1Over the past few decades, two movements have rallied to the cause of digital tools that serve to make information a common good.  [1] The first lobbies for the release of information held by the state, a demand voiced since the end of the 2000s by advocates of open data.  [2] In many countries, this has led to legislative reforms and to data produced by public authorities being made available in digital form.  [3] The second protests against the commodification of knowledge that has gone hand in hand with the advent of so-called information capitalism.  [4] It brings together defenders of hacker culture and free software  [5] who wish to take back control of the production, dissemination, and use of information from the market and the state through the creation of a “digital commons.” A “digital commons,” of which Wikipedia is the paradigmatic example, may be defined as a digital resource produced and maintained by a community according to its own rules of governance and characterized by a regime of shared ownership that is neither private nor public.  [6]

2This article describes a campaign that has brought together both of these movements. Its origin lies in a critique of the exclusive possession and use, by the Ministry of Economy and Finance (known as “Bercy”), of a particular information resource that economists call a microsimulator. This piece of software plays an important role in the ex ante economic evaluation of any reform of the tax and welfare system­that is to say, before it is actually implemented among the population.  [7] The microsimulator makes it possible for decision makers to envisage the costs of the reform, as well as to predict who will benefit and who will lose out as a result­evaluations that until now have been the sole monopoly of agents of Bercy, a few exceptions aside.  [8] Technically, the microsimulator consists of a computer model of most of the rules of the French tax and welfare system, a model of the reform to be evaluated, and a representative database of household tax and welfare data. Although the tax and welfare rules are all available online on the legifrance.fr website, they are not available in this integrated, meticulously coded form; as for the tax and welfare data, these are not accessible outside of the civil service, owing to the principle of government secrecy. This campaign, then, sought to open up tax data to actors involved in the evaluation of public policy other than Bercy’s agents and to code, law by law, the entire tax and welfare system into a microsimulator that would then be made a digital commons accessible to all. The movement began in 2011 and resulted in the building of the free software package Openfisca.  [9] Initially, it brought together various protagonists: researchers in economics who wanted to be able to come back to Bercy with legitimate counterexpertise; lawyers and developers from the new public agency Etalab,  [10] responsible for the government’s open data policy; members of parliament acting in the name of their democratic prerogatives to engage in reasoned legislative debate and to act as a check on government action; and citizen associations involved in causes related to tax justice.

3It should be noted that these actors were widely dispersed and rather marginal in comparison to a Ministry of Economy and Finance widely considered the most powerful in Europe.  [11] However, this did not prevent them from succeeding, within a few years, in making tax data accessible to researchers and producing, on the basis of this information extracted from the heart of the state machinery, a shared information resource of recognized legitimacy that is now challenging Bercy’s institutional monopoly on tax expertise. Openfisca is now used daily by thousands of citizens to assess their social rights.  [12] Members of parliament are considering using it as their own evaluation tool.  [13] In 2018, it enabled researchers from the Institut des politiques publiques (IPP) to produce the first independent ex ante economic evaluation of a draft budget law, the conclusions of which were presented to the chair of the Finance Committee, reported by most major French media outlets, and, more recently, taken up by social justice campaigns (see Figure 3).  [14] Today, Openfisca brings together some twenty partners­government departments, research institutes, town councils, students, startups, and associations­in France and elsewhere.  [15]

4In this article, we will try to explain how actors who at first sight seem marginal to the “Bercy fortress” have managed to access data of high political value and, in collaboration, have begun the process of making tax and welfare information a digital commons.

5The social science and political science literature offers several avenues for understanding the success of such a movement. First of all, the still-incomplete success of Openfisca might be seen as the product of a historical trend that has seen transparency become a cardinal political value.  [16] Pierre Rosanvallon, a French sociologist and historian of political theory, shows that the advent of representative democracy went hand in hand with the circulation of budgetary figures in the public sphere.  [17] Several laws have given legal expression to this moral duty of the state to make itself accountable, from article 15 of the Declaration of the Rights of Man and of the Citizen to France’s 2016 Lemaire law, which stipulates open access to public data by default. However, an explanation in terms of political and legal principles has several limitations. On one hand, the implementation of such regulations tends to encounter bureaucratic barriers, particularly where the Ministry of Economy and Finance is concerned.  [18] On the other hand, these principles of transparency are counterbalanced by divergent state norms, including the protection of government secrecy.  [19]

6Another approach would be to attribute the success of this movement to technological factors. According to some researchers and activists alike, digital technologies naturally facilitate the free flow and sharing of information.  [20] By driving work on free software and open data forward, they are thought to stimulate collective action  [21] and citizen action respectively.  [22] However, these kinds of explanations, often somewhat deterministic, are not entirely adequate either. On the one hand, these same digital technologies can also be used to restrict the free flow of information.  [23] On the other, as recent sociological studies show, such explanations neglect the practical and cognitive work that is necessary in order to “open up” data, put it into circulation, and reduce certain informational asymmetries.  [24]

7Finally, another approach would be to turn to the literature on commons. In her seminal work, Elinor Ostrom analyzes a large number of cases of the collective management of natural resources that are neither private nor public and identifies eight principles of governance found in all of the examples where this management is considered successful.  [25] Research on Wikipedia and open-source software since the late 2000s also seems to confirm the importance of these principles of governance in explaining the success of digital commons.  [26] As far as Openfisca is concerned, however, the limitations of any such explanation lie first of all in the fact that its governance involves public actors who are subject to particular duties, a situation that has enjoyed little attention in a field of research more focused on exploring a third way “beyond market and state.”  [27] Moreover, unlike natural resources, open-source software, or an online encyclopedia, the very constitution of the resource poses a problem here: The information used to build Openfisca is in fact difficult to identify, opaque, elusive, and politically sensitive. It seems to require a very specific kind of work in order to assemble and utilize it. This kind of assembly process has to date been little explored in research that has focused more on resources made up of more easily identifiable and less sensitive information.

8This existing literature, therefore, may provide us with interesting leads but, in isolation, is inadequate to explain the early success of the campaign around Openfisca. Distancing ourselves from any normative, technical, or organizational determinism, we have sought to investigate the Openfisca experience as it actually happened. The theoretical framework of pragmatic sociology  [28] combined with the sociology of public action  [29] seems to us best suited to track the sociotechnical, legal, and administrative challenges this movement has faced. Our research is based on twenty-eight interviews conducted with the set of actors involved in Openfisca (founders, economists, developers, MPs, Bercy agents, etc.)  [30] as well as making use of a large corpus of documentation (parliamentary reports, legislative clauses, newspaper articles, blog posts, etc.).

9Unlike the existing literature on transparency, open data, and the digital commons, this article documents a movement that aims to turn an opaque and elusive public dataset into a digital commons open to all. We show that, before setting up shared governance of this resource, substantial technical, cognitive, and institutional work had to be carried out in order to bring it together.  [31] This process involved a series of “tests” that we categorize below. We conclude by hypothesizing that Openfisca represents an institutional experiment carried out by political “hackers” on the margins of the administration, an experiment concerning a particular way in which the state might adapt to the demands of the digital commons­and that in this sense it represents a “state test” that remains in progress.  [32]

10In the first section, we will see how Bercy’s power over information is dependent on digital infrastructure that it owns, and to which it prevents access on the part of those actors who have gradually mobilized to critique this monopoly. In the second section, we trace the beginnings of this movement, paying attention to the technical, legal, and institutional choices made by the founders of Openfisca as they attempted to import certain digital-commons-inspired practices into the civil service. In the final section, we categorize three types of “tests” that the Openfisca community had to face and examine the ways in which it attempted to resolve them.

A campaign against the opacity of the tax and welfare system

11To begin, we will review how political scientists have analyzed the opacity of Bercy’s actions. As we shall see, this opacification is partly enabled by its agents’ exclusive access to microsimulators. Finally, we will present the criticisms of the actors (MPs, associations, researchers, and Etalab developers) who identified this exclusivity as a problem and advocated the possibility of producing and exploiting their own digital resource.

Disempowered MPs

17These criticisms were first voiced by MPs. It is they who receive a draft law once it has been approved by the Council of Ministers for debate, amendment, and voting. Rapporteur MPs from the Committees on Social Affairs and Economic Affairs and in particular the Finance Committee, assisted by their civil servants, translate texts written in abstruse legal language so as to render explicit and assess their consequences for tax and welfare. According to Pierre Rosanvallon, it is this process of explication that makes possible and fulfils the principle of reasoned democratic debate and parliament’s function of government oversight.  [43]

18Since the 1990s, however, MPs, rapporteurs, and civil servants have complained that they no longer have the means to evaluate tax and welfare texts.  [44] The evaluation tools available to the Finance Committee are few and far between, if not entirely non-existent in the case of ex ante economic evaluation. Indeed, the first report of the Finance Committee’s Evaluation and Control Group (Mission d’evaluation et de contrôle, MEC), as early as the turn of the 2000s, deplored the fact that

19

Draft laws coming up for discussion in our assembly are rarely subject to ex ante evaluation. The notes on potential impact that ought to accompany draft laws, when they exist, are very brief. Likewise, parliament generally has no access to any real simulation of proposals for the reform of tax or social contributions presented to it.  [45]

20MPs are then often obliged to rely on Bercy’s expertise, “without parliament being in any position to make a critical assessment.”  [46] This is why some members of the Finance Committee, including current vice chair Jean-Noël Barrot, campaigned to create a specific evaluation body that, among other things, would be equipped with a microsimulator and access to tax data to run on it.  [47]

Researchers without tools

21Bercy’s “monopoly on legitimate expertise” has also been problematized by economists from independent and academic institutes, such as Thomas Piketty.  [48] For the most part, these are actors who produce recommendations for public institutions. When we asked Antoine Bozio, director of the IPP and a former doctoral student of Piketty’s, how much power Bercy had in the drafting of laws, he answered:

22

100 percent [...] even if the political request is almost always made by the president and the prime minister, it is the minister of finance and the secretary of state for the budget who draft the law. It’s them who say: “We’ll put x amount into the prime d’activité [a French welfare benefit], and who get to see all the indicators and simulations behind it. The force, the power of expertise lies in the fact that the text is incomprehensible. [...] Bercy has an almost total monopoly on the analysis and production of budgetary data [...] and therefore wields really radical power”. (October 2018)

23The IPP, which currently includes some twenty researchers, was created in 2012 by the Paris School of Economics and the Center for Research in Economics and Statistics (Centre de recherche en économie et statistique, CREST), two major French centers for economics research. Notably, it was founded by Thomas Piketty along with François Bourguignon, an international pioneer in microsimulation research, with the aim of developing the quantitative evaluation of public policy in France and contributing to public debate. As we shall see in the second half of this article, the IPP was to play an important role in the development of Openfisca.

A segment of civil society left bewildered

24As for civil society, as the saying goes, “all are presumed to know the law.” Legal experts, public information sites, associations, and other interest groups continually obtain and decipher legislative texts in order to make their economic and social implications known and to inform citizens. As the ultimate recipients of these measures, citizens must indeed be kept informed of their rights in this way. However, as we pointed out in the introduction, such dissemination of legal information has become an increasingly complex affair over the last thirty years. According to some associations, such as the Observatory on Non-Take Up of Social Rights and Public Services (Observatoire des non-recours aux droits et services, Odenore), the increase in the rate of failure to take advantage of welfare rights shows that the poorest citizens have little grasp of the tax and welfare system compared with other dominant interest groups  [49]­an inequality that would serve to rally activists to the Openfisca cause.

Etalab, champion of open access to public data

25One final player to join the movement against Bercy’s opacity was the young agency Etalab, responsible for “open data and open government.” Since its creation in 2011, Etalab has reported directly to the prime minister, but for the past eight years it has had an unstable relationship with the government departments, testifying, according to Samuel Goëta, to “the shifting moorings of open data in government practices.”  [50] Today it is part of the Interministerial Directorate for Digital Technology and the State Information and Communication System (Direction interministérielle du numérique et du système d’information et de communication de l’État, DINSIC) and is made up of some thirty jurists, developers, data scientists, and project managers. Its principal mission is to “implement the principles of open government” by encouraging government departments to “open up” their data,  [51] sometimes leading to tense, if publicly cordial, relations with some of these departments, including Bercy.

26The Ministry of Economy and Finance handles a large amount of data, and in particular a huge number of lines of code. For example, Bercy’s tax legislation department breaks down legal texts into operational circulars that are sent to its technicians, who enter the laws into the source code of the software that calculates­and today deducts at source­a number of taxes. It may be the Journal officiel (JO) that gives us the definitive version of the law, but it is this code that actually implements the legislative texts. Lawrence Lessig’s famous phrase “code is law” takes on its full meaning here. Similarly, the agents who maintain Bercy’s microsimulators will also translate the legislative texts into computer code in order to update their tool. However, these codes and these simulators, considered public data, were not publicly available before the passing of the 2016 “Digital Republic” law (Pour une République numérique, PRN), which stipulates open access by default to all source code used by government departments. As we shall see, even before this legislation was passed, members of Etalab campaigned to have access to this software as a resource for building Openfisca.

27In the following section tracing the history of Openfisca, we will show how a number of these actors critical of the opacity of the tax and welfare system coalesced around common goals and a shared digital resource, in the hope of counteracting the asymmetry between themselves and “the Bercy Fortress.”

Openfisca: Building a collaborative infrastructure on the margins of the state

28In this section we will see how Openfisca was developed as collaborative, freely available software on the margins of the state, before enlisting a prime minister to its cause and finally becoming a resource straddling the worlds of the state, research, and the digital commons.

Openfisca enlists public actors

34However, at this stage there were still no actual contributors, and the number of laws modelled in the software was limited. As the two founders had no access to Bercy, INSEE, or CNAF’s exhaustive data, they had to run the model with “test cases,” which are useful for evaluating the impact of a reform on a hypothetical individual (for example, a single mother earning €24,000 per year), but not on the real population. Nicolas A. left France Stratégie in 2012 and, following the change of government, joined the forecasting department of the Treasury Directorate General at Bercy, where he operated the very microsimulator to which Openfisca had set itself up as a countermodel. He did try to disseminate his thinking on the opening up of digital resources among his new colleagues, organizing informal meetings with all of the departmental actors involved in microsimulation so as to “start a dialogue” (Nicolas A., October 2018). Until 2013, therefore, there was little opportunity for Openfisca to be developed further.

35Nevertheless, in December of the same year, Mahdi B. managed to capture the interest of a major actor. The sociology of innovation teaches us that a technology succeeds in developing once its defenders manage to convince a group of actors that it is a necessary vehicle, a relevant means of expression, to meet their interests.  [55] At that time, the prime minister wished to carry out a major tax reform and wanted to develop an online simulator as an educational tool to explain the reform to citizens. Mahdi B. thus found himself in Jean-Marc Ayrault’s office in his capacity as an economist for France Stratégie. He was invited by Fabien Dell, economic advisor to the prime minister, who also summoned Henri Verdier, then director of Etalab, and Antoine Bozio, director of the IPP, whom Dell knew from working with Piketty. During the meeting, Mahdi B. presented Openfisca, using a visualization tool he had built to simulate and portray the impact of a particular tax change on the monthly income and outgoings of a test case (see Figure 1).

36Recall that, toward the end of 2012, Jean-Marc Ayrault had penned a report on the use of free software within the civil service.  [56] So: the tax system was to be reformed, this reform had to be explained to citizens, and that required an online simulator, the open-source software Openfisca being the ideal tool. This chain of reasoning on the part of Mahdi B. convinced Dell and enlisted him to the cause and, through him, the prime minister. Bozio proposed to draft in the IPP by using Openfisca to model the upcoming reform, and Etalab was to be in charge of making the model accessible online. This second step in the construction of Openfisca would give it a new script oriented toward understanding the law: the goal was no longer research or decision support, but enabling citizens to find their way through the complex tangle of the tax and welfare system.

Figure 1: Breakdown of the impact of the tax and welfare system on a typical test case

Figure 1: Breakdown of the impact of the tax and welfare system on a typical test case

Explanation: Screenshot of the online simulator permitting visualization of what a typical test case (whose characteristics are indicated in the left-hand column) stands to gain (light gray) or lose (dark gray) from various welfare benefits, taxes, and other contributions (shown on the horizontal axis of the graph and detailed in the right-hand column). Source: www.youtube.com/watch?v=bEgNZsrcX54.

The state put to the test by Openfisca, and vice versa

42The first test, which we will call the test of openness, consisted in creating a practical, legal, and digital network to access Bercy’s software and tax data. The second, which we will call the legitimacy test, consisted in securing legitimacy for Openfisca among the scientific, government, and media communities. The third test, which remains unresolved, raises the question of Openfisca’s institutional positioning on the boundary of the state.

The legitimacy test

57The second test related to the process of obtaining institutional legitimacy for Openfisca as a tool for determining the official position on what the law is, in relation to other government departments, and developing recognized expertise, in relation to Bercy.

Acquiring the legitimacy to officially state what the law is

58Between 2014 and 2015, two projects aimed at making tax and welfare law intelligible were developed by the IPP and Etalab using Openfisca. The first was the “simplified pay slip [fiche de paie simplifiée]” project, which aimed to simplify, automate, and make free of charge the issuing of pay slips, a process that used to be a complex affair requiring legal expertise. The second was the “mes-aides [my helpers]” project: a website developed in connection with certain associations, designed to give citizens personalized information on what welfare benefits they are entitled to as a way to counter the non-take-up of benefits.

59Despite the apparent simplicity of these projects, making this information visible raised the problem of its official status. The question arose as to who could legitimately produce and disseminate tax and welfare information:  [67]

60

[In regard to the simplified pay slip,] people said: “It’s not for the IPP or Etalab to do the math, what we want is a guarantee from the state”. The state [...] the government departments, said: “They’re going to attack me if there’s an error, [I] can’t commit myself to saying that the tax and welfare system is like that.” So, there’s no commitment on the part of the state, there’s no guarantee, there’s just: “Go and look at the Journal officiel and if you don’t agree then sue us.” (A. Bozio, October 2018)

61This appropriation by other actors on the margins of the administration of tools and competences to advise on how the law should be applied, was experienced by government departments historically seen as the legitimate authorities on tax and welfare issues as an attack on their institution. On one hand, this new voice confronted them with their own inability to state with any certainty the rules they were supposed to be responsible for applying. On the other hand, as with the example of the tax source code discussed above, it threatened to reveal errors on their part, or at least to call their account of things into question. If “mes-aides” calculates a benefit amount that differs from the amount that their system is actually allocating to an individual, then that individual can make a claim, which would raise the question of who is right­their own system or Openfisca. Agents were therefore reluctant to see other administrators­particularly Etalab developers who did not necessarily have any competence in tax and welfare issues­take on the legitimacy to officially state the amount of such and such a state benefit or tax in their place. Certain key elements toward an analysis of this situation are provided by the sociology of the state, which seeks to explain how the state produces government documents and an “official” point of view.  [68] Knowing who speaks officially in the name of the state is indeed a major problem­the institution (or official) with this authority enjoys a legitimacy on which it will try to retain the monopoly; but, in turn, it runs the risk of losing face if its word is contradicted by other sources of truth.  [69]

62“Mes-aides” therefore met with great resistance, particularly from the Ministry of Social Affairs and Health’s Directorate General for Social Cohesion (Direction générale de la cohésion sociale, DGCS). This resistance resulted in institutional blockages and constant checks for more than a year and a half, so as to test the resource before finally validating it, reassured by the accuracy of the results and the advice of certain MPs.  [70] This allowed “mes-aides” to remove the “beta” banner from its site and to speak officially on behalf of the state.  [71] The site is now used daily by more than 10,000 citizens to assess their social rights.

Gaining legitimacy for expertise

63As well as being able to say how the tax and welfare system should be applied to individuals, actors mobilized around Openfisca, in particular IPP researchers, had to obtain recognition of the quality of their expertise in order to weigh in on the public debate in the face of Bercy’s monopoly on legitimate speech.

64In 2014, Mahdi B. left France Stratégie to join the IPP in order to work on the development of Openfisca. This research institute was founded in 2011 by various researchers from the Paris School of Economics (including Thomas Piketty and François Bourguignon, the director at the time and a pioneer in the field of microsimulation) and modelled on the Institute for Fiscal Studies (IFS) in London where Bozio worked for five years. The IFS is regarded as a major scientific and institutional authority in public debate, particularly during the vote on the UK budget.  [72]

65

Researchers don’t have access to Bercy’s models. But to measure the effects of taxes, we need to identify the real effects; we need these tools. What’s striking about the IFS is that it is research that fueled public debate, so that by doing one good thing, we could do the other as well. (A. Bozio, October 2018)

66Bozio, Mahdi B., and other economists who would later join Openfisca understood a microsimulator as a statistical tool for both description and prescription; but in order to succeed in making Openfisca a legitimate tool for prescription in public debates, it first had to be made into an exhaustive descriptive tool recognized by fellow researchers.

67With this in mind, the IPP set itself the task of reconstructing all of the French barème scales since 1917. The barème scales represent all the fixed data of the tax and welfare system­for example, the rates of income tax or the yearly livret A (a tax-free instant-access savings account) rates. One would think that this information would be available in digital form somewhere within the administration. In fact, considerable cognitive and exegetical work on the legal texts proved necessary in order to reconstruct the past reality of the tax and welfare system:

68

We bought all the legislative briefs, the income tax and social tax brochures; there are whole libraries of this stuff, and we tried to find the descriptions for each year and pull out the barème rates. Ideally, we wanted to find the references in the briefs and then look them up in the Journal officiel. [...] [But in fact] this is information that has never been kept; the administration itself doesn’t know! [...] Even for social contributions, it’s a large amount, it’s simple, and you’d think you could find it, but no. [...] Now, we’ve done a lot, and with Etalab, who code for “mes-aides,” it’s growing. (A. Bozio, October 2018)

69As far as current legislation is concerned, Etalab, along with other bodies that have joined the Openfisca community, such as Mutualité sociale agricole (MSA) and the city councils of Paris, Rennes, and Brest, are contributing to the updating of national and local law and barème scales. Each actor has channels of access to current legal data in its field of expertise, and encodes the legislation it needs into Openfisca. For example, Clara M., an economist at the IPP, was commissioned by thirteen French départements to prepare and simulate a Universal Basic Income (UBI) experiment. As a part of this, she had to update the income tax formulas:

70

I updated the 2014-2016 income tax. I had to do this because it was needed to do the UBI simulation. And I did it by changing the legislation line by line. For example, tax credits vary every year [we go onto Github, the platform that allows researchers to work on and store the Openfisca code, so she can show us a specific case. See Figure 2 below]. So, if you look there [she indicates the address bar at the top of the screen], here we’re in the model of the French tax and welfare system, in the models of compulsory deductions and, more specifically, tax reductions. This is the Scellier tax reduction for real estate investment [she indicates the “label” line 2697]. And if you look here, you’ll see the changes that I deleted in red [with a – in front] and those that I added in green [with a + in front]: I extended the reference period until December 31, 2014, and below I wrote the formula for the legislation, updated, along with the actual calculations [she shows me about fifty lines below with the formulas and the code for the law].

71This contribution, made to meet a specific requirement, enriches the resource by updating it and ultimately benefits the entire community. All of the successive contributions made by various actors, then, contribute toward achieving exhaustiveness, precision, and annual updating of this data, including the barème scales, which are now a data resource for the academic and administrative spheres.

72As mentioned in the introduction, thanks to Openfisca the IPP produced the first independent ex ante evaluation of the 2019 draft finance law, which was delivered and presented to the members of the Finance Committee before being taken up by the majority of the French media and in recent social protests (see Figure 3). This newly acquired legitimacy can be partly attributed to the collaborative nature of Openfisca, and to the heterogeneity of the network that fuels it and has made it a sufficiently exhaustive resource to be taken seriously. However, it also derives from the interest of other actors­in particular governmental bodies, MPs, journalists, and citizens­in availing themselves of a resource that disturbs Bercy’s monopoly on tax and welfare expertise.

Figure 2. Updating legislation in the Openfisca code on the Github platform

Figure 2. Updating legislation in the Openfisca code on the Github platform

Figure 3: A graph produced by the IPP via Openfisca and used by activists and journalists.

Figure 3: A graph produced by the IPP via Openfisca and used by activists and journalists.

Description: On the left, a photo taken during a demonstration for climate justice in December 2018. On the right, a still from the “Arrêt sur image” television program on November 30, 2018. In both cases, we see a graph showing the inequalities caused by the Macron government’s reforms, taken from IPP research carried out using Openfisca.

Conclusion

87In the introduction, we asked how actors on the margins of the state had managed to make a legitimate shared resource out of tax and welfare information that was once the sole preserve of Bercy. We showed, on one hand, that these actors had to carry out some substantial institutional labor and, on the other hand, that their heterogeneity proved to be a strength in carrying out this labor along with all the cumulative cognitive work that was necessary to build the resource. Finally, the desire to make Openfisca a digital commons, defined as a digital resource produced and maintained by a community according to its own rules of governance and characterized by a regime of shared ownership, called into question the porous boundaries of the state in a test that has not yet been resolved.

88Firstly, we recalled that some of the literature explains the success of open data in terms of the conjunction of the growing importance of the principle of transparency and the almost natural facility with which digital tools fluidify the circulation of information and the opening up of public data. We then confirmed the results of earlier research that has emphasized the cognitive and practical labor demanded of certain agents, often out of sight, in order to open up non-sensitive public data.  [80] However, our investigation seems to reveal another type of labor, an institutional labor, that has to be carried out when it comes to information with high symbolic, political, and legal value. Institutional labor is defined as an intentional action aimed at creating, maintaining, or destabilizing an institution. In the case of the opening up of the source code used to calculate taxes, this work took the concrete form of a lawsuit and legal reforms within the legal sphere, and campaigning by MPs, journalists, and digital activists in the public sphere. In the case of the opening up of the tax data, it took the form of more discreet administrative lobbying activity. This led to the destabilization of Bercy’s institutional monopoly by making some of the tax and welfare data accessible.

89Secondly, our research shows that the question of governance, although a subject of debate within the Openfisca project, is not, as we might have expected from the literature, the primary explanation for its early success. Our research shows that there were two more decisive factors at work: It was possible to build Openfisca thanks to the heterogeneous nature of the coalition that came together around it, and in particular the fact that some of its actors were situated on the margins of the administration, their collaboration facilitated by a digital system open to collaborative contributions. On the first point, some of Openfisca’s founding institutions, such as Etalab and France Stratégie, were close enough to the government to obtain a certain legitimacy, to enlist insiders, to know the details of opaque tax and welfare information, and to have access to windows of opportunity such as the request from Ayrault. Other actors were far enough away from it to be able, for example, to take Bercy to court to access this information. It is also this heterogeneity of positions and authorities that made it possible to build and to continue enriching the resource through the aggregation of complementary contributions. This seems to us to be a central finding of our investigation: The task of creating a shared resource built on government information is facilitated if public and non-public actors, on the periphery of this government, enter into collaboration. A second point is that all of these practices could be aggregated around Openfisca thanks to the collaborative nature of this digital tool that was part of it from the beginning: an open license, a collaborative work platform and communication tools accessible online without restriction, and modular software pliant enough to accommodate different scripts (oriented toward an understanding of the law, research, and decision support).

90However, as we have also seen, this unclear situation also led to a test of institutional qualification that has not yet been successfully negotiated. In closing, we would like to propose a hypothesis. Over the past two decades, several models of the state, i.e. configurations of the state, have been envisaged in the context of its digital modernization to make it more “open” and “democratic.”  [81] The idea of e-government has flourished,  [82] and more recently some have proposed the development of a “platform state.”  [83] Openfisca seems to represent an institutional experiment that presents an alternative to these models. It has been led by the work of hackers on the margins of the state,  [84] testing out a particular configuration of the state in relation to the digital commons. Yet, this configuration, played out at the sociotechnical, legal, and institutional levels, has not yet been stabilized. What role does the state play in the development of this original digital commons? Is it a partner, a guarantor, or an entirely separate participant in the governance of what could then be called a “public digital commons”? How does the pooling of public information redistribute knowledge/power relationships within and outside of the administration? Some doubt remains, and seems to point toward what Linhardt calls a state test: a test where the “state quality” of an arrangement such as Openfisca is uncertain and becomes an object of dispute.  [85] In our view, this opens up stimulating perspectives for further research, which we have tried to indicate in this article, and more broadly in the context of a doctoral thesis.

91A first limitation of our work lies in the tests we have had to set aside in the interests of clarity, namely the test of the economic sustainability of the project, which was often called into question, as well as that of the collaboration of heterogeneous actors who wish to develop the resource in divergent and sometimes antagonistic directions­something that has already led to the departure of some members. A second limitation of our research is that of any case study that takes a contemporary event as its object. Openfisca serves us here as an observatory, helping us grasp the still very recent tests undertaken in the construction of a new political and sociotechnical configuration­one that must be situated within a broader movement coming together around digital tools in order to make public information a commons and perhaps to transform the state. From this point on, we will need to continue our investigation with other case studies in order to confirm, invalidate, or deepen our hypotheses.