Libertarian Paternalism, 20 Years After.
What To Do With Behavioral Insights in a Liberal Democracy?
Almost twenty years ago, the legal scholar Cass Sunstein and the economist Richard Thaler published an article called “Libertarian Paternalism Is Not an Oxymoron”.[1] This article provides the philosophical and theoretical background for Thaler and Sunstein’s subsequent work on nudges[2] and more generally “behavioral public policies.” It has triggered a vast literature on what Sunstein later called the “ethics of influence.”[3] I will, in this post, briefly reflect on what we have learned after two decades of discussions about the economics and ethics of behavioral public policies.
To have an idea of the importance taken at the policy level of behavioral sciences and especially behavioral economics, a good place to start is the 2017 OECD report on “behavioral insights.” The report lists and discusses around a hundred behavioral public policies implemented at different scales in the world in a range of areas (consumer protection, education, energy, environment, financial products…). Empirically, it indicates that behavioral public policies are effective under a large range of circumstances. It also shows that the use of behavioral insights in policymaking is not restricted to nudges in the narrow sense that Thaler and Sunstein gave to the concept in the 2000s, i.e., interventions that exploit individuals’ cognitive biases to steer their behavior in a direction that the policymaker estimates the individuals would judge to be good for themselves. What appears however is that all these behavioral policies have in common to affect people’s behavior through influence, rather than based on coercion or economic incentives.
This is the core idea of Sunstein and Thaler’s 2003 paper. Public policies are generally viewed as coercive. Even when they rely on economic incentives instead of direct coercion, they only work because the change in opportunity costs operates through people’s beliefs that property rights will be enforced if needed (think of the way taxes work). This is unproblematic in many cases. The question is always what makes the use of coercion legitimate. Externalities and collective goods are among the main reasons that may justify that public policies interfere with people’s choices and eventually coerce them. Things are different however when policies target people because the policymaker considers that they make suboptimal if not plainly wrong choices for themselves. This becomes contentious for well-known reasons, many identified by John Stuart Mill in his classic On Liberty:[4] such policies prevent people to learn, they substitute the judgment of someone else for the judgment of the individuals directly concerned about what is good for them, they infringe into their privacy rights, they undermine diversity and social innovation. That explains why paternalism generally has bad press. It tends to be viewed as inefficient (people tend to better know than anyone else what is good for themselves) and unethical (it disrespects people’s autonomy and individuality). However, behavioral economics may be interpreted as showing that the former is false – people do make choices that adversely affect their wellbeing. What about the latter?
The ethical issue can be framed in terms of a dilemma.[5] If it happens that behavioral public policies are effective and efficient (i.e., they indeed have a significant effect on people’s behavior and tend, everything else equals, to improve their wellbeing), then this will be in a large number of cases through mechanisms of influence the targeted individuals do not have any knowledge and are not aware of. In this case, we may wonder if the influence is justified and legitimate as it seems to disregard people’s autonomy and freedom from covert interference. On the other hand, making people aware of these interferences may run the risk of diminishing the effectiveness and thus the efficiency of behavioral public policies.[6] So, a choice may have to be made between the pursuit of effectiveness and efficiency and the respect of personal autonomy and freedom without which the legitimacy of public policies in a liberal-democratic political culture cannot be secured.
It is interesting to note the discrepancy between the views of policymakers and academic scholars regarding the importance of this ethical issue. When you look at the OECD report I referred to above (p. 39), you can see that a large majority of the respondents (public policymakers or organizations mandated to act on behalf of state authorities) consider that there is no significant ethical issues related to the use of behavioral insights for policymaking.
This contrasts with the large academic literature (especially in philosophy, but also in economics) that addresses various worries all related to the ethical issue as I’ve formulated above. Sunstein and Thaler’s 2003 paper is obviously motivated by some of those concerns. The very title of the article is clear: the point is to claim that we should not be worried by the kind of paternalism behavioral policies implement because it is fully compatible with liberal and democratic values. Sunstein and Thaler deploy a variety of arguments to defend this claim. The main one is that it is basically impossible for a policymaker not to frame people’s choices. The policymaker could indeed, in principle, opt for a frame such that people ultimately choose based on what they actually want. But such a frame is very difficult to design in practice. Most of the time, the policymaker has no choice but to design a non-neutral frame that will influence people’s choices. In the most common kind of cases, behavioral sciences give the opportunity to improve individuals’ situations through the implementation of nudges or other behavioral devices. Why should the policymaker refrain to do so?
As it has been noted multiple times in the literature, it is a relatively weak argument. Even if we grant the truth of the claim that influence through framing is unavoidable, it doesn’t follow that the policymaker should intentionally frame individuals’ choices such as to influence them in a specific way, even if it is for their own good. There is first the obvious problem of determining what individuals “really want”. It is unclear where this knowledge can come from. Behavioral economists have relied on an array of proxies such as laundered or true preferences, subjective happiness judgments, or “folk” goodness judgments based on intuition. None of them is convincing. Second, the argument completely sidesteps the dimension of intentionality. Influence is arguably unavoidable; intentional influence is another thing because it expresses contempt for others’ ability to judge and make choices. As far as liberal values are the expression of respect for persons and their autonomy, paternalism expresses a form of authority that seems hardly compatible with them. Finally, there is the problem that, at least for those nudges that work by exploiting people’s cognitive biases, we have a form of covert influence that seems antithetical to the publicity requirement that grounds liberal political morality, at least according to a widespread philosophical view.
It is therefore not surprising that most of the commentators have been skeptical, if not very critical of the idea of libertarian paternalism. Economists like Robert Sugden[7] and, in a more dramatic fashion, Gilles Saint-Paul,[8] have well expressed the problems with Sunstein and Thaler’s original claim. Alternative defenses have progressively been proposed. A very sensible argument is that nudges and other kinds of behavioral public policies are necessary to counter the use of behavioral insights by private companies that are ostensibly directed at profit maximization but not necessarily consumers’ wellbeing.[9] It doesn’t however establish that behavioral paternalism can be compatible with liberal values, only that the latter might be inconsistent and the corresponding society unstable. Sunstein and co-authors have conducted surveys suggesting that, as a matter of fact, people in Western liberal democracies are not necessarily opposed to being nudged, including in the most covert manners, as long as the goals of the corresponding policies are perceived as legitimate.[10] Finally, and relatedly, there is the claim that paternalism is not ethically problematical as long as it targets only the means people use to achieve their ends, not people’s ends themselves.
None of this is fully convincing. After all, we may doubt that people, in light of their cognitive biases, are able to appropriately judge the goals pursued by behavioral public policies. Surveys also indicate that they still prefer more overt policies, rather than covert ones. Moreover, the means/ends distinction, though useful in theory, is hard to use in practice. If you consider for instance Thaler and Sunstein’s famous 401(k) plans example, the motivation for this nudge was that people display through their choices a higher discount rate that the one that is best for them. The question is whether their “real” discount rates correspond to the one assumed by the policymaker to set the nudges (in which case, the policy interferes only with the means) or if individuals’ actual discount rates is higher than the policymaker judges it should be – in which case, the policy interferes also with the ends pursued by individuals. Obviously, the distinction is very hard to sustain in practice and is quite often meaningless.
All this however doesn’t mean that liberal paternalism is ultimately an oxymoron. After all, there is a liberal tradition that is friendly to paternalist policies on perfectionist grounds.[11] Liberal perfectionists claim that state authorities are legitimate in coercing people if such coercion contributes to developing their personal autonomy – for instance, collecting taxes to finance museums. Of course, not everyone agrees with perfectionist morality but strictly speaking, if one accepts that liberal perfectionism is a consistent moral and political view, then libertarian paternalism is not a logical impossibility. A problem with this approach – if this is one – is that it can have profoundly nondemocratic implications. Indeed, it seems that liberal perfectionism can encourage a form of “political paternalism” that may legitimize epistocratic political institutions.[12] But this is anyway an issue that proponents of behavioral public policies must face upfront, for if you show that individuals are generally biased and make bad decisions because of that, then the same must apply when they play their role as citizens and vote for candidates and policies.[13]
There is also a more democratic approach to partially vindicate behavioral public policies, even if they have a paternalistic character. This approach has essentially contractualist and deliberative roots. As John Rawls briefly noted in A Theory of Justice, individuals in the original position may perfectly consent to paternalistic policies.[14] This depends however on the satisfaction of at least two requirements. First, the public justification of behavioral public policies must be grounded on a general and shared knowledge of the biases that affect individual behavior and their consequences. Such knowledge gives a reason (not necessarily decisive though) to accept the principle of public institutions relying on and using behavioral insights to influence people’s behavior. Second, the local design and implementation of behavioral public policies should, when possible, be conducted in such a way that people can potentially deliberate (individually and collectively) on the acceptability of the mechanisms of influence used by the policies. This second requirement is arguably vague and difficult to put into practice. What policymakers should aim for, in this perspective, is not individuals are always aware of the mechanisms that are influencing them, but that they have the resources at their disposal to learn about a specific policy that may influence them is supposed to work. In an ideal setup, this may imply for instance giving people access to information explaining the rationale of the policy and how it is supposed to work. Behavioral interventions that only work in the most covert way will be eliminated (or will not work) but those that remain efficient when they are more overt will likely pass the test of public justification.
[1] Cass R. Sunstein and Richard H. Thaler, “Libertarian Paternalism Is Not an Oxymoron,” The University of Chicago Law Review 70, no. 4 (2003): 1159–1202, https://doi.org/10.2307/1600573.
[2] Richard H. Thaler and Cass R. Sunstein, Nudge: Improving Decisions About Health, Wealth, and Happiness, Revised&Expanded (Penguin Books, 2009).
[3] Cass R. Sunstein, The Ethics of Influence: Government in the Age of Behavioral Science (Cambridge University Press, 2016).
[4] John Stuart Mill, On Liberty (Harper Collins, 2013).
[5] I expose and build on this dilemma in my paper “Consent and Behavioral Public Policies: A Social Choice Perspective,” Res Publica 29, no. 1 (2023): 141–63.
[6] This is not always the case, though. See for instance Hendrik Bruns et al., “Can Nudges Be Transparent and yet Effective?,” Journal of Economic Psychology 65 (April 1, 2018): 41–59.
[7] Robert Sugden, The Community of Advantage: A Behavioural Economist’s Defence of the Market (Oxford, United Kingdom ; New York, NY: OUP Oxford, 2018).
[8] Gilles Saint-Paul, The Tyranny of Utility: Behavioral Social Science and the Rise of Paternalism (Princeton University Press, 2011).
[9] Adam Oliver, “Nudging, Shoving, and Budging: Behavioural Economic-Informed Policy,” Public Administration 93, no. 3 (2015): 700–714, https://doi.org/10.1111/padm.12165.
[10] Lucia A. Reisch and Cass R. Sunstein, “Do Europeans like Nudges?,” Judgment and Decision Making 11, no. 4 (July 2016): 310–25. Lucia A. Reisch, Cass R. Sunstein, and Wencke Gwozdz, “Viewpoint: Beyond Carrots and Sticks: Europeans Support Health Nudges,” Food Policy 69 (May 1, 2017): 1–10.
[11] Joseph Raz, The Morality of Freedom (Clarendon Press, 1986). Wall, Liberalism, Perfectionism Restraint, (Cambridge (GB): Cambridge University Press, 2008).
[12] I make such a claim in a forthcoming paper in Public Affairs Quarterly. You can email me for the preprint.
[13] Christopher H. Achen and Larry M. Bartels, Democracy for Realists: Why Elections Do Not Produce Responsive Government (Princeton: Princeton University Press, 2016).
[14] John Rawls, A Theory of Justice (Oxford University Press, 1971), pp. 248-9.