The multiple institutional logics of innovation
Department of Political Science
Department of Public Administration
Department of Politics and History
Liverpool Hope University
Department of Political Science
University of California, Riverside
Department of Political Science
The Ohio State University
How do decentralized systems deal with innovation? In particular, how do they aggregate the
myriad experiences of their component parts, facilitate diffusion of information, and encourage
investments in innovation? This is a classic problem in the study of human institutions. It is also
one of the biggest challenges that exists in the governance of decentralized systems: how do
institutions shape individual behavior around solving problems and sharing information in a
fashion that is reasonably compatible with collective well being? We use a particular
decentralized institution (the House of Representatives), wrestling with a novel problem (how to
utilize the Internet) to explore the implications of three archetypical principles for organizing
collective problem solving: market, network, and hierarchy.
Institutional Entrepreneurship, Collaborative Learning, Information Technology, Congress, Diffusion, intra-, inter-
and extra-organizational influence, online practices.
Acknowledgements: We gratefully acknowledge the support of NSF grant No. 0429452. Any opinions, findings, and
conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of
the National Science Foundation (NSF).
How do decentralized systems deal with innovation? In particular, how do they aggregate
the myriad experiences of their component parts? This is a classic problem in the study of
human institutions— at all scales, from small groups of individuals wrestling with common
problems to policy learning across national boundaries. It is one of the biggest issues in collective
action and thus governance: how do institutions shape individual behavior around solving
problems and sharing information in a fashion that is reasonably compatible with collective well
being? In fact, much of human history can be viewed through the lens of information production
and sharing: how different ways of organizing humans provided reasons for individuals to solve
other people‟s problems (Diamond 1998). Innovation presents a particular governance challenge
within the public sector because the market-based mechanisms (e.g., intellectual property rights)
that have evolved to encourage innovation seemingly do not apply to the public sector. There is
no way for an innovative jurisdiction to profit from the successful adoption by another of its
We use a particular decentralized institution (the House of Representatives), wrestling
with a novel problem (how to utilize the Internet) to explore the implications of three
archetypical principles for organizing collective problem solving: market, network, and
hierarchy (Powell 1990). We find that these three institutional logics intertwine in this particular
case. The market turns out to be a powerful mechanism for fueling innovation, because an array
of vendors has emerged that supplies web services to House offices. These vendors provide the
scale necessary for the development of cutting edge features to House websites. Information
about vendors, in turn, flows through interpersonal networks, although there is relatively little
consultation among House offices about actual features on websites work. There are multiple
hierarchical mechanisms that direct and limit what Members do with their websites. There are
House rules that limit the content of websites, and parties play a limited role in subsidizing
innovative practices. We do find, however, evidence of underinvestment in innovative practice
within offices, with little critical examination of what constituents want and need from Member
websites. That is, while there are powerful drivers of conformity (DiMaggio and Powell 1983)
with respect to Internet practices, the system fails at harnessing the collective capacity of these
offices for problem solving.
PROBLEM SOLVING IN DECENTRALIZED SYSTEMS
Imagine a system where the agents in the system are each playing the same game against
the environment. Because these games are similar to one another, the lessons learned in one
game have implications for the best strategy in the other games. These potential informational
externalities create a governance challenge: how do different principles for organizing collective
human efforts affect the pooling of experiences, create incentives to innovate, and to share the
lessons learned from a particular experiment?
Here we explore the role that three institutional mechanisms for organizing collective
human effort play in the aggregation of problem solving: market, hierarchy, and network (cf
Powell 1990). By market we mean that the problem is solved through an arms-length transaction
between actors with a problem (consumers) and actors who can address that problem
(producers). By hierarchy we mean that the problem is solved through authoritative fiat or
through concentration of power and resources in a system. And we define problem solving in
networks to refer to the informal exchange and flow of favors and information within a given
system, outside the existing hierarchical reporting structures.
The governance challenges with respect to decentralized innovation revolve around the
production, sharing, and aggregation of information:
Challenge 1—Information production: The very possibility that the lessons learned by
one actor would be useful to another means that there may be a mismatch between private
incentives and public benefits to experimentation. Innovations that will diffuse and benefit
others may be underproduced or not shared, because actors do not incorporate those benefits to
others in their calculations. The incentive issue creates a subsidiary issue of scale—where there
may be some innovations that only produce net benefits if there are multiple adopters. To take an
extreme example, imagine an actor could invest c to yield benefits less than c, but where many
other actors could then receive those benefits for free. In the absence of some type of return from
the benefits others receive, it will not be in the interest of any single actor to produce them.
Challenge 2—Information sharing: The fact that innovative practices can be public
goods—where the innovative practices of one actor do not reduce the value those practices offer
other actors, means that communicating those practices can greatly increase collective welfare.
However, sharing sometimes comes with a cost—perhaps quite high, depending on the quality of
information being conveyed. A key question, then, is whether actors share private information
when it would increase overall welfare.
Challenge 3—Information aggregation: Even if information is being shared, it does not
mean that there will be a convergence to optimal array of practices. For any given domain, there
are likely many alternative practices, and a key question will be whether the system of
communication facilitates a convergence toward optimal practices. As the information cascade
literature highlights (Bikhchandani et al. 1992), it is eminently possible that “thin”
communication processes, that do not convey all private information, may result in the
convergence on practices that, if all private information were properly pooled, would be
understood to be suboptimal.
A key question, then, is how do these various institutional mechanisms succeed or fail at
addressing these governance issues. We consider each institutional archetype in turn.
The issue of market failure and information production has been well explored in
economics. Economic models emphasize a few mechanisms to address the issue of
informational externalities. The first is intellectual property protection. That is, if an actor
produces an innovation that produces value for others, intellectual property rights provide a
mechanism for the innovator to gain some rents from other beneficiaries. In principle,
intellectual property protection addresses much of the issue of information production and
sharing, because actors internalize some of the benefits other actors receive from an innovation.
Apple invested in the iPhone because it gained substantial profits from its sales; similarly, it has
not been shy about telling the world about the benefits of the iPhone. Finally, one of the
potentially compelling features of markets is that they aggregate information through price
signals (Hayek 1945).
These mechanisms of governance are seemingly inapplicable to the public sector, because
there is limited capacity to gain rents from successful policies that are emulated elsewhere (e.g.,
Massachusetts cannot charge other states for lessons learned from its health coverage reforms),
and constitutional limits on consolidation (e.g., New York cannot launch a hostile takeover of
New Jersey). However, as the case study below illustrates, for some policies, there is significant
potential for market actors to “join” the informational ecosystem that includes the relevant
government actors, developing and offering innovations that can then be transferred to other
government actors through payment. Indeed, much production of various kinds in the public
sector has been outsourced (sometimes called the hollowing out of the state — Milward &
Provan 2000), where the same private firm serves multiple jurisdictions. The potential of this
market mechanism depends on:
(1) The capacity of actors to contract out production. Certain types of policies can be
contracted out and others not, e.g., because of their complexity (Brown, Potoski, and Van Slyke
forthcoming). For example, implementation of IT policy, to some extent, can be outsourced,
which means that lessons learned from implementation in one location can be utilized by the
contractor elsewhere, providing the incentive to the contractor to innovate if they can capture
some of the resulting rents. This example also highlights that non-government consumers on the
demand side are also part of the broader information ecosystem, because the same contractors are
providing services to both government and nongovernment entities.
(2) Formal or informal mechanisms of intellectual property rights. There will only be an
incentive for contractors to innovate if those innovations are not easily emulated. If another actor
could easily copy what was done in one location, then it is not possible for the original innovator
to capture rents from successful innovations. Such emulation can be blocked through formal
property protection (e.g., through the development of proprietary systems), or through informal
mechanisms (the development of systems that are difficult to copy because of their complexity).
Market-based approaches create their own dysfunctions. The existence of intellectual
property rights creates market power, and necessarily shifts surpluses from consumers to
producers (contractors). An exclusive market arrangement also means that innovation will be
biased toward the development of systems that are protected by formal or informal intellectual
property rights. Some of the support for the development of open source software for
government use has derived from this dual critique (Hamel and Schweik 2009). That is, the open
source software movement may be seen as support for outsourcing to the “commons” rather than
to private sector entities.
As the US system of dual sovereignty (and the case study below) highlights, systems can
have strands of both hierarchy and decentralization. That is, one can embed certain elements of
hierarchy in an essentially decentralized system. Hierarchical governance resolves the scale issue
by, essentially, removing (at least partly) “decentralized” from decentralized decision-making.
There are no informational externalities for a centralized decision maker (Strumpf 1999). Thus, a
hierarchy can consolidate systemic production, potentially eliminating the issues of externalities.
A hierarchy can also subsidize innovation, providing resources to subsidiary actors for
experimentation. Alternatively, a centralized decision maker can authoritatively constrain or
mandate particular behaviors—e.g., mandating that particular units attempt innovative policies
that might not be in their self-interest to pursue. In the US many policy domains have a blend of
these relationships between the federal and state governments, where, for example, with
Medicaid there is a mix of some flexibility with federal mandates and dollars.1
Hierarchies, of course, have their own failings in the creation and dissemination of
information. First, if production is consolidated, the hierarchy is creating an internal monopolist
that may have the same issues around pricing (or efficiency) that a monopolist in the market
would have (Williamson 1971). Second, even in a world where agents confront identical
problems, they might have different information (or perspectives) on what are promising
solutions, thus potentially reducing the exploration of alternative policies. As Brandeis pointed
out long ago, one of the presumed benefits of a decentralized system is that it promotes
experimentation, and subsequent dissemination of successful policies.2 Constraints on the
behaviors of actors would, from this perspective, reduce innovation.3 Third, the central actor
may confront vast informational overload in terms of its capacity to aggregate different
information, especially if peripheral actors have subtly different problems. In fact, this was the
primary Hayekian critique of central planning.
1 Of course, federal-state relationships entail an array of interdependencies beyond policy information (Esterling
2 New State Ice Company v. Liebmann (285 U.S. 262, 311).
3 Analytically, this is a tricky question, since the central actor would presumably have greater capacity to extensively
evaluate policy alternatives than peripheral actors.
The premise of a network approach to innovation is that agents have differentiated
connections and/or finite attention. That is, A can emulate B only if it sees what B is doing.
Visibility depends on the nature of the innovation. Some “innovations,” by their nature, are
publicly visible. This is part of a larger pattern of systems where units can cheaply observe or
refer to particular other units, where that link requires no reciprocal effort.
The literature on networks and governance has emerged in the breach—where market or
hierarchy are clearly missing and an informal system has emerged that apparently addresses the
underlying governance problem. Notably, Ostrom (1990, 2009), for example, has focused on the
role that networks (among other things) regulate behavior in the provision of collective goods in
the absence of a hierarchy. This research has focused on the provision of common pool
resources (e.g. Ostrom et al. 1994), but, as discussed below, much of the logic could apply to the
creation and dissemination of innovative information. The emergence of enforced norms is
likely one key component of information production and sharing—e.g., recognition of helpful
behavior, punishment of actors who do not share information (Mergel et al. 2008).
The economic sociology literature, in contrast, has focused on the role of networks where
markets fail. Much of the work on social capital (in particular, the vein of work following from
Bourdieu (2001) and Coleman (1988) is really about how networks govern dimensions of
exchange that would otherwise fail because of asymmetric information. A substantial literature
has emerged around “network industries”—e.g., Broadway, construction, apparel— where
economic actors are repeatedly reconfigured around existing projects (Uzzi 1999; Jones 1996;
Jones et al. 1997). This literature has focused on the role that (1) relational embeddedness
(repeat interactions with the focal actor); and (2) structural embeddedness (reputation vis-à-vis
third parties) play in regulating the behavior of individuals (Uzzi 1996).
There is also a rapidly emerging literature on public organizations, which has focused, for
example, on the role that networks play in the success of public managers (Meier & O'Toole
2001); as well as the role that network structure plays in the provision of public services (Provan
& Kenis 2008; Milward et al. 2010; Provan & Milward 2001) . Most of this literature, we would
note, focuses on formal networks, rather than emergent, informal networks (Isett et al. 2011) on
which we focus on in this paper.
How do informal networks address the three governance challenges outlined above? Part
of the answer may lie in reciprocity (Axelrod & Hamilton 1981), where dyads of actors with long
run relationships develop cooperative relationships. However, at most, reciprocity can be a small
part of the answer, because the benefits of exchange within a dyad will not reflect the broader
systemic benefits resulting from information sharing. It is not clear how dyadic reciprocity will
encourage investments in innovation that reflect the benefits to the whole system. It is unclear
whether dyadic reciprocity will also reflect the benefits that third parties (tied to one or both
members of the dyad) get when that information is subsequently shared with them. It is possible
that information exchanged within a dyad offers currency for exchange in other dyads, but it is
not at all obvious that this should lead to a healthy equilibrium of information sharing. Finally,
as the information cascade literature highlights, it is not clear that dyadic informational exchange
should yield effective information aggregation. Rather than reciprocity, more promising, is that
informal networks might foster the emergence of pro-social norms that encourages information
sharing (Mergel et al. 2008). Norms that, for example, reward risk taking and innovation with
status; or where information sharing is publicly recognized and seen as appropriate; or where
sharing otherwise private information regarding failure as well as success is encouraged, would
certainly increase innovation, information sharing, and information aggregation, respectively.
As a general proposition, all of these organizational forms coexist, in different degrees.
For example, if you look at one of the classic studies of diffusion—tetracycline (Coleman, Katz,
and Menzel 1957)—all three mechanisms were at work within the system. At its core, this was a
study of the decentralized decision-making of doctors, whose objective was to take care of their
patients. Coleman and collaborators focused on how doctors learned through their network
whether tetracycline was effective. However, not too far in the background was the fact that
there was a market for drugs, with powerful intellectual property rights, which facilitated the
creation of the drug in the first place, and which created an incentive to share information about
the drug (in fact, a recent re-analysis—Van den Bulte and Lilien (2001) — of the Coleman data
suggest that the network had little effect on behavior once one controlled for who was targeted
for marketing). Further, in the background was a legal and regulatory regime (i.e., hierarchy) that
(1) protected those intellectual property rights, and (2) approved of the drug for circulation.
Our objective is to use the emergence of a particular “problem”, the use of the Internet by
members of Congress, to explore how these mechanisms for organizing collective problem
solving coexist. We now turn to a discussion of our methodology.
Case selection and context
The empirical focus of this paper is on the use of the official websites by congressional
offices. There are two primary reasons that we chose this as a setting in which to study
innovation. First, is the novelty of the medium, and second is the decentralized management of
Congressional offices. We discuss each in turn.
Novelty of the medium: Every member of the House currently has an official website.
These websites present a virtual representation of the member to the world, and most notably, to
his/her constituents. The existing rules, such as the Franking rules, focus on the use of paper-
based communication with constituents, so that the Internet is still a relatively novel medium for
offices with a lot of uncertainty to what is acceptable practice.
Given the wide reach of the Internet, these official websites offer a potentially powerful
means to communicate with constituents and the public in general. However, it is also a
relatively novel medium for communication—as of the late 1990s, only about half of the
Members even had a website, and a few years before that the World Wide Web did not even
exist. A casual perusal of websites suggests a substantial degree of convergence, but also
nontrivial variation and experimentation. From our perspective, learning about the power of the
medium is still fresh. This provides the opportunity for us to examine this learning process.
The following screenshots show a) an example of one of the first websites of a Senator
(in this case Senator Edward Kennedy in 1993) and b) an example of a highly developed website
by Representative Mike Honda who used a crowdsourcing approach to include his constituents‟
feedback into the web-design process:
Edward Kennedy: First senator with a website 1993
Figure 1: Sample Congressional websites
Mike Honda‟s crowd-sourced website in 2009
Decentralized nature of Congress: Congressional offices can be considered as 440 small,
functionally identical, and independent public organizations (Salisbury and Shepsle 1981). As
one staffer we interviewed stated: “There‟re 435 small businesses here, and each „CEO‟ can do
what they want.”4 There are some House rules that limit how offices can use their websites, but
within those constraints, Congressional offices collectively come close to matching the
assumption that decision-making is truly decentralized.
There is clearly some interdependence of payoffs in the success of Members—e.g.,
Democrats have a stake in the success of other Democrats and in the failure of Republicans, and
vice versa. However, the payoff a particular Member gets out of the effective use of the official
website by a specific other Member is surely fairly tiny.
We would also note that Members confront somewhat different challenges in
communicating with constituents. The communication needs of a Member from a rural Colorado
district differs from the needs of a Member representing Manhattan. However, all offices are
operating under similar resource constraints, and similar desires to satisfy constituents and
portray a positive image of the Member. Further, much of this heterogeneity is visible both to
the offices and to the researcher; and one key subsidiary question we pursue is how this
heterogeneity affects the search process.
4 There are 435 voting Members, and five non-voting delegates.
We conducted interviews with the Congressional staff person who had primary
responsibility for the official Member website of 99 Members of Congress in the summer of
2006. We were assisted in recruitment by the Congressional Management Foundation, a small
nonprofit organization dedicated to helping Members of Congress better manage their offices.
The sample was constructed purposively to be roughly reflective of the body as a whole, but, due
to the vagaries of who was willing to cooperate, is biased toward affluent urban districts,
Democrats, and offices with above average websites (Table A1 in the appendix compares key
descriptives for the body as a whole to the sample). In our analyses below we examine whether
our observations are robust across subsamples, focusing, in particular, on under-represented
strata. We note where we find significant deviations across subsamples.
The semi-structured interviews each lasted about 45 minutes. The interviews were
transcribed and each statement in every interview coded in an iterative process using the
qualitative data analysis program NVivo 2 (QSR 2002).
At approximately the same time we conducted the interviews we also conducted a survey
of the Communications Directors in House offices. This survey was more broadly about
communication strategy, but did include a few items regarding where offices got information
about what to do with their websites. We received 100 responses from the 440 offices, for a
response rate of 23%.5
Our core research question is to understand the ways in which Congress pools the
experiences of Congressional offices vis-à-vis the use of the Internet. We break this down into
three subsidiary research questions, matching the three governance challenges laid out above:
(1) How do offices learn about which of their practices are successful and which are
How do offices learn from other offices? In particular, how do they learn which
of the practices of other offices are successful and which are not successful?
What institutions outside of offices have a major impact on aggregating
information, provide incentives for certain practices, and set the normative
5 There was an overlap of 53 offices between our two samples.
The first question focuses on feedback from the environment—e.g., do offices gather data
on constituent preferences? Do they track what information web-surfers look at, and what
proportion of hits are from the district? This information has possible positive externalities,
because what is learned by one office potentially has relevance for others. For example, if an
office finds that particular content viewed more than others, this could be beneficial to other
offices if this observation has relevance to what issues they might place on the website. This
information might be transmitted through interpersonal communication, or through the simple
observation of what that office is and is not doing with its website. Finally, what role in
aggregating information do the institutional structures within Congress play? The two obvious
institutions with an interest in the successful aggregation of information are (a) the parties, and
and (b) the administrative infrastructure of Congress.
Our analytic interest is how the system pools together the experiences of various offices‟
use of their official websites. Viewing the Member‟s office as the relevant locus of decisions
about how to use the website, we split our analysis up into (1) environmental feedback; (2)
market-driven influences on decisions regarding websites; (3) hierarchical drivers of practices;
and (4) inter-office (network) flows of experiences.
Innovation: the role of environmental feedback
The Internet is a communication medium. The primary function of an official website is
to facilitate communication with constituents. A key part of viewing the innovations around the
official websites is to understand how offices assess whether a particular intervention was
effective. In this case, do offices find out what works and does not work through feedback from
their constituents? The Internet offers a particular promise for learning from the environment,
because it is possible to track with some detail what visitors to the website are doing (e.g.,
number of hits, pages visited, referrers, etc.).
We therefore evaluate two questions: to what extent do offices (1) proactively assess
what their target audience (generally, constituents) wanted from the website; and (2) ex post
assess whether their website was hitting the target?
The data definitively show that there was remarkably little effort by offices to proactively
assess what constituents wanted. In fact, only two out of 99 offices (both representing districts
with above average educations and incomes) indicated any type of research into what their
constituents wanted, as one of them describes:
We … sent out a survey to about 40,000 constituents… asking whether podcasting was a
feature they‟d use, and whether tele-video conferences, online town halls, all [that] stuff,
what did they want, what should our very limited resources be devoted to. The website was
one of the top ones, without a doubt. Podcasting, in contrast, had a very limited response…
The lack of active proactive research on what constituents want is, perhaps, unsurprising,
given the expense of surveys, focus groups, etc. The monitoring of the use of the website,
however, is much cheaper. Readily available data which are usually automatically collected
include: number of unique visitors and page views; what parts of the website are viewed; the
approximate geographic location of each visitor to the website; who is linking to the website;
where traffic to the website is coming from. It should therefore be easy to produce regular
reports indicating how much traffic the website is getting, and from what part of the district,
looking at which web pages. It is therefore surprising that few offices reported looking regularly
(or even recently) at these data (although the offices, when asked about this, very often noted this
lack with regret). Only four of the interviewees specifically stated that they used web traffic
reports on a regular basis to help determine how they should operate their websites, and two
other interviewees indicated that they collected information about their web traffic but did not
state how they used this information. All other offices stated that they did not get the
information, did not use the information, had not looked at it recently, or did not know the
current specifics about their website‟s traffic. Even those offices that reported tracking what
parts of the website got hits did not examine the data carefully, as this one exchange highlights:
[Q: Do you look at what pages on your website get the most hits?]
[Q: Do you do anything with that?]
We used to do it more, I actually need to find out. (Sigh)
From another office:
We don‟t monitor as closely as I think would be helpful in terms of the Web
traffic. We do an email program that runs separately that I can tell what links
people are clicking on, but we don‟t often go back to the website to see, okay, this
page was the most visited.
In short, assessment of practices with respect to these websites was quite limited. There
was certainly evidence over time of the emergence of novel practices—e.g., new features to a
website—but there was little assessment of whether those practices were effective, even in the
minimal sense of a count of how many times they were used.
As noted above, markets resolve collective problem solving through the development of
proprietary information, which is then monetized and sold (often times in the form of services).
In this case, we did observe that a population of small companies that caters to the web needs of
Congressional offices has emerged during the past decade. Congressional offices are provided
with a fixed budget, with which they have a fair degree of discretion. For some services, they
therefore confront a build or buy decision—do they use staff to build a website, or purchase
expertise to do so? Many offices have pursued the latter route, where a small cottage industry of
firms have sprung up to offer web services to Members, where those firms are divided along
Democratic and Republican (e.g., one Republican oriented firm was called “Right click”) lines.
Fifty-eight percent of the offices in our survey reported hiring an outside consultant, where the
following were typical comments from interviews:
We had a great vendor that allowed us to change our whole front page. We could
do a lot of things internally.
[O]ur vendor came up with the ideas for how exactly to make the tour pages, a lot
of pages, automated. Because when we did that, that was almost five years ago,
so that was when nobody was doing it. Now it‟s pretty standard to have some of
The essential market logic behind the outsourcing of the website is that there are
economies of scale in the problem solving process, where the unit cost of production for many
websites is lower than the cost of producing a single website. In principle, the resulting surplus
can be divided among profit, quality of website, and the ability to offer lower prices to offices.
15 Download full-text
The effectiveness of the market in part rests on how well information can be protected.
That is, if one vendor comes up with an innovative solution at some cost, and either offices or
other vendors are able to copy that solution cheaply, there will be an underinvestment in
innovation. In this particular domain, protection likely does not come in the form of intellectual
property rights, but through the development of expertise that cannot be easily copied or built
internally. However, that may still leave much relevant information unprotected, as highlighted
by the statement of one clever staffer made use of vendor services in another way:
One of the things that I‟ve seen several external consultants do to pursue our
business is send us a review and a report on our website, sort of analyzed by their
staff and saying this really should be updated, this really should be changed to
meet the standard government configuration or whatever… I then see that and I
say, okay, I think I‟m going to implement these things myself.
In this case, the staffer is essentially free riding on the insights provided by a vendor.
An additional issue with respect to vendors is that there will be a natural push from
vendors to homogenize their products, because it is cheaper to offer a limited set of options, and
in part because they are using off-the-shelf software designed for commercial purposes.
Individual solutions for the different offices are then adapted from the standard set of solutions
plus a personalized configuration. As one individual noted:
Looking at some of the other companies, it sounded like they were trying to sell us
more of like a template of like, what a website would look like, instead of… just
completely designed to us, and how we wanted it.
In short, vendors play a critical role in aggregating experiences and standardizing practice
through their provision of services to multiple offices.
There are two layers of hierarchy within the House; one based on authority, and one based
on power and resources. With respect to the former, the Committee on House Administration
(CHA) sets budgets and rules constraining off-budget behavior (e.g., postings to official websites
were constrained in the last 30 days before the election). They also run the technical