A Millian Case against Epistemic Arguments for Federalism

Abstract

John Stuart Mill is central to parallel debates in mainstream contemporary political epistemology and philosophy of federalism concerning the epistemic dimension(s) of legitimate authority. Many scholars invoke Mill to support epistemic arguments for democratic decision-making and decentralized federalism as a means of conferring democratic legitimacy. This article argues that Millian considerations instead provide reason to reject common epistemic arguments for decentralized federalism. Combining Mill's own insights about the epistemic costs of decentralization and recent work in philosophy, politics, and economics undermines purportedly Millian arguments for federalism focused on political experimentation, diversity and participation. Contrary to many interpretations, Millian considerations weaken, rather than strengthen, arguments for federalism. Any valid justification for federalism must instead rest on non-epistemic considerations. This conclusion is notable regardless of how one interprets Mill. But it also supports Mill's stated preference for local decisions subject to central oversight.

This article develops a Millian argument against common epistemic arguments for decentralized federalism that invoke Mill. Contemporary philosophical debates about the epistemic dimensions of democratic legitimacy and the benefits of federalism focus on questions concerning authority. Many invoke Mill to support epistemic arguments for democratic decision-making and decentralized federalism as a means of conferring democratic legitimacy.¹ Yet federalism does not fill epistemic deficiencies with democracy and raises new epistemic issues. Contrary to many interpretations, Millian considerations weaken, rather than strengthen, cases for federalism.

To demonstrate this, I first outline core issues in political epistemology and philosophy of federalism (Section 1) and Mill's positioning in each (Section 2). I then rationally reconstruct the strongest avowedly “Millian” arguments for federalism (Section 3) before drawing on Mill and contemporary work in philosophy, politics, and economics to explain why they do not succeed (Section 4). I conclude by examining my findings' implications for distinct literatures (Section 5). My arguments are primarily philosophical, not exegetical, developing an original account of epistemic authority and its implications for federalism inspired by, but distinct from, Mill.² However, they jointly explain why Mill reached his conclusions. They accordingly not only help demonstrate why many contemporary appeals to Mill do not work as intended but incidentally help vindicate Mill.

1. Authority and federalism: the epistemic dimensions

Distinct debates in mainstream contemporary political epistemology and philosophy of federalism in which Mill is positioned in central roles focus on similar issues of political authority and raise similar analytic burdens. Insights from one help identify options for and constraints on plausible arguments in the other. Yet these domains rarely intersect. With few exceptions below, core works in each domain minimize Mill's complicated understanding of political epistemology's implications for federalism. The following demonstrates the benefits of analyzing the domains together.

A central question in political epistemology asks whether democratic decision-making meets epistemic standards for legitimate authority. Democracy's proponents (e.g., Estlund 2008) and critics (e.g., Brennan 2016; Somin 2016) agree that legitimate authority has an epistemic component. Standard accounts focus on decisions, though larger debates concern decision-makers or forms of governance. Authority's epistemic dimension is sometimes couched in terms of voter, decision-maker, or institutional knowledge of particular facts, preferences, values, etc. (“informational” accounts) and sometimes in terms of the probability that each can identify and make the “correct” choice (“dispositional” accounts) (Méndez 2022). Yet even informational accounts suggest decision-makers who fail to act on relevant knowledge cannot meet epistemic standards for authority. The epistemic condition on legitimate authority could require relevant information, dispositions, or both (Viehoff 2016).

Whether relevant standards should be indexed to subjective preferences or an objective standard is another concern. A prominent account distinguishes epistemic competence as a disposition to select preferred policies from competence as an ability to affect certain outcomes (Méndez 2022).³ However those interested in policy preferences also desire particular “outcomes,” namely those in which specific policies are in place. Most work accordingly focuses on outcomes, which are sometimes characterized in terms of preferences (e.g., voters want policy X in place or interest rates to be at level Y) and sometimes in terms of particular goods (e.g., economic health is objectively best characterized as Z and decision-makers should know how to achieve Z).

These philosophers accept an epistemic condition on political authority (with different thresholds for each component): a decision-maker lacks legitimate authority if it lacks knowledge on how to bring about relevant outcomes or does not use knowledge in an epistemically responsible manner to affect the same at acceptable rates. Basic epistemic competence is thus, minimally, a “disqualifier” (Brennan 2016) for authority claims. Some further accept a qualifying variant: one has a greater claim to legitimate authority where one has more relevant information or better reflects the relevant epistemic disposition (see summaries in Estlund 2008; Viehoff 2016).⁴

Philosophers of federalism then seek to justify federal “divisions of power” on which at least two entities (e.g., federal governments, provinces, cities) possess final decision-making authority over at least one subject (e.g., immigration, crime) each and seek to identify which entities should have authority over which subjects and when. Epistemic considerations are often raised for both purposes. While federalism admits many definitions,⁵ this division of powers is common to and arguably constitutive of most definitions in contemporary philosophy (Da Silva 2022; Føllesdal 2003/2022) and so central here. Federalism purportedly leverages the goods of centralized decision-making on matters impacting all parties while simultaneously permitting local bodies to leverage knowledge of not only local preferences and values but also of factual matters that can impact implementation of policies. For instance, federalism can simultaneously leverage centralized knowledge of public health norms to set vaccination schedules and use local demographic knowledge to establish local hospital networks that maximize ease of access to vaccinations.

Both projects view epistemic concerns as key to identifying conditions for legitimate authority and plausible candidate wielders thereof. Possibilities that work in one debate can illuminate issues in the other and warrant greater interactions between political epistemologists and federalism scholars. Analyzing Mill's positioning in both provides proof of concept for this tack.

2. Mill in contemporary debates

Mill is seen as a bridge between political epistemology and federalism in the few works discussing their interaction. Scholars invoke Mill to support diverse (and even contradictory) views on authority and federalism. While this text is not primarily exegetical, background information on Mill's positioning in existing scholarship is useful for assessing purportedly “Millian” arguments below.

Mill is oft-invoked to link democratic and epistemic concerns. Estlund (2008) and Brennan (2016) quote Mill in epigraphs. Many position Mill as a predecessor for arguments that democracy inculcates cultures where individual and collective epistemic conditions on legitimate authority can be met.⁶ Mill's (in)famous suggestion that knowledgeable individuals could possess multiple votes to secure a collective knowledge condition is inconsistent with many contemporary egalitarian democratic views but also speaks to epistemic conditions on authority.⁷ Mill is, in turn, one of the most commonly invoked historical figures in federalism studies, where he is oft-considered a champion of decentralized governance. “Of Federal Representative Governments,” Chapter 17 of Considerations on Representative Government (1861/2010), is reproduced in full in the leading philosophy of federalism collection (Karmis/Norman 2005). Mill is the subject of one of only two philosophical entries in a recent collection on federalism (Heidemann/Stoppenbrink 2016). Many exegetical works recognize the complexities (or “ambivalence”; Porter 1977: 114) of Mill's views. Interpreters emphasize different aspects of Mill, leading to variety in their attributions. Yet epistemic considerations are central in each.⁸

Federalist interpretations admit that Mill first provides an instrumental case for federalism but also identifies purportedly normative arguments therefor. Many emphasize Chapter 17 of Considerations, which contains Mill's most explicit comments on federalism.⁹ They unsurprisingly interpret Mill in light of its interest in conditions under which federalism is “advisable.” They then find positive arguments for federalism. Stoppenbrink (2016: 223), for example, first notes that Mill thinks the strongest arguments for federalism stem from its “pacifying effects.” Conditions for “advisable” federalism are also those for a stable state consisting of parts that would be too weak on their own. But Stoppenbrink then identifies normative arguments for federalism focused on “(1) the chain of legitimacy under conditions of representative government is closer to citizens if ‘mediated’ by territorial sub-units” and “(2) the sub-units constitute each an individual ‘object of identification’ for their citizenry and population and thus better allow for and foster a ‘sense of belonging’ to the political entity” (210). The “closeness” in (1) could make decisions more likely to reflect local knowledge. Stoppenbrink's Mill views (1) and (2) as valuable for fostering civic participation and a civic virtue requiring knowledge. Føllesdal's (2003/2022) discussion of Mill on federalism likewise first notes Mill's concern with peace but highlights that Mill also suggests federalism can be desirable for protecting minority rights, ensuring responsive or efficient governance, or providing for “experiments-in-living.”

Contemporary epistemic defenses of decentralized federalism also invoke Mill. Kelly (2021: 52) suggests “Mill intended the central government to be focused on a narrow set of great issues which would relate to political economy, trade, and national security” with local government exercising “most responsibility for social and public policy.” Contemporary thinkers invoke Mill to promote such decentralization. Weinstock (2001: 77), for example, highlights the ways in which federalism makes decision-making “cognitively more accessible” to citizens as key to a Millian “democracy argument” for federalism. Somin (2016) appeals to Millian experiments in his epistemic argument for federalism: decentralized federalism can address contemporary democracies' epistemic deficiencies by creating conditions under which people are incentivized to learn more about political issues. If decentralization leads to policy experimentation and people are able to choose between experiments by moving to different jurisdictions, they should be suitably incentivized to learn more about the options and their (likely) consequences.

A less common but persuasive interpretation reads Mill as an advocate for centralized governance and a federalism skeptic. For instance, Porter (1977) surveys Mill's wider corpus, including “Municipal Institutions” (1833/1986), and identifies a preference for centralization.¹⁰ Porter's (1977) Mill believed central governments should coordinate policies. Local governments should be empowered to administer policies in distinct ways. But central governments should maintain authority to set policies and substitute decisions. This position had multiple supports discussed further below; “considerations of corruption, competence, efficiency, popular control, and the need for periodic impetus motivated Mill” to defend centralization.¹¹

Levy's (2014) book on pluralism then opens with a quotation from Mill's “Centralisation” (1862/1986: 606): “Any despotism is preferable to local despotism. If we are to be ridden over by authority, if our affairs are to be managed for us at the pleasure of other people, heaven forefend that it should be at that of our nearest neighbours.” Levy (2014) repeats this quotation twice in later chapters. Levy's Mill recognized local politicians lacked knowledge necessary to make good decisions and were more prone to irrational decision-making that reflected local mores but were contrary to good governance. The Mill passage thus continues (1862/1986: 606):

To be under the control, or have to wait for the sanction, of a Minister or a Parliament, is bad enough; but defend us from the leading-strings of a Board of Guardians or a Common Council. In the former authorities there would be some knowledge, some general cultivation, some attention and habitual deference to the opinions of the more instructed minds. To be under the latter, would be in most localities … to be the slave of the vulgar prejudices, the cramped, distorted, and short-sighted views, of the public of a small town or a group of villages.

Members of smaller groups are more likely to be similar and subject to local pressures. A “local despotism of custom and opinion” (Levy 2014: 263) can further undermine the pursuit of Millian experiments-in-living by making persons more likely to bow to social pressures. Such persons are also more likely to elect local charismatics that lack knowledge required for one to govern well.

Porter and Levy persuasively outline Mill's epistemic concerns about federalism. Some may be historically contingent. The corruption and “incompetence and lack of expertise” in local government of Mill's time (Porter 1977: 114–15) may not exist today.¹² Mill himself outlined aspects of contemporary federalism that may have addressed some concerns. Porter (118–19) notes that “visible elected officials” in local government could have addressed concerns about a lack of accountability in municipal bodies of those days. One should further note that federal governance could be less “fragmented” (115) than devolution to municipal government. Yet Mill's skepticism of federalism remains remarkable given contemporary uses of Mill above.

Each interpretation above has textual support, and one could adopt aspects of Mill without adopting his settled position. A Millian experiments-in-living argument could, for example, be severed from Mill's own concerns in “Centralisation” and “Municipal Institutions.” With this background in mind, I next explicate leading contemporary “Millian” arguments for federalism.

3. “Millian” epistemic arguments for federalism

Epistemic arguments for federalism take several purportedly Millian forms that address distinct problems mirroring the different epistemic approaches to authority above.¹³ The Individual Knowledge Problem holds that individual voters or decision-makers they elect (in democratic states) lack knowledge necessary to fulfill the epistemic condition for legitimate authority. Contemporary democracy skeptics suggest individual voters not only lack technical knowledge about how policies relate to outcomes but also basic knowledge about who is responsible for which decisions necessary to correctly select them (e.g., Brennan 2016; Somin 2016). They accordingly cannot properly select outcomes in referenda/elections that would fulfill their self-defined ends or hold decision-makers to account. Those elected may lack pertinent expertise, presenting another source of illegitimacy. The Institutional Knowledge Problem suggests that wider decision-making bodies (e.g., legislatures) lack knowledge necessary to fulfill the epistemic condition.

The Individual Dispositional Problem posits that voters or decision-makers they elect fail to meet the threshold for correct decision-making necessary to fulfill the epistemic condition. Even where voters know basic political facts, for instance, they may suffer from irrationality or biases that lead them to vote in ways that do not further intended ends. These worries stem not only from concerns about “groupthink” whereby people are likely to vote like their neighbors but also from concerns of other social pressures and basic difficulties of being able to reliably perform means-end reasoning in the oft-contentious political domain.¹⁴ The Institutional Dispositional Problem suggests the wider decision-making body suffers the same defect(s). This can stem from various sources, including a need to be responsive to epistemically problematic publics or incentive structures within institutions creating path dependencies toward inferior solutions.¹⁵ Any legislature that consistently makes “wrong” choices given citizen preferences is non-responsive. One that consistently does so respecting topics impacting their vital interests, like climate change or inflation control, fails its constitutive purpose. Both risk illegitimacy.

Millian concerns with experimentation, diversity, and participation purportedly address each problem. Some arguments feature multiple concerns. Mill's experiments-in-living primarily concerned individuals: people should feel free to exercise their autonomy and learn how best to live (thereby further testing their considered moral views). However, Mill also suggested that different policy options can be conducive to such pursuits.¹⁶ Policy experiments are thus plausibly described as experiments-in-living in contemporary Millian views on which they are to offer legitimacy-conferring epistemic benefits. I present each argument before critiquing them.

Experimental arguments take two general forms, focusing on moral incentives and options, respectively. Incentive-based arguments suggest federal structures will lead relevant parties to be adequately informed and make epistemically desirable choices. Somin (2016), for example, suggests decentralized federalism fosters individual (often voter) competence by providing a set of local policy options that will be individually decisive for each citizen. If decentralization leads to policy experimentation and people can choose between experiments by moving to different jurisdictions, they should, again, be suitably incentivized to learn more about those options and their consequences (or at least the outcomes they are likely to produce). Federalism accordingly permits experiments-in-living that increase the chances of epistemic conditions being met.

An analogous argument suggests legislators and other elected officials will be incentivized to learn how to produce desirable outcomes to maintain an electorate and to experiment in ways that make such outcomes more likely. Federal experimentation can thereby address the Individual Knowledge Problem. It also helps address legitimacy concerns for those who understand epistemic dispositions in terms of preference satisfaction: Dispositional Problems are less acute with respect to issues voters each view as individually decisive in terms of overall well-being (as in Somin's account). This adds an epistemic dimension to the view that federalism aims to efficiently match individual preferences and public policy regimes (e.g., Tiebout 1956).

Options-based experimental arguments suggest experiments produce knowledge or make correct choices more likely. Individual experiments-in-living are partly beneficial because individuals can test moral views and provide models from which others can learn.¹⁷ Policy experiments can likewise provide new knowledge about ways to address issues from which experimenters and nearby parties can learn. Learning leads to better choices, thereby addressing Dispositional Problems; any costs attendant to particular experiments are confined to one place. Experimental conditions further present environments where individuals can experiment. More distinct contexts for pursuing conceptions of the good provide more opportunities for testing one's aims.

This insight connects to related works raising Mill's concern with diversity.¹⁸ Jewkes (2016) invokes Mill in a recent example linking diversity and experimental arguments. Jewkes believes societal diversity is prerequisite for individual moral cultivation and federalism best ensures such diversity. Per Jewkes, one must have desirable alternatives to exercise the basic autonomy central to Mill. Where few “blaze a truly and entirely original trail through life,” societal diversity is necessary to identify possible ways of living and make them salient as “a realistic possibility” others have proven possible (188). Federalism provides conditions for such diversity. Provinces possessing distinct domains of authority lead to variety of policy regimes and possible ways of living. Even differences in terms of funding for social goods, for example, can lead to widely divergent life opportunities. Jurisdictions with low taxes and a laissez-faire approach to health regulation not only produce different outcomes than those with higher taxes and higher regulation of healthcare. They also distribute goods and reflect values in ways that are likely to produce different lifestyles and present different opportunities to pursue particular lifestyles.

Group-based diversity that is prominent in one major epistemic defense of democracy could supplement epistemic arguments for federalism. Goodin and Spiekermann (2018) suggest greater group diversity can improve group competence absent improved individual competence. Under specific conditions, democratic decision-making can establish a form of institutional dispositional competence attenuating deficiencies in individual knowledge or even dispositional competence. Per Condorcet's Jury Theorem, a group of a sufficient size whose members are independent and whose mean individual competence is “appreciably above random” will eventually reach correct conclusions. As Estlund (2008: 223) notes, the Theorem can work at the “moderately-sized town” level: a sufficiently diverse and independent set of voters at the municipal level will at least be disposed to reach the correct conclusions. Federalism offers means of recognizing differences across jurisdictions. If each municipality has distinct needs or implementation challenges for a given issue and the Theorem holds for each, this provides reason for each municipality, let alone province, to possess formal authority over it.¹⁹ While many suggest the Theorem's basic assumptions do not hold, the idea that a suitably diverse body of even minimally informed persons could approximate its conditions has plausibility.²⁰ If something similar holds, the standard for avoiding Knowledge Problems is lower than critics of democracy expect and Dispositional Problems can be addressed. Less formal appeals to the Millian concept of a marketplace of ideas, in which intellectual diversity converges on good outcomes, play a similar discursive role, albeit without comparable mathematical support.²¹

Where local knowledge is relevant, federalism can ensure decision-makers possess or make correct decisions conditional upon it. Education, a domain in which Mill sought policy diversity (Jewkes 2016; Porter 1977; etc.), is a useful example. If the provincial government in a Francophone-majority province is unaware of the distinct educational needs of an Anglophone majority in a city with minimal representatives in the provincial legislature, Knowledge and Dispositional Problems can easily appear. To put the point more broadly, if democracy alone cannot reliably guarantee that decision-makers possess, let alone use, relevant local knowledge and federalism provides means of fulfilling this epistemic gap, this at least suggests a plausible epistemic case for federalism. A complete case must further explain when local knowledge is relevant and how federalism will better leverage it. But the basic idea is intuitive. A local politician is more likely to know about local geographic and demographic conditions that will impact what policies will look like “on the ground.” This person will have more formal power if serving on a smaller city council than in a larger provincial, let alone national, parliament. That more local body will thus be more likely to make decisions that incorporate the local knowledge.

Finally, Mill's interest in political participation can connect to arguments for federalism. Mill famously suggests political participation fosters civic virtue and knowledge (Føllesdal 2003/2022; Stoppenbrink 2016).²² Persons must participate in government to hold government to account. When they do, they develop knowledge and talents that can address Individual Problems above. As they gain competence, they will plausibly increase chances of addressing Institutional Problems. Federalism offers additional fora for political participation. One can get involved at multiple levels. And one may face incentives to learn more to understand how they relate and thus how to impact decisions at each. If, in turn, federalism offers no distinct participation advantages, it makes forms of participation and political knowledge more cognitively accessible. This is true not only of decisions (Weinstock 2001) but also of democratically elected decision-makers (Porter 1977). Federalism makes some kinds of participation less epistemically challenging, thereby limiting institutional barriers to addressing problems above.

4. Mill and the case against federalism

The arguments in Section 3 have merit and plausible claims to continuing the spirit of Mill's work. However, a mix of insights from the contemporary political epistemology and social theory on which the arguments are purportedly constructed and Mill's own concerns suggest the arguments, whatever their merits, cannot sufficiently establish an epistemic case for federalism. Millian concerns are unlikely to establish federalism as a cure for epistemic deficiencies in democratic states or address other epistemic problems with establishing authority. Rather, Millian concerns are part of the mix of factors undermining any plausible epistemic case.

First consider experiments-based arguments. Many incentives-based arguments only apply under idealized circumstances whereby people can easily move. Somin (2016) suggests this is a reason to permit free movement across borders. He adds that people have clear incentive to set their priorities straight and learn enough about what jurisdictions will likely get it “right” on what matters most to them where costs of movement are inevitable. Both points are fair. Yet the former response limits application of his view to idealized conditions at odds with much of contemporary political epistemology.²³ And the latter relies on a vision of competence many reject: the epistemic condition on legitimate authority could require more than the information required or ability to make rational choices between policies in an individually decisive domain.

There is, additionally, reason to question whether persons respond rationally to such incentives. Somin's positive arguments for decentralization rely on posits from public choice theory. Whether people are less likely to submit heuristics, biases, and other irrationalities outside its idealized conditions is questionable. Real-world individual voters may prove equally epistemically incompetent and may not “learn” enough about experimental outcomes and their applicability to their case or prove capable of applying those outcomes to their case.

The larger issue with experimental arguments is that federal design does not reliably produce the incentives to create distinct policies or to adopt policies that are successfully tested elsewhere. Experimental conditions are unlikely to provide individuals with the choice set necessary to create incentives to learn more or increase the domain of acceptable choices in ways that will produce better outcomes. If so, incentive- and outcome-based experimental arguments fail to solve Section 2's epistemic problems. Policy laboratories are unlikely to fill epistemic deficits in democratic governance. A paradigmatic case example clarifies the issue. The USA has been exemplary of federalism since Mill's time.²⁴ Modern U.S. states are supposed to be paradigmatic laboratories and are well-resourced compared to sub-national bodies elsewhere. But recent work by Tyler and Gerken (2022) reveals that U.S. states do not experiment in epistemically valuable ways and that this partly results from the higher epistemic standards for understanding governance in federal countries and federal governance's inherent incentive structure. Even well-resourced state governments lack the time and knowledge necessary to evaluate, let alone create, new policies. State legislative sessions have a short duration and cover many issues. Even large political staffs are unlikely to develop expertise on several issues and propose new work. The need to understand the contours of a state's constitutional powers over the topic exacerbates difficulties: one needs time and resources to understand not only the subject of a piece of legislation, whether it reflects the will of one's constituents, and its potential implications but also whether it is within one's competence and its potential impact on the division of powers.

Tyler and Gerken further demonstrate that state incentives to innovate are limited. Experiments are technically and politically risky. Chances of success on either score are mixed. And the public may not reward even an effective policy. This is particularly so, I add, where people do not understand the connection between policies and outcomes. In the USA, the federal government's power to override many state decisions negates potential political gains from innovations, further lessening incentives to innovate. The same circumstances and dynamics will apply wherever there is concurrency of federal and state powers and the federal governments' decisions are hierarchically superior, a result that also features in other major federal jurisdictions, like Canada.²⁵ Where a new policy proves successful, in turn, Tyler and Gerken highlight that a given state can free-ride on others' successful, politically acceptable experiments.

Beyond Tyler and Gerken, the pool of potential decision-makers is smaller for any state than for the country, decreasing the chances of finding an expert on a topic or a polymath capable of true innovation in any state. Mill (1862/1986: 605–6) noted that many central government officials throughout time have proven incompetent. However, the chances any single central legislator will be competent are higher. Capacity to innovate in these circumstances is severely constrained. It is thus unsurprising that federal and state governments outsource policy-making features to other bodies. Legislators at either level do not develop laws or even set the terms of discussion.

If the foregoing is true, sub-state bodies are unlikely to experiment in ways that address epistemic deficiencies above. Where states do experiment, jurisdictions do not clearly “learn” from one another in ways that support outcome-based responses to the Dispositional Problems or offset epistemic losses arising from experimental failures or otherwise acceptable deviations from central norms. Evidence on whether states “learn” from one another is mixed. Work in comparative healthcare policy, for example, suggests policy migration can follow political, rather than epidemiological, patterns.²⁶ One may prove more likely to adopt a policy from another sub-state unit or even country with a similar legislative profile rather than one that proved effective.²⁷ Basic political expediency may produce widespread migration of policy failures. Desires to “do something” about a problem can lead one follow a perceived policy “leader” absent evidence a policy is effective.²⁸ This leads to the migration of policy ideas that have not yet proved effective or whose effectiveness is already questionable. Consider the questionable success and heavy punitive burdens occasioned by widespread adoption of broken-window policing or, on the other end of the ideological spectrum, the public health impact of more permissive drug policies.²⁹

Federal structure and political economy produce further challenges. Federalism places barriers on the ability to successfully export successful experiments. Having to successfully pass through many state legislatures is a higher bar than having to pass it in a central legislature. Splitting identities between provinces can, moreover, create resentments that make people less likely to learn from some provinces.³⁰ Even competition meant to incentivize experimentation can forestall innovation: recent work in economics suggests sub-state units' aversion to free-riding leads them to avoid positive developments if they would permit informational spillover (Callander/Harstad 2015). Options for potentially addressing these challenges within federal boundaries present their own issues. If, for example, Tyler and Gerken are correct, giving a federal government the authority to set an overarching policy in the same area will undermine state incentives to adopt distinctive ones. Another proposal whereby an overarching central government “harmonizes” policies after experimentation (Callander/Harstad 2015) is, in turn, no longer recognizably federal even if practicable: if a central government maintains authority to harmonize policy, one is in a situation like Mill's oversight model of devolution in a unitary state.

Both problematic outcome differences across states and attempts to rectify them could, moreover, undermine the federal project and its potential benefits. Recall that Mill only considered federalism “advisable” to maintain stability (Stoppenbrink 2016: 213). Differences in outcomes across states and a strong federal hand may create resentments across state boundaries and splinter national ties. Differences in health outcomes cause resentments in major federal states and yet many sub-state units protest where federal governments seek to use their constitutional powers to equalize outcomes (e.g., Béland et al. 2017). Some uses draw claims of federal “overreach” that can itself undermine state solidarity and stability (ibid.).³¹ Concerns others will act outside their authority can also undermine solidarity and even stability: “the advantages of decentralization are realizable … only if there are good reasons for the players … to believe that others will generally abide by the terms of the federation” (Bednar et al. 2001: 223).

Complexities of federal governance provide further challenges. One may, for instance, believe federal and state governments can work together to ensure proper learning occurs. However, the possibility of doing so will depend on a given constitution's rules. Recent work suggests that parties will only be incentivized to act together in areas within areas of “overlapping jurisdictions” under limited conditions where doing so is politically favorable (Scholtz/Munteanu 2023). If this is so, the mere possibility of cooperation will not foster experiments. Some successful cooperative efforts then undermine federal constitutional orders: federal and state governments have incentives to let each other deviate from the formal federal division of powers to bring about politically desirable outcomes but those still constitute deviations.³² Such agreement can, in turn, limit the possibility of experimentation by permitting federal oversight by a different name. If joint federal–provincial agreements provide detailed options, they will create path dependencies toward uniformity.³³ If the path leads to poor outcomes, opportunities to change may be limited by the complexity of the arrangement. Any possible solution to the epistemic problem here may accordingly come at the cost of a commitment to the basic federal division of powers that epistemic concerns were meant to support – while also raising the possibility of limiting future beneficial opportunities to learn.

This helps explain why Mill did not take the case for experimentation to entail one for federalism. If federalism is an “instrument to achieve … self-determination, high rates of participation and a government close (closer) to citizens than distant, centralized, unitary political authorities” (Stoppenbrink 2016: 228) and those means can be fulfilled through local administration or devolution, there is little reason to adopt federalism. Mill's antipathy towards federalism is unsurprising where that instrument also has epistemic costs that other forms of governance lack.

Porter (1977) provides a compelling case that Mill himself took experimentation's benefits as supports for local administration of policies. Considerations and “Centralisation” distinguish “policy-making and policy-execution or administration suggesting that the first is more appropriate to the central government and that the second is more appropriate to local governments” (119–20). Central governments should, on Porter's reading, “collect and disseminate information for use by the localities,” including information about policy experimentation results; advise local governments; and ensure local administration is consistent with central policy (120–21). Such an arrangement combines central “knowledge, expertise, and enlightened public opinion with local understanding of details … and the diligent attendance of those with a direct interest in a matter” (120). “Excessive centralization” is then partly defined by central inability to play these necessary roles: Mill not only desired “administrative efficiency” but sought “to ensure that central departments did not neglect their primary function of policy-making over the whole range of departmental concerns” (107).

Mill thus recognized that other forms of governance can leverage any potential epistemic benefits of policy experimentation and likely preferred those alternatives as means of remedying federalism's particular epistemic deficiencies. Mill permitted decisions on “purely local matters” to be made locally (though details on what qualifies as “purely local” are lacking (120)). Yet he believed central supervision of local bodies was necessary to address decentralization's epistemic deficiencies even in the education sector. Mill desired diverse locally run educational programs supervised by central governments (Levy 2014: 193–95; Porter 1977: 118–19). Mill would have denied sub-state entities (provinces, cities, etc.) the independent authority characteristic of federalism in “all fields of any significance” (Kurer 1989: 294). Local authorities “ought not to be supreme and absolute” (Mill 1862/1986: 606) in most domains.³⁴

Mill's conclusions remain plausible. Local bodies can successfully experiment without full authority and full authority limits possibilities to correct errors. If Mill's belief that central governments are epistemically superior to local ones proves true, local administration can leverage relevant local knowledge and incorporate it into broader policies. They can, moreover, more easily transfer policy lessons so more persons benefit. If a central government maintains final authority over a subject, it can always require localities use a newly discovered successful method for producing an outcome. One may wish for sparing use of his power to permit continued experimentation. The power remains desirable given issues with learning above.

Non-federal modes of governance could more broadly leverage the benefits of diversity while minimizing costs of local epistemic deficits, including informational gaps, irrationalities, and poor experiments. This is true not only of Millian “local administration” but also of forms of devolved authority in which a central government lets its constituent parts legislate over given subjects while maintaining power to revoke that authority where it is misused or substitute for inapt decisions.³⁵ Indeed, as we will now see, the epistemic problems that federalism was meant to address provide further reason to question whether sub-state polities will experiment well or learn from one another's experiments and whether individual voters can judge experiments.

Political epistemologists, recall, worry about biases and irrationalities that track social connections. The biases and irrationalities have been observed among those who use party preference as a proxy for apt decision-making.³⁶ There is little reason to think such tendencies will not occur across party lines. There is, again, evidence that state governments with similar political persuasions are more likely to adopt each other's policies. Polarization across multiple dimensions significantly decreases the chances people will rationally respond to policy outcomes.³⁷ Where politicians begin to sell ideologies first, rather than policies, this threat is even greater; recent work suggests ideologies can guide voter choice and serve as the true source of competition (Izzo et al. 2023). This problem is not unique to federalism but challenges claims people will “learn” within federal states. The aforementioned Millian concern that people will be more likely to conform to local social pressure suggests irrationality could be greater in any decentralized, let alone federal, state. This repeats, rather than solves, bias/irrationality concerns.

Demographic sorting further minimizes opportunities and incentives to learn from others. If voters irrationally commit to a set of policies, the costs of accepting contrary experimental results will be too high. Purportedly free competition of policy ideas will not provide laboratories of learning but sites for developing a common denominator among shared irrationalities.

Demographic sorting within federal states additionally undermines individuals' capacity to benefit from policy experiments where learning otherwise would occur. Consider, for example, recent “sorting” of liberals into large cities and non-liberals into rural communities (Rodden 2019). In such circumstances, liberal voters are unlikely to be able to affect change outside city boundaries unless larger cities have many seats in broader legislatures. These conditions do not obtain in most modern federations, resulting in conditions where cities often “lose” (ibid.). That is, provincial or federal governments do not adopt policies preferred in municipalities and cities are subject to rules created by those with other views. This can also result in a pattern with fewer liberal jurisdictions and thus sites of liberal experimentation. And residents of cities cannot benefit from many experiments they would desire without moving. This creates incentives to move to areas where liberal policies are more likely, but also produces still further sorting and still fewer sites for genuine experimentation. These results are avoidable. One can divide federal sub-units in ways permitting greater experimentation. Many contemporary demographic sorting problems may resolve when cities possess genuine constitutional authority (Hirschl 2020; Weinstock 2014). However, persistent demographic sorting broadly challenges Somin's “free movement” as a simple solution to the challenge of ensuring all persons can benefit from policy experiments. The bar for justifying federalism is again higher than many propose.

There is further reason to question whether individual decision-makers in sub-state units can meet an epistemic threshold for legitimate rule. Incentives for sub-state experimentation come with aforementioned trade-offs attendant to smaller pools of candidate decision-makers. Local decision-making takes place in conditions of less diversity and greater prospects of close group identification. When one turns to assess political figures, rather than voters, a further challenge emerges: each person elected has a smaller group of persons to whom they must be accountable and that group is more likely to be homogenous.³⁸ The possibility of bias and irrationality for any decision is thus much higher, particularly where the smaller voter base increases the relative cost of each lost vote. The chances of solving the Individual Problems above thus remain low, suggesting any individual competence condition on legitimate authority will remain unfulfilled.

Non-politicians face further pressures to conform to local mores even where contrary to individual desires, creating further path dependencies away from plausible epistemic norms. Local living makes difference much more salient. Everyday interactions create social pressures to conform that increase the cost of public discourse for others, limiting the domain of available policy choices discussed. Secret ballots may offset some costs.³⁹ But much of the decision space is determined long before any ballot. If persons do not feel free to express desires for more immigration in public debate, political candidates will lack signals this is something worth pursuing. This limits opportunities to vote for candidates likely to bring about such change.

Mill was, again, alive to these issues. Levy (2014: 212) highlights concerns with “local and in-group tyranny” (212) as central to Mill's critique of decentralization. Mill's worries about local governance were not limited to historically contingent worries about non-intellectual “local elites” like “aristocrats or their descendants, priests, wealthy businessmen” dominating local politics. He further worried that pressures to conform to local mores would limit rational political action. In fact, the kind of “local” knowledge found in decentralized governance units often includes personal knowledge that makes it easier to make people conform. Levy further describes the “busybody” who “is all too aware of the particularity of those over whom he or she exercises power, and is all too likely to use the power in personal rather than impersonal ways” (79–82). Where, in turn, busybodies pose no genuine threats, local communities still create their own social pressures that exert a distinct normative force. Government is meant to combat these irrational social pressures. Central governments far removed from the local group and its mores are more likely to provide the rational clarity necessary to combat them (219).

Federalism, then, is likely to result in outcome differences without epistemic justification. Federal political structure creates conditions that limit good policy migration (and can lead to inapt policy migration) and thus limits many persons' ability to enjoy any goods of experimentation. Where, in turn, federalism does not raise unique epistemic concerns, it appears subject to and even exacerbates standard problems meeting the epistemic condition on legitimate authority.

One may, of course, respond that providing for differences is the point of federalism. Millian interests in diversity also challenge this response. The diversity required for the Condorcet Jury Theorem to obtain does not require federalism. Mill instead suggested that policy differences and experiments can be valuable for safeguarding a diversity of individual interests.⁴⁰ However, federalism is not only unnecessary for that type of diversity but also characteristically produces conditions that limit particular individuals' opportunities to experiment. Absent easy movement, many will be stuck in policy regimes contrary to their aims. Federal governance increases the costs of voicing contrary political opinions while limiting potential benefits. This, in turn, limits the diversity of opinion on which many epistemic arguments for federalism rest. Evidence that local decision-makers make better choices might blunt these concerns. Unfortunately, such evidence is lacking. Considerations above suggest they are at best equally irrational.

Critics may further suggest outcome differences are justified reflections of local preferences. That argument is not clearly epistemic. Arguments for federalism often appeal to the value of local knowledge. But “local knowledge” is not always epistemically on a par, let alone superior to, more general knowledge. Insofar as it is important, evidence that a central government cannot access or leverage it is often lacking. And local decision-makers do not clearly use local knowledge in epistemically desirable ways, or even in ways that are epistemically preferable to federal governments. “Local” choices are at best sub-optimal in many core cases. Consider provincial reluctance to adopt basic climate measures (e.g., References re Greenhouse Gas Pollution Pricing Act, 2021 SCC 11) or provide basic healthcare funding in many states (Michener 2018) absent strong federal oversight. Costs can offset any epistemic benefits.

Where different outcomes in core cases appear legitimate, this is unlikely to be due to their epistemic benefits. Many differences will instead be justified in spite of their epistemic issues, highlighting a trade-off. Accepting economically adverse limitations on immigration or soda bans without clear evidential support is more plausible as an acceptance of the importance of furthering non-ideal but acceptable local values. One can, of course, define “correct” decision-making in terms of reflecting local values. But this is question-begging here and independently implausible. If epistemic value cashes out in picking what constituents prefer, epistemic considerations add little to our evaluative toolkit: the epistemic condition becomes a basic democratic norm and one loses one's ability to evaluative democratic choices on epistemic grounds, a main goal of political epistemology. A “right” to sub-optimal decision-making suggests one can accept epistemic costs in the name of other goods. If the decision does not fall below the standard of the epistemic condition on legitimate authority, this is acceptable. But epistemic interests provide no additional justification. If the foregoing is correct, values-based arguments are likely to create epistemic problems. Commitments to local mores can be a source of irrationality. This does not render federalism unjustifiable. But one may need to accept a trade-off if/where local values should be decisive. Epistemic considerations are then only going to be non-disqualifying for, rather than supportive of, federal governance justified by local values.

Appeals to participation are unlikely to greatly aid federalists. Participation in political processes prone to polarization is unlikely to foster good character. Indeed, Kurer (1989) suggests Mill's case for centralization is rooted in distrust of mass participation in government. Kurer's Mill sought balance between competence and participation as competing principles.⁴¹ There is, it seems, a limit to the amount of character-building that participation can produce: “deficiencies of mental improvement then justify a great deal of restrictions” (ibid.: 289). If one rejects this general skepticism about mass participation, possibilities of participatory processes going awry due to polarization and other distorting effects detailed above still cannot be easily discounted. Mill's call for another body to provide expert course correction appears epistemically apt.⁴² Federalism remains no better than and ultimately appears worse in a key respect than forms of devolution.

Jewkes's (2016: 196) diversity- and participation-based argument for a federal “public sphere” produces a similar result. Jewkes suggests a “unified public sphere” will produce a smaller number of political leaders and concomitant “alternative and potentially valuable viewpoints.” Leaders will largely speak to one another, producing a kind of “incestuous” influence that results in “groupthink” whereby “key public figures converge upon a broad consensus of collective moral value, leaving little space for alternate conceptions.” Federalism's division of authority among multiple entities will, Jewkes thinks, produce multiple “publics” and avoid this result.

Other forms of governance could, however, equally fulfill many benefits of multiple decision-making publics and “local” governance seems equally, if not more, subject to the threat of “groupthink.” Mill, again, believed “groupthink” was more likely locally. I take no stand on which level of government is most susceptible. But the disqualifying variant of the epistemic condition on legitimate authority suggests federalism is problematic absent someone who can ensure local bodies do not violate epistemic limits on permissible policies. Forms of devolution can provide backstops against many egregious errors. A federal government cannot substitute for problematic provincial decisions where federal and provincial legislative powers are truly distinct. A central government that can replace entities with devolved powers' decisions is better-positioned to be a necessary “check.” Central governments also err. Decision-makers in countries exercising devolution can fail to meet epistemic conditions on legitimacy authority. But federalism fares no better than alternatives here and eliminates useful epistemic safeguards.

5. Conclusion and future directions

Purportedly Millian epistemic arguments for federalism cannot establish the authority of a federal liberal democracy if democracy alone cannot do so. Millian considerations instead illuminate federalism's epistemic costs. These results incidentally help vindicate Mill's settled position on federalism and seeming skepticism toward decentralization free from central oversight. At minimum, concerns with “experiments-in-living,” diversity, and democratic participation do not require federal governance, freeing Mill from inconsistency challenges.

The arguments above are primarily non-exegetical but may also aid work on Mill. If, e.g., one adopts a principle of interpretation under which one should seek to render a historical figure's positions as plausible as possible, the above supports anti-federalist readings of Mill. Even proponents of that interpretative principle accept that it can be defeated by clear textual/historical evidence that the author held different views.⁴³ The foregoing can inform new interpretations even if such evidence exists and help scholars adjudicate textual ambiguities.

Considerations above may, for instance, help illuminate interpretations of the aforementioned Chapter 17 of Considerations on Representative Government (1861/2010). Mill writes: “When the conditions exist for the formation of efficient and durable federal unions, the multiplication of them is always a benefit to the world.” This suggests federalism is all-things-considered preferable to other forms of governance. But questions about whether the chapter as a whole endorses federalism and the nature of Mill's arguments in support of federalism remain. Chapter 17 notably begins by noting that those otherwise “fitted or not disposed to live under the same internal government may often” advantageously federate to avoid military aggression. The sentence following the “multiplication” line then states that federalism “has the same salutary effect as any other extension of the practice of co-operation, through which the weak, by uniting, can meet on equal terms with the strong.” Yet “federalism” can be preferable to the independence of many small states for the purposes of peace, stability, etc. without being superior to other forms of governance. Whether and when Mill's use of the term “federalism” reflects contemporary usage remains contested.⁴⁴ And Mill does not provide a full-throated defense of federalism when discussing whether Italy should be a federal state. Chapter 17's concluding lines setting out when a formal division of legislative powers is desirable may then have clearer application to contemporary federalism but admits distinct interpretations.⁴⁵ Mill writes:

if there is a real desire on all hands to make the experiment successful, there needs seldom be any difficulty in not only preserving … diversities, but giving them the guarantee of a constitutional provision against any attempt at assimilation, except by the voluntary act of those who would be affected by the change.

Yet whether, when, and why constitutional guarantees should occur is debatable. This passage appears in a discussion of the maintenance of Civil Law in a sub-state unit of a country otherwise adopting Common Law norms. Its “diversities” concern the details of sub-state “forms of governance,” like Civil Law. And the passage follows discussion of how administrative decentralization should suffice to address many of the peace- and stability-related concerns Mill takes as central. The passage likely supports the kind of legislative division of powers characteristic of federalism. But if Mill only considers federalism desirable to foster peace and stability under set conditions, federalism may prove a second- or third-best mode of governance for non-ideal settings and normative arguments therefore may be secondary to a practical one.

Analyzing competing interpretations of Chapter 17 is beyond my scope of inquiry. But arguments above can inform interpretative debates. It is, for instance, notable that neither quoted passage from Chapter 17 contains epistemic or other normative reasons to favor federalism. Considerations above provide reason to consider this intentional and favor interpretations of Mill as at best ambivalent about federalism. If one finds that result problematic, in turn, the preceding still helps identify new research questions. For instance, the preceding raises clear questions about how epistemic norms figure in Mill's arguments about federalism. Even if, for example, Mill “really” believed that federal unions of smaller states are all-things-considered desirable, whether their “coming together” will have epistemic benefits or costs is important.

Arguments above more clearly illuminate the normative status of federalism. The foregoing does not establish an epistemic disqualifier on federal rule. Each level of government in a federal country could meet relevant epistemic thresholds. But federalism as such is now unlikely to fill any gaps in epistemic justification that may otherwise exist. Epistemic considerations are at best neutral respecting the justification of federalism and unlikely to favor it over alternative modes of governance, including unitary ones. If federalism is desirable, it is for non-epistemic reasons.

One natural path for future research would examine alternative justifications for federalism and roles epistemic conditions on authority can play therein. Epistemic concerns may be relevant to justifications for federalism even if they do not favor federalism on their own. If, for example, federalism is necessary to ensure those most impacted by certain kinds of decisions make those decisions, impacted parties' knowledge of how that impact affects the local community can play a role in a democratic argument for federalism.⁴⁶ However, any such justification will be primarily grounded on democratic concerns. Likewise, one may believe that federalism is useful for respecting self-determination of peoples.⁴⁷ Such peoples are well-placed epistemically respecting their members' interests and this contributes to an explanation of why self-determination is also instrumentally valuable. But if the people did not have independent grounds to rule, it is not clear that local knowledge alone provides them with a claim. Plausible authority claims are not clearly well-realized within a federal state. Future research should recognize the nature of these arguments and merely supplementary role that epistemic concerns play within.

If federalism is justified, epistemic considerations could additionally help identify who can plausibly possess authority in federal countries. Future work may analyze whether and when this is so. The foregoing suggests epistemic approaches to allocating authority within federal boundaries are likely to reach varied results. Nationwide, sub-state level, and municipal governments each have epistemic vices and virtues, which occur to different degrees based on a host of factors that will not generalize across all countries. Identifying what knowledge is most relevant to an issue and how to allocate authority to maximize the chances that the level of government most likely to have that knowledge will exercise it requires careful tailoring. Mill's lack of details on what qualifies as “purely local concerns” is then unsurprising. The class of goods that will be consistently best addressed by a local body may not submit to easy definition.

The epistemic case for federalism is, in any case, ultimately weaker than many suppose. Stronger arguments rely primarily on non-epistemic considerations. Recognizing this need not cause despair among federalists. The foregoing helps us better understand which arguments are available. At the same time, it provides moral clarity on why arguments for federalism do not submit to easy solutions. Arguments above also highlight the trade-offs that can attend even strong arguments for democracy. One should attend to federalism's epistemic losses even if they do not defeat a case for federalism. Analyses of federalism should account for the trade-offs and examine whether and when federalism's non-epistemic benefits offset attendant epistemic losses.