat to the bureaucratic status quo.
How more substantive improvements in public service provision might arise via technology
poses another important question. Internet-based tools for more dynamic and dialogic
communications that have emerged since the mid-to-late-2000s (i.e., “Web 2.0”) provide reform
opportunities requiring fundamental changes to organizational forms and behaviors. These changes
informed DEG theory, which argues that their use will improve service delivery by leveraging
returns to scale and using ICT to “disintermediate” preexisting institutional arrangements that
determine how the public interact with government (Dunleavy et al., 2006; Fishenden & Thompson,
2013; Margetts & Dunleavy, 2013). This potential to alter public agencies’ existing practices is the
central concern of my dissertation: (1) how do bureaucracies respond to pressure to implement
5
technologies that entail disruptive internal and external change; and (2) does that change improve
outcomes? This dissertation addresses these questions in turn across the three essays.
Overview and Contributions
The first essay, “The Promise and the Peril: Open Data in Local Government,” develops a
theoretical framework for understanding the opportunities and challenges associated with
implementing open data policies and systems. It grounds this implementation within the context of
DEG and disintermediation, and identifies the opportunities and threats faced by organizations
during the implementation process. It then identifies ways public managers can manage the risks
open data poses during implementation. This analysis leads to the conclusion that while technology
reforms like open data that disintermediate government-public relationships have the potential to
improve service delivery, they also carry risks. If left unaddressed, these risks may result in: (1) goal
displacement; (2) data use that may be embarrassing for organizations and officials; (3) a loss of
organizational control over service delivery; and (4) the privileging of technocratic elites.
This essay makes a theoretical contribution by developing a conceptual framework for
evaluating potential benefits and risks from implementing open data. Much has already been said
about the benefits open data can offer government, especially at the local level (Ganapati & Reddick,
2014; Thorsby, Stowers, Wolslegel, & Tumbuan, 2016; Wang & Feeney, 2016). This research
provides a countervailing discussion of the risks that public managers face from the sort of changes
open data in particular and technological innovations in general, and provides several possible
courses of action managers can take to mitigate these risks. Understanding these risks and how to
account for them is crucial for both the elected or senior executive officials who decide whether or
not to adopt open data and for the public managers who find themselves responsible for its
implementation.
6
The second essay, “Open Data Implementation in US Cities,” examines how municipal
departments (e.g., planning, transportation) implement open data policies and systems, with
particular attention to institutional and organizational factors that may influence departments’
decisions to provide data. It provides an empirical analysis of open data implementation by
departments in large cities throughout the US
1
. I extend my argument that cities are likely to
prioritize economic development goals in implementing open data. I also argue that the level of
implementation – measured as the number of discrete files made available through the city’s open
data site by each department – is also a function of department- and city-level characteristics,
including administrative capacity, institutional features, and demand pressure from the public.
The second essay builds on the first by adding a systematic empirical evaluation of
how city governments have implemented open data in practice. The findings carry several
implications for both research and practice. First, they make a crucial point: relative
implementation is enabled or inhibited at the department level, not by citywide measures.
Neither centralized IT infrastructure nor executive-level leadership has an appreciable impact
on how many files departments make available through open data. Rather, department
characteristics – including their relative administrative capacity as measured by the ratio of
managers to staff – are what appear to drive open data implementation. The implication for
public managers is that there is unlikely to be a “one size fits all” solution for implementing
open data. Instead, particular care and attention is needed on a department-by-department
level. Finally, the data required for this analysis represents a contribution to public
management research in its own right. These data will be used in the future to move beyond
a cross-sectional snapshot of open data implementation to, for example, understand how
1
The sample for this essay consists of cities with 100,000 or more residents as of the 2010 Census that have adopted and
7
changes to existing open data policies affect implementation rates over time as well as across
departments.
Second, and directly related to the first contribution, they paint a picture of a highly
uneven degree of implementation across different types of departments: economic
development-oriented departments dominate the landscape, with other service-oriented
departments still outpacing the administrative core. For practitioners, this begs the question
of what sorts of data public managers want the public to access and use. If the answer lies
somewhere beyond development and economic growth and includes data with more
practical and democratic uses for the average resident, then public managers must invest
additional time and resources in working with the appropriate departments to help them
publish these data.
The third and final essay, “The Medium is the Message: Innovation in Municipal
Coproduction Systems,” analyzes how mobile and web technologies ‘disrupt’ a preexisting municipal
coproduction
2
system, via efficiency gains and changes to service distribution. It exploits the case of
two separate, and fundamentally different, innovations in a single preexisting coproduction platform,
known as 311, in the City of San Francisco. It measures outcomes as the time it takes the city to fix
problems voluntarily reported by residents via 311, and compare the time to resolve problems
reported through two different technologies – a purpose-built mobile phone application called
Open311; and social media platform Twitter – with preexisting phone and website interfaces.
The third essay pivots from open data to a different technological innovation that, through
service digitization, is another example of disintermediation between public organizations
and the public. Here the contributions include robust causal estimates of the impact that
DEG-style innovations can have on local government public service delivery. The analysis
2
Coproduction in this context refers to a system of voluntary citizen participation in the production of publically
provided goods and services (Ferris, 1984).
8
also shows how different approaches to service digitization implementation lead, in turn, to
different outcomes. Specifically, the mobile application built specifically for the 311 system
outperformed all other mediums for making requests, while the social media platform
Twitter performed worse that not only Open311 but also the default telephone call center.
However, the analysis also showed that while Twitter underperformed in general, it
in fact outperformed other mediums, including Open311, in areas of the city with the
highest proportion of low-income and minority residents. In explicitly considering the
distributional consequences of these innovations in terms of their performance in minority-
majority and low-income areas, the findings also suggest that the public nature of reports
made via social media may, in fact, make this medium more effective for disadvantaged
residents. Because public managers are frequently caught between political pressure to
change and bureaucratic resistance to change, this research will also be useful for those
looking to effectively implement innovative technologies for the public good. While the
citywide results would suggest that implementing social media reporting is not worth the
effort, it is in fact the most efficient way for the least well-off to report public service
breakdowns.
Taken together, the three essays address the motivating question behind this
dissertation by (1) critically examining a case of technological innovation using theories of
public management to identify its opportunities and threats and their implications for
implementation; (2) empirically assessing how this innovation is implemented in practice;
and (3) measuring the impact of innovation implementation on service delivery. The three
essays are unified by the theoretical contexts of Digital Era Governance, e-government, and
coproduction. The goal, however, is to ground the investigation in a consideration of the
organizational context of technology, opening up the “black box” of technology
9
implementation to better understand the organizational and institutional factors that shape
public sector technology use at the level of the public manager and his or her constituents.
The result is a body of work that adds to our understanding of public sector innovation both
theoretically and empirically.
The Promise and the Peril:
Open Data In Local Government
Matt Young
Ph.D. Candidate, Public Policy and Management
Sol Price School of Public Policy
University of Southern California
Abstract
Public organizations are under increasing pressure to use data to motivate, inform,
and shape their policy- and decision-making. Public organizations are expected to
collect and use more data, and face pressure to disclose those data to the public
under the auspice of transparency and accountability. Open data is both a policy and
technological response to these pressures; it makes government data freely available
to the public. This paper critically examines the phenomenon of open data adoption
by local governments. It draws upon the concept of Digital Era Governance (DEG)
proposed by Dunleavy et al. (2006) to argue that open data is a case study in the type
of “radical disintermediation” expected by DEG, and that this framework is useful
for understanding how open data can be both beneficial and detrimental to local
governments. Potential benefits include service digitization; increased transparency
and accountability; and enhanced economic development. Risks include the use of
data to shame organizations and individuals; loss of control over service delivery;
goal displacement; and replacing public employees with technocratic elites. The
paper concludes by identifying how public managers can mitigate the risks posed by
open data during its implementation.
11
INTRODUCTION
Public organizations are under increasing pressure to use data to motivate, inform, and shape
their policy- and decision-making. The exponential growth of computing power, sharp decline in
storage costs, and growth of networked systems make it easier and easier to generate and capture
large volumes of data (e.g., vehicle traffic or climatological data) or organizational processes (e.g.,
information collected for business licensing). At the same time that public organizations are
expected to collect and use more data, they face additional pressure to disclose those data to the
public under the auspice of transparency and accountability (Dawes, 2010; Ganapati & Reddick,
2012, 2014; Kassen, 2013; Welch, Feeney, & Park, 2016). Demands for government transparency are
not new, but as technology has changed, so have the policies around transparency and disclosure.
The most recent example of these changes is the Obama administration’s open government
initiative of 2009. Its purpose was to extend transparency policies to incorporate new technologies
that facilitate two-way communication and make it easier to share data. One of the technical and
policy components of this initiative, known as open data, is the subject of this essay. Specifically, the
essay focuses on open data in the context of local governments in the United States. While open
data was first introduced at the federal level, it was quickly championed by city governments and
adoption rapidly diffused throughout the United States.
Open data is a systematic combination of policies that require departments
3
to make their
data publically available and technology that centralizes and facilitates public access. Open data’s use
as a policy and platform for increasing government transparency is commonly understood. There are
other arguments for open data, though, beyond transparency. Open data differs from previous
transparency-oriented policies, and those differences make it useful in new ways. Open data is
unique because it makes previously restricted data available in a format that is free, ready to use, and
3
This essay will use ‘department’ for consistency, but it is interchangeable with agency, office, bureau, etc.
12
platform independent. Open data encourage and enable the public to engage with data to produce
information about government activities and the state of the world.
Traditionally, the ability to generate information from government data has been largely the
purview of public managers and administrators due to the high costs of accessing raw data from
government. Public use of government data is not without precedent; examples range from the
federal census to city business permitting data. But open data standardizes, codifies, and, most
importantly, expands the range of data available for public use, while lowering both fiscal and
transactions costs associated with access. As with previous technological innovations, the open data
movement has raised hopes among reformers that there will be a sea change in the way the
government and public interact (Jun & Weare, 2010; J. Musso et al., 2000; J. A. Musso & Weare,
2005; D. F. Norris & Reddick, 2013). While these hopes are likely overambitious, there is value in
considering how the promise of this new technology can be realized in practice.
This essay critically examines open data’s potential benefits and risks for public
organizations. It employs a theory of modern public management called Digital Era Governance
(DEG) to contextualize open data’s capacity to alter institutional norms that govern how the public
interacts with government. Broadly speaking, Digital Era Governance refers to a series of
administrative and technological changes to public organizations to improve operations and service
delivery (Dunleavy et al., 2006). I argue that the DEG framework is useful for understanding how
open data can either meet or fail to live up to its advocates’ expectations. I also argue that open data
exemplifies the process of “radical disintermediation” – the removal of gatekeepers, in the form of
public sector employees and bureaucratic processes, to allow the public to interface directly with
government systems – expected by DEG. I further argue that this ‘disintermediation’ effect is not
guaranteed to produce positive outcomes; instead its relative benefits and costs are determined in
large part by how open data are implemented and used in practice.
13
While open data policies and systems exist at every level of government in the United States,
this essay focuses on open data’s potential effects for cities and their constitutive departments, and
the implementation approaches that can promote its use. Cities throughout the United States have
been early and frequent open data adopters. Indeed, local government is a frequent unit of analysis
for research on public sector innovation and change (Walker, 2014).
This essay is organized as follows. The next section provides a detailed explanation of open
data, its connection to and departure from previous transparency-promoting policies, and how
Digital Era Governance provides a useful framework for understanding how open data’s unique
characteristics create both opportunities and risks for public managers. It then critically examines
these opportunities and risks in detail. Finally, it concludes with a discussion of how public managers
can make the most of open data by managing its implementation with these characteristics in mind.
OPEN DATA AND DIGITAL ERA GOVERNANCE
Open data is implemented by combining internal-facing information technology (IT)
systems and external-facing “e-government” components to make data previously considered
internal to the department available to the public through a centralized system or “portal.” Though
there is no legal or even unified definition for the term, standards exist for what can be considered
open data. In order for data to be open, there should be no technical or legal barriers to the
acquisition or use of open data by actors external to the organization (Open Knowledge
International, 2017).
One potentially significant barrier is the need for proprietary software to access or use the
data. Open data must use a non-proprietary “machine-readable” format, which means that the data
are structured logically and explicitly, and can be “read” by any computer. For example, a table of
data stored as a comma separated value (.csv) file is both machine-readable and open; the same table
saved as an Adobe Portable Document Format (PDF) file is not. Legal barriers include monetary
14
cost and copyright and licensing limitations on use. Open data must be licensed in such a way that it
can be used and re-used for commercial or non-commercial purposes without restrictions (Open
Knowledge International, 2017). Thus, open data refers to data that machines can process that do
not require proprietary software to read or manipulate; are usable for any legal purpose without
restriction; and are available for free. The public is then left to access, interpret, and use these data as
they please.
Administrative and Policy Context
Open data was introduced in the United States as part of the Office of Management and
Budget's Open Government Directive (OGD) of 2009. From the very beginning, OGD grounded
its design and the need for open data based on the normative argument that public organizations
should be transparent. According to this argument, the public are the true owners of government
information. President Obama's Memorandum on Transparency and Open Government (2009), for
example, refers to government information at the federal level as “a national asset.” Thus,
transparency improves government-as-agent’s accountability to the public-as-principal.
While policies to improve transparency have a long history, the OGD is a new addition to a
series of policies requiring the government to collect and share information about its activities that
has grown and incorporated new technologies over time. The Federal Register Act of 1935 and the
Administrative Procedure Act of 1946 introduced systematic regular recordkeeping and reporting
requirements for the federal government in general and the executive branch in particular. In 1966
Congress passed the Freedom of Information Act (FOIA) largely in response to perceived
shortcomings in the Administrative Procedure Act. In response to the advent of the internet and
World Wide Web, Congress passed the E-Government Act of 2002, in order to require departments
to make “essential” information, such as staff contact information, available on their websites. A
15
similar process has played out at the state and local level over time as well, especially during the
progressive reform era of the late 19th and early 20th centuries (Dawes, 2010).
Open data is another step in the evolution of transparency-enhancing policies. Open data
policies both digitize and centralize the information exchange between government and the public.
Centralization both reduces search costs for the recipient and unifies data sharing procedures that
may have otherwise been fragmented across and within constituent departments (Jaeger, 2007). So
far, empirical evidence suggests that open data is an effective transparency tool. Surveys of senior
managers at state level found consistent affirmation of open data’s usefulness in advancing
organizational transparency-related goals (Ganapati & Reddick, 2012). A survey of government at
the local level found similar results: more than 80% of responding city managers reported being very
or extremely satisfied with how these systems increased transparency (Ganapati & Reddick, 2014).
In the United Kingdom, increased transparency as a result of implementing open data was found to
improve local evidence-based policymaking (Weerakkody, Irani, Lee, Osman, & Hindi, 2015).
Members of the public have also used open data to improve transparency. In Chicago, a nonprofit
organization called Open City created a website, chicagolobbyists.org, that uses open data to identify
relationships between lobbyists and politicians, including donation amounts and the number of
lobbying actions taken for all city departments (Kassen, 2013).
But open data has the potential to do more than just improve transparency. It can also
change the relationship between government and the public in ways that substantively differ from
previous information-focused policies like the E-Government Act and FOIA. Open data has the
potential to radically reshape the relationship between government and outside actors by giving the
latter raw material in the form of data used to create information instead of information as finished
or intermediate product. Open data also creates a single access point for data previously siloed in
different constituent departments across the government. These innovative possibilities of open data
16
are consistent with a theory of public management originally advanced by Dunleavy et al. (2006)
called “Digital-Era Governance” (DEG). DEG provides a useful lens for understanding what makes
open data different from other transparency policies. DEG can also provide insights on how
differences between open data and other transparency policies create both new opportunities and
risks for public managers and the publics they seek to serve.
Open Data as Digital-Era Governance
Digital-Era Governance, a proposed successor to the paradigm of New Public Management,
is predicated on public sector adoption of new internet-based technologies as platforms for both
internal operations and, most importantly, public service delivery (Margetts & Dunleavy, 2013). The
DEG framework has three principal themes: (1) service delivery digitization; (2) reintegration of
public sector processes previously fragmented under New Public Management (NPM) reforms; and
(3) reversing organizational fragmentation enacted under NPM to leverage returns to scale and
improve processes used to address public (expressed as “client”) needs (Dunleavy et al., 2006;
Margetts & Dunleavy, 2013). Open data’s innovative potential bridges all three themes of DEG.
First, open data radically alters the preexisting boundaries between citizens and government, leading
to new possibilities for digitizing services. Open data addresses the second and third themes by
centralizing organizational data resources in a client-focused “one-stop shop” system that breaks
down department-level IT silos, both reintegrating and refocusing organizational tasks.
Changes to Organizational Structure and Processes
The innovative potential for open data is not limited to the boundary space between public
organizations and their environment. It is also a process-oriented innovation. Open Data alters
existing institutions that govern information technology and management within public
organizations and replaces them to an extent with its own, new processes (Damanpour, 1991;
17
Walker, 2008). Operational changes such as these are often more difficult for public organizations
than adapting to new technology itself (Fountain, 2001). Conceptually, the logic of open data seeks
to counter to expected bureaucratic behavior, such as the tendency of complex hierarchies to
compartmentalize and even hoard information (Posner, 2010).
The institutional logic of control over information has direct consequences for the design of
public sector IT systems: more information may be made available within a department, but
information sharing between them is unlikely absent targeted policy interventions (Gil-Garcia, 2012;
Peled, 2016). Information systems are therefore a political concern for public managers (J. Lee,
2008). Open data challenges preexisting IT design and institutional norms that support data
compartmentalization. By requiring managers to make their data available to outside observers –
including their peers and superiors within the government hierarchy, their peer/competitor
managers, and the public – open data systems can serve as the “new institutional checks and
balances” called for by Bovens and Zouridis (2002).
Open Data as Disintermediation
Digital-Era Governance’s proposed ability to reshape the preexisting relationships between
government and technology rests in part on leveraging open technology standards in order to
digitize services efficiently (Fishenden & Thompson, 2013). For open data specifically, this sort of
change to service delivery also constitutes “radical disintermediation,” a process by which private
actors are allowed to interface directly with state systems (Dunleavy et al., 2006).
Data differ from information not in terms of their structure, but their function (Rowley,
2007). In information systems theory, data are symbols of objects and their properties; they are
representative constructs of observations of the state of the world (Ackoff, 1989). Information
involves the use of data to answer questions about the state of the world. This dependency forms
18
the base of the “wisdom hierarchy” proposed by Ackoff (1989), in which data allows for
information, which then allows for knowledge, then understanding, and finally wisdom.
The difference between making data vs. information available is crucial. Creating
information from data involves a suite of choices on the part of the creator, ranging from the
questions to be answered, to which data are appropriate, to what the results of the data processing
itself means. With respect to government data, public managers traditionally controlled these
choices. Open data effectively revokes this privilege – anyone outside of government can generate
his or her own information from the same sources used by public sector workers. Instead of a
memorandum about a project or policy as the result of a FOIA request, open data gives users access
to the source material used to inform the summary or position detailed in the memorandum itself.
Whoever accesses these data are free to draw their own conclusions from them. This represents a
complete end-around traditional interfaces for public interactions with government that involve
administrative agents (Jakobsen, James, Moynihan, & Nabatchi, 2016; D. Moynihan, Herd, &
Harvey, 2014; Nabatchi, 2012).
As part of altering the relationship between government and outside actors, the technical
characteristics of open data make it useful as building blocks for software applications (apps) and
services that automate the information generation process. One example of this use of open data
involves a public transit department making both fixed schedule and real-time transit routing data
available through open data. These data are then used to create an app that transit riders can use to
not only plan their trip, but have relatively precise information about when the next bus or train is
going to arrive at a given stop instead of just the scheduled time. Not only does this create digitized
service delivery in the form of the app, it has the added benefit of not requiring any capital or labor
inputs from the transportation department beyond those used to generate the data. A second
example is mi parque, a publically created project that lets residents of the Little Village neighborhood
19
in Chicago coordinate volunteer activities to maintain a new park (Kassen, 2013). Another parks-
related project in San Francisco resulted in an officially-sanctioned mobile app
(http://sfrecpark.org/san-francisco-recreation-parks-official-mobile-app) that lets users find
recreation space based on amenities, location, and other characteristics.
Motivation for Adopting Open Data
Given the range of open data’s potential impacts, the motivations behind its adoption are
likely to vary. Increasing transparency is perhaps an obvious justification. But public sector
organizations differ in their views on the value of sharing data. Some are more risk averse, avoiding
data sharing in general and often citing security concerns as limiting factors, while others use the
sharing of data to improve their relationships and/or perceived legitimacy with stakeholders and
external constituents (Dawes, 2010; Dawes, Cresswell, & Pardo, 2009; Ganapati & Reddick, 2012;
Welch et al., 2016). Moreover, providing data (instead of information) gives outsiders the capacity to
generate information that can have a broad array of uses for both government and the public at
large. But whatever value proposition for open data, its change potential also presents risks for
public managers and citizens alike. The next section of this essay explores the promise and the peril
of open data as they affect both public agents and their constituents. It begins by theorizing how
open data’s benefits might differ within local governments, and then goes into the potential benefits
in detail. It then turns to open data’s potential negative unintended consequences for both public
managers and potential end-users.
THE PROMISE AND THE PERIL OF OPEN DATA
Open data’s capacity to change (a) the institutions governing internal data management and
(b) the interactions between government and the public creates opportunities to realize the benefits
of Digital-Era Governance. Open data can increase transparency while also adding value to the
20
public in the form of access to the building blocks of information, which can in turn promote
digitized service delivery. For public managers, open data’s transparency-enhancing features make it
potentially more efficient than other options, and could be used to help legitimate departments
whose efforts usually go unnoticed by the public. Open data’s effect on information resource
sharing can enhance existing managerial programs like performance management. And open data’s
capacity for digitizing service delivery may be especially attractive given that the government need
not shoulder all of the development costs, as many private sector developers may have an interest in
data supply and services.
Open Data’s Promise
The manner in which open data are used can be categorized roughly along two dimensions:
accountability and usability. Accountability involves using data to draw inferences about
organizational performance or value. This is synonymous with transparency when those outside of
the organization have access to the data or information necessary to draw such inferences. Usability
refers to the use of data as a component in the production of a good or service. Data can be used
for either of these ends in ways that benefit public sector organizations, the public, or both. Table
1.1 illustrates how these use categories map onto different value foci in matrix form. Note however
that these categories are not necessarily mutually exclusive. They are meant to characterize ways in
which public managers may view open data in terms of its value for their respective departments
and/or the public. It could easily be the case that open data’s value for a given department spans
multiple categories.
Table 1.1. Matrix of Perceived Value From Open Data by Use Type
Value: Internal Value: External
Use: Accountability Performance Management Transparency
Use: Usability Economic Development Service Digitization
21
Internal Accountability: Promoting Performance Management
Open data can complement and enhance performance management. Senior management
and elected and appointed officials use data to create efficiency and effectiveness measures, inform
the creation of organizational goals, and evaluate performance post hoc. To the extent that
departments are subject to preexisting performance management regimes, open data policies may
seem redundant. But open data can improve performance management programs in several ways.
Open data breaks down the fiefdoms of data and information that constitute a central challenge for
performance management’s need for sufficient and appropriate data to measure performance (Behn,
2003; D. P. Moynihan, 2008; Radin, 2006). Because open data spans across all bureaus irrespective
of the governmental unit of analysis (e.g., federal, state, county, or municipal), it requires the
commitment of high-level leadership, which is also crucial for performance management systems
that challenge existing institutional norms (Sanger, 2008). Of course, open data could easily be
replaced by a direct mandate for managers to provide data for use in performance management
systems. However, this leaves the system open to gaming on the part of the manager. For example,
she may end up prioritizing work that generates the data used in the performance management
system to the detriment of other tasks that may be of equal or even greater value to overall social
welfare (Heinrich & Marschke, 2010).
Open data mitigates this problem in two ways. First, it creates the infrastructure for a
multifaceted performance management system in the sense that if all data are open then it is possible
to assign proportional weights across different task-specific data streams. Thus open data supports
performance management without directly signaling its construction to bureau managers in a way
that allows gaming. This strategy gives principals a chance to learn about agent behavior faster than
the agent can learn about the principal’s performance measuring instrument, which should result in
an increase in the performance measure’s usefulness (Heinrich & Marschke, 2010). Second, open
22
data contributes to the “numbering of the modern world” (Radin, 2006). This creates institutional
isomorphic pressures by which departments are compelled to increase the amount of data
surrounding their activities that they capture, in order to conform to the norms of sharing and
transparency. This form of pressure has been shown to predict adoption of other e-governance
systems, particularly municipal-level websites (DiMaggio & Powell, 1983; Jun & Weare, 2010; J.
Musso et al., 2000).
External Accountability: Advancing Transparency
External accountability is synonymous with governmental transparency; the idea that the
public should be able to hold government responsible for its decisions and actions drives the
normative argument for government transparency. Under the (strong) assumption that available data
accurately reflects departmental processes and outputs, users of public open data may uncover
operational inefficiencies or injustices and convey these findings to political principals. Open data
can also reduce the transaction costs associated with organizational transparency in general. An
example of this benefit is the use of open data to meet disclosure requirements for budgetary and
other financial documents. Because finance data are effectively tailored to meet open data
requirements, making them freely and openly available requires a minimal amount of effort on the
department’s part. Thus the department can, in theory, meet its disclosure requirements, whether
imposed voluntarily or by mandate, with fewer resources expended in the process compared to
traditional transparency processes.
Internal Usability: Toward Economic Development
Economic development is a dominating concern for local governments, often at the expense
of redistribution efforts and other social programs (Peterson, 1981). The economic development
potential of open data is frequently, though not uniformly, cited as a motivation for policy adoption
23
by both advocacy groups and governments. One example of how this maps to open data is the
possibility of a city planning department making data necessary for developers to evaluate potential
infill sites. By doing so, they reduce the transaction costs associated with the developer’s due
diligence process and, ideally, make their city more attractive for development projects compared to
its local or regional competitors. Indeed, open data has the potential to drastically lower the
transactions costs associated with such analysis by both eliminating fee-for-data regimes and the
potential need for physical media (e.g., DVDs or, in some cases to this day, hand-scanned digital
copies of documents in department archives) either hand-collected or with associated shipping costs
for firms not otherwise present in the city. The internal use value proposition may be a strong
motivator for senior management and elected officials in the city to implement open data. They, in
turn, are likely to encourage departments with data that can be leveraged to promote economic
development to make it available. Moreover, managers of these departments likely share this
perspective irrespective of administrative pressures due to professional norms (Moon & Norris,
2005).
A second argument for open data’s economic development benefits is its use in developing
commercial applications. For-profit firms can use open data to create applications that lower
information costs for users (e.g., mobile applications for logistical planning of public transit use) and
are then monetized through fees, ads, data about user activity, or some combination thereof. This
effectively monetizes the process of transforming data into information. Under the assumption that
the firm creating these apps is profitable and located within the city’s jurisdiction, open data can
therefore promote economic development both in terms of job growth and increases in tax revenue.
But this use of open data has potential value both to public managers and the public beyond
economic growth. These apps are also privately-provided digitized public services. These services
can be both standalone and complementary to existing public services. To return to the example of
24
developing mobile applications for public transit users that lower their information costs, this
product makes public transit more attractive by providing up-to-date arrival information and/or
simplifying the process of figuring out which lines and modes of service to transfer between to
complete their trip.
External Usability: Service Digitization
Creating digital applications with open data can improve service delivery and lead to the
creation of new services that were not previously possible. Improvements in service delivery include
the use of open data to create mobile apps that facilitate resident (or tourist) use of department
services or facilities – apps like the one in San Francisco discussed earlier. Making it easier to use
department services is particularly valuable if the department relies on resident and visitor patronage
(e.g., transit ridership) either directly or indirectly for their funding. Creating new services with open
data-powered apps can improve on existing services, but it can also help legitimate departments
whose efforts often go unobserved by voters and taxpayers. It is much easier for the average person
to notice when a public health department fails to do its job and people are sickened than when
outbreaks are avoided in the first place. By pushing data to outside users, such departments can
make their proactive work more visible. For example, public health departments in San Francisco
and New York did just that in 2012, partnering with crowdsourced restaurant rating company Yelp
to integrate restaurant inspection data into Yelp’s app (https://www.yelp.com/healthscores).
Open data-driven service digitization has a particular advantage for public managers: it is
very cheap. The government is not responsible for any development or operational costs beyond
those associated with the data’s creation and maintenance. For cash-strapped public organizations,
this may be the only realistic way that they can hope to offer these services to the public. And even
those organizations with fiscal capacity are unlikely to have staff with the technical expertise
necessary to design, write, and validate software. But beyond resource constraints, market efficiency
25
arguments suggest that private individuals and organizations are more capable of identifying user
(read: consumer) needs and developing services to meet them. To the extent that decision-makers in
public organizations believe those arguments, they may be inclined to “let the market decide” what
digitized services should be created based on open data (Goldsmith & Crawford, 2014; Robinson,
Yu, Zeller, & Felten, 2009).
To summarize, open data can improve accountability and provide use values that are both
internally- and externally-oriented. Internal accountability consists of open data’s capacity to
improve performance management programs by increasing data access and volume. External
accountability refers to open data’s impact as a transparency-improving system. Internal use value
for government comes primarily in open data’s potential for promoting economic development,
while external value is added through open-sourcing and digitizing service delivery. The next section
turns to open data’s risks for public organizations.
The Perils of Open Data
Opportunity and risk are inextricably linked. Ultimately, all of the risks open data poses for
public managers involve misusing data; different forms of misuse give rise to distinct types of risk.
These types of risk in also stem from the same features of open data, and of the DEG framework
more broadly, that drive its potential benefits. Allowing the public direct access to state systems –
the disintermediation proposed in DEG – represents, fundamentally, a loss of control on the part of
public managers and administrators. This loss of control has consequences for how the data are used
both to digitize service delivery and advance the city’s broader economic development goals. The
technical and subject area expertise necessary to use open data properly magnifies and augments
these risks as well. Table 1.2 summarizes how the risks open data poses map onto use types and
users in the same way as its prospective benefits.
26
Table 1.2. Matrix of Perceived Risks From Open Data by Use Type
Risk: Internal Risk: External
Use: Accountability Goal Displacement Public Shaming
Use: Usability No Control Over Service Delivery Privileging Technocratic Elites
The drive to generate and publish data carries the risk of goal displacement; data may be created and
made available simply for the sake of satisfying requirements. On the other hand, some managers
may worry that those same data will be used to generate performance measures irrespective of
whether they represent accurate or appropriate measures of organizational performance in practice,
potentially causing them to actively resist making data available. External transparency can lead to
the shaming, deliberately or otherwise, of both organizations and public sector employees. And the
same disintermediation that allows departments to advance city goals and digitize service delivery
comes with a loss of control over how those data are used and for what purpose, and leads to the
empowerment of technocratic elites at the expense of public managers and those without the
necessary skills and experience to make sense of raw data.
Internal Accountability: Goal Displacement
Goal displacement takes place when incentive structures lead organizations to focus on
generating outputs that improve their evaluation at the expense of activities that advance their
broader mission (Bohte & Meier, 2000). It is particularly difficult to assess organizational
performance when the relationship between its outputs and outcomes is difficult to establish – or,
indeed, when there are few measurable outputs at all. As with performance management, open data
may present vexing challenges to departments with low output measurability (Wilson, 1989). It may
be difficult for these departments to generate data useful for inclusion in open data sites. Moreover,
especially when paired with performance management programs, open data may engender heel-
dragging and other informal forms of noncompliance from departments that see the same problems
27
with open data that they have with performance management (D. P. Moynihan, 2008; Radin, 2006,
2009; D. C. Smith & Bratton, 2001).
If, however, elected officials and/or senior administrators apply sufficient pressure to
departments and agencies in their desire to increase the volume and breadth of files available
through their jurisdiction’s open data site, those organizations whose normal operations do not lend
themselves to data generation and retention may find themselves scrambling to adopt new
processes, policies, or systems to meet their performance objectives with respect to open data. In
organizations with more general performance management systems in place, this behavior carries the
additional risk of becoming self-reinforcing; as departments generate data where none previously
existed, the temptation arises to use those data as quantifiable outputs to assess organizational
performance. Once this happens, the cycle of goal displacement can become locked in, to the
detriment of organizational effectiveness in practice.
Goal displacement from open data is not limited to departments or agencies within local
government. For both analysis and development, elected officials and executive managers’ incentives
for prioritizing specific data can also be a problem. With few exceptions, unlike other levels of
government in the United States, local governments are constantly vying for commercial investment
and development. To be sure, economic development is not the only concern of local governments,
but it dominates, such that local officials are often willing to cooperate across political and social
divides to advance development despite other differences (Ferman, 1996; Molotch, 1976; Peterson,
1981; Stone, 1989, 2005). The profit motive on the private side and the growth machine forces
within local governments can also lead local governments to focus most if not all of their attention
on economic development data, to the detriment of other, democratic public sector interests.
28
External Accountability: Public Shaming
One of the biggest risks open data presents public managers is that public access to data will
lead to embarrassment either for their city government at large, department in particular, themselves
individually, or some combination thereof. At a basic level, the very nature of some data can be
embarrassing when made public. Consider the case of the City of Baltimore, which makes the salary
of all City employees publicly available through its open data site, data.baltimorecity.gov. An
anonymous comment posted on the 2011 salary data file five years ago – presumably though not
definitively from a City employee – reads, “I am not in favor of my salary being displayed as a public
spectacle. Government transparency is one thing, but this is reaching a bit too deep. Can we rethink
this unnecessary invasion into public servants privacy please?”
Even seemingly innocuous data may be used to embarrass public organizations and
employees or twisted to advance a cause or agenda. This misuse could take the form of anti-
government activists using open data out of context to advocate for budget or personnel cuts.
Another possibility is unintentional embarrassment that occurs when the public uses data without
the insider and contextual knowledge to grasp its true meaning. Consider the case where a user
downloads the data used to evaluate a department under a performance management program.
Independent of the metrics developed as part of the organization’s performance management
process, she can use these data to generate her own measures of performance. These measures may
provide a novel perspective on how the given department is performing. They may also fail to
consider any number of logistical, technical, administrative, or other factors that are unknown to
government outsiders.
The recent rollout of USAFacts
4
, a project funded by former Microsoft CEO Steve Ballmer,
is an example of how such external use of data may help or hurt public organizations. USAFacts
4
https://usafacts.org
29
uses public financial and demographic data – including data available through open data sites at
every level of government in the US – to create visualizations and reports that are meant to inform
users of how the public sector receives and spends money, and what the associated outcomes are
(USAFacts Institute, 2017a). On the positive side, these data can lead the public to understand
government’s value in ways they had not previously. As an article on USAFacts in The New York
Times notes, “Suddenly, he [Ballmer] explained, the faceless bureaucrats who are often pilloried as
symbols of government waste start to look like the people in our neighborhood whom we’re very
glad to have” (Sorkin, 2017).
However, arriving at these sorts of realizations requires users of USAFacts to actively “drill
down” into the data. For those with less time or focus, the site makes summary reports of
government financial positions available. These use the same format as the annual reports – known
as Form 10-Ks – that the Securities and Exchange Commission (SEC) requires of large, publically
held private firms. The logic behind using this template is that it offers a consistent, standardized
format that is already established and familiar to investors. But the difference in organizational
mission and other institutional frameworks between private and public organizations may lead
readers to draw conclusions about public sector performance that are not appropriate. By using the
format adopted by the SEC to shape data about private firms into information useful for investors
in order to transform public sector data into information useful for citizens, USAFacts risks
implicitly leading the reader to compare government to private firms, consciously or not. The
organization does provide a conditional caveat in their 2017 “10-K” report for the US government,
noting “Of course, our Government is not a corporation; its purpose is not to make a profit but to
provide services to its citizens that improve the quality of life. But this Form 10-K format does have
the advantage of providing a thorough account of government finances, structure, and activities”
30
(USAFacts Institute, 2017b). Exactly how these differences are accounted for, however, is unclear
from reading the report.
The case of USAFacts is particularly illustrative for understanding how open data can be
both a boon and a bane for public managers. By making data available, public organizations can help
inform debate on the value of the services they provide, and do so without shouldering the costs of
facilitating the necessary conversion of data to information that the public can understand –
USAFacts’ founder estimates the startup costs to be approximately $10 million with another $3-5
million in annual operational costs (Sorkin, 2017). But in trusting outside actors with the process of
turning public data into information, public managers risk being set up for comparisons to the
private sector in ways that may cause them to be held to inappropriate standards and expectations.
Internal Usability: Loss of Control Over Service Delivery
Innovations to public service delivery can take many forms. The disintermediation effect
with respect to public services that open data makes possible contains elements of two previously
established methods for delivering services, but does not neatly fit into either – a fact that has
implications for how it causes departments to lose control over service delivery. Its voluntary use by
nonprofits and civic-minded individuals to produce services without compensation is consistent
with definitions of coproduction (Bovaird, 2007; Brandsen & Honingh, 2015; Parks et al., 1981).
Using Bovaird (2007)’s typology, coproduction of apps or other services without any direction or
input from data suppliers is most consistent with “community self-organization,” where the public is
responsible for both planning (app or project design) and production (programming and distribution
of the app or project). Classifying open data’s use as coproduction is less clear, however, when
considering the for-profit use case by private firms. In this case, open data more closely resembles
contracting out to private firms for service delivery. The skills necessary to transform bureaucratic
data into apps that deliver user-friendly value to the public are exemplary of the asset specificity
31
theorized and empirically associated with the decision to contract out for services (T. L. Brown &
Potoski, 2003; J. Ferris & Graddy, 1986).
Coproduction’s clearest recognized advantage for governments comes from efficiency
improvements combined with cost savings arising from the use of free labor (Kiser, 1984; Parks et
al., 1981). This argument is also used to justify contracting out for services, though efficiency gains
through labor substitution in that case rest on the argument that private firms facing market
competition should be more efficient (J. Ferris & Graddy, 1986; Hefetz & Warner, 2012; Joassart-
Marcelli & Musso, 2005). The fiscal constraints most US cities face make the near-zero cost of
digitized services via open data particularly attractive whether they are coproduced or privately
provided (Kiewiet & McCubbins, 2014). The institutional characteristics of open data, however,
complicate this picture.
The use of open data as laissez-faire, community self-organization style coproduction leaves
public managers with no recourse for addressing the free rider problem endemic to coproducing
public goods that can lead to insufficient service delivery (Parks et al., 1981; Powers & Thompson,
1994). At the same time, open data’s use by private firms is the ultimate edge-case of an incomplete
contract: there is no contract at all. On the upside, this effectively eliminates the transaction costs
arising from contracting while potentially lowering search costs for application users (T. L. Brown &
Potoski, 2003; Girth, Hefetz, Johnston, & Warner, 2012; Kiser, 1984; J. Musso, Young, & Thom,
2017). That the data themselves are free helps to eliminate the hold-up problem, increasing the
likelihood of use by private firms (Hart & Holmstrm, 1986; Hoppe & Schmitz, 2010). And there is
less potential for any one developer to use strategic capacity building to crowd out others (Bertelli,
2012). But without the structured relationship that a coordinated volunteer program or contract
provides between a supplier and public managers, the latter are left to effectively trust “the market”
as an amorphous whole to supply information-producing applications that meet the public’s needs.
32
In the case of municipal governments, the assumption that both voluntary and for-profit application
developers are likely to be located within the community may bolster this trust, but whether this is
true in practice remains to be seen (Lamothe & Lamothe, 2012).
A direct result of this loss of control over service delivery is that the public sector is left with
no enforceable mechanism, whether contractual or through the supervision of volunteer activities,
to ensure that equity concerns are addressed with respect to disadvantaged and underserved
communities. For example, a firm may choose to develop an application to assist users in using
public transit, but decide that the opportunity cost of translation into languages other than English is
too high. Given the empirical evidence calling the presumed benefits of private sector provision into
question – especially when there are direct implications for equity – open data may result in less
effective and equitable service delivery (Amirkhanyan, Kim, & Lambright, 2008; Bel, Fageda, &
Warner, 2010). Similarly, the unsupervised use of open data by volunteers is likely to be subject to
philanthropic particularism, leading to the privileging of subpopulations that have sufficient
organizational and technical capacities to use open data for their benefit (Salamon, 1995).
Finally, while these apps may help legitimize public organizations (as in the case of Yelp’s
use of public health department data) not every developer may be willing to share the spotlight.
While the licensing agreement for open data used by most cities requires some form of attribution, it
does not set guidelines for how obvious or accessible that attribution must be. At the same time,
some argue that the future of the public sector consists of “government as a platform” whose
responsibilities are primarily centered on making data available for private individuals and
organizations to use (Goldsmith & Crawford, 2014). This private use and branding of public data
may obscure the role and effort made by public organizations in collecting or generating those data
and making them available. Public misconceptions about the role government plays in providing
33
services can lead to depressed levels of political support for public sector expenditures and trust in
the government (Mettler, 2011).
External Usability: Privileging Technocratic Elites
In theory open data is highly democratic, but in practice its use requires gatekeepers in the
form of the technical elite. Data may be made available to anyone with computer hardware and an
internet connection, but relatively few are capable of using the data in a meaningful way. There are
several reasons for this. First, most people do not possess the technical skills necessary to use these
data. Cleaning and analyzing large-scale spreadsheet or relational database data requires both
hardware powerful enough to handle “big data-” sized files and knowledge of how these files and
the data they contain work. Hacker hobbyists exist, but they are likely to be a subset of those already
employed in computer and software engineering. In 2015 less than 3% of the US workforce were
employed in computer or mathematical occupations. Moreover, White men dominate this field as
the result of both the systematic and unintended exclusion of women and people of color. Less than
a quarter of computer or mathematical workers are women, and Black and Latino workers each
constitute less than 10% of these employees (Bureau of Labor Statistics, 2017).
Viewed in this light, DEG’s argument that disintermediation eliminates the traditional
gatekeepers in the form of agency personnel may be technically correct, but these personnel are
replaced by new gatekeepers in the form of these technocratic elites. To the extent that agency
personnel are representative of the communities they serve in ways that private sector technocrats
are not, and are afforded the discretion necessary to act on the public’s behalf, this has direct
implications for service provision equity (Bradbury & Kellough, 2007; Marvel & Resh, 2015). And as
private actors, these new gatekeepers are unlikely to share the same sense of purpose or mission as
the public organizations that they supplant, and – though this is by no means guaranteed – may be
34
less likely to be driven by the same prosocial motivations as public sector employees (Bozeman &
Su, 2015; Houston, 2000; Perry, Hondeghem, & Wise, 2010; Rainey, 1982).
Beyond technical expertise, poor data curation leads to the need for subject area expertise to
make use of open data, which also privileges elites. Not all data are created the same, and even
someone familiar with website server administration may not know much at all about traffic data
involving vehicles. While standards such as W3C’s Semantic Web require data to be bundled with
metadata – nested information about the data themselves – as well as links to other, related data,
much of the data available through city open data systems fails to include more than the title and file
itself (Young, 2017b). The need for subject area expertise is not as limiting as technical expertise,
though its violation may be more dangerous; the presence of technical but not subject area expertise
increases the risk that the data will be misused.
The set of potential open data users that possess both technical and subject area expertise is
likely to be small. It is also likely to underrepresent women, minorities, and the poor. The
implication is that those deciding whether and how to use open data are likely to be
disproportionately white, male, and middle- to upper-class. Their decisions may still lead to equitable
outcomes, but it is certainly less likely when other members of the public are systematically excluded
from participating. This systematic exclusion also precludes any hope of democratic outcomes. And
of those White middle- and upper-class men, those using poorly curated data without possessing
subject area expertise are also likely to misuse those data. On top of issues of representation, misuse
is more likely because efficiency-oriented non-governmental actors lack formal responsibility to
further the public good or individuals’ constitutional rights (Rosenbloom, 1983).
IMPLEMENTING OPEN DATA
The potential benefits of open data are compelling. Its risks are sufficient to give serious
pause. As a political matter, elected or appointed officials usually have the responsibility for deciding
35
to adopt open data as a government-spanning policy.
5
Implementation might therefore be expected
to follow in a top-down manner (Pressman & Wildavsky, 1973; Sabatier & Mazmanian, 1980;
Weible, Sabatier, & Lubell, 2004). However, that open data not only crosses intra-governmental
boundaries across departments but also leads to DEG-predicted forms of disintermediation between
the public and government means that implementation is likely to be more collaborative in practice
(Emerson, Nabatchi, & Balogh, 2012; May & Winter, 2007; Meier & O'Toole, 2003; O'Toole, 1986).
Yet the ultimate responsibility for making data available invariably falls to public managers within
the departments that generate them. The good news is that management matters: how public
managers choose to implement open data in practice conditions the likelihood of success (enjoying
its benefits) or failure (suffering its adverse effects) (Hupe, 2014; Riccucci, Meyers, Lurie, & Han,
2004). This section of the essay addresses how different implementation approaches can help public
managers make the most of open data by addressing its risks head-on.
Addressing Goal Displacement
Goal displacement is the most internally-oriented of open data’s identified risks for public
organizations. Two factors that motivate goal displacement are task complexity and insufficient
resources (Bohte & Meier, 2000). Addressing the former requires paying particular attention to how
performance metrics map to organizational tasks and mission. The latter requires investing in the
labor and capital resources necessary to ensure implementing departments or agencies are not forced
to sacrifice operational efficiency in order to meet compliance requirements or performance
standards with respect to making data available.
5 In the case of local governments, these appointed officials are usually city managers.
36
Appropriately Structure and Manage Incentives
As the research on performance management shows, setting and managing objectives for
organizational performance is difficult to do well (D. P. Moynihan, 2008; Radin, 2006, 2009). Public
managers overseeing open data implementation should avoid measures and metrics like blanket
publication quotas that do not acknowledge institutional differences between departments with
respect to data generation and collection. At the same time, focusing implementation on “low
hanging fruit” or data that directly promote city interests like economic development at the expense
of other types of data reduce the policy’s innovative impact. Laissez-faire approaches to
implementation are also problematic; relying entirely on department managers’ intrinsic motivations
heightens the risk of reducing content depth, and standalone one-and-done style file quotas are likely
to result in tokenistic implementation. Targeting implementation expectations and evaluations may
be the best approach, but it is also the most time and labor intensive. One way to reduce these costs
is to solicit input from the public in identifying the data of greatest interest. Engaging public
stakeholders to identify focus areas for implementation can also spur interest in open data,
improving the public’s trust in government.
Provide Administrative Support
Open data’s capacity to alter preexisting structures both within government and between
government and the public makes it a complicated innovation that bridges both internally- and
externally-oriented processes (Damanpour, 1987; Damanpour & Schneider, 2009; Walker, 2014;
Walker et al., 2011). Organizations with more administrative capacity perform better in general, but
are also better able to adapt to and manage change, making innovation adoption and implementation
easier (Andrews, Boyne, & Mostafa, 2017; Fernandez & Wise, 2010). Some governments, have taken
steps to increase administrative capacity at the executive level by creating “Chief Data” and “Chief
Innovation” Officer positions.
37
Adopting innovations like open data, however, likely requires more distributed, broad-based
support within and across departments (Meijer, 2014). Providing additional support and resources
for implementation to smaller and “flatter” departments may help overcome the risk of resistance to
implementation based on fears of goal displacement. Those overseeing open data’s implementation
should understand that not all departments will be able to make large amounts of data available and
set their expectations accordingly – or they should be prepared to pick up the slack themselves.
Addressing Shame, Loss of Service Delivery Control, and Privileging Technocratic Elites
The remaining three risks – public shaming, loss of control over service delivery, and
privileging technocratic elite interests – are all grounded in the risk of data misuse by those outside
the organization that comes with disintermediation. In all cases, but especially for the risk of public
shaming, there is no way to eliminate the risk while still making meaningful data available under
open standards. Rather, public managers responsible for implementing open data must settle for risk
mitigation. Two strategies that are applicable for all three risks are (1) engaging in data stewardship;
and (2) engaging directly with open data’s potential and actual users. Both measures involve making
those outside of the organization more aware of the data’s context, appropriate use, and prospective
goals or requirements that departments might have for service digitization in practice.
Engage in Data Stewardship
Taking steps to make data easier to understand and use properly helps democratize their use
by lowering or eliminating expertise requirements; guides those using the data to avoid accidental
misuse that paints the data provider in a bad light; and provides context that can help shape how the
data are used to digitize service delivery. One such step is including contextual information, also
known as “metadata,” with data available through an open data site. These metadata can include
information about the data’s structure, size, and row and column headings. Metadata can also
38
provide more in-depth context for the data. Examples include describing how the data are collected
or generated; how they are used at the department and/or city level; whether they are a component
of a larger data array; and how they may relate to other data, including data external to the
department or government altogether.
Explicitly identifying and linking to related data also adds value to the user beyond what is
contained in the data themselves; including such links is the final stage of Berners-Lee (2011)’s
proposed five-star value ranking of online data
6
. Including metadata therefore not only makes it less
likely that its users will inadvertently misuse open data, but increases its usefulness as well. Both of
these advantages also serve to level the playing field between subject area experts and laypeople, thus
also reducing the privileging of technocratic elites as gatekeepers for using open data effectively.
Engage with Open Data Users
Open data necessitates ceding control over data. But it does not prohibit public sector
involvement in application development or in helping open data users generate information from
the data by hand. Cities with open data can also contract out for app development. A number of
firms already offer services for governments that use open data. For example, buildingeye is a private
firm that sells an open data-driven app for visualizing planning and development data. Through their
service users can see a spatial representation of development patterns, and configure notifications
for new permit requests within range of a given address or intersection. In this particular case
buildingeye’s app is a turnkey product with relatively little customization across cities. But cities can
also contract for app development that is specific to their needs and data.
Contracting out, however, obviates the cost benefits associated with open data. Those
hoping to use open data to spur service digitization without paying for it can still take proactive
measures to make sure their organizational needs are met without exacerbating service inequality.
6 Open data’s definitional standards meet the four preceding stage’s requirements.
39
This involves engaging with private and nonprofit organizations interested in using open data (or
those that might be interested but are unaware of the city’s open data policy). Getting out ahead of
(or spearheading) the development process allows public managers to advocate for features and
functionality that benefit underprivileged groups that might otherwise not receive attention.
Taken together, these two approaches have the potential to resolve the challenges of open
data’s use by both for-profit and volunteer organizations and individuals. In the case of
coproductive open data use, structuring the relationship between government and volunteer users
gives public managers more control over service design while leaving delivery to the public (Bovaird,
2007). Similarly, explicitly identifying requirements and structuring rewards and incentives for firms
via a contract reestablishes some degree of control over open data’s use, irrespective of how
imperfect the contract may be. Of course, these advantages come with the tradeoff of giving up the
reduction in transactions and fiscal costs associated with using open data to digitize services.
Whether the benefits outweigh the costs depends on contextual factors that public managers will
need to consider on a case-by-case basis.
CONCLUSION
Open data’s characteristics set it apart from other transparency policies and technologies
because it gives the public the ability to construct information from the raw material of data instead
of receiving pre-generated information from the government. This directly challenges both internal
organizational processes around data storage and sharing, and the institutions that govern how local
governments and the public interact. In this way open data represents the sort of intra-
organizational changes and disintermediation between government and the public that are hallmarks
of Digital Era Governance. These changes offer both promise and peril for pubic organizations.
Open data can improve both internal and external organizational accountability; stimulate economic
development; and promote the digitization of both new and existing service delivery. At the same
40
time, open data can lead to goal displacement; cost departments control over how digital services are
designed and distributed; lead to organizational and professional embarrassment; and privilege
technocratic elites to the detriment of the public good.
While the ultimate decision to adopt open data in cities is the responsibility of elected
officials or city managers, its implementation falls on the shoulders of department managers and
staff. If they view open data as a threat on account of the loss of control over data use and the
potential for misuse that follows, the rational choice is to delay or inhibit implementation by
whatever means are available. Departments with responsibilities that do not lend themselves to
producing large volumes of machine-readable data as well as those whose data are subject to privacy
or security restrictions are also likely to either under-provide data or sacrifice effectiveness without
adequate administrative support to help identify and enact ways to bridge these limitations. And
even if data with no direct benefit for economic development are made available, they may be
underutilized and/or misused if the public does not understand their value and context. Without
careful attention to implementation, many if not most roads lead to open data systems that are either
tokenistic or anemic beyond those data that can support growth machine politics.
Whether those responsible for implementing open data make the effort to address its risks
remains to be seen. Prior public sector technology adoption and implementation has frequently
failed to live up to expectations, whether on account of tokenistic implementation or bureaucratic
resistance to change (Jun & Weare, 2010; D. F. Norris & Reddick, 2013; Reddick, 2004). In the end,
open data’s success or failure hinges on its implementation, a particularly crucial process for
innovative policies and systems (Damanpour, 1991; Damanpour & Evan, 1984; Moldogaziev &
Resh, 2016; Walker, 2014). Those responsible for implementing open data can take measures to
mititgate risk and improve the likelihood that it is useful for both the government and the public.
41
The first of these measures is acknowledging that departments will differ in their ability to
comply with quotas or other uniformly applied requirements for open data; some will be data-rich
while others will have less to contribute. This acknowledgement should inform implementation
requirements. It can also help senior managers identify where to focus additional administrative
support and resources for departments’ efforts to comply with and contribute to the city’s open data
system. Managers should also work closely with department staff to make sure that data are well-
curated and are accompanied by metadata that contextualize their meaning and purpose. Finally,
taking a proactive approach to guiding how data are used by external actors by either specifying
requirements as part of a contract for application development or through establishing relationships
with third parties interested in using open data can help ensure that data are used properly and in
ways that promote equity and the public good.
It is worth noting that this last recommendation appears to be at odds with one of the
principal benefits of open data, namely the value of disintermediation as a way to promote public
engagement at almost no cost to the government. What this means in practice is that open data does
not, in fact, offer a free lunch for public managers. The price of engagement-via-disintermediation is
the risk of both misuse and privileging the technologically savvy and wealthy. If these risks are
unpalatable, the only meaningful solution is to make some investment in outreach and engagement
to help inform and shape data usage. This can come in the form of direct engagement with
individuals, groups, and nonprofit and private organizations. But it can also include improvements
to existing processes around data documentation and management, which carry positive externalities
for organizational performance beyond open data. These suggestions are, of course, not panacea:
change is always accompanied by risk. Identifying these risks as early as possible and adapting to
them when implementing open data gives local government the best shot at creating a system that is
a valuable resource for everyone involved.
Open Data Implementation in US Cities
Matt Young
Ph.D. Candidate, Public Policy and Management
Sol Price School of Public Policy
University of Southern California
Abstract
Since President Obama signed the open government initiative in January 2009,
governments across the United States at every level have adopted and implemented
open data systems. The motivations for open data vary: it is expected to improve
transparency and engagement, reduce administrative costs, and support performance
management systems in a similar manner to previous e-government initiatives
(Petychakis, Vasileiou, Georgis, Mouzakitis, & Psarras, 2014; Thorsby et al., 2016;
Tolbert & Mossberger, 2006; Zuiderwijk & Janssen, 2014). This paper examines
implementation of open data policies at the department level within and across city
governments, and analyzes the institutional, organizational, and supply- and demand-
side factors that promote or inhibit implementation as measured by the number of
datasets made available that meet the definitional standards for open data. The
findings suggest that implementation is influenced by department type,
administrative capacity, city characteristics, and demand-side pressures from
residents.
43
INTRODUCTION
The open data movement is a relatively new phenomenon, starting in the United States in
2009 as one of the first directives issued by the Obama administration. The premise of open data is
to establish policies that make administrative data available to the general public. Open data is a
recent addition to a rich institutional legacy of transparency policies that includes the federal
Freedom of Information Act and numerous sunshine ordinances in state and local governments
from the progressive era to the present day (Bimber, 2001, 2003; Ganapati & Reddick, 2012; J. A.
Musso & Weare, 2005; Verba, Schlozman, & Brady, 1995). Open data is, however, more than a
transparency policy – it is also an e-government system involving substantive information
technology (IT) use to produce the policy’s desired outcome. History suggests that e-government
innovation is fraught and often incomplete, with little appreciable impact on municipal
governments’ functions (Coursey & Norris, 2008; Fountain, 2004; Jun & Weare, 2010; D. F. Norris
& Moon, 2005; Reddick, 2004; West, 2005). Without careful implementation by departments
responsible for generating or collecting data, open data systems will be tokenistic: if the system
provides no little to no useful content, open data’s benefits will not be realized.
This essay contributes to the literature on public sector innovation by empirically assessing
the institutional and organizational factors that shape open data’s implementation. Implementation
is an under-studied aspect of the innovation adoption process (Moldogaziev & Resh, 2016). The
innovation adoption process includes three phases: initiation, when the innovative idea is either
generated or discovered; decision, when the innovation is formally adopted; and implementation,
when the innovation is put into use by the organization and its staff (Walker, 2014). Much of the
empirical literature on public sector innovation examines factors that affect the decision to adopt
and how those decisions diffuse across organizations (Hansen, 2011; Jun & Weare, 2010;
Moldogaziev & Resh, 2016; Walker et al., 2011). These factors are generally recognized as the
44
innovation’s characteristics, the organization, and the organization’s environment (Borins, 2001;
Damanpour & Schneider, 2006; Fernandez & Wise, 2010; Walker, 2008, 2014). Less is known about
the implementation phase, despite its status as arguably the most important part of the adoption
process (Damanpour, 1987, 1991; Walker, 2014). Without effective implementation, innovations
cannot affect substantive change in processes, outputs, or outcomes. Studies that focus on
implementation frequently employ perceptions of implementation as measured through surveys
administered to managers (Ganapati & Reddick, 2014; Welch et al., 2016; Zuiderwijk & Janssen,
2014) rather than observations of implementation processes or outputs.
This study exploits open data’s nature and the functional separation of municipal
government by departments to observe and assess the factors that shape its implementation. In
particular, this study uses a count of discrete files made available through open data sites by
individual departments. The strength of this approach is that the public-facing nature of open data
sites allows for directly observable measures of implementation. Because open data allows outsiders
to observe data made available by departments, it offers a way to measure implementation across
both departments and cities through revealed preferences rather than managerial perceptions. This
approach helps to eliminate the potential for own source bias present in survey data (Meier &
O’Toole, 2013). The readily-observable nature of open data makes it a particularly useful subject for
theory generation and testing in addition to its topical relevance for public managers and citizens
alike. This research also illuminates the process of organizational change, because while the decision
to adopt open data is made at the organizational level, its implementation relies on both supra-level
institutional factors and department-level decisions on what – if any – data are made available for
outside use.
The next section of this essay discusses the open data movement and positions it within the
literature on public sector innovation. This context is then used to generate hypotheses about the
45
factors that should affect open data’s implementation. The next section explains the data and
methodology used to test the hypotheses, followed by the analytic results. The essay then concludes
by discussing the implications of this analysis and directions for further research.
OPEN DATA IN CONTEXT
Open data systems are hybridized, internal-facing information technology (IT) systems and
external-facing “e-government” interfaces. Within the government’s IT system, they require
constituent bureaus or agencies to supply data previously ‘siloed’ within the agency to a central
repository. These data are in turn made available for free to the public through the internet, most
often through websites but also through Application Programming Interfaces (APIs) that allow
applications (e.g., smartphone ‘apps’) to dynamically request and retrieve information. Open data are
provided in “machine-readable” format, defined by OMB (2014) as “[f]ormat[ted] in a standard
computer language (not English text) that can be read automatically by a web browser or computer
system. (e.g.; xml).” A PDF or Word document, for example, would not be considered machine-
readable by this standard because in addition to the requirement for proprietary software from
Microsoft or Adobe, machines cannot easily interpret the information contained within them. This
distinction is crucial, because open data systems are explicitly designed to facilitate external actors
interpreting large amounts of public data to create meaningful digital information.
Open data as policy in the United States began on President Obama’s first day in office
when he signed a memorandum on transparency and open government. It declared that his
administration was “committed to creating an unprecedented level of openness in Government”
through “…a system of transparency, public participation, and collaboration” (Obama, 2009). That
December, the Office of Management and Budget (OMB) issued an open government directive to
46
all executive departments and agencies. It created a federal open data site, data.gov
7
, and set
timetables for agencies to publish previously-unavailable datasets for public consumption. It also
called for agencies to create their own open government webpages with implementation plans for
open government programs specific to their organizations, as well as requirements for data quality
improvements. This directive only applied to federal agencies within the executive branch, but it
served as a clarion call to action for others. Within two years, open data sites were operational at
every level of government across the US. In early 2017, however, the newly established Trump
administration shuttered the White House open data site, data.whitehouse.gov, though as of June
2017 the at-large federal government site, data.gov, remains open. The motives behind the Trump
administration’s decision remain unknown.
Proponents of open data argue that it has three principal benefits. First, as with many e-
government technologies it is meant to increase transparency and thus increase public trust
(Baldwin, Gauld, & Goldfinch, 2012; Dawes, 2010; Kim & Lee, 2012; J. A. Musso & Weare, 2005;
Tolbert & Mossberger, 2006). Second, open data advocates argue that open access to data makes
government more efficient (Petychakis et al., 2014; Thorsby et al., 2016). Implicit in this argument is
the idea that those outside of government are able to use public data more efficiently than those
within, presumably on the grounds that the private sector possesses greater technical competency
and is more efficient because it faces market pressures. Third, open data’s proponents argue that
government data constitute a public good that can promote economic development through mobile
application creation and other marketable goods by the private sector (Thorsby et al., 2016;
Zuiderwijk & Janssen, 2014).
Economic development arguments for open data hinge on two premises. The first premise
is that the data collected by public organizations possesses use value that can be effectively
7
These are often referred to as ‘portals;’ this work will refer to them hereafter simply as ‘sites.’
47
monetized in some way. The second premise logically follows from the outsider-efficiency premise
used in the efficiency argument. If there are uses for public sector data that can meet market needs
so that cities with open data are more competitive in attracting and promoting new and existing
technology companies, then giving private actors unfettered access to that data will reduce
uncertainty and search costs and lead to local investment. The diffusion of contracting out for
service provision among cities and other governments over the past three decades makes this
premise familiar and possibly attractive to public managers.
Open data challenges the status quo by requiring managers to make their data available to
outside observers – including not just for-profit firms but voters and activists as well. The innovative
potential for open data systems lies less in the system’s technical components themselves than in the
implementation process, through which existing administrative and technological processes may be
substantively modified, refined, or abandoned. This potential, however, may be viewed negatively by
risk-averse public managers (Wang & Feeney, 2016; Wood, Bernt, & Ting, 2009). These challenges
are not unique to open data. They share commonalities with other innovations in general, and public
sector innovations in particular. Considering open data through this theoretical lens is useful for
understanding the factors that may promote or inhibit its implementation in practice.
Open Data as Innovation
This research uses the definition of innovation proposed by Rogers (2010), namely that
innovations is “an idea, practice, or object that is perceived as new by an individual or other unit of
adoption” (De Vries, Bekkers, & Tummers, 2015; Rogers, 2010). Prior research has also identified
two distinctive temporal features of innovation. First, adopting such ideas, practices and objects is
only innovative the first time it is adopted by a given organization; subsequent adoptions by the
same organization in different contexts are not considered innovative behavior (Borins, 2000).
Second, innovations differ from incremental change to the extent that they represent a
48
discontinuous rather than incremental break from the status quo (L. Brown & Osborne, 2013).
Open data meets both criteria. The decision to adopt an open data policy is made once at the city
level, and it is a distinct change from prior transparency regimes by requiring raw data rather than
information.
The difference between data and information is subtle but important. Information consists
of data that have been given structure and meaning by the information’s creator. When the creator is
a public administrator within the department that created or collected the data, that structure and
meaning is informed by both personal and institutional knowledge, expertise, and perspective
(Ackoff, 1989; Rowley, 2007). When an outside actor downloads these data and analyzes or uses
them to create a product or service, the public agency providing the data cedes control over the
information’s structure and meaning. Relinquishing control has several consequences. One is that
data are much more versatile and useful for those with the technical expertise necessary to use them
to create information. Another consequence is that those without expertise, or with preferences that
run counter to the supplying agency, are free to misuse the data (Nam, 2015; Williams, 2009).
Relinquishing control of data also risks the abdication of responsibility for democratic engagement
on the part of government; agencies may come to see the release of data as the boundary of their
responsibility to the public, with outside actors responsible for the ‘last mile’ of adapting data for
civic engagement. This broad increase in the potential for both use and misuse value associated with
open data sets it fundamentally apart from prior transparency regimes.
Innovation can occur in various ways. Conceptual innovations include introducing new ways
of viewing or framing a problem or situation (De Vries et al., 2015). Service (or Product) innovations
are the creation of previously non-extant goods or services (Damanpour & Evan, 1984; Damanpour
& Schneider, 2009). Governance innovations represent changes not only in what an organization
produces, but also in the relationship to means of production themselves (Moore & Hartley, 2008).
49
Process innovations are those that affect the manner in which an organization carries out the activities
necessary to generate the goods or services they provide (Damanpour, 1991; Walker, 2014). Process
innovations are further categorized into two subtypes: administrative and technological. The former
deal with changes to the management functions or work processes of an organization. The latter
include technologies introduced to alter existing or create new processes within the organization –
such as frequently-studied e-government technologies (De Vries et al., 2015; Jun & Weare, 2010;
Walker, 2008, 2014). Finally, organizational theory distinguishes between intra- and inter-
organizational, or ancillary innovations. Ancillary innovations differ from others because they affect
the way in which the organization interfaces with its environment, and are therefore dependent to
one degree or another on factors outside of an organization’s immediate control for their success or
failure (Damanpour, 1991; Walker, 2008).
Open data can be classified as process innovation, but one that spans both the
administrative and technological subcategories. Beginning with the federal open data policy laid out
by the Obama administration and continuing throughout its diffusion across other levels of
government in the United States, open data has involved substantive changes to information
administration and management. This includes creating one or more new senior-level management
positions (e.g., Chief Data Officers), creating new committees tasked with formulating
implementation plans, and often requiring constituent agencies and departments to designate staff
for assignment as ‘open data liaisons’ that report directly to these new positions.
But open data is also fundamentally technological. The principal difference between open
data and preceding transparency-related policies and e-government systems are its data structure
requirements. This difference has important consequences for the recipients’ sense-making
processes. Instead of information presented in a framing developed by elected officials and public
sector employees, end-users are given data from which they may construct information to suit
50
personal, organizational, political, or market needs. For example, a parks and recreation department
could use the city website to post a map with the location of various amenities for residents and
visitors to download and print. When the same data used to create such a map were made available
through the City of San Francisco’s open data site, private corporations used it to create an app
where users can search for parks according to various criteria (location, amenities, etc.), navigate to
the park of their choice using GPS, and volunteer for or donate to the city’s parks and recreation
department (see http://sfrecpark.org/san-francisco-recreation-parks-official-mobile-app/).
Open data is used in this way by design; it is an ancillary innovation that also involves
substantive intra-organizational administrative changes. Open data requires substantive,
discontinuous changes to both information management and the technology with which it is made
available, but its ultimate goal is access to and use of these data by external actors. Thus, open data’s
very nature presents a dilemma for the public managers charged with its implementation: a promise
of improved trust from residents, increased efficiency, and economic growth on the one hand and
the peril of being supplanted by the private sector or facing heightened scrutiny from hostile
external constituents on the other (Williams, 2009).
Factors Affecting Implementation
Numerous factors influence a department’s ability to implement open data. These include an
array of organizational factors, such as the motivation of departmental leaders responsible for
implementation, and whether the department’s tasks and responsibilities naturally lend themselves to
data generation. Innovations such as the open data movement require organizational changes that
may be perceived as risky in the public sector (Ae Chun et al., 2012; Borins, 2001; Damanpour &
Schneider, 2009; Klein & Sorra, 1996). In the face of these risks, the decision to innovate is often
motivated by concerns about organizational problems that warrant risk-taking to improve efficiency
and/or effectiveness (Damanpour, 1987, 1991; De Vries et al., 2015; Walker, 2008, 2014). However,
51
research has shown that motivations for innovation in public organizations can also include
organizational factors such as employee public service motivation, leadership quality, and
perceptions of organizational legitimacy (Jun & Weare, 2010). In open data’s case, it is likely that its
perceived value will vary across managers and departments. Some may see open data as an
opportunity to improve constituent relations or reduce the costs of using data to produce new
services or administrative procedures. Others, in contrast, may view releasing data as a threat to their
security, legitimacy, or operations (Wang & Feeney, 2016; Wood et al., 2009).
In short, open data implementation will be affected by several factors. The first set of these
factors has to do with the responsible department’s characteristics. These include whether the
department is service or administratively focused, is likely to have preexisting data that are
appropriate for open data, and whether the department’s core mission is related to economic
growth. A second set of factors includes institutional characteristics of both the department and the
city in which it is nested. These include organization size, administrative capacity, the degree of IT
centralization, the municipality’s form of government, and demand-side pressures from residents.
The rest of this section addresses these factors in turn in order to generate testable hypotheses about
how cities implement open data in practice.
Department Orientation
Interaction between government and the public is both frequent and varied at the local level
(Christensen & Lægreid, 2005; Fung, 2006). But not every department within a municipal
government has tasks that require this interaction. Core administrative departments focus on
logistical and operational support for the government as a whole. Examples include purchasing, risk
management, operations, and human resources departments. Executive management in the form of
mayoral, council, and city manager departments are also administratively-oriented. The argument for
implementing open data in these departments is likely to be primarily driven by external
52
stakeholders’ desires for increased transparency. However, these administrators are likely to be
reticent to share data, as they are often unaccustomed to taking a forward-facing approach to
constituents (Ho & Ni, 2004; Wang & Feeney, 2016; Wood et al., 2009). Furthermore, it may be
difficult for these administrators to identify a public use value for their data, especially compared to
those managing service-oriented departments where interaction with the public is more common.
The public may also have difficulty identifying the value of these data, because they are likely to be
more familiar with the workings of service delivery-oriented departments and may therefore focus
their energy on pressuring them rather than the administrative core.
H1: Administrative departments will make less data available than service-oriented
departments.
Available Data
While all bureaucracies generate some data through their administrative functions, not every
agency can easily measure and quantify their outputs or outcomes (Wilson, 1989). Tasks like paving
streets, arresting criminals, and issuing permits and licenses naturally lend themselves to
measurement and, by extension, representation as data. Departments with tasks that are not as
conducive to data collection should therefore have less data to begin with, complicating open data’s
implementation process (Wilson, 1989). Moreover, departments that are responsible for social
services delivery, where outputs can already be difficult to measure likely face an additional barrier to
open data implementation. Because they serve vulnerable populations, whatever data is likely to be
collected may be sensitive in nature and therefore withheld from open data due to privacy concerns.
For departments that have observable outputs, public managers have found several ways to
put their data to work. Performance management (PM) practices’ broad diffusion has put public
sector output quantification into overdrive, as departments are required to set benchmarks, measure
their progress towards them, and report on the extent to which they were met in a given period
53
(Behn, 2003; Hood, 2012). For example, in San Francisco the Public Works department is expected
to address 90% of all potholes within 72 hours of identification, and 95% of all street cleaning
requests within 48 hours. These and similar evaluations require significant amounts of data – data
that are machine readable and tabular to facilitate analysis. They are also frequently shared with the
public through city websites and other channels (e.g., financial reports) as part of both performance
management and efforts to promote accountability/transparency (Gerrish, 2016; D. P. Moynihan,
2008; Radin, 2006). Returning to the earlier example, a visitor to San Francisco’s city website can
quickly determine whether the department is meeting their goals via user-friendly ‘dashboard’
visualizations. Thus, the presence of observable outputs and the likelihood of preexisting
performance management regimes give service-oriented departments a head start on implementing
open data, since they already tend to be required to divulge data which are likely to qualify as open
data if made available through the city’s site.
H2: Service-oriented departments with more easily observable outputs will make
more data available than those with outputs that are more difficult to observe.
Advancing Growth
Cities place a significant emphasis on economic development, often at other activities’
expense (Peterson, 1981). Departments responsible for tasks directly related to economic growth
tend to have observable outputs both due to the nature of their tasks (e.g., issuing permits and
licenses) and because such data are used to evaluate economic development efforts. Over half of all
respondents to a 2009 ICMA survey on municipal economic development practices said that they
had explicit performance management programs in place. But beyond having preexisting institutions
around data collection, open data presents real potential for these departments to advance their
mission. Making economic development data easily available in open data format drastically lowers
search and other transaction costs for businesses, developers, and other investors. This in turn may
54
facilitate their decision to establish new units within the city, or expand their existing presence.
Given the premium that municipal public managers place on economic development as well as their
predisposition to have large data archives that meet open data standards, these departments should
have more data available on average than all other department types.
H3: Service-oriented departments involved with economic development and growth
will make more data available than other departments.
Table 2.1 summarizes how the characteristics described above map to different types of
service-oriented departments across municipalities. These mappings do not – cannot – account for
every discrete subdivision or task for a given department. Instead, they are meant to classify
departments broadly along the dimensions discussed earlier. Budget and finance departments have
readily observable outputs: the city budget and fiscal positions. Health and human services
departments include social service and public health departments. They have difficult-to-measure
outputs, often deal with sensitive information, and are redistributive rather than growth-focused
(Peterson, 1981; Wilson, 1989). Economic and community development includes economic
development departments, community development departments, housing authorities, and
departments responsible for convention centers, small business resources, and the like. They are
growth-focused departments with observable outputs such as licenses issued, units built, etc. Public
works and utilities departments have easily identifiable outputs. Departments responsible for parks,
recreation and cultural activities have some outputs that are easily measured, but they are not
principally motivated by economic development. Planning and permitting includes planning and
zoning, code, and permitting departments. These departments have outputs that are easily measured
55
and they serve to promote growth and economic development
8
. Public Safety includes police and
sheriff departments, fire and emergency medical services/departments, emergency management, and
homeland security. Their outputs are readily measurable. Finally, transportation departments have
observable outputs in the form of station/bus stop counts and geolocations, schedules, ridership
rates, ratios of on-time to delayed trips, etc.
Table 2.1. Mapping Service-Oriented Department Types to Factors Affecting
Implementation
Organizational Size
Organizational size and administrative capacity are associated with the likelihood of
innovation adoption and utilization in general (Damanpour, 1991; Fernandez & Wise, 2010) and
process innovation adoption in particular (Walker, 2014). Organization theory predicts a positive
8
Some might argue that planning and zoning can just as easily work against the urban growth machine. But whether
their actions are pro- or anti-development, the data that these departments generate are useful for both retrospective and
prospective analysis for developers, businesses and other potential investors.
Department Type
Easily Observable
Outputs?
Growth
Focused?
Predicted Level of
Implementation
Budget and Finance Yes No Medium
Health and Human Services No No Low
Economic and Community Development Yes Yes High
Public Works and Utilities Yes No Medium
Parks, Recreation, and Culture Yes No Medium
Planning and Permitting Yes Yes High
Public Safety Yes No Medium
Transportation Yes No Medium
Table 1. Mapping Service-Oriented Department Types to Factors Affecting Implementation
56
relationship between organization size and innovation. Larger organizations have more resources at
their disposal, tend to be more complex, and are more capable of absorbing the risk and shocks
associated with innovation adoption (Damanpour, 1991). Alternatively, public choice theory
suggests that size is negatively associated with innovation due to inefficiencies that make them
unable to respond quickly to change (Walker, 2014). Empirical evidence on public sector innovation,
however, lends support to the positive association between organizational size and innovation
(Damanpour & Schneider, 2006, 2009; Fernandez & Wise, 2010; Walker, 2014; Walker et al., 2011).
City governments should be subject to the same returns to scale as their constituent departments.
Larger cities should be able provide more resources and weather more risk than smaller ones.
H4: Organization size is positively associated with the amount of open data
departments make available.
Administrative Capacity
Administrative capacity is the amount of managerial time and experience that an
organization has at its disposal. Increased administrative capacity is theorized to provide the ability
to cope with new realities, responsibilities, and other changes in the organizational environment by
providing leadership, developing strategy, and overseeing task execution (Damanpour, 1991;
Fernandez & Wise, 2010). This capacity exists at both the department and city level. As with
organizational size, empirical evidence for innovation adoption in the public sector supports this
association (De Vries et al., 2015; Walker, 2014).
H5: Administrative capacity is positively associated with the amount of open data
departments make available.
While department characteristics are undeniably important, they are not standalone
organizations. Rather, they are specialized units tasked with executing particular duties to support
57
the city’s broader mission. They are overseen by higher-level managers, as well as elected and
appointed officials. Furthermore, the decision to adopt open data policies is made at the city level,
and it is likely that differences in managerial and administrative pressure on departments will lead to
differential open data implementation rates. Key city-level institutional characteristics that are likely
to affect open data implementation include their form of government, information technology
centralization, and whether demand for open data exists on the part of city residents and
stakeholders.
Form of Government
Government form has been shown to affect the rate and level of innovation adoption in US
cities (Nelson & Svara, 2011). Council-manager governments have been associated with increased
levels of innovation and e-government system adoption, possibly owing to an increased level of
professionalization (Walker et al., 2011). A plausible counter-argument is that cities with strong
mayors could face additional political pressure to make more data available across departments
(Pardo, Cresswell, & Thompson, 2001). But political pressure is an uneven tool for promoting
innovation implementation, with mixed empirical results (Choi & Chandler, 2015; Wood et al.,
2009). The relative stability and professionalization associated with council-manager governments
increase the chance for genuine innovation implementation (Rivera, Streib, & Willoughby, 2000;
Streib & Willoughby, 2005).
H6: Cities with council-manager forms of government will make more open data
available across departments than strong mayor cities.
IT Centralization
Research has demonstrated that public organizations with centralized IT departments tend
to have higher e-government system adoption rates (Moon, 2002; Moon & Norris, 2005; D. F.
58
Norris & Moon, 2005; D. F. Norris & Reddick, 2013; Tolbert, Mossberger, & McNeal, 2008). A
citywide IT department makes it more likely that IT systems interface and communicate both within
and across departments, potentially reducing technical implementation barriers. Centralized IT may
also help city administrators circumvent departmental objections to open data if the IT department
hosts and maintains – and therefore has access to and control over – the data systems used by
departments across the city.
H7: Cities with centralized IT departments will make more open data available across
departments than those without.
Demand-Side Pressures
Finally, demand-side factors are also likely to drive open data implementation in cities
(Grimmelikhuijsen & Welch, 2012; Wang & Feeney, 2016). Prior work on public sector innovation
has shown that higher socioeconomic status and levels of wealth among residents are associated
with increased innovation adoption rates in US municipalities (Walker, 2014). This pressure is most
likely to come from residents with the technical skills to use and make sense of open data and/or a
business use for it. There is no clear relationship between racial and ethnic composition of a city and
demand for innovation (Walker, 2014).
H8: Cities with higher proportions of professionalized workers will make more open
data available across departments.
In conclusion, the relative rate of open data implementation is likely associated with whether
a department is service- or administration-focused; its level with output observability; its size; and its
administrative capacity. City-level characteristics likely also affect implementation rates, including
government form and size; administrative capacity; IT infrastructure and process centralization; and
demand side pressures for innovation approximated by resident demographic characteristics.
59
DATA AND METHODS
This study considers departments within US cities with populations of 100,000 or more
9
that
have adopted open data policies. Cities with open data were identified in several stages. A baseline
list was developed from preexisting databases identifying cities with open data from both the federal
government at data.gov and Sunlight Foundation, a nonprofit focused on government transparency.
This was augmented with (1) client lists from firms that sell open data technology solutions to public
organizations; (2) brute force searches of city website sitemaps; and (3) a series of Google queries for
combinations of city government names with terms related to open data (e.g. ‘open data’ ‘data
portal’, etc.). To qualify as an open data site, at least two different data types (e.g., shapefiles,
spreadsheets) had to be available in a format meeting open data’s definitional standards.
As a result, the dataset excludes sixteen cities that only make GIS shapefiles available and
nine other cities whose sites did not meet open data standards. The search identified 76 out of 274
cities (28%) as having an open data site. Among the 274 cities sampled, the 76 cities with open data
differed demographically from the 198 that did not have open data along several lines. Cities with
open data tend, on average, to be larger and denser, with more college-education residents employed
in professional settings (defined by the Census as business, sciences, and the arts) compared to cities
that do not have open data. Cities with open data also tend, on average, to have a greater proportion
of non-Hispanic Black residents, and a lower proportion of Hispanic residents, compared to cities
without open data. Table 2.2 contrasts the demographic characteristics for cities with and without
open data.
9
As of the 2010 census
60
Table 2.2. Difference in Means Test for Demographic Characteristics
For Cities With and Without Open Data
Data
The data used in the multivariate analysis presented here are drawn from cities that either
developed their open data site entirely in-house, used an open source API standard called CKAN,
10
or contracted with a private technology vendor for open data site development and maintenance,
and their constituent departments.
Outcome Variable
Counts of available open data sets represent the implementation level. Data collection
involved systematically scraping each city’s open data site using scripts developed in the Python
programming language. A count of total files represents a first-order approximation of open data
10
https://ckan.org/
Demographic Characteristic
Mean,
Cities with
Open Data
(n = 78)
Mean,
Cities without
Open Data
(n = 198) t-value
Total population 601781 200553 -3.30
**
Population density 5146.22 3867.37 -2.29
**
Median household income 1518.64 1218.05 0.99
Proportion college educated 36.09 28.74 -4.61
***
Proportion employed in business, science, and arts 39.90 34.52 -4.37
***
Proportion of civilian labor force unemployed 6.72 6.60 -0.48
Proportion of population non-Hispanic Black 21.06 15.41 -2.35
*
Proportion of population Hispanic 19.66 26.14 2.79
**
Proportion of population Asian and Pacific Islander 8.49 7.24 -1.03
* = p < 0.05 | ** = p < 0.01 | *** = p < 0.001 (two tailed value)
61
implementation, in that it is the most straightforward output measure that can be observed (Hill &
Hupe, 2002). While this limitation is discussed in more detail later, the number of files is also
the principal standard by which departments at all levels of government, including municipalities,
are currently evaluated (see data.gov/metrics). The count excludes nontabular data and other files
(e.g., PDFs) that do not meet the definitional standard of open data. It also excludes observations
that cannot be attributed to a department; duplicate files that contain the same data points; and
datasets that originated from outside of the department or city (e.g., federal or state government,
nonprofits, etc.).
Independent Variables
To test the hypotheses about departments’ open data implementation, a categorical
variable, department type, clusters departments by type to account for the fact that the names of
departments with the same general set of tasks and responsibilities vary considerably across
jurisdictions. For example, a department responsible for urban planning may be called ‘planning’ in
city A, ‘community planning’ in city B, ‘planning and design’ in city C, etc. Table 2.3 presents the
departmental types, an example of a department within the typology, and the number and
proportion of departments clustered within them.
62
Table 2.3. Department Types, Examples, and Proportions
Department characteristic variables include department size and administrative capacity.
Data for these variables were collected individually from each city’s budget and, where available,
databases of city employees by department and job title. The number of full-time equivalent (FTE)
employees in a given department in 2015 is used to measure department size. This approach avoids
overweighting departments responsible for large-scale capital resources and projects. However, the
large number of workers necessary in public safety and public works departments makes them
outliers. To correct for this problem, the variable is log-transformed. Administrative capacity is
measured as the ratio of managers to the total full-time-equivalent employees (FTEs) by department
(Fernandez & Wise, 2010). A position is flagged as managerial if its title consists of or includes
‘Director’ or ‘Manager.’ These are further refined to exclude positions that include these terms but
do not have managerial authority over a given department or its constituent divisions, e.g.
Department Type Example Departments N
Proportion of
All Departments
Administration HR; IT; Mayor's Office 667 42%
Budget and Finance Finance 119 8%
Health and Human Services Public Health; Veterans Services 84 5%
Economic and Community Development Economic and Workforce Development 139 9%
Public Works and Utilities Public Works; Austin Energy 126 8%
Parks, Recreation, and Culture Youth Recreation 134 8%
Planning and Permitting Planning; License Commission 99 6%
Public Safety Police 161 10%
Transportation Traffic Parking and Transportation 50 3%
1579 100%
Note: 10 cities in the sample did not have transportation departments.
63
office/case/program/project/system managers, and assistants to department directors and
managers. In the case of police departments officers at the rank of Captain or above are considered
managerial. Battalion Chiefs and above are flagged as managerial for fire departments.
Turning to city-level characteristics, form of government is measured via an indicator
variable for whether the city uses a council-manager form of government. Fifty-two percent (32 out
of 60) of cities in the sample are council-manager cities. Government size is operationalized as the
ratio of total city FTEs in 2015 per 1,000 residents. As with department size, this variable is log-
transformed. An indicator variable is included to identify cities with centralized IT departments.
City-level administrative capacity is operationalized via an indicator for whether a city has either a
Chief Data Officer or Chief Innovation Officer. These positions were initially created as part of the
open data implementation process; San Francisco was the first city to create a Chief Data Officer as
a revision to its open data policy. Their responsibilities include city-wide oversight of data-driven
and technological innovations, including managing the adoption and implementation process.
Whether a city contracted out for developing their open data architecture is also measured via an
indicator. Finally, the model includes a dummy variable to control for whether the city contracted
out with a private firm for their open data system.
Demographics are drawn from 2010-2014 ACS 5-year data. The proportion of residents
employed in management, business, science, and the arts is included to model potential demand-side
pressures from residents who are likely to possess the skills necessary to make use of open data and
for whom it may be beneficial for personal or commercial use (e.g., the subset of workers employed
in computer and engineering jobs)
11
. Several other demographic variables are included as controls.
These include population and its proportional breakdown by racial and ethnic groups (with
nonwhite Hispanics serving as the base case), population density, median household income in 2014
11
Educational attainment is not included because it is nearly perfectly correlated with this measure (c = 0.97).
64
dollars, and the unemployment rate. Population is log-transformed. The model includes a
polynomial (squared) form of median household income. The final sample includes 1,579
departments in 60 cities. Table 2.4 provides summary statistics for all continuous variables.
Table 2.4. Descriptions, Means, Standard Deviations, and Observations
for All Continuous Variables
Model Specification
While open data implementation falls to department managers and staff who must supply
data, individual departments share certain institutional characteristics in that they are nested within
the same city government. A simple solution to this clustering would be to include a fixed effect that
soaks up unobserved city-level characteristics. However, this approach does not allow the model to
produce estimations for observed city characteristics with theoretically or empirically predicted
Variable Level of Analysis Description Mean SD N
total files Department Number of open data files published 6.2 25.4 1579
department size Department Number of full-time employees (FTEs) 446.0 2893.6 1579
admin capacity Department Proportion of managers to total FTEs 9.6 12.3 1579
city size City Total FTEs per 1,000 residents 16488.4 66390.5 60
population City Total population 648143.8 1202150.4 60
population density City Population density 5544.7 4703.2 60
income City Median household income 52393 14246.2 60
% professional City Proportion employed in business, science, and arts 40.1 10 60
% unemployed City Proportion of civilian labor force unemployed 6.8 2 60
% black City Proportion of population non-Hispanic Black 21 18.1 60
% hispanic City Proportion of population Hispanic 20 15.7 60
% api City Proportion of population Asian and Pacific Islander 9 10.1 60
Note: All dollar figures are in thousands of 2014 inflation-adjusted dollars.
Department size, city size, and population are log-transformed for analysis.
65
implications for open data implementation. This paper adopts a hierarchical count modeling strategy
to explicitly account for the fact that departments are nested within cities while minimizing
correlation between city- and department-level variable standard errors (Raudenbush & Bryk, 2002).
Because the dependent variable is both a count variable and overdispersed (M = 6.2, S
2
= 644), the
model is specified as a two-level negative binominal regression. A random effects parameter is
included at the city level to account for unobserved heterogeneity between them. Furthermore,
because cities adopted open data policies at different times, departments in early adopter cities have
had more time to make files available, introducing potential estimation bias. The model includes an
exposure parameter to correct for this effect; it is operationalized as the numbers of days that a
given city’s open data site has been online up to the date the data were collected for analysis.
Departments of education are excluded from the model, both because they are frequently quasi-
autonomous and their size – especially in Boston and New York at ~ 13,000 and 250,000
respectively – makes them extreme outliers. The final sample includes 1,579 departments nested
within 60 cities (N = 60, n = 1,579).
RESULTS
Table 2.5 reports the estimates from the two-level negative binomial model. The likelihood
ratio test comparing the results to a pooled estimator is significant at the p < 0.001 level, suggesting
that overdispersion is present and the negative binomial specification was appropriate. The model
coefficients are reported as incidence risk ratios (IRRs).
12
All discussion of coefficients is conditional
on holding their covariates constant. Figures 2.1 and 2.2 provide a graphical representation of these
results by city- and department-level variables, respectively. The incidence rate ratios are expressed
as a horizontal bar chart centered on 1 (i.e., no change from the baseline incidence rate). Bars to the
12 This is equivalent to exp(β), and in this model is equivalent to the relative proportion of files made available over a
given period. In this case, the period is log-transformed with a coefficient constrained to 1 by the model to control for
the age of each city's open data site.
66
left of 1 represent a lower incidence rate for a given variable; bars to the right of 1 represent a higher
incidence rate. Each bar includes 95% confidence intervals. Statistically significant variables have
blue-colored bars and the significance level associated with their IRR value (* = p < 0.05; ** = p <
0.01; *** = p < 0.001).
The results offer mixed support for the first two hypotheses regarding department type.
With the exception of parks, recreation, and culture, service-oriented departments as well as budget
and finance departments all have IRRs in excess of one, indicating that on average they all make
more data available than administrative departments. However, several department types – health
and human services; parks, recreation, and culture; public works and utilities; public safety; and
transportation – have IRRs that are not statistically different from administrative departments. In the
case of parks, recreation, and culture-type departments, this may be because their easily measured
outputs only superficially link to outcomes, like total park acreage, the number of programmatic
services offered in a year, etc. These data are likely to exist, but they may be relatively sparse
compared to other departments.
67
Table 2.5. Estimated incident rate ratios from multilevel negative binomial regression
models for the effect of city- and department-level characteristics on the number of datasets
available on a city’s open data site
Independent Variable IRR S.E.
Department Level Variables
department type
budget and finance 1.894
***
0.266
health and human services 1.009 0.182
economic and community development 1.600
***
0.223
public works and utilities 1.189 0.174
parks, recreation, and culture 0.959 0.145
planning and licensing 3.139
***
0.416
public safety 1.202 0.166
transportation 1.363 0.263
administration (ommitted as reference)
department size (logged) 1.440
***
0.045
administrative capacity 1.011
**
0.004
City Level Variables
city manager 1.175 0.166
chief data or innovation officer 1.050 0.129
centralized IT 1.226 0.209
private vendor for open data site 1.278 0.197
city government size (logged) 0.913 0.131
population (logged) 1.078 0.086
population density 1.000
*
0.000
median household income (thousands of dollars) 0.863
***
0.029
median household income (thousands of dollars, squared) 1.001
***
0.000
proportion of workforce in business, science, and the arts 1.021
*
0.011
proportion unemployed 1.118
*
0.059
proportion nonhispanic black 0.976
***
0.005
proportion hispanic 1.002 0.006
proportion asian and pacific islander 1.002 0.007
constant 0.000
***
0.000
ln(period) 1 (exposure)
χ
2
374.59
***
* = p < 0.05
** = p < 0.01
*** = p < 0.001
n = 1,579
N = 60
Note: Reported χ2 value is Wald
Likelihood Ratio test vs. pooled estimates: 98.21 (p < 0.001)
68
Figure 2.1. Estimated incident rate ratios from multilevel negative binomial regression
models for the effect of city-level characteristics on the number of datasets available on a
city’s open data site
Figure 2.2. Estimated incident rate ratios from multilevel negative binomial regression
models for the effect of department-level characteristics on the number of datasets available
on a city’s open data site
1.18
1.05
1.23
1.28
0.91
1.08
1.00*
0.86***
1.00***
1.02*
1.12*
0.98***
1.00
1.00
0.75 1.00 1.25 1.50
City Manager
Adm. Capacity (CDO or CIO)
Centralized IT
Contracted Out
Government Size (logged)
Population (logged)
Population Density (mi2)
Median HH Income (logged)
Median HH Income (squared)
% Workforce Professionalized
% Unemployed
% Non-Hispanic Black
% Hispanic
% Asian and Pacific Islander
1.44***
1.01**
1.89***
1.60***
1.01
0.96
3.14***
1.20
1.19
1.36
0.50 1.00 1.50 2.00 2.50 3.00 3.50
Department Size (logged)
Adm. Capacity (Manager:Staff)
Department Type
Budget and Finance
Econ. and Com. Development
Health and Human Services
Parks, Recreation, and Culture
Planning and Licensing
Public Safety
Public Works and Utilities
Transportation
69
The three other service department types with outputs that are easy to measure are
statistically different from both administrative departments and service departments with less-
measurable outputs. Consistent with the third hypotheses about economic development-oriented
service departments, economic and community development departments make 60% more files
available on average than administrative departments (p < 0.001). Still, economic and community
development departments’ data files are dwarfed by planning and permitting departments, which on
average make 213% more data available than administrative departments (p < 0.001).
Support for the hypotheses on organization size (4) and administrative capacity (5) is mixed
depending on the level of observation. In both cases the department-level measure was significant.
A one percent increase in department size is associated with a 44% increase in the number of open
data files available (p < 0.001). A one percent increase in department administrative capacity –
operationalized as the ratio of senior management to staff – is associated with a 1% increase in the
number of files available (p < 0.01). But at the city level, neither the size of government nor
administrative capacity in the form of a chief data or innovation officer were statistically significant.
The estimates do not support either the sixth hypothesis, that council-manager cities will
have more files than those with strong mayor governments, or the seventh about the presence of
centralized IT departments; neither indicator were statistically different from zero. Finally, the
results support the eighth hypothesis that cities with higher proportions of professionalized workers
tend, on average, to have more data available. A one percent increase in the proportion of
professionalized workers is associated with a 2% increase in the number of open data files available
at the city level (p < 0.05).
The control for contracting versus developing an open data system in-house is not
significant. Several estimates of demographic controls, however, were statistically significant. Median
household income had a ‘U’ shaped relationship to the number of open data files available (p <
70
0.001). The estimated IRR remains below one for all observed values of median household income
(that is to say, increases in median household income are associated with fewer open data files
available for all city departments, on average), but the marginal effect of an additional thousand
dollars of income turns positive at around $70,000. A one percent increase in a city’s unemployment
rate was associated with a 12% increase in the average number of files available (p < 0.05).
Population density was statistically significant (p < 0.05), but its effect has no practical change at the
margin. Population itself was not statistically significant. Finally, while the proportion of Hispanic
and Asian and Pacific Islander residents had no statistically identifiable effect, a one percent increase
in the proportion of African American residents was associated with a 2% decrease in the average
number of open data files available by city (p < 0.001).
DISCUSSION AND IMPLICATIONS
This paper considers open data implementation in city governments; a recent phenomenon
that proponents argue will make radical improvements in public sector transparency, efficiency, and
economic development. It contributes to the literature on innovation adoption in the public sector
by asking and answering the question: how is open data implemented in practice by local
governments? Implementation is operationalized as the number of files made available by city
departments that meet open data standards. Several departmental and city institutional features are
identified as likely to carry implications for observed implementation levels both within and across
city governments. These factors as well as others previously found to affect innovation adoption are
then empirically assessed using a two-level hierarchical count model to account for the fact that
departments are clustered within city governments.
The reported findings suggest higher levels of open data implementation in departments
with tasks related to economic development; some departments with easily observable outputs; and
departments focused on service rather than administration. Administrative capacity as measured by
71
the ratio of managers to staff is positively associated with data availability at the department level,
but the addition of “C-level” executive managers responsible for implementation at the city level had
no discernible effect. Neither council-manager cities nor those with centralized IT functions had
more open data files on average than others. Demand side pressure as measured by the proportion
of professional workers was also positively associated with the number of files on a city open data
site.
This analysis offers the first direct observation of how cities’ open data policies are being
implemented in practice. Several findings are worth noting as possible indications of larger overall
trends. First, the findings provide support for the argument that open data implementation is likely
to be uneven across different types of departments. Increased transparency is one of the principal
benefits ascribed to open data. Yet administrative departments whose data are least likely to already
be in the public domain are systematically underrepresented compared to service-oriented
departments. The argument that administrative departments are lacking data that meet open data’s
definitional standards is not credible on its face. Rather, it may be that these managers and
administrators are skeptical of open data’s value proposition for their departments.
Among service-oriented departments, economic development-focused departments like
planning and permitting dominated in expected open data file counts. This could be indicative of
city leadership using open data to actively promote economic development by lowering search costs
for investment. It could also reflect a prior, general emphasis on data collection for tasks and
activities related to economic development; these departments may simply have the most “ready to
go” data in a given city. This might also serve to explain the higher than expected average number of
files made available by transportation departments. Not only do these departments tend to collect
and generate a large volume of data, they may be particularly motivated to share these data both for
intrinsic reasons and as the result of demand-side pressure from the public. Relatedly, the
72
relationship between output observability and the amount of data made available by service-oriented
departments highlights the difficulties inherent to administrative policies involving data and
measurement in general. As the aphorism goes, “not all that matters can be measured, and not all
that is measured matters” (Eisner, 2002).
One way to frame an implication for this finding is that without making a concerted effort to
facilitate implementation among other department types, public managers risk creating open data
systems that are little more than incremental changes to the status quo. This result, of course, may
not be accidental. One should appreciate that making even something as simple as budget data
available in an analysis-ready tabular format instead of a PDF document is, in fact, a substantive
change.
13
But for public managers interested in substantive change, this distinction is important. To
the extent that open data can change the relationship between government and its constituents,
providing the same data as before but in a different format is at most a first step.
The positive relationship between department size and the number of open data files made
available is unsurprising given the increased scope and resulting data-generating possibilities that
larger organizations have relative to smaller ones. This effect is not mirrored at the city level, which
may illustrate how open data implementation is fundamentally a department-level process. This
perspective is further supported by the significance of administrative capacity within departments
and the lack of any detectable effect from creating “C-level” senior management positions at the city
level. Chief data officers were purpose-hired to manage data in general and implement open data in
particular, yet only relative administrative capacity within a department affects the average rate of
data availability. Future research should examine the development and impact of data and
innovation officer positions in more detail.
13
The functional difference between PDF and machine-readable budget data was highlighted in this project, which
required combing through the budgets of 49 cities for the number of FTEs and managers in each department.
73
Cities with more workers in professional fields – which include software development and
computer engineering – make more data available on average. It is unlikely that this association
exists due to the higher socioeconomic conditions associated with professional occupations, as the
number of files available was negatively associated with median household income and positively
associated with the unemployment rate.
One explanation is that all three factors may indicate demand-side pressures for data for two
separate uses. Professional workers (and those who employ them) are more likely to possess the
technical capacity to use raw data in generating information and services and products that automate
that process. Commonly found examples of this phenomenon are mobile apps that make it easier to
use public transit through real-time scheduling updates and functions that help plan transfers
between lines, modes, and even providers. This form of demand-side pressure is also likely to be
greeted enthusiastically by public managers interested in economic development or, as with transit
agencies, facilitating service use, leading to a virtuous cycle of increased data availability. In the case
of household income and unemployment, it may be that residents in cities with poor economic
conditions are generating demand-side pressure for increased transparency and accountability
and/or resources for economic development from “underperforming” city governments. At the
same time, it may be that poor economic conditions prompt public managers to make more data
available in the hopes of spurring growth and/or legitimation in the face of public displeasure.
A negative relationship between the proportion of Black residents and the number of open
data files available is perhaps the most vexing result from the model. That cities with high
proportions of Black residents make less data available on average may be indicative of racial biases
in transparency and innovation implementation in local government. Further research should
explicitly focus on the potential equity implications of open data’s adoption and implementation. As
the opportunity gulf between those proficient with technology and data continues to widen,
74
increasing opportunities for disadvantaged groups to use and engage with public data could act as a
plank in the needed bridge.
One limitation of the analysis is that the dependent variable is a first-order approximation of
open data implementation. Estimating the volume of files available in open data format is a first step
towards differentiating open data sites from preceding information sharing regimes like FOIA
requests or more localized sunshine ordinances, but not all data are created equally. Not only can
they vary substantially by size, but also the quality of the data – in terms of their fidelity,
completeness, and use value – is unlikely to be consistent across all files. A count model, while it has
a high degree of reliability, cannot account for these characteristics. Future research should consider
applying measures of data quality used in the study of information systems and science to the study
of open data.
Managerial perceptions of open data – its value, opportunities, threats, etc. – are not
accounted for in this analysis. While the dependent variable assures against issues of own-source
bias, the attitude of public managers towards open data is likely to play a large role in their decision-
making around its implementation. Ganapati and Reddick (2014) use data from a survey of city and
county chief administrative officers, but data on department-level manager perceptions of open data
is sparse. Jackman and Young (2017) take a first step in surveying all staff in the City of Los Angeles
on their perceptions of both open data and performance management. Unfortunately however,
these data do not yet exist for all cities with open data policies, which limits external validity.
External validity is also constrained to relatively large cities. Some mid-sized cities have open
data systems in place, but the study’s sampling strategy limits its generalization to those with 100,000
or more residents. Another constraint is the model’s inability to properly account for the time-
varying characteristics of implementation. Despite controlling for the length of time that a given
city’s open data site has been available for departments to contribute to, the current model cannot
75
assess how institutional changes over time affect departments’ implementation rates. Examples of
possible institutional changes include merging two previously independent departments as a cost
control measure, or the lag between open data policy adoption and the city creating new managerial
(e.g., CDOs) positions dedicated to implementation. Future analysis will employ a longitudinal
approach to model these discontinuities.
Nevertheless, these findings carry practical importance for public managers. While citywide
professionalization, as proxied by the use of a council-manager form of government, is associated
with increased implementation overall, neither IT centralization nor the creation of executive-level
managers appear to have an appreciable effect. Instead, the results suggest that those looking to
promote implementation should focus their efforts directly at the department level. The positive
association between department size and administrative capacity suggest that implementation is in
part a function of department resources. This is likely to be especially true for smaller departments
without observable outputs like health and human services. Holding these departments to the same
standards as others – especially economic development-oriented departments – could set them up
for failure and even lead to negative perceptions of open data by chastised managers and staff.
The Medium is the Message:
Innovation in Municipal Coproduction Systems
Matt Young
Ph.D. Candidate, Public Policy and Management
Sol Price School of Public Policy
University of Southern California
Abstract
This study analyzes whether technology adoption in a coproduction strategy can
improve efficiency, and considers whether such efficiency improvements come at the
expense of worsened inequality in service provision. Coproduction and digitizing
service delivery are both innovative strategies that municipal public managers use to
improve service delivery in part through disintermediation that devolves tasks from
public administrators. The study uses survival analysis to assess how the introduction
of a smartphone application called Open311 and the integration of social media
platform Twitter affect the time required to resolve reported issues in San
Francisco’s 311 system. Results suggest that effects vary by technology. In general,
reports entered through Open311 get resolved faster than others, while Twitter
reports take longer. Efficiency gains associated with Open311, however, diminish
over time. Both technologies offer improvements over traditional reporting methods
for resolution of service issues in historically disadvantaged communities, though
unlike Open311, Twitter’s effect remains consistent over time. Twitter also appears
to advantage higher income areas.
77
INTRODUCTION
Local governments now confront both increasingly complex service environments and
severe fiscal austerity characterized by Kiewiet and McCubbins (2014) as a “new fiscal ice age” (J.
Musso et al., 2017). In this setting, coproduction has drawn renewed interest as a tool for local
governments to provide public services with the voluntary assistance of citizens. Innovative
information and communication technologies enable the public sector to support a range of citizen
engagement efforts, including coproduction. Information systems and technologies also have the
potential to change public sector management and even culture, which has led to claims of a new era
of Digital Era Governance (DEG). DEG is a theory of public management focused on “the central
role that IT and information system changes now play in a wide-ranging series of alterations to how
public services are organized as business processes and delivered to citizens or customers”
(Dunleavy et al., 2006; Margetts & Dunleavy, 2013).
Digital Era Governance and coproduction are conceived of as ways for governments
(especially municipal governments) to “do more with less” and maintain service levels in the modern
era of dwindling budgets. Moreover, both seek to improve service delivery through
disintermediation that can radically alter both the citizen-government interface and the internal
logics of government itself (Fishenden & Thompson, 2013; Margetts & Dunleavy, 2013).
Understanding how emergent technologies affect the coproduction of public goods is important for
several reasons. First, given the normative argument for efficiency that motivates both coproduction
and DEG/e-government adoption, it is crucial to understand whether these strategies can work
together to yield efficiency gains predicted by theory. Second, the question of whether innovative
approaches to service delivery may have distributive impacts on service provision remains
unanswered. There has been a concern that the digital divide constrains involvement of lower
income individuals in e-government platforms (P. Norris, 2001; van Deursen & Van Dijk, 2014).
78
However, recent research has suggested that the implementation of emergent technologies in
coproduction systems may increase relative usage rates by disadvantaged groups (Clark & Brudney,
2014; Clark, Brudney, & Jang, 2013; Clark & Guzman, 2016).
This study analyzes how implementing two new technologies, the social media platform
Twitter and a mobile application (app) called Open311, affects the length of time necessary to
resolve problems reported through a municipal coproduction system known as ‘311’ in the City of
San Francisco. Resolution time is measured as the number of hours it takes for the city to mark a
request finished from when it was first made, and is analyzed using survival analysis with panel data
from 2008 through 2015. Briefly stated, the findings indicate that both new technologies differ from
previous reporting methods with respect to the average time it takes to resolve a report. In general,
reports entered through Open311 are resolved more quickly than phone and web reports, while
Twitter reports are resolved more slowly. This analysis also provides evidence that both Twitter and
Open311 offer improvements over traditional reporting methods for those in historically
disadvantaged communities and, in the case of Twitter, high-income areas as well. However, the
improvement in response time associated with Open311 use diminishes the longer a report remains
unresolved.
The rest of the paper is organized as follows. The next section summarizes the theoretical
and empirical research that informs the study, describes the case that serves as its empirical focus,
and presents hypotheses. This is followed by a discussion of the data and methods used to test the
hypotheses. The paper then turns to the findings and discusses their implications. The paper
concludes with a summary of the results as well as an evaluation of its limitations and directions for
further research.
79
DISINTERMEDIATION IN SERVICE DELIVERY:
SAN FRANCISCO’S 311 SYSTEM
In March 2007 the City of San Francisco rolled out its new 311 system, providing residents
with a single point of contact for questions, concerns, and requests for services twenty-four hours a
day, seven days a week (Vega, 2007). Generally speaking, a 311 call center employs trained operators
who determine which department should address a service complaint and forward the call
accordingly. The City of Baltimore conceived the first 311 system in the mid-1990s in an effort to
mitigate the misuse of its 911 emergency system for complaints about city services. Since its
inception, many cities have adopted 311, including New York, San Francisco, Chicago, and others
across the United States.
14
After introducing the 311 system, San Francisco moved to integrate two
different technology innovations: a mobile phone application called Open311 that sent information
directly to service departments, bypassing the call center; and Twitter integration, which had the
effect of cataloguing the complaint not only with the 311 call center, but with the broader Twitter
community. This essay analyzes San Francisco’s 311 system as a case of administrative
disintermediation, in which coproduction (using citizens as service monitors) is joined with
technology innovation. It assesses the extent to which these innovations improve service delivery by
reducing the time required to resolve complaints, and also investigates the distribution of requests
across neighborhoods by input method.
Coproduction is defined as voluntary citizen participation in the production of publically
provided goods and services (J. M. Ferris, 1984; Kiser, 1984; Parks et al., 1981; Percy, 1984).
Inasmuch as coproduction disrupts traditional patterns of administrative control, it raises questions
regarding the motivations that lead citizens and administrative entities to cooperate and the resulting
consequences for the effectiveness, efficiency, and fairness of service arrangements (Bovaird, 2007;
14
Some cities, such as Boston, have systems that function identically to 311 but do not employ the use of the number to
direct users to a call center (Clark & Guzman, 2016).
80
Chen, Tsou, & Ching, 2011; Clark & Guzman, 2014; Isett & Miranda, 2015; Jakobsen & Andersen,
2013). In the case of San Francisco’s 311 system, coproduction resembles a shared planning
relationship, where the citizen/user participates in the planning process but the delivery is managed
and executed wholly by the government. Individuals reporting problems are coproducers in the
sense that they identify problems in need of attention, thereby divesting the government agency
from the task. Once a report enters the system, it is assigned to the appropriate agency or
department for resolution. Prioritization of particular reports becomes a function of administrative
discretion within the municipal government; citizens notify the government that a problem exists
but have no means to directly affect the priority any one report receives.
15
In this sense, 311 users
“coplan” the service in the sense that they are providing critical information about community
needs. However, service delivery is determined entirely by professionals, such that 311 is an
administratively executed form of coproduction (Bovaird, 2007).
Coproduction as Disintermediation
Critical to the concept of coproduction is disintermediation: the blurring of boundaries
between the administrative state and citizens. By shifting choices or tasks away from administrators,
coproduction disintermediates traditional relationships between government and citizens. Bovaird’s
(2007) framework implicitly acknowledges this disintermediation in defining coproduction along two
dimensions: (1) the planning process that determines what is to be produced; and (2) the provision
of services or public goods. Within this matrix, the degree of disintermediation ranges from
effectively nil (either traditional service provision or self-organized community provision) to “full”
coproduction, involving equal planning and provision efforts on the part of citizens and government
(Bovaird, 2007). In the middle ground lie arrangements where either the planning or production
15
This leads, of course, to indirect attempts to draw attention to particular problems, e.g., multiple reports for the same
problem or complaints to department administrators and/or elected officials.
81
processes (but not both) are shared. Examples of the latter type include programs designed to train
immigrant parents how to assist their children in learning the local language (Jakobsen & Andersen,
2013). Other examples include participatory urban planning regimes, or participatory budgeting
(again, see Bovaird (2007), but also Clark and Brudney (2014); Clark and Guzman (2016)). In all of
these cases, coproduction extends beyond the New Public Management conception of the citizen-
government relationship as a market. New Public Management reframes citizens as consumers but
leaves institutional barriers intact. In coproduction, citizens play an active role in service delivery.
For 311 systems, where participants act as “eyes and ears” for identifying service and
infrastructure breakdowns, citizens are given the power to call attention to issues that city
bureaucrats may (or may not) have overlooked or otherwise deprioritized. This difference is
important, because administrative rule- or process-based decision making for allocating street-level
services can lead to unintended consequences. Levy, Meltsner, and Wildavsky (1974) illustrate this
problem in their classic work examining service distribution in Oakland, California. The paradox of
“urban outcomes” was that processes based on efficiency norms ultimately resulted in structural
biases in the level of services and goods provided to different socioeconomic groups, an effect that
has continued to the present day (Bovaird, 2014; Jun & Musso, 2013; Levy et al., 1974). The
engagement of users in coproduction inevitably raises distributional questions, as studies of civic
engagement have consistently found socioeconomic biases in participation rates (Verba, Schlozman,
Brady, & Brady, 1995).
Research to date on potential biases in coproduction has produced mixed findings. Early
work in coproduction recognized the potential for distributional issues in coproduction; for
example, Brudney (1985) noted that political economy concerns could lead governments to offload
costs to coproducing citizens, with disproportionate negative effects for neighborhoods with lower
socioeconomic status. Minkoff (2015) finds no evidence of inequality across socioeconomic status in
82
New York City’s 311 system. Clark et al. (2013) use data from Boston’s 311 system to test for
distributive bias, finding that areas with lower incomes and higher proportions of Hispanic residents
are less likely to engage in 311-based coproduction, although other racial and ethnic groups use the
system in proportion to their share of the population. A subsequent study employed survey data
from San Francisco to examine the system’s representativeness, again finding no substantive pattern
of bias (Clark & Brudney, 2014).
E-government, Digital Era Governance, and Coproduction
Over time, 311 systems have transitioned from phone-only systems to systems incorporating
web- and smartphone-based technologies. These additions are attractive to governments, as they
reduce reliance on expensive call centers, and attractive to citizens, who often prefer these
technologies over the telephone. As such, these “smarter” approaches to 311 exemplify the
“Government 2.0” technological advances that arguably allow for more participatory, collaborative
two-way communication by lowering the technical, knowledge, and resource barriers to application
design and content production (Chang & Kannan, 2008; Chun, Shulman, Sandoval, & Hovy, 2010;
Nam, 2012; Reddick & Aikins, 2012). Some public managers consider such technologies essential
for engaging the millennial generation (Mancini, 2012). Both the theoretical and empirical literature
on specific social media platforms largely focuses on the technology’s dialogic capability, which in
theory should promote coproduction by fostering dialogue between government and citizens (Ae
Chun et al., 2012; Chadwick, 2008; Linders, 2012; Mossberger, Wu, & Crawford, 2013; A. Smith,
2010; Thomas, 2013). In essence, the disintermediation made possible by Digital Era Governance
turns on its head the classic dilemma of technology use in public administration. While the difficulty
of adapting new technologies to existing institutional structures and behaviors has long been
acknowledged, Digital Era Governance seeks not adaptation but disruption, employing new
83
technologies to facilitate more holistically disintermediated practices and governance approaches
(Fishenden & Thompson, 2013; Margetts & Dunleavy, 2013).
Implementing DEG necessarily carries both political (citizen-facing) and managerial
(internal) consequences, a familiar combination in public sector ICT management. Well before the
modern smartphone made apps ubiquitous, the World Wide Web (web) created new opportunities
to develop innovative approaches to information and service provision with the potential to
challenge both the political and managerial status quo. Politically, web-based approaches were
hypothesized to facilitate civic participation and engagement among previously underrepresented
groups within the polity (Jun & Weare, 2011; Moon, 2002; J. Musso et al., 2000; Weare et al., 1999;
West, 2004). These changes could, in theory, result in more democratic processes and outcomes
(Moon, Lee, & Roh, 2014; J. Musso et al., 2000), as well as increasing public trust in government
(Kim & Lee, 2012). Moreover, as J. Musso et al. (2000) argue, web-based ICT systems can generate
managerial as well as political effects. The potential managerial impacts include changes to local
government’s peripheral and core managerial and corporate functions. These may in turn improve
public sector performance by, for example, lowering public sector transaction costs (Berry, Berry, &
Foster, 1998; Bozeman & Bretschneider, 1986; Coursey & Norris, 2008; G. Lee & Perry, 2002;
Moon & Bretschneiber, 2002; Moon & Norris, 2005).
From a technical perspective there is no reason to believe that either web- or app/social
media-based ICTs’ theorized impacts require their implementation as brand new, standalone
systems. Yet, many early studies predicted that technological change would manifest rapidly as
entirely new, disruptive innovations (Coursey & Norris, 2008). These expectations, however, failed
to live up to empirical scrutiny: E-government adoption followed a much more incremental path,
and evidence of its transformative nature was mixed, at best (Coursey & Norris, 2008; Jun & Weare,
2011; Moon, 2002; J. Musso et al., 2000; D. F. Norris & Reddick, 2013; Weare et al., 1999).
84
Coproduction in particular presents the possibility to maximize the potential for technologically
driven radical disintermediation between citizens and government because coproduction generates
disintermediation by construction. The question, then, is whether such combinations create a
positive feedback loop of disintermediation that improves the system’s performance and/or service
inequality, or if they end up overwhelming the public sector’s capacity for disintermediation, leading
to inefficiencies or exacerbating existing inequalities. The case of San Francisco’s 311 program and
its implementation of two distinct new communications technologies to facilitate coproductive
behavior in citizens reporting breakdowns in city services and public spaces offers a compelling
empirical environment to consider the effects of differentiated media use.
Research suggests that these new technologies may help reduce disparities while improving
public service delivery. For example, in the case of Boston, Clark et al. (2013) find that smartphone-
based 311 applications are used for submitting reports at a higher rate in lower-income than higher-
income areas of the city and argue that this may be evidence of the application’s ability to bridge the
participation gap across income strata. Similarly, Clark and Brudney (2014) use data from city-
administered surveys and find that smartphone-based 311 applications serve to increase
coproduction rates among minority groups in San Francisco. However, it is still an open question
whether changes in the pattern of 311 use actually correspond to changes in the city’s patterns of
service provision. If demands for services in wealthier neighborhoods are given higher priority (and
vice versa), then the increased participation rate may ultimately fall on deaf ears or tied hands. There
is little in the way of research that evaluates the impact of the e-government evolution in 311
systems on system performance.
Modern ICT systems potentially affect urban inequality and service delivery bias through
both access to physical resources and infrastructure as well as the skills and socialization necessary to
use them – i.e., the “digital divide” (P. Norris, 2001; van Deursen & Van Dijk, 2014; Warschauer,
85
2004). Still, the potential for such technologies to expedite service delivery by producing higher-
quality report information and/or making reports publically viewable may act to overcome these
challenges at the aggregate level and lead to improvements in distributional equity.
Disintermediation and Innovation in 311: The Case of San Francisco
As Moldogaziev and Resh (2016) note, the relationship between innovation characteristics
and adoption is largely underexplored, especially as it relates to the innovation’s success once in use.
San Francisco’s 311 system offers a unique opportunity to study the effect of adopting innovative e-
government technologies within the context of coproduction. San Francisco launched its 311 system
in March 2007 with both phone and email capabilities. A custom website for submitting reports
followed shortly, with its first report submitted eight months later in November. In 2009, then-
Mayor Gavin Newsom announced that Twitter would be incorporated as a medium for submitting
311 reports and following up with the city. The idea arose during an unrelated meeting between
Newsom and executives from Twitter, and the Mayor directed the city’s technology department to
begin scoping the project shortly thereafter.
That same year, the city also began working with other municipalities, nonprofits, and
private developers to create an open standard for accessing 311 systems through Application
Program Interfaces (APIs). This standard, eventually dubbed Open311, led to the city launching a
smartphone application and publically accessible API for use with 311 that went live in 2013
(Mayor's Office of Communication, 2013).
16
These applications use smartphones’ built-in geospatial
system to automatically submit the geocoordinates associated with the reported issue. They also
allow the user to take and attach a photograph of the issue with the device’s integrated camera. Two
applications written by private developers are also available.
16
Privately-developed applications such as SeeClickFix had been using APIs to pass requests to the 311 system on a
more limited scale since June 1, 2008.
86
It is important to qualify the differences between Open311 and Twitter by first noting their
similarities. Both involve the adaptation of preexisting technologies and platforms designed by
actors external to the organization, similar to the bottom-up technocratic process described by
Andersen (2008) (albeit in fundamentally different ways discussed below). Both also enjoyed
consistent support from elected officials across two different administrations. The principal
difference, however, lies in the relationship between their locus of origin within the organization and
their technical characteristics vis-à-vis 311 as a preexisting organizational technology.
The seemingly ad-hoc nature of Mayor Newsom’s decision to adopt Twitter for use with 311
and the technocratic and deliberate process of creating the Open311 architecture provide a sharply
etched contrast between innovations originating from two different organizational ‘cores:’
administrative on the part of Twitter and technical on the part of Open311 (Borins, 2000; Daft,
1978; Damanpour, 1991; Richard, 1992). The degree of interconnectedness or ‘coupling’ between
these cores is thought to be a function of how professionalized the technical core of an organization
is, with more professionalization resulting in looser coupling and vice versa (Daft, 1978). In the case
of San Francisco, a strong professionalization exists amongst the city departments, and its weak
mayor system of government tends to cluster technical competency, when it exists in the
administrative core, within the Board of Supervisors.
311 is a coproductive system that pairs volunteer reporters with frontline workers tasked
with making the necessary repairs or removing problems. By nature, 311 is (1) technically complex,
in no small part because it is (2) dual-facing between external stakeholders (coproducers) and
internal processes (departmental task acquisition and assignment), and (3) firmly situated within the
core technical functioning of many city departments. Thus, successful adoption of new processes for
receiving requests in the 311 system requires some degree of technical specialization and
sophistication. Viewed in this light, the question becomes how innovations championed from a
87
loosely-coupled administrative core might differ in technical feasibility and goodness of fit from
those championed from within the technical core.
Substantive technical and administrative differences between Open311 and Twitter suggest
they should perform differently. Open311’s architecture is designed to minimize human error in
every stage of the reporting process. Automatic geocoding minimizes the risk of citizens
misreporting the location of the problem, and attached visual evidence both reduces the risk that
city staff will fail to locate the problem and allows staff to identify ‘false positives’ or duplicate
reports. Open311 also removes the human intermediary in the form of call center staff, eliminating
this vector for errors that could delay resolution. Furthermore, the mobile nature of Open311 may
reduce the lag time between observing a problem and reporting it. Thus,
Hypothesis 1. Service reports made via Open311 will be resolved faster compared to
other submission methods because Open311 reduces error generation and lag time
between submission and integration into a department’s task queue.
The politicized character of the Twitter innovation suggests that it may not have
incorporated “adequate feedback from the organizational members with the institutional
competence necessary to identify potential obstacles to implementation success” (Moldogaziev &
Resh, 2016). Twitter was designed as a short-length rapid communication system similar to texting.
While it is true that the platform allows for both in situ reporting and attaching photographs of the
problem of interest, much like Open311-based applications, the level of detail involved in producing
311 reports is arguably inconsistent with a 140-character limit on messages. Moreover, the logistical
challenge of managing and tracking a stream of incoming tweets in real time led the City ultimately
to introduce an intermediary technology between Twitter and 311: CoTweet,
17
a platform purpose-
17
CoTweet was eventually acquired by ExactTarget in 2010, which was in turn acquired by Salesforce in 2013. The
uncertainty introduced by these takeovers should not go unappreciated in considering the differences between the
adoption of Twitter and Open311 in this case.
88
built to facilitate using Twitter as a Customer Relationship Management (CRM) system. The 311 call
center staff act as intermediaries for creating a report in the 311 system based on the information
supplied in the message. 311 staff also follow-up with the user in the case of ambiguous or
incomplete information. Twitter as a 311 input technology does not involve much disintermediation
compared to the telephone; city employees remain intermediaries between users and the 311 system.
At the same time, it adds new ways for miscommunication to arise between 311 staff and users,
making errors more likely. Thus,
Hypothesis 2a. Reports made via Twitter will be resolved slower than both Open311
and the telephone because Twitter is both unsuited for making 311 reports and
introduces new error vectors compared to the telephone.
However, one feature of Twitter-based reporting distinguishes it from Open311 or other
reporting methods. Reports made via Twitter are observable by anyone either logged in to a Twitter
account or visiting twitter.com by searching for the appropriate hashtag (#SF311). These reports
and any subsequent comments or discussion have the potential to become public relations issues,
and in that sense may represent a path for political considerations to affect response time. For
example, in a study of social media usage surrounding public transportation agencies in US cities,
Schweitzer (2014) finds evidence of follow-up and direct attention via Twitter on the part of transit
staff in response to negative and hostile mentions by transit-riding Twitter users. In the context of
urban coproduction, failing to resolve these reports in a timely fashion could lead to public shaming
and agitation for increased accountability within the 311 department and the city government
overall. Thus,
Hypothesis 2b. Reports made via Twitter will be resolved faster because they are
public and can therefore draw the attention and ire of others if the reports are not
fixed quickly.
89
The potential distributive impacts of Open311 and Twitter are similarly ambiguous. On the
one hand, prior research on the use of smartphone-based reporting in 311 in general and in San
Francisco in particular suggests that despite the digital divide issues around the high fixed and
variable costs of these technologies, they may promote participation rates among historically
disadvantaged citizens, which in turn may lead to greater levels of effort invested by the city in these
neighborhoods. On the other hand, research on the digital divide as well as Levy et al. (1974) remind
us that systems and processes designed with an eye towards efficiency can inadvertently create
biased outcomes. It may, therefore, be the case that even if both Open311 and Twitter lead to
decreased times to resolution (which, of course, may not be true), their effects are insufficient to
overcome the existing systemic biases in service provision. Thus,
Hypothesis 3a. Reports made via either Open311 or Twitter are associated with a
decrease in the time to resolution for reports made in historically disadvantaged
and/or minority neighborhoods due to the efficiency gains that these technologies
provide relative to traditional reporting methods.
Hypothesis 3b. Reports made via either Open311 or Twitter are not associated with
any change in the time to resolution for reports made in historically disadvantaged
and/or minority neighborhoods, because the causal processes behind the differences
in resolution time by socioeconomic position are not affected by the quality or
transparency of 311 reports.
DATA AND METHODS
This study utilizes public information from the City of San Francisco’s 311 database made
publically available through its open data portal, data.sf.gov. This database contains detailed
information on every report (or, otherwise framed, incident of coproduction) made through the
city’s 311 system since June 1, 2008. The database is updated in near real-time with new records,
with multiple changes every hour. Report-based information includes the technological source (e.g.,
phone, twitter, or Open311); the type of report (e.g., street/sidewalk cleaning, graffiti, or abandoned
90
vehicles); the department responsible for resolution; its geographic coordinates; the link to an image
of the problem if one is submitted;
18
its status (open or closed); and the time that the report was
made and when it was closed. As of May 11, 2016 the database consisted of 1,276,549 reports
generated before November 1, 2015.
The geographic coordinates provided by the city support the association of reports with
demographic characteristics of neighborhoods previously used in analyzing 311 reporting (Clark &
Brudney, 2014; Clark et al., 2013). These are drawn at the census tract level from the 2008-2012 five-
year American Community Survey dataset provided by the Census Bureau via Geographic
Information Systems (GIS) software using TIGER/Line census tract boundaries. It is important to
note that these characteristics are place-based, and cannot be used to make inferences about the
individual making the report. Instead, they are socioeconomic characteristics of the place within San
Francisco where the problem was experienced. Thus, they are indicators of the characteristics of
community residents who are likely exposed to the reported problem. These characteristics include
population, median household income, homeownership rates, and race and ethnicity. Population
data are log-transformed to allow for analysis of relative changes.
The sample is restricted in several ways. First, it excludes reports that have been identified as
closed but did not involve the actual resolution of the reported problem. This includes, but is not
limited to, instances where the report was forwarded on to other public agencies (e.g., the USPS);
where there was insufficient information provided in the report to assign it to a given department;
where the individual making the report later cancelled their report; and where the report is flagged as
a duplicate. Importantly, the sample does include reports where city staff were dispatched to the
location of the problem but found no problem to be fixed – this constitutes an investiture of labor
and capital by the city irrespective of whether the observed problem remained once they were on the
18
The presence of images is so strongly correlated with the use of Open311 (>.80) that its inclusion in the final model is
not warranted.
91
scene. Reports involving graffiti on private property are also excluded; resolution is the responsibility
of the property owner and not the city. The analysis also excludes reports that consist of feedback
(e.g., a complaint of a police officer being rude or a bus driver being helpful), reports assigned to
external agencies (such as a state agency), reports that are opened to track callbacks or follow-ups,
interdepartmental reports, cab-related complaints, 311 volunteer program-related reports, and
reports for construction permits and temporary signs. In the case of permit and sign reports these
are both subject to an exogenous, unobserved review process and are often made far in advance,
introducing bias in their time to resolution. The other excluded reports do not represent a material
problem that can be fixed. Observations are further restricted to reports made within the limits of
the City of San Francisco. The County of San Francisco operates two parks outside of the city where
reports were made in 2013: Sharp Park in Pacifica, and Camp Mather in Tuolume (Yosemite
National Park). The County also operates San Francisco International Airport, which is located in
San Bruno. Additionally, reports made outside of both the City and County of San Francisco’s
jurisdictions (e.g., a few blocks into the City of Colma) are excluded. The final sample consists of
953,164 observations. Descriptive statistics for all continuous variables can be found in Table 3.1.
Table 3.1. Summary Statistics For All Continuous Variables
Survival analysis is employed to assess the factors that influence the time-to-event for the
city to resolve 311 reports. This time is measured in days, rounded to the nearest whole day. In the
Variable Mean Std. Dev. Median
Lower
Quartile
Upper
Quartile Min Max
Time to closure (days) 3 1 11 0.50 182
Population (log) 8.28 0.72 8.38 3.30 9.31
Median Household Income 75,469 59,007 99,861 12,240 156,613
% Homeowners 0.38 0.23 0.35 0.20 0.59 0 0.89
% African-American 0.07 0.10 0.03 0.01 0.07 0 0.70
% Hispanic 0.15 0.12 0.11 0.06 0.18 0 0.59
% Asian & Pacific Islander 0.31 0.20 0.27 0.13 0.44 0 0.90
92
cases where the report was resolved in less than 12 hours, the period is set to one half of a day.
19
The 311 data can be conceptualized as a classic follow-up observational study, where observations
may enter the study at any point between its start date (July 1, 2008) and the last follow-up date
(October 31, 2015). In order to weight reports equally, each is followed for a period of 6 months;
reports that are not closed within that time period are censored at 182 days. Of the 953,164
observations, 80,862 (8%) are censored. This level of censoring is relatively low compared to other
fields of social science that frequently employ survival analysis, most notably international relations,
where censoring rates ranging from 25% to 50% are not uncommon (Box-Steffensmeier & Zorn,
2001; Park & Hendry, 2015). In every case this censoring consists of non-informative (i.e., the
censoring mechanism was not conditional on properties of the report) right censoring (Hosmer,
Lemeshow, and May, 2008; Klein and Moeschberger, 2003). The model uses a Cox proportional
specification because there is insufficient theory on 311 systems’ effectiveness to justify an a priori
specification of baseline hazard function; its shape is also of no direct consequence for the questions
under consideration. The Cox proportional hazard model relaxes assumptions about the hazard
function through semiparametric estimation (Cox 1972).
The dependent variable is a dichotomous indicator for whether the observation was closed
(i.e., experienced the ‘event’). The model also contains a continuous measure of the time in days
between when a report was input into the system and when its status was changed to closed. Days
were chosen as the period for ease of interpretation. Thus, the coefficients associated with
independent variables in the model reflect their conditional effect on the length of time it takes for
an average report to transition from open to closed. The independent variables of interest are
dichotomous indicators for whether the report was made using Open311 or Twitter, and the
socioeconomic indicators listed above. To analyze the distributional impacts of the different
19
This choice was made both to account for the fact that many (>40%) reports are resolved in less than one day, while
reducing the computational cost of estimating time-varying covariates across time periods.
93
technologies, socioeconomic variables are stratified by their distribution to facilitate analysis of
interaction effects. For income and homeownership, the model includes the upper and lower
quartiles separately, with the second and third quartiles serving as the base case. For three different
racial groups – African American (non-hispanic), Asian and Pacific Islander, and Hispanic – the
model includes indicator variables identifying neighborhoods where their proportional share of the
population exceeds the upper quartile of their distribution across the city. For example, a
neighborhood where more than 44% of the population are Asian and Pacific Islanders would be
included in the Asian and Pacific Islander indicator variable. The model also includes an interaction
effect for both Open311 and Twitter and each of these socioeconomic indicators. By doing so, it is
possible to estimate the relative impact of using different technologies for making 311 reports across
these socioeconomic strata within the city. Thus the basis of comparison is to reports made by
telephone or web browser in (1) middle-income neighborhoods with (2) neither low nor high
relative rates of homeownership (i.e., predominantly renter-populated), that are (3) relatively
heterogeneous in their racial composition, as no one group is over-represented relative to the rest of
the city.
Because the Cox model relies on the condition that hazards remain proportional over time,
postestimation tests for proportionality are an important component of analyzing model fit (Box-
Steffensmeier & Zorn, 2001; Hosmer, Lemeshow, & May, 2008). Several variables were found to
violate the assumption of proportionality (rejecting the null hypothesis of proportionality with p <
0.001), including both indicators for the submission technologies used; most demographic strata; the
category of report (e.g., graffiti removal); the department responsible for the report; most of the
interactions between demographics and Open311; and three interactions between demographics and
94
Twitter.
20
Two different steps are taken to correct these violations. First, the final estimation
stratifies observations across both categories of reports and the responsible department. While this
controls for proportionality violations, it does not allow for interpretable findings for those variables
(Box-Steffensmeier & Zorn, 2001; Hosmer et al., 2008). The strategy produces a more parsimonious
model – while still controlling for these factors – as the variation across types of reports and
responsible departments are not central to this research’s questions of interest. Second, the model
includes an interaction of each violating variable with the natural log of survival time at each point in
time when an observation has yet to either be closed or censored at the end of the observation
period. This specification not only eliminates the bias and loss of power introduced by violations of
the proportional hazard assumption, it allows the researcher to identify how the relative hazard in
each report changes over time, providing an additional level of nuance to the findings where
appropriate.
RESULTS
Figure 3.1 shows how the introduction of these different technologies has changed the
patterns of 311 use over time through October 31, 2015. Open311-based reports have rapidly grown
in overall number and as a proportion of total reports since its official roll-out in 2013, while Twitter
usage has remained relatively low in absolute and relative terms and phone has consistently lost
share of total reports over time.
20
Because the data exhibit low rates of censoring, the test for proportionality was conducted using the rank
transformation of analysis time to avoid both false positives and false negatives (Park & Hendry, 2015).
95
Figure 3.1. 311 Usage Rates By Different Reporting Technologies Over Time
311 Usage Patterns By Technology
Figure 3.2 summarizes the top five types of report made through each technology, and
suggests that there is differential use of technology to make certain types of service complaints.
Reports made by phone are predominantly for street and sidewalk cleaning (39%) followed by
abandoned vehicles and general reports at a distant second (8% each). Three of the most reported
categories via phone – abandoned vehicles, damaged property (8%) and housing authority reports
(4%) – are associated with repeat interaction with a fixed location, suggesting that these reports may
be made from a residence or place of work. Graffiti-based reports are not a common report category
for phone submissions (4%), whereas they are either the first or second most common category for
all other technologies.
Phone
Web
Twitter
Open311
Internal
0
50
100
150
200
250
2008 2009 2010 2011 2012 2013 2014 2015
Thousands
Request Medium Usage Over Time
96
Figure 3.2. Share of Reports by Technology (Top 5 Categories)
Use of the web submission technology is strongly associated with offensive graffiti – 28% of
all web submissions. Street and sidewalk cleaning is the second most common report at 19%. Web
reports also have the highest proportion of abandoned vehicle reports (16%) and illegal postings
(9%) of any technology. Both web and Open311 have streetlight issues as their fifth-most common
report type at 4%. While Open311 also closely resembles the web with regards to graffiti (27%), the
type of graffiti is divided between offensive (16%) and non-offensive (11%). However, the dominant
report type for Open311 is street and sidewalk cleaning at 41%. Twitter also has a large proportion
of graffiti reports (29%), though they are all categorized as non-offensive. Street and sidewalk
28%
19%
16%
9%
4%
0% 10% 20% 30% 40%
Graffiti - Offensive
Street and Sidewalk
Abandoned Vehicle
Illegal Postings
Streetlights
Web
39%
8%
8%
7%
6%
0% 10% 20% 30% 40%
Street and Sidewalk
Abandoned Vehicle
General Requests
SFHA Requests
Damaged Property
Phone
29%
25%
8%
6%
5%
0% 10% 20% 30% 40%
Graffiti - Not Offensive
Street and Sidewalk
Sidewalk or Curb
311 External Request
General Requests
Twitter
41%
16%
11%
6%
4%
0% 10% 20% 30% 40%
Street and Sidewalk
Graffiti - Offensive
Graffiti - Not Offensive
Illegal Postings
Streetlights
Open311
97
cleaning reports closely follow at 25%. Twitter also has the largest proportion of reports involving a
broken or damaged sidewalk or curb (8%).
There also appear to be demographic differences in the types of technology used to report
service needs, although it must be noted that these are characteristics of the neighborhood in which
the complaints are made, not the user. Figure 3.3 shows the mean demographic characteristics of the
census tracts in which 311 reports are made, categorized by the type of technology used. Usage of
the telephone call center is associated with places with higher proportions of Asian and Pacific
Islander residents, as well as senior citizens, which may reflect the technology preferences of
residents in those areas. Both phone and web-based usage is associated with areas that have more
homeowners. Reports generated through the web are associated with census tracts that have a
greater proportion of non-Hispanic Whites and college graduates. Web usage is lowest in areas with
higher proportions of Hispanic, foreign-born, and African American residents. Twitter is associated
with census tracts with higher proportions of Hispanic and foreign-born residents, but has the
lowest rate of association of any technology among neighborhoods with high levels of non-Hispanic
Whites, Asian and Pacific Islanders, homeowners, and college graduates. Open311’s use relative to
other technologies is higher in areas with increased proportions of Hispanic, Asian and Pacific
Islander, and college graduate residents.
98
Figure 3.3. Demographic Characteristics of Report Locations by Technology
0% 10% 20% 30% 40% 50%
Phone
Web
Twitter
Open311
Mean % Non-Hispanic White
0% 10% 20% 30% 40% 50%
Phone
Web
Twitter
Open311
Mean % Hispanic
0% 10% 20% 30% 40% 50%
Phone
Web
Twitter
Open311
Mean % African American
0% 10% 20% 30% 40% 50%
Phone
Web
Twitter
Open311
Mean % Asian & Pacific Islander
0% 10% 20% 30% 40%
Phone
Web
Twitter
Open311
Mean % Foreign Born
0% 20% 40% 60%
Phone
Web
Twitter
Open311
Mean % College Graduates
0% 5% 10% 15%
Phone
Web
Twitter
Open311
Mean % Seniors
0% 10% 20% 30% 40%
Phone
Web
Twitter
Open311
Mean % Homeowners
99
Model Results
Table 3.2 summarizes results from the Cox proportional hazards model, where again, the
dependent variable is time to resolution for 311 reports, and the critical variable of interest is
whether the report was using Twitter or the Open311 app. The reported coefficients can be
interpreted as whether the time to resolution is faster (positive) or slower (negative) relative to
reports submitted using the telephone or web. As discussed previously, the stratification technique
for dealing with proportional hazards violations for problem type and responsible department
prohibit the presentation or analysis of their relative differences. Instead, they are controlled for
directly but implicitly in the model. The reported likelihood ratio test for the chi-square distribution
is highly statistically significant (p < 0.001), indicating that the model is a good fit for the data. Most
of the variables and interactions are statistically significant as well, with the exception of the
indicator for reports made in the areas with the lower rates of homeownership; the indicator for the
use of Open311 in areas with the highest proportion of Asian and Pacific Islanders; and the
indicator for the use of Twitter in areas with the lowest median income.
General Effects of Technology Innovation.
The results offer nuanced support for the hypothesis that reports made through Open311
are resolved faster than those made through preexisting mediums as well as Twitter both across the
city overall and in disadvantaged areas. In heterogeneous, middle-income neighborhoods with an
average mix of rental and owner-occupied housing units
21
, reports submitted using Open311 are
resolved faster on average than those using the telephone or web when controlling for other factors.
However, this effect appears to diminish over time. As reports stay open longer, there is a narrowing
of the average difference in closure times between Open311 and traditional 311 technologies. In
21
This interpretive caveat reflects the
100
Table 3.2. Hazard Model Results On Time To Resolution For 311 Reports
Standard
Error
Census Tract Population (log) -.016
***
.002
Lowest Income Tracts (Lower Quartile) -.026
***
.005
Highest Income Tracts (Upper Quartile) .011
**
.004
Lowest Homeownership Tracts (Lower Quartile) .004 .004
Highest Homeownership Tracts (Upper Quartile) -.042
***
.003
Highest % African-American Tracts (Upper Quartile) .014
***
.004
Highest % Hispanic Tracts (Upper Quartile) .065
***
.004
Highest % Asian & Pacific Islander Tracts (Upper Quartile) -.018
***
.004
Report Submitted Using Open311 .304
***
.009
Open311 * Lowest Income Tracts .056
***
.012
Open311 * Highest Income Tracts .025
*
.010
Open311 * Lowest Homeownership Tracts -.062
***
.011
Open311 * Highest Homeownership Tracts -.102
***
.016
Open311 * Highest % African-American Tracts .054
***
.012
Open311 * Highest % Hispanic Tracts .106
***
.011
Open311 * Highest % Asian & Pacific Islander Tracts .001 .013
Report Submitted Using Twitter -.428
***
.030
Twitter * Lowest Income Tracts .018 .026
Twitter * Highest Income Tracts .438
***
.036
Twitter * Lowest Homeownership Tracts .187
***
.029
Twitter * Highest Homeownership Tracts .115
***
.032
Twitter * Highest % African-American Tracts .200
***
.026
Twitter * Highest % Hispanic Tracts .373
***
.031
Twitter * Highest % Asian & Pacific Islander Tracts .139
***
.029
Interactions with log-time
Lowest Income Tracts .022
***
.002
Highest Income Tracts -.010
***
.002
Lowest Homeownership Tracts -.009
***
.002
Highest % African-American Tracts -.013
***
.002
Highest % Hispanic Tracts -.031
*
.002
Highest % Asian & Pacific Islander Tracts .005
***
.002
Report Submitted Using Open311 -.218
***
.004
Open311 * Lowest Income Tracts -.106
***
.007
Open311 * Lowest Homeownership Tracts .069
***
.006
Open311 * Highest Homeownership Tracts .055
***
.009
Open311 * Highest % African-American Tracts -.059
***
.007
Open311 * Highest % Hispanic Tracts -.027
***
.006
Report Submitted Using Twitter -.037
**
.012
Twitter * Highest Income Tracts -.049
**
.017
Twitter * Lowest Homeownership Tracts -.050
***
.013
Twitter * Highest % Hispanic Tracts -.053
***
.013
Likelihood Ratio Chi-Square 11,469.58
***
N 953,164
* p<0.05, ** p<0.01, *** p<0.001
Coefficient
101
most cases, the positive effect of Open311 holds true; it remains superior to traditional mediums for
more than 90% of all reports. This suggests that for most cases Open311 improves efficiency, but
the lack of interaction with 311 staff may make it more difficult to identify and resolve difficult
reports.
Twitter, in contrast, appears to be associated with longer time to resolution compared to
traditional reporting mediums. Holding other variables constant, reports submitted using Twitter
require significantly longer times to resolve compared to the telephone and web mediums. In
contrast to reports submitted using Open311, this effect does not diminish over time. Rather, the
differential increases, such that reports generated through Twitter become increasingly less likely to
be closed the longer they stay unresolved. This suggests that not only is Twitter inefficient for low-
to medium-complex tasks, it further complicates the resolution of more challenging reports.
Influence of Neighborhood Demographic Characteristics.
Prior research on urban service delivery and coproduction suggests that city response times
ought to be faster in areas with greater concentrations of wealth and slower in areas with greater
proportions of minorities (Brudney, 1985; Levy et al., 1974). Evaluations of 311 systems in
particular, however, have found both its use and outcomes to be more equitable (Clark & Brudney,
2014; Minkoff, 2015). Among San Francisco’s 311 reports on the whole, time to resolution does
appear to vary across demographic and economic strata. Controlling for other factors, reports made
in low-income areas take longer to resolve on average than those in middle-income areas, while
high-income areas experience relatively faster resolution compared to their middle-income
counterparts. This appears to be consistent with the argument that privileged groups within the city
are able to use their economic and political influence to receive preferential treatment.
The association of other demographic characteristics with time to resolution show
inconsistent patterns of bias. For example, considering that homeowners are more likely to vote and
102
otherwise be active politically, the patterns with respect to homeownership rates are somewhat
counterintuitive. Reports made in areas with low homeownership rates are not statistically different
from those in areas with a more even distribution of renters and homeowners initially, although they
do take longer to resolve as reports stay open longer. Areas with relatively high rates of
homeownership, on the other hand, have longer average resolution times than those with average
rates, counter to expectation. Areas with either a high proportion of African-American or Hispanic
residents are initially associated with shorter times to report resolution when holding other variables
constant, although this advantage declines the longer a report remains open. In contrast, places with
higher proportions of Asian and Pacific Islander residents take longer to resolve on average, though
this effect also diminishes as reports stay open longer in the 311 system. While these results are
robust to controls for regional effects within the city, the fact that most Asian and Pacific Islander-
heavy areas are in the western edges of San Francisco relatively far from the city center may account
for some of this effect.
Interaction of Technology Type and Demographic Characteristics.
The effects of Open311 and Twitter on time to resolution across these socioeconomic strata
are also mixed. Reports generated using Open311 in low-income areas are resolved faster, suggesting
that Open311 in this case has some ameliorative effect on service inequality as measured by time to
resolution. Once again, this effect diminishes over time: as reports stay open longer the relative
difference between Open311 and traditional technologies decreases within lower income
neighborhoods. The use of Twitter, meanwhile, is statistically no different than the use of the
telephone or the web in low-income areas relative to middle-income counterparts. At the same time,
Open311- and Twitter-submitted reports are on average resolved faster than phone or web in high-
income areas relative to middle-income peers. Interestingly, this effect does not vary over time for
Open311 reports, while it diminishes over time for Twitter.
103
Comparing areas with relatively low or high levels of homeownership against areas with
middling rates, the relative effects of using new technologies flip, with Twitter outperforming
Open311. Reports made in both lower and higher homeownership areas using Open311 take, on
average, longer to resolve compared to the original reporting technologies within these areas. In both
cases the average increase in resolution time for Open311 requests diminishes over time, such that
long-lived reports end up more likely to be closed if they were submitted using Open311 in those
areas when holding other factors constant. For reports submitted via Twitter the effect is reversed.
Compared to areas with average homeownership levels, Twitter-based reports in both low and high
homeownership areas are resolved faster than web or telephone-based. In the instance of low
homeownership areas this effect is constant over time, while in areas with high rates of
homeownership Twitter’s relative advantage decreases as reports stay open longer.
Both Open311 and Twitter outperform the telephone and web-based reporting technologies
in areas with the highest proportion of African-American and Hispanic residents when measured
against more heterogeneous (or White-majority) places. Interestingly, while this effect diminishes
over time for both Open311 and Twitter in areas with a high proportion of Hispanic residents, the
association of Twitter use with more rapid closure in high African-American areas remains constant
over time. Open311 is not significantly faster or slower than the web or telephone in places with a
high proportion of Asian and Pacific Islander residents, although Twitter does seem associated with
faster resolution times within these neighborhoods compared to counterparts with average or lower
amounts of Asian and Pacific Islander residents. This effect is also consistent across time.
DISCUSSION AND IMPLICATIONS
With respect to the use of Open311, the results suggest that reports submitted through
purpose-built mobile phone applications are more likely on average to be resolved faster than those
that were phoned in or made through the 311 website, both in general and across income and
104
demographic strata. Using Open311 is associated with faster average closure times in low-income
areas as well as those with greater proportions of Hispanic and African-American residents, although
it has no discernable impact for reports made in areas with high proportions of Asian and Pacific
Islanders. While it is important to note that this decrease in average time to resolution diminishes as
reports remain open longer, the fact that 50% of all reports are closed within 3 days suggests
Open311’s impact remains relevant for the plurality of reports. Indeed, the effective regression to
the mean as survival time increases for Open311 reports is likely indicative of the unobservable
characteristics of hard-to-resolve reports rather than a shortcoming of the technology. The singular
exception to Open311’s superior performance relative to the telephone and website is in areas with
relatively low or high rates of homeownership. For areas with high homeownership, it may be that
residents are able to lobby over the phone for expedition in addressing their concerns as property
owners, either due to a perceived right as (property) taxpayers or the fact that they are more likely to
vote in local elections than renters (or both). These possibilities suggest the need for additional
research into the way in which 311 call center employees are able to exercise bureaucratic discretion
in assisting individuals making reports.
The findings for the effect of the use of Twitter are substantially more contingent. In
general, the use of Twitter is associated with longer average times to resolution compared to
preexisting submission technologies, which is congruent with hypothesis 2a. Moreover, this effect
increases over time, implying that difficult-to-resolve reports made by Twitter take longer to resolve
than do difficult reports made by other means. That reports made through Twitter tend, on average,
to have decreasing relative odds of being resolved may be attributable to the difficulties inherent to
Twitter in communicating with high levels of detail. If initial reports lack sufficient information or
details for city workers to correctly identify and fix problems, the 311 operators are required to ‘re-
Tweet’ the original reporter to ask for clarification. It is possible that this process leads to dialogic
105
communication that does not directly address the city’s need for information, and/or instances
where the time between the initial report and the follow-up leave the reporter unable (or unwilling)
to recall the necessary information.
However, the differences in average time to resolution associated with the use of Twitter
relative to the internet and telephone in different demographic strata offer support for the
hypotheses that the public nature of Twitter-based reports could lead to faster resolution time both
in general and particularly in disadvantaged areas. With the exception of low-income areas, where
the use of Twitter was not statistically different than the web or phone, reports made through
Twitter are associated with faster average closure times for all three racial and ethnic strata, as well as
for those made in high-income areas and places with both low- and high-homeownership rates. In
areas with low homeownership or high proportions of historically disadvantaged groups, this effect
may be attributable to the salience provided by social media. Breakdowns in the urban environment
in disadvantaged areas may go unnoticed by politicians, bureaucrats, and the median voter when
reported through the telephone or internet.
With Twitter, however, the fact that these reports are observable by anyone who searches by
the 311 hashtag and can include photographic evidence of problems may make these reports hard to
ignore. This is especially true given the political climate in San Francisco during the observed period,
as the city has wrestled (and continues to do so) with significant concerns over inequality in all facets
of public life. At the same time, the associated decrease in average times to resolution for Twitter-
submitted reports in high-income and high-homeownership areas may be a result of their public
nature, raising the risk – whether potential or actualized – of political action by the residents of these
more wealthy parts of the city. It may also be simply a magnification of the preexisting difference in
average time to resolution for reports made by telephone or internet in high-income relative to mid-
income areas.
106
While the reported results are almost universally highly statistically significant, their
interpretation is not without caution or limitations regarding generalizability. First, this study
considers usage and outcomes in one jurisdiction that arguably is not highly representative of other
major cities. Moreover, while it is often useful as a case study for early adopter phenomena, San
Francisco’s idiosyncrasies call for caution in generalizing findings to other major metropolitan areas,
although it should be noted that the findings largely comport with previous work done regarding
311 systems in Boston. It would be helpful to replicate the results from Boston and San Francisco,
both cities that have a relatively concentrated population of highly educated and higher income
residents, with cities that have lower levels of education and higher concentrated poverty.
As mentioned earlier, another key limitation is that the demographic data are limited to the
2010 census tract level and not associated with discrete 311 reports. This is a common issue with
research on 311 systems, as they do not collect demographic information on users by design. Clark
and Brudney (2014) circumvent this problem by using survey data on 311 users from the City of San
Francisco, but this comes at the expense of not being able to employ the report data, making it
inappropriate for answering the questions motivating this study. This analysis would also benefit
from a distinction between phone reports made via landline and those made via cellular phone, but
San Francisco’s 311 database does not support such granularity. Finally, while this study analyzes the
differences in resolution time associated with different reporting technologies, it does not address
questions about the specific impact of departments and types of reports on these times. Future
research will be devoted to examining whether there is variation in ‘buy in’ on coproduction across
different departments and agencies.
Despite these limitations, this research advances our understanding of both coproduction
and innovation through an empirical assessment of the efficiency and equity impacts of introducing
two substantively different new technologies to 311, a preexisting coproduction regime in San
107
Francisco. It finds that the use of mobile applications in the form of the Open311 standard is
associated with decreases in the average time to resolution for reports compared to those made
through the telephone or the 311 website. This effect holds true for reports made in most
disadvantaged areas, with the exception of those with low homeownership rates. The majority of
reports generated through Open311 experience the decrease in survival time on average, although
these effects diminish over time. The research further finds evidence that the use of Twitter for
submitting 311 reports is associated with an average increase in the time taken for a report to be
resolved compared to preexisting technologies. However, the opposite effect is found when
examining reports made in both disadvantaged and privileged communities. This may be due to the
effect of social media highlighting distributional concerns in a city facing rapidly growing inequality,
as well as acting as focal point for mobilizing residents to demand expeditious service at either end
of the socioeconomic distribution.
These findings improve the body of knowledge on coproduction by showing how the
disintermediation between citizen and government generated through new technological systems can
alter the operation of coproductive arrangements in practice. They also provide evidence to support
the theory that the source and characteristics of innovation in public sector organizations matter for
their performance. While San Francisco is in may ways an edge case, especially as it relates to
technology, over time these technologies will – and in many cases already have – diffused across the
municipal landscape. It is crucial that public managers understand the potential effects on both
coproduction systems in general and in particular 311 systems’ distributive outcomes and overall
efficiencies as they consider whether and how to introduce new technologies for engaging citizens in
planning and producing public goods.
108
Conclusion
The normative concerns underlying this dissertation deal with the importance of local
government service provision, public sector innovation, and how information and communication
technologies are used by public organizations and affect them in turn. While proponents of Digital
Era Governance are right to claim that this second wave of ICT innovations brings significant
possibilities to enhance citizen-government interactions through disintermediation, the research has
not paid sufficient attention to organizational factors and institutional constraints on
implementation. This dissertation seeks to answer two questions: (1) how do bureaucracies respond
to pressure to implement technologies that entail disruptive internal and external change; and (2)
does that change improve outcomes?
The case of open data in US local governments helps to illuminate the potential benefits and
risks that would likely drive or inhibit implementation by highlighting the paired nature of potential
benefits and attendant risks facing public managers who seek to innovate using open data. As a
result, it augments and ultimately strengthens the Digital Era Governance framework proposed by
Dunleavy et al. (2006) by calling attention to the calculus that public managers face in terms of
tradeoffs when implementing open data. Indeed, implementing open data offers local governments
the potential to augment and improve performance management and promote economic
development, while giving constituents increased capacity to observe and monitor their government
and new digitally-delivered services at little to no cost for the city. I argue that at the same time,
however, local governments face the risk of goal displacement from departments focusing on data
generation and sharing at the expense of organizational objectives, the embarrassment that can come
from public transparency, replacing public employees with unaccountable technocratic elites in
determining how and what data-driven applications are developed, and, as a result, losing control
over service delivery. I conclude the first essay by proposing a set of internally- and externally-
109
oriented steps public managers can take to mitigate these risks. Within their organization, managers
can invest the time and effort necessary to develop implementation plans and requirements that
account for differences in how departments collect and use data, and provide additional
administrative support to minimize the impact of implementing open data on other organizational
duties and responsibilities.
The second essay builds on the first by developing and testing hypotheses regarding the
organizational and institutional characteristics that should affect the relative level of open data
implementation in departments within US cities with populations of 100,000 or more as of the 2010
Census that have city open data sites. It operationalizes implementation levels as the count of files
that meet open data standards and are made available by departments through the city's open data
site, and argues that relative implementation rates are likely to be a function of both department- and
city-level characteristics. At the department level, its size and degree of administrative capacity –
measured as the ratio of managers to staff – should be positively associated with the number of files
made available; service-oriented departments should make more data available than those in the
administrative core; and departments with data useful for economic development should have the
largest relative number of files available. At the city level, implementation is likely to hinge on
administrative capacity in the form of either chief data or information officers, having a centralized
IT department, having a professionalized (i.e., a city manager) administration, choosing to contract
out for developing the open data system itself, and demand pressures, operationalized as the
proportion professionally employed residents (defined by the Census Bureau as working in business,
science, and the arts).
The findings suggest that the factors driving open data implementation in US cities are
largely department-level characteristics: both hypotheses about department size and degree of
administrative capacity were supported. While the findings on department type were mixed to the
110
extent that several service-oriented departments were not statistically significantly different from
administrative departments, economic development-focused departments were likely to have more
than twice the number of files available than those within the administrative core. Meanwhile, the
only city-level characteristics that were statistically different from zero were the proxy for demand-
side pressures to make data available and demographic controls for median household income,
unemployment, and the proportion of non-Hispanic Black residents. Form of government,
contracting out for open data development, having centralized IT, and adding executive-level
administrative capacity had no detectable effect on the number of files available through city open
data sites.
The third and final essay contributes to the literature on technology innovation by
considering the actual impact of implementation on service delivery. It considers this question both
in terms of efficiency, measured in terms of time to resolve public service quality issues, and in terms
of distributive equity. The essay exploits a quasi-natural experiment in the City of San Francisco's
311 system, a service for residents to report breakdowns in public services and other nuisances,
which introduced both Twitter and a mobile phone application called Open311 as new ways to file
reports. Using panel data of more than a quarter of a million reports made over 7 years, I use the
technology employed to file a report as the treatment and, by controlling for confounding factors,
establish causal estimates of the effect of each technology relative to each other and the preexisting
reporting methods. I test for distributive implications by measuring the degree to which these effects
differ according to the racial and socioeconomic demographics of the vicinity where the report was
made.
The evidence shows that using Open311 results in a faster resolution time for the average
report across the city, while using Twitter to make a report results in longer resolution times than
both Open311 and the call center. Twitter's underwhelming performance, however, was reversed in
111
areas of the city with the highest median incomes as well as the highest proportion of minorities,
suggesting that the public nature of reporting public service breakdowns on a social media platform
may add a political dimension to the city's response times.
The three essays taken together make several contributions to the field of public
management. First, they identify and detail the potential benefits and risks that public managers face
when implementing open data. They also make several empirical and methodological contributions.
The project has led to a novel, nationwide database of open data implementation in large US cities,
as well as a cross-sectional database of staffing data for over 1,500 departments across 60 cities.
These data are then used to identify how institutional and organizational characteristics promote or
inhibit implementation in practice. This dissertation makes a further empirical contribution through
its causal estimates of how implementing new technology affects organizational performance in
responding to requests for government services.
EXTENSIONS AND FUTURE RESEARCH AGENDA
Going forward, my research agenda continues to focus on service provision and innovation,
including several extensions of this dissertation. Some of these will leverage the data collected for
and used here; others will build on the theoretical contributions. The following are brief summaries
of research questions and ideas for these future projects.
Understanding How Open Data Are Used In Practice
One future project is a logical extension of the second essay: an exploratory study of what
happens once data are made available through an open data system. It will answer key questions that
have yet to be addressed beyond individual case studies, specifically: Who uses open data? How are
the data used in practice? Do those uses correspond to the goals set by the decision makers who
chose to adopt open data? I am particularly interested in understanding whether and why users and
112
uses vary systematically across cities, and the degree to which open data are being used in ways that
either directly meet or complement organizational as well as user needs. Another facet of this
research is determining the degree to which the services that use these data make their dependence
on public sector information known to the end-user. This has potential implications for the public's
understanding of and appreciation for the government's role in making such services available,
which in turn has direct consequences for democratic decision-making (Mettler, 2011).
Comparing Open Data to Other Service Digitization Arrangements
A related empirical extension will be an analysis of how service digitization differs by service
provision arrangement. Open data represents one way for creating mobile applications and other
forms of digital service delivery, but it is not alone. Another option which predates open data is the
concept of a “hackathon,” where amateurs compete in a fixed period of time to create a program or
application that meets an organization's needs. These needs are generally expressed in terms of
software functionality (e.g., create an application that identifies park facilities based on reservation
availability, programmatic functions, sports fields, etc.) or measures of efficiency (e.g., read and
optimize transportation patterns and routes as fast as possible while minimizing travel time or
distance). The public sector can also either contract out directly for such services or provide them in
house. In the latter case, the City of Boston in particular has made significant investments in civic
application development; former Mayor Menino's New Urban Mechanics office has digitized many
services and created new ones, and has been directly copied by the City of Philadelphia.
Compared to these options, open data represents more of a 'Field of Dreams' approach
where digital service development is expected to happen by virtue of source material being freely
available. But even across instances of city-level open data implementation, some cities are more
proactive with respect to engaging private and non-profit developers and advocacy groups to
promote their data and spur its use. This research project would assess whether these approaches
113
yield consistently different outcomes in terms of the quality of digitial service provision; the types of
services provided; the level of adoption and/or use by the public; and the degree to which they
address concerns about access and equity (e.g. Does it offer multilingual support? Does it provide a
service that is of use to disadvantaged residents or visitors?). Both internal and external validity of
these measures will likely entail collaboration with Computer Science and/or Information Systems
faculty.
Measuring The Impact of Fiscal Shocks on Open Data and Digital Service Provision
Open data and service digitization are relatively new phenomena. Since their adoption and
implementation, cities have not faced a systematic economic downtown of any magnitude, let alone
that of the 2008 economic crash, that would lead to program downsizing and other austerity
measures. Whether either or both of these innovations would survive the political maelstrom of
budget cuts remains an open question, and one that speaks directly to their potential for long-term
impact in public organizations. Building on the data collection methodology that I will develop for
the projects on open data use and digital service provision, I intend to build a longitudinal dataset of
the state of open data and digital service delivery in US cities that will, eventually, have both pre- and
post-shock observations. This will allow me to employ a discontinuity design to determine whether
such innovations are vulnerable to fiscal pressures, are unaffected by them altogether, or if local
governments turn to them as substitutes for more labor-intensive modes of service delivery.
Identifying the Effect of Managerial Changes in Open Data Implementation
Building on the second essay's cross-sectional analysis of open data implementation in US
cities, I plan to exploit and expand the dataset to account for changes that cities have made to their
open data policies and administrative staffing over time. These changes include creating new
executive-level managerial positions in the form of Chief Data Officers (CDOs) and/or requiring
114
departments to designate staff for assignment to work directly with senior city-level management on
making data available on the city's open data site. These changes were explicitly made to address
perceived under-implementation by city departments. While the results from the second essay of this
dissertation do not find a statistically significant effect from having a CDO, the model is cross-
sectional and therefore cannot account for whether the introduction of this position had an effect
on the rate of implementation – again, measured as the volume of files made available by
department – over time in the given city.
Open Data and Performance Management: Complements or Complications?
Another new research project is based on one of the potential benefits of open data
identified in the first essay: its ability to complement performance management systems. This project
uses the City of Los Angeles' joint implementation of performance management and open data in
2013 as its research setting. Through a combination of multiple survey waves of 1,500 managers
across all city departments, in-depth case studies of 10 departments, and documentary analysis of
both performance management and open data as implemented, the study examines how the two
data-driven initiatives affect each other in practice. This includes the extent to which the data used
to evaluate departments and/or the evaluations themselves are made available as open data;
managerial perceptions of making evaluative data public; the degree to which managers consider
open data and performance management related and/or an asset or liability; and whether making
data available through the city's open data site is an explicit component of performance management
plans.
Conclusion
This dissertation and the work that will follow contribute to our understanding of public
administration and management in several key ways. First, these technological innovations carry
115
fiscal, administrative, and opportunity costs; understanding their potential and demonstrable value
and risks is important for democratic governance on its face. But beyond that, they are quintessential
examples of how technology is “embedded” in organizational structures and hierarchies, making
their implementation as much a social challenge as a technical one (Fountain, 2001). They also serve
to break down preexisting structures for how the public interfaces with government, replacing
street-level bureaucrats with relational databases – what Digital Era Governance refers to as
disintermediation – to a degree that simply was not possible with previous technology (Margetts &
Dunleavy, 2013). This potential for change carries significant opportunities and risks. Many of the
normative arguments for such innovations have been documented (Fishenden & Thompson, 2013;
Goldsmith & Crawford, 2014). With this research, we now have a more complete picture of the
tradeoffs public managers face in implementing them, what factors promote or inhibit their
implementation, and whether they offer demonstrable improvements over the status quo.
References
Ackoff, R. L. (1989). From data to wisdom. Journal of applied systems analysis, 16(1), 3-9.
Ae Chun, S., Luna-Reyes, L. F., Sandoval-Almazán, R., Carlo Bertot, J., Jaeger, P. T., & Grimes, J.
M. (2012). Promoting transparency and accountability through ICTs, social media, and
collaborative e-government. Transforming Government: People, Process and Policy, 6(1), 78-91.
Amirkhanyan, A. A., Kim, H. J., & Lambright, K. T. (2008). Does the public sector outperform the
nonprofit and for‐profit sectors? Evidence from a national panel study on nursing home
quality and access. Journal of Policy Analysis and Management, 27(2), 326-353.
Andersen, O. J. (2008). A bottom-up perspective on innovations mobilizing knowledge and social
capital through innovative processes of bricolage. Administration & Society, 40(1), 54-78.
Andrews, R., Boyne, G., & Mostafa, A. M. S. (2017). When bureaucracy matters for organizational
performance: exploring the benefits of administrative intensity in big and complex
organizations. Public Administration, 95(1), 115-139.
Baldwin, J. N., Gauld, R., & Goldfinch, S. (2012). What public servants really think of e-
government. Public Management Review, 14(1), 105-127.
Behn, R. D. (2003). Why measure performance? Different purposes require different measures.
Public Administration Review, 63(5), 586-606.
Bel, G., Fageda, X., & Warner, M. E. (2010). Is private production of public services cheaper than
public production? A meta‐regression analysis of solid waste and water services. Journal of
Policy Analysis and Management, 29(3), 553-577.
Berners-Lee, T. (2011). Linked data-design issues (2006).
https://www.w3.org/DesignIssues/LinkedData.html, 10, 11.
Berry, F. S., Berry, W. D., & Foster, S. K. (1998). The determinants of success in implementing an
expert system in state government. Public Administration Review, 293-305.
Bertelli, A. M. (2012). The political economy of public sector governance: Cambridge University Press.
Bimber, B. (2001). Information and political engagement in America: The search for effects of
information technology at the individual level. Political Research Quarterly, 54(1), 53-67.
Bimber, B. (2003). Information and American democracy: Technology in the evolution of political power:
Cambridge University Press.
Bohte, J., & Meier, K. J. (2000). Goal displacement: Assessing the motivation for organizational
cheating. Public Administration Review, 60(2), 173-182.
117
Borins, S. (2000). Loose cannons and rule breakers, or enterprising leaders? Some evidence about
innovative public managers. Public Administration Review, 60(6), 498-507.
Borins, S. (2001). Innovation, success and failure in public management research: some
methodological reflections. Public Management Review, 3(1), 3-17.
Bovaird, T. (2007). Beyond engagement and participation: User and community coproduction of
public services. Public Administration Review, 67(5), 846-860.
Bovaird, T. (2014). Attributing outcomes to social policy interventions–‘Gold standard’or ‘fool's
gold’in public policy and management? Social Policy & Administration, 48(1), 1-23.
Bovens, M., & Zouridis, S. (2002). From street‐level to system‐level bureaucracies: how information
and communication technology is transforming administrative discretion and constitutional
control. Public Administration Review, 62(2), 174-184.
Box-Steffensmeier, J. M., & Zorn, C. J. (2001). Duration models and proportional hazards in
political science. American Journal of Political Science, 972-988.
Bozeman, B., & Bretschneider, S. (1986). Public management information systems: Theory and
prescription. Public Administration Review, 475-487.
Bozeman, B., & Su, X. (2015). Public service motivation concepts and theory: A critique. Public
Administration Review, 75(5), 700-710.
Bradbury, M. D., & Kellough, J. E. (2007). Representative bureaucracy: Exploring the potential for
active representation in local government. Journal of public administration research and theory,
18(4), 697-714.
Brandsen, T., & Honingh, M. (2015). Distinguishing different types of coproduction: A conceptual
analysis based on the classical definitions. Public Administration Review.
Brown, L., & Osborne, S. P. (2013). Risk and innovation: Towards a framework for risk governance
in public services. Public Management Review, 15(2), 186-208.
Brown, T. L., & Potoski, M. (2003). Transaction costs and institutional explanations for government
service production decisions. Journal of public administration research and theory, 13(4), 441-468.
Brudney, J. L. (1985). Coproduction Issues in Implementation. Administration & Society, 17(3), 243-
256.
Bureau of Labor Statistics. (2017). Current Employment Statistics for 2015. Retrieved from:
https://www.bls.gov/ces/ - data
Chadwick, A. (2008). Web 2.0: New challenges for the study of e-democracy in an era of
informational exuberance. Isjlp, 5, 9.
Chang, A.-M., & Kannan, P. (2008). Leveraging Web 2.0 in government: IBM Center for the Business of
Government Washington, DC.
118
Chen, J.-S., Tsou, H.-T., & Ching, R. K. (2011). Co-production and its effects on service innovation.
Industrial Marketing Management, 40(8), 1331-1346.
Choi, T., & Chandler, S. M. (2015). Exploration, exploitation, and public sector innovation: an
organizational learning perspective for the public sector. Human Service Organizations:
Management, Leadership & Governance, 39(2), 139-151.
Christensen, T., & Lægreid, P. (2005). Trust in government: The relative importance of service
satisfaction, political factors, and demography. Public Performance & Management Review, 28(4),
487-511.
Chun, S. A., Shulman, S., Sandoval, R., & Hovy, E. (2010). Government 2.0: Making connections
between citizens, data and government. Information Polity, 15(1), 1.
Clark, B. Y., & Brudney, J. L. (2014). Representation and Beyond: Coproduction and the New
Information Technology. Available at SSRN 2473543.
Clark, B. Y., Brudney, J. L., & Jang, S. G. (2013). Coproduction of government services and the new
information technology: Investigating the distributional biases. Public Administration Review,
73(5), 687-701.
Clark, B. Y., & Guzman, T. (2014). Does Local Government Coproduction Lead to Budget
Adjustments? An Investigation of Boston, MA and San Francisco, CA. An Investigation of
Boston, MA and San Francisco, CA (August 20, 2014).
Clark, B. Y., & Guzman, T. S. (2016). Does Technologically Enabled Citizen Participation Lead to
Budget Adjustments? An Investigation of Boston, MA, and San Francisco, CA. The American
Review of Public Administration, 0275074016642568.
Coursey, D., & Norris, D. F. (2008). Models of e‐government: Are they correct? An empirical
assessment. Public Administration Review, 68(3), 523-536.
Daft, R. L. (1978). A dual-core model of organizational innovation. Academy of management journal,
21(2), 193-210.
Damanpour, F. (1987). The adoption of technological, administrative, and ancillary innovations:
Impact of organizational factors. Journal of management, 13(4), 675-688.
Damanpour, F. (1991). Organizational innovation: A meta-analysis of effects of determinants and
moderators. Academy of management journal, 34(3), 555-590.
Damanpour, F., & Evan, W. M. (1984). Organizational innovation and performance: the problem
of" organizational lag". Administrative science quarterly, 392-409.
Damanpour, F., & Schneider, M. (2006). Phases of the adoption of innovation in organizations:
Effects of environment, organization and top Managers1. British Journal of Management, 17(3),
215-236.
119
Damanpour, F., & Schneider, M. (2009). Characteristics of innovation and innovation adoption in
public organizations: Assessing the role of managers. Journal of public administration research and
theory, 19(3), 495-522.
Dawes, S. S. (2010). Stewardship and usefulness: Policy principles for information-based
transparency. Government Information Quarterly, 27(4), 377-383.
Dawes, S. S., Cresswell, A. M., & Pardo, T. A. (2009). From “need to know” to “need to share”:
Tangled problems, information boundaries, and the building of public sector knowledge
networks. Public Administration Review, 69(3), 392-402.
De Vries, H., Bekkers, V., & Tummers, L. (2015). Innovation in the public sector: A systematic
review and future research agenda. Public Administration.
DiMaggio, P., & Powell, W. W. (1983). The iron cage revisited: Collective rationality and
institutional isomorphism in organizational fields. American Sociological Review, 48(2), 147-160.
Dunleavy, P., Margetts, H., Bastow, S., & Tinkler, J. (2006). New public management is dead—long
live digital-era governance. Journal of public administration research and theory, 16(3), 467-494.
Eisner, E. W. (2002). The arts and the creation of mind: Yale University Press.
Emerson, K., Nabatchi, T., & Balogh, S. (2012). An integrative framework for collaborative
governance. Journal of public administration research and theory, 22(1), 1-29.
Ferman, B. (1996). Challenging the growth machine: Neighborhood politics in Chicago and Pittsburgh: Univ Pr
of Kansas.
Fernandez, S., & Wise, L. R. (2010). An exploration of why public organizations ‘Ingest’innovations.
Public Administration, 88(4), 979-998.
Ferris, J., & Graddy, E. (1986). Contracting out: For what? With whom? Public Administration Review,
332-344.
Ferris, J. M. (1984). Coprovision: Citizen time and money donations in public service provision.
Public Administration Review, 324-333.
Fishenden, J., & Thompson, M. (2013). Digital government, open architecture, and innovation: why
public sector IT will never be the same again. Journal of public administration research and theory,
23(4), 977-1004.
Fountain, J. E. (2001). Building the virtual state: Information technology and institutional change: Brookings
Institution Press.
Fountain, J. E. (2004). Building the virtual state: Information technology and institutional change: Brookings
Institution Press.
Fung, A. (2006). Varieties of participation in complex governance. Public Administration Review, 66(s1),
66-75.
120
Ganapati, S., & Reddick, C. G. (2012). Open e-government in US state governments: Survey
evidence from Chief Information Officers. Government Information Quarterly, 29(2), 115-122.
Ganapati, S., & Reddick, C. G. (2014). The use of ICT for open government in US municipalities:
Perceptions of chief administrative officers. Public Performance & Management Review, 37(3),
365-387.
Gerrish, E. (2016). The Impact of Performance Management on Performance in Public
Organizations: A Meta‐Analysis. Public Administration Review, 76(1), 48-66.
Gil-Garcia, J. R. (2012). Towards a smart State? Inter-agency collaboration, information integration,
and beyond. Information Polity, 17(3, 4), 269-280.
Girth, A. M., Hefetz, A., Johnston, J. M., & Warner, M. E. (2012). Outsourcing public service
delivery: Management responses in noncompetitive markets. Public Administration Review,
72(6), 887-900.
Goldsmith, S., & Crawford, S. (2014). The responsive city: Wiley.
Grimmelikhuijsen, S. G., & Welch, E. W. (2012). Developing and testing a theoretical framework for
computer‐mediated transparency of local governments. Public Administration Review, 72(4),
562-571.
Gunn, L. (1988). Public management: a third approach? Public Money & Management, 8(1-2), 21-25.
Hansen, M. B. (2011). Antecedents of organizational innovation: The diffusion of new public
management into Danish local government. Public Administration, 89(2), 285-306.
Hart, O. D., & Holmstrm, B. (1986). The theory of contracts.
Heeks, R. (2002). Reinventing government in the information age. Reinventing government in the
information age: International practice in IT-enabled public sector reform, 9-21.
Hefetz, A., & Warner, M. E. (2012). Contracting or public delivery? The importance of service,
market, and management characteristics. Journal of public administration research and theory, 22(2),
289-317.
Heinrich, C. J., & Marschke, G. (2010). Incentives and their dynamics in public sector performance
management systems. Journal of Policy Analysis and Management, 29(1), 183-208.
Hill, M., & Hupe, P. (2002). Implementing public policy: Governance in theory and in practice: Sage.
Ho, A. T.-K., & Ni, A. Y. (2004). Explaining the adoption of e-government features a case study of
Iowa county treasurers’ offices. The American Review of Public Administration, 34(2), 164-180.
Hood, C. (1991). A public management for all seasons. Public Administration, 69(1), 3-19.
Hood, C. (2012). Public Management by Numbers as a Performance‐Enhancing Drug: Two
Hypotheses. Public Administration Review, 72(s1), S85-S92.
121
Hoppe, E. I., & Schmitz, P. W. (2010). Public versus private ownership: quantity contracts and the
allocation of investment tasks. Journal of Public Economics, 94(3), 258-268.
Hosmer, D. W., Lemeshow, S., & May, S. (2008). Applied survival analysis. Hoboken: NJ, Wiley.
Houston, D. J. (2000). Public-service motivation: A multivariate test. Journal of Public Administration
Research and Theory: J-PART, 713-727.
Hupe, P. (2014). What happens on the ground: Persistent issues in implementation research. Public
Policy and Administration, 29(2), 164-182.
Isett, K. R., & Miranda, J. (2015). Watching Sausage Being Made: Lessons learned from the co-
production of governance in a behavioural health system. Public Management Review, 17(1), 35-
56.
Jaeger, P. T. (2007). Information policy, information access, and democratic participation: The
national and international implications of the Bush administration’s information politics.
Government Information Quarterly, 24(4), 840-859.
Jakobsen, M., & Andersen, S. C. (2013). Coproduction and equity in public service delivery. Public
Administration Review, 73(5), 704-713.
Jakobsen, M., James, O., Moynihan, D., & Nabatchi, T. (2016). JPART Virtual Issue on Citizen-
State Interactions in Public Administration Research. Journal of public administration research and
theory, muw031.
Joassart-Marcelli, P., & Musso, J. (2005). Municipal service provision choices within a metropolitan
area. Urban Affairs Review, 40(4), 492-519.
Jun, K.-N., & Musso, J. (2013). Participatory governance and the spatial representation of
neighborhood issues. Urban Affairs Review, 49(1), 71-110.
Jun, K.-N., & Weare, C. (2010). Institutional motivations in the adoption of innovations: The case
of e-government. Journal of public administration research and theory, muq020.
Jun, K.-N., & Weare, C. (2011). Institutional motivations in the adoption of innovations: The case
of e-government. Journal of public administration research and theory, 21(3), 495-519.
Kassen, M. (2013). A promising phenomenon of open data: A case study of the Chicago open data
project. Government Information Quarterly, 30(4), 508-513.
Kiewiet, D. R., & McCubbins, M. D. (2014). State and local government finance: The new fiscal ice
age. Annual Review of Political Science, 17, 105-122.
Kim, S., & Lee, J. (2012). E‐participation, transparency, and trust in local government. Public
Administration Review, 72(6), 819-828.
Kiser, L. L. (1984). Toward an institutional theory of citizen coproduction. Urban Affairs Review,
19(4), 485-510.
122
Klein, K. J., & Sorra, J. S. (1996). The challenge of innovation implementation. Academy of management
review, 21(4), 1055-1080.
Lamothe, M., & Lamothe, S. (2012). To trust or not to trust? What matters in local government-
vendor relationships? Journal of public administration research and theory, 22(4), 867-892.
Lee, G., & Perry, J. L. (2002). Are computers boosting productivity? A test of the paradox in state
governments. Journal of public administration research and theory, 12(1), 77-102.
Lee, J. (2008). Determinants of Government Bureaucrats' New PMIS Adoption The Role of
Organizational Power, IT Capability, Administrative Role, and Attitude. The American Review
of Public Administration, 38(2), 180-202.
Levy, F., Meltsner, A. J., & Wildavsky, A. B. (1974). Urban outcomes: Schools, streets, and libraries: Univ of
California Press.
Linders, D. (2012). From e-government to we-government: Defining a typology for citizen
coproduction in the age of social media. Government Information Quarterly, 29(4), 446-454.
Mancini, R. (2012). Raising the Bar for E‐Government. Public Administration Review, 72(6), 829-829.
Margetts, H., & Dunleavy, P. (2013). The second wave of digital-era governance: a quasi-paradigm
for government on the Web. Philosophical Transactions of the Royal Society of London A:
Mathematical, Physical and Engineering Sciences, 371(1987), 20120382.
Marvel, J. D., & Resh, W. G. (2015). Bureaucratic discretion, client demographics, and representative
bureaucracy. The American Review of Public Administration, 45(3), 281-310.
May, P. J., & Winter, S. C. (2007). Politicians, managers, and street-level bureaucrats: Influences on
policy implementation. Journal of public administration research and theory, 19(3), 453-476.
Mayor's Office of Communication, S. F. (2013). PRESS RELEASE: MAYOR NEWSOM
ANNOUNCES INNOVATIVE NEW TOOL TO ACCESS CITY SERVICES WITH A
“TWEET” [Press release]. Retrieved from
http://sf311.org/Modules/ShowDocument.aspx?documentid=273
Meier, K. J., & O'Toole, L. J. (2003). Public management and educational performance: The impact
of managerial networking. Public Administration Review, 63(6), 689-699.
Meier, K. J., & O’Toole, L. J. (2013). Subjective organizational performance and measurement error:
Common source bias and spurious relationships. Journal of public administration research and
theory, 23(2), 429-456.
Meijer, A. J. (2014). From Hero-Innovators to Distributed Heroism: An in-depth analysis of the role
of individuals in public sector innovation. Public Management Review, 16(2), 199-216.
Merton, R. K. (1968). Social theory and social structure: Simon and Schuster.
123
Mettler, S. (2011). The submerged state: How invisible government policies undermine American democracy:
University of Chicago Press.
Minkoff, S. L. (2015). NYC 311 A Tract-Level Analysis of Citizen–Government Contacting in New
York City. Urban Affairs Review, 1078087415577796.
Moldogaziev, T. T., & Resh, W. G. (2016). A Systems Theory Approach to Innovation
Implementation: Why Organizational Location Matters. Journal of public administration research
and theory, muv047.
Molotch, H. (1976). The city as a growth machine: Toward a political economy of place. American
journal of sociology, 82(2), 309-332.
Moon, M. J. (2002). The evolution of e‐government among municipalities: rhetoric or reality? Public
Administration Review, 62(4), 424-433.
Moon, M. J., & Bretschneiber, S. (2002). Does the perception of red tape constrain IT
innovativeness in organizations? Unexpected results from a simultaneous equation model
and implications. Journal of public administration research and theory, 12(2), 273-292.
Moon, M. J., Lee, J., & Roh, C.-Y. (2014). The Evolution of Internal IT Applications and e-
Government Studies in Public Administration Research Themes and Methods. Administration
& Society, 46(1), 3-36.
Moon, M. J., & Norris, D. F. (2005). Does managerial orientation matter? The adoption of
reinventing government and e‐government at the municipal level. Information Systems Journal,
15(1), 43-60.
Moore, M., & Hartley, J. (2008). Innovations in governance. Public Management Review, 10(1), 3-20.
Mossberger, K., Wu, Y., & Crawford, J. (2013). Connecting citizens and local governments? Social
media and interactivity in major US cities. Government Information Quarterly, 30(4), 351-358.
Moynihan, D., Herd, P., & Harvey, H. (2014). Administrative burden: Learning, psychological, and
compliance costs in citizen-state interactions. Journal of public administration research and theory,
muu009.
Moynihan, D. P. (2008). The dynamics of performance management: Constructing information and reform:
Georgetown University Press.
Musso, J., Weare, C., & Hale, M. (2000). Designing Web technologies for local governance reform:
Good management or good democracy? Political Communication, 17(1), 1-19.
Musso, J., Young, M. M., & Thom, M. (2017). The Political Economy of Co-Production: Voluntary
Police and Fire in California Cities. under review at Public Management Review.
Musso, J. A., & Weare, C. (2005). Implementing electronic notification in Los Angeles: citizen
participation politics by other means. International Journal of Public Administration, 28(7-8), 599-
620.
124
Nabatchi, T. (2012). Putting the “public” back in public values research: Designing participation to
identify and respond to values. Public Administration Review, 72(5), 699-708.
Nam, T. (2012). Suggesting frameworks of citizen-sourcing via Government 2.0. Government
Information Quarterly, 29(1), 12-20.
Nam, T. (2015). Challenges and Concerns of Open Government: A Case of Government 3.0 in
Korea. Social Science Computer Review, 33(5), 556-570.
Nelson, K. L., & Svara, J. H. (2011). Form of government still matters: Fostering innovation in US
municipal governments. The American Review of Public Administration, 0275074011399898.
Norris, D. F., & Moon, M. J. (2005). Advancing e-government at the grassroots: tortoise or hare?
Public Administration Review, 64-75.
Norris, D. F., & Reddick, C. G. (2013). Local e‐government in the United States: Transformation or
incremental change? Public Administration Review, 73(1), 165-175.
Norris, P. (2001). Digital divide: Civic engagement, information poverty, and the Internet worldwide: Cambridge
University Press.
O'Toole, L. J. (1986). Policy recommendations for multi-actor implementation: An assessment of
the field. Journal of public policy, 6(02), 181-210.
Obama, B. H. (2009). MEMORANDUM FOR THE HEADS OF EXECUTIVE
DEPARTMENTS AND AGENCIES. Retrieved from
https://obamawhitehouse.archives.gov/realitycheck/the-press-office/transparency-and-
open-government
Open Knowledge International. (2017). The Open Definition. Retrieved from
http://opendefinition.org/
Pardo, T. A., Cresswell, A. M., & Thompson, F. (2001). INTERORGANIZATIONAL
KNOWLEDGE SHARING IN PUBLIC SECTOR INNOVATIONS. Paper presented at
the Academy of Management Proceedings.
Park, S., & Hendry, D. J. (2015). Reassessing Schoenfeld residual tests of proportional hazards in
political science event history analyses. American Journal of Political Science, 59(4), 1072-1087.
Parks, R. B., Baker, P. C., Kiser, L., Oakerson, R., Ostrom, E., Ostrom, V., . . . Wilson, R. (1981).
Consumers as coproducers of public services: Some economic and institutional
considerations. Policy Studies Journal, 9(7), 1001-1011.
Peled, A. (2016). Coerce, Consent, and Coax: A Review of US Congressional Efforts to Improve
Federal Counterterrorism Information Sharing. Terrorism and Political Violence, 28(4), 674-691.
Percy, S. L. (1984). Citizen participation in the coproduction of urban services. Urban Affairs Review,
19(4), 431-446.
125
Perry, J. L., Hondeghem, A., & Wise, L. R. (2010). Revisiting the motivational bases of public
service: Twenty years of research and an agenda for the future. Public Administration Review,
70(5), 681-690.
Peterson, P. E. (1981). City limits: University of Chicago Press.
Petychakis, M., Vasileiou, O., Georgis, C., Mouzakitis, S., & Psarras, J. (2014). A state-of-the-art
analysis of the current public data landscape from a functional, semantic and technical
perspective. Journal of theoretical and applied electronic commerce research, 9(2), 34-47.
Posner, R. A. (2010). From the new institutional economics to organization economics: with
applications to corporate governance, government agencies, and legal institutions. Journal of
Institutional Economics, 6(1), 1.
Powers, K. J., & Thompson, F. (1994). Managing coprovision: Using expectancy theory to
overcome the free-rider problem. Journal of public administration research and theory, 4(2), 179-
196.
Pressman, J. L., & Wildavsky, A. B. (1973). Implementation: How Grert Expectations in Washington are
Dashed in Oakland; Or, why It's Amazing that Federal Programs Work at All: University of
California Press.
Radin, B. (2006). Challenging the performance movement: Accountability, complexity, and democratic values:
Georgetown University Press.
Radin, B. (2009). Response to Dennis Smith. Journal of Policy Analysis and Management, 28(3), 514-516.
Rainey, H. G. (1982). Reward preferences among public and private managers: In search of the
service ethic. The American Review of Public Administration, 16(4), 288-302.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods
(Vol. 1): Sage.
Reddick, C. G. (2004). A two-stage model of e-government growth: Theories and empirical evidence
for US cities. Government Information Quarterly, 21(1), 51-64.
Reddick, C. G., & Aikins, S. K. (2012). Web 2.0 Technologies and Democratic Governance: Springer.
Riccucci, N. M., Meyers, M. K., Lurie, I., & Han, J. S. (2004). The Implementation of Welfare
Reform Policy: The Role of Public Managers in Front‐Line Practices. Public Administration
Review, 64(4), 438-448.
Richard, S. W. (1992). Organizations: Rational, natural, and open systems. Aufl., Englewood Cliffs (NJ).
Rivera, M. D., Streib, G., & Willoughby, K. G. (2000). Reinventing government in council-manager
cities: Examining the role of city managers. Public Performance & Management Review, 121-132.
Robinson, D. G., Yu, H., Zeller, W. P., & Felten, E. W. (2009). Government data and the invisible
hand.
126
Rogers, E. M. (2010). Diffusion of innovations: Simon and Schuster.
Rosenbloom, D. H. (1983). Public administrative theory and the separation of powers. Public
Administration Review, 219-227.
Rowley, J. (2007). The wisdom hierarchy: representations of the DIKW hierarchy. Journal of
information science, 33(2), 163-180.
Sabatier, P., & Mazmanian, D. (1980). The implementation of public policy: A framework of
analysis. Policy Studies Journal, 8(4), 538-560.
Salamon, L. M. (1995). Partners in public service: Government-nonprofit relations in the modern welfare state:
JHU Press.
Sanger, M. B. (2008). From measurement to management: Breaking through the barriers to state and
local performance. Public Administration Review, 68(s1).
Schweitzer, L. (2014). Planning and social media: a case study of public transit and stigma on
Twitter. Journal of the American Planning Association, 80(3), 218-238.
Smith, A. (2010). Government Online. Retrieved from
http://www.pewinternet.org/2010/04/27/government-online/
Smith, D. C., & Bratton, W. J. (2001). Performance management in New York City: Compstat and
the revolution in police management. Quicker, better, cheaper, 453-482.
Sorkin, A. R. (2017, April 17 2017). Steve Ballmer Serves Up a Fascinating Data Trove. The New
York Times. Retrieved from
https://www.nytimes.com/2017/04/17/business/dealbook/steve-ballmer-serves-up-a-
fascinating-data-trove.html
Stone, C. (1989). Regime politics: governing Atlanta, 1946-1988: Univ Pr of Kansas.
Stone, C. (2005). Looking back to look forward: Reflections on urban regime analysis. Urban Affairs
Review, 40(3), 309-341.
Streib, G. D., & Willoughby, K. G. (2005). Local governments as e-governments: Meeting the
implementation challenge. Public Administration Quarterly, 77-109.
Thomas, J. C. (2013). Citizen, customer, partner: Rethinking the place of the public in public
management. Public Administration Review, 73(6), 786-796.
Thorsby, J., Stowers, G. N., Wolslegel, K., & Tumbuan, E. (2016). Understanding the content and
features of open data portals in American cities. Government Information Quarterly.
Tolbert, C. J., & Mossberger, K. (2006). The effects of e‐government on trust and confidence in
government. Public Administration Review, 66(3), 354-369.
127
Tolbert, C. J., Mossberger, K., & McNeal, R. (2008). Institutions, policy innovation, and E‐
Government in the American States. Public Administration Review, 68(3), 549-563.
USAFacts Institute. (2017a). About Us. Retrieved from https://usafacts.org/about
USAFacts Institute. (2017b). United States Government Annual Report For The Fiscal Year Ended September
30, 2014. Retrieved from https://usafacts.org/resources/USAFacts_10-K_2017.PDF
van Deursen, A. J., & Van Dijk, J. A. (2014). The digital divide shifts to differences in usage. New
media & society, 16(3), 507-526.
Vega, C. (2007). San Francisco / Dial 311 — for everything but emergencies. San Francisco Chronicle.
Retrieved from www.sfgage.com/bayarea/article/SAN-FRANCISCO-Dial-311-for-
everything-but-2567298.php
Verba, S., Schlozman, K. L., & Brady, H. E. (1995). Voice and equality: Civic voluntarism in American
politics: Harvard University Press.
Verba, S., Schlozman, K. L., Brady, H. E., & Brady, H. E. (1995). Voice and equality: Civic voluntarism in
American politics (Vol. 4): Cambridge Univ Press.
Walker, R. M. (2008). An empirical evaluation of innovation types and organizational and
environmental characteristics: Towards a configuration framework. Journal of public
administration research and theory, 18(4), 591-615.
Walker, R. M. (2014). Internal and External Antecedents of Process Innovation: A review and
extension. Public Management Review, 16(1), 21-44.
Walker, R. M., Avellaneda, C. N., & Berry, F. S. (2011). Exploring the diffusion of innovation
among high and low innovative localities: a test of the Berry and Berry model. Public
Management Review, 13(1), 95-125.
Wang, S., & Feeney, M. K. (2016). Determinants of information and communication technology
adoption in municipalities. The American Review of Public Administration, 46(3), 292-313.
Warschauer, M. (2004). Technology and social inclusion: Rethinking the digital divide: MIT press.
Weare, C., Musso, J. A., & Hale, M. L. (1999). Electronic democracy and the diffusion of municipal
web pages in California. Administration & Society, 31(1), 3-27.
Weber, M. (2009). From Max Weber: essays in sociology: Routledge.
Weerakkody, V., Irani, Z., Lee, H., Osman, I., & Hindi, N. (2015). E-government implementation: A
bird’s eye view of issues relating to costs, opportunities, benefits and risks. Information systems
frontiers, 17(4), 889-915.
Weible, C., Sabatier, P. A., & Lubell, M. (2004). A Comparison of a Collaborative and Top‐Down
Approach to the Use of Science in Policy: Establishing Marine Protected Areas in California.
Policy Studies Journal, 32(2), 187-207.
128
Welch, E. W., Feeney, M. K., & Park, C. H. (2016). Determinants of data sharing in US city
governments. Government Information Quarterly, 33(3), 393-403.
West, D. M. (2004). E‐government and the transformation of service delivery and citizen attitudes.
Public Administration Review, 64(1), 15-27.
West, D. M. (2005). Digital government: Technology and public sector performance: Princeton University
Press.
Williams, A. (2009). On the release of information by governments: Causes and consequences.
Journal of Development Economics, 89(1), 124-138.
Wilson, J. Q. (1989). Bureaucracy: What government agencies do and why they do it: Basic Books.
Wood, L., Bernt, P., & Ting, C. (2009). Implementing public utility commission web sites: Targeting
audiences, missing opportunities. Public Administration Review, 69(4), 753-763.
Zuiderwijk, A., & Janssen, M. (2014). Open data policies, their implementation and impact: A
framework for comparison. Government Information Quarterly, 31(1), 17-29.
Abstract (if available)
Abstract
This dissertation considers how technological innovations are implemented in local governments. Public organizations are under increasing pressure to implement new, innovative technologies in order to be more transparent, efficient, and improve service delivery. But both the challenges public managers face in implementing such innovations as well as the question of whether they improve outcomes in practice remain under-examined. This work uses two recent innovations—open data and the introduction of mobile applications and social media in municipal 311 systems—to directly address these issues. ❧ The first essay critically examines open data implementation by local governments. It draws upon the theoretical framework of Digital Era Governance (DEG) to argue that open data is a case study in the type of “radical disintermediation” expected by DEG, and that this framework is useful for understanding how open data can be both beneficial and detrimental. Open data’s potential benefits include service digitization
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
How can metrics matter: performance management reforms in the City of Los Angeles
PDF
The context of leadership in the development of California’s innovation hubs
PDF
Community development agreements: addressing inequality through urban development projects
PDF
Emerging technology and its impact on security policy
PDF
Reproducing inequity in organizations: gendered and racialized emotional labor in pubic organizations
PDF
Lessons learned from the FHWA State Infrastructure Bank (SIB) program, 1995 to 2016
PDF
University campus expansion: an analysis of Harvard University’s expansion into Boston, Massachusetts
PDF
A learning model of policy implementation: implementing technology in response to policy requirements
PDF
Governing public goods: how representation and political power in local and regional institutions shape inequalities
PDF
Community, empowerment, and the city: sources of capacity in local governance
PDF
A framework for good local governance: achieving prosperity in an increasingly complex environment
PDF
Choice neighborhoods: a spatial and exploratory analysis of Housing Authority City of Los Angeles public housing
PDF
Essays on mitigation, adaptation, and resilience to urban fiscal crises
PDF
The crisis of potable water in Mexico City: institutional factors and water property rights as conditions for creating adequate metropolitan water governance
PDF
Diverging pathways to citizenship and immigrant integration in the U.S.
PDF
Innovation and good intentions: evaluations of three cross-sectoral programs for at-risk populations in Southern California
PDF
Elements of a successful multi-sectoral collaborative from a local government perspective: a framework for collaborative governance – dimensions shared by award winning multi-sectoral partnerships
PDF
The functions of the middleman: how intermediary nonprofit organizations support the sector and society
PDF
How does collaborative governance work? The experience of collaborative community-building practices in Korea
PDF
Environmental justice in real estate, public services, and policy
Asset Metadata
Creator
Young, Matthew Michael
(author)
Core Title
Technological innovation in public organizations
School
School of Policy, Planning and Development
Degree
Doctor of Philosophy
Degree Program
Public Policy and Management
Publication Date
07/15/2019
Defense Date
04/12/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
311,coproduction,digital era governance,innovation,OAI-PMH Harvest,open data,Public Management,Technology
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Musso, Juliet (
committee chair
), Resh, William (
committee member
), Schweitzer, Lisa (
committee member
)
Creator Email
matthew.m.young@usc.edu,mmyoung@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-402745
Unique identifier
UC11265806
Identifier
etd-YoungMatth-5540.pdf (filename),usctheses-c40-402745 (legacy record id)
Legacy Identifier
etd-YoungMatth-5540.pdf
Dmrecord
402745
Document Type
Dissertation
Rights
Young, Matthew Michael
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
311
coproduction
digital era governance
innovation
open data