Sage Journals: Discover world-class research

Abstract

Consider a society where the prestige of orthodox views is linked to the prestige of the elite. Heterodox individuals are less likely to express their views if other peers refrain from doing so and if the elite is prestigious. In turn, corruption by the elite is less easily detected if orthodox views dominate. We characterize equilibrium self-denial and corruption and show that an exogenous increase in the range of orthodox views may result in a decrease in the total number of individuals truthfully expressing their views. Some features of the model are shown to be compatible with US data.

Keywords

Conformity Overton window political correctness preference falsification social pressure

1. Introduction

The purpose of this article is to develop a theory of equilibrium ‘political correctness’, that is, a situation whereby individuals express views which are not actually theirs because they are well-received within society (Bernheim, 1994; Loury, 1994; Michaeli and Spiro, 2015). As in any theory of conformity, individuals with politically incorrect views have to choose whether to truthfully express their heterodox view making them an outsider to society or instead to self-deny by expressing an orthodox view to be accepted by society. Peer pressure and the prestige of orthodox views are therefore decisive ingredients in this choice. What is new to our paper is that we link the prestige of orthodox views to the prestige or power of the social group representing these orthodox views (see Bourdieu and Boltanski, 1976; Susen, 2013). Infringements of the orthodoxy should be more likely the weaker the power/prestige of the elite defending these views. For instance, liberation of speech from the monopoly of experts/economists and criticism of ‘excessive’ political correctness were important themes put forward by respectively defendants of Brexit and supporters of Trump in 2016,¹ and these political movements rose after the prestige of elites was damaged by the Great Recession (Hopkin, 2020).

To illustrate the relevance of the novel mechanism that links the prestige of orthodox views to the prestige of the elite representing these views our paper answers three questions:

When will heterodox individuals express themselves in a politically correct way, if their choices only depend on what other individuals do?

How will the results change in the presence of an elite that influences the prestige of the politically correct views?

If more views become politically correct, will this always lead to more individuals truthfully expressing their views, and how does this depend on elite influence?

To answer the first question, we develop a benchmark model without an elite which we refer to as the simple model. In this model, all individuals simultaneously decide which view to express. They gain a benefit from expressing their view truthfully; but in case this view is not politically correct, they suffer from a reputation loss which is increasing in the proportion of self-deniers due to peer pressure. If they express a politically correct view, they gain utility from belonging to society, but this requires self-denial which entails a loss increasing in the distance between their true view and their expressed view. We show that truthfully expressing their view is a dominant strategy for all orthodox individuals, while the amount of peer pressure is crucial for the behavior of heterodox individuals. When peer pressure is low, nobody chooses self-denial at equilibrium, as being truthful generates a utility and any self-denial by other heterodox individuals would be heavily discounted due to the low peer pressure. Instead, when peer pressure becomes larger, the interaction among heterodox individuals becomes relevant, generating multiple equilibria: while no self-denial remains the best option for everybody if nobody else self-denials, full self-denial by heterodox individuals now becomes an equilibrium as self-denial by sufficiently many peers renders the expression of heterodox views more costly than self-denial.

To answer the second question, we develop our main model by adding an orthodox ‘elite’ (a dominant group) to our simple model and we link the prestige of orthodox views with the prestige of the elite. Specifically, we assume that a high level of self-denial among heterodox individuals renders orthodox views more prestigious, which in turn enhances the prestige of the elites and makes it easier for them to engage in rent-seeking activities (corruption) without being detected. At the same time, we assume that if elite members choose not to engage in corruption, their prestige is higher. To the extent to which masses care about elite behavior, better behavior by the elite translates into a higher prestige of orthodox views, and thus raises the value of self-denial for heterodox individuals. In our main model, masses and elite members move simultaneously, the former choosing which view to express and the latter whether or not to be corrupt. Since orthodox individuals always reveal their views from now on when we talk about the masses we refer to heterodox individuals only.

We show that the existence of the elite leads to two new results. Firstly, self-denial can arise as an equilibrium outcome even for low levels of peer-pressure whenever the good behavior of the elite sufficiently raises the value of self-denial. In the simple model, the externality created by self-denial when peer pressure is low is very weak and for this reason no self-denial is the unique equilibrium; instead, if heterodox individuals care about the good behavior of the elite, the latter generates individual incentives for self-denial, which increase the externality created by self-denial even for lower levels of peer-pressure.²

Secondly, if the masses care sufficiently about the elite’s behavior and peer pressure is sufficiently high, the presence of the elite results in full self-denial becoming the unique equilibrium in situations where there was multiplicity in the simple model. Interestingly, full self-denial always holds in this case no matter the elite’s payoff from corruption, and the corruption payoff only affects the observed level of corruption characterizing the equilibrium. Specifically, if corruption pays little, full corruption is never a best reply for the elite, but at the same time the prestige of orthodox ideas stemming from self-denial lowers the corruption detection probability and generates some corruption. In turn, if corruption pays a lot, the same mechanism leads to a situation in which full adhesion of the masses to the ideas of the elite results in full corruption by the elite.

The final question we examine is how an exogenous increase in the range of socially acceptable views affects the proportion of agents expressing their views in equilibrium. Naturally, those individuals whose views have (exogenously) become orthodox can now freely express them, so initially less individuals self-deny. Consider however the situation of an individual holding a ‘very heterodox’ view: this individual was initially choosing to express her views because self-denial was too costly, as her own view was too different from any orthodox view. However, when more views become socially acceptable, her view is less distant from the limit orthodox view, she then may choose to self-denial, which in turn lowers the number of individuals expressing their views. In the simple model, we show that this second effect is always dominated and that as result a wider set of orthodox views always leads to less self-denial. This does not necessarily hold in the model with elite, as the initial fall in the proportion of self-deniers boosts the corruption detection probability, which improves the behavior of the elite, and further reinforces self-denial. When the masses care a lot about elite behavior, an increase in the range of socially acceptable views might end up lowering the number of individuals expressing their views in equilibrium.

We empirically illustrate some of the features of the model using the 2016 Survey of American Political Culture (Hunter and Bowman, 2016) conducted in August 2016 two months ahead of Donald Trump’s election. Consistently with the model, individuals who view themselves as more distant from society are more likely to believe there is a political correctness problem in the US. In addition, individuals considering that the elites are self-interested are also more likely to believe political correctness to be a problem. This is also in accordance with our model.

Our paper relates to a growing literature on political correctness. Loury (1994) shows how in the presence of norms on acceptable views within a group, the interaction between sender and receiver becomes strategic and can generate a link between ways of expression and bad qualities of senders, resulting in the avoidance of such ways of expression. Bernheim (1994) shows that if intrinsic preferences are not observable and individuals care sufficiently about status, heterogeneous individuals may all choose the same action in order to avoid being characterized as somebody with uncommon intrinsic preferences. Hallman and Spiro (2023) modified the curvature of Bernheim’s (1994) intrinsic preferences to develop a theory of hypocrisy. Morris (2001) differs from the above papers and ours in that there is a true state of the world, with the analysis focusing on the extent to which an advisor will be able or not to convey this state of the world to the policy maker when the advisor cares about his own reputation in the eyes of the policy maker. The general message of our simple model is similar to that of Michaeli and Spiro (2015) except that in the latter individuals can choose the extent to which they self-deny, implying that conformity to the social norm does not only depend on the number of self-deniers.³

Overall, the contribution of our paper to the literature is to link the prestige of orthodox ideas and the behavior of the elites representing them, which can be related to Bourdieu’s (1979) notion of symbolic power arguing that while debates of views tend to be presented as belonging to an autonomous asocial sphere, the power of orthodox ideas is also dependent on the power of the associated dominant groups.

The remainder of the article is organized as follows. In Section 2, we analyze a simple political correctness model which is extended in Section 3 to incorporate an elite. Section 4 describes the equilibria of this full model and emphasizes the new results with respect to the benchmark. Section 5 provides an empirical illustration of the main mechanism in the model. Section 6 concludes. Most technical details are relegated to the Appendix.

2. A simple model

2.1. The model ingredients

Consider a society with a continuum of individuals uniformly distributed on the line $[0, \bar{k}]$ where $k_{i} \in [0, \bar{k}]$ represents individual i’s view. This distribution of preferences is common knowledge but each individual’s real preference is private information. Among the existing views, only those located in $[0, k_{R}]$ with $k_{R} < \bar{k}$ are thought of as acceptable by society. We will refer to these views as ‘orthodox’ or ‘politically correct’ views and to $[0, k_{R}]$ as the Overton window. Views outside this Overton window namely $k_{i} \in (k_{R}, \bar{k}]$ are heterodox and rejected by society. Expressing a politically correct view implies being accepted by society which results in a utility gain of $v_{A}$ . Individuals enjoy a positive value $v_{S}$ from being truthful to their own self, that is, from expressing their own view truthfully, but are also subject to peer pressure. In particular, there is peer pressure towards political correctness within the group of heterodox individuals, and thus a heterodox individual expressing her views enjoys utility

U_{h}^{S} (μ) = v_{S} - α μ

(1)

where the positive constant

α

represents the intensity of peer pressure towards political correctness⁴ and

μ

captures the proportion of heterodox individuals expressing a politically correct view. For orthodox individuals there is no peer pressure to express a non politically correct view, hence their utility from being truthful towards themselves and revealing their view truthfully is always

U_{o}^{S} (μ) = v_{S} + v_{A}

Falsifying one’s preferences comes at a cost which is linear in the distance between the expressed view

k_{i}^{e}

and the truthful view

k_{i}

, that is,

c (k_{i}^{e}) = - | k_{i} - k_{i}^{e} |

. All individuals play a simultaneous move Bayesian game in which they decide which view to express.

Consider first the group of orthodox individuals with $k_{i} \in [0, k_{R}]$ . Clearly, their utility is always maximized by expressing their view truthfully. Not doing so results in the utility loss of $v_{S} + v_{A}$ and entails the cost of misrepresenting their preferences. Therefore being truthful is a strictly dominant strategy for orthodox individuals. Given this result, we suppress from now on the subindex h and concentrate on the choices of the heterodox individuals, who have to decide whether to express their views truthfully or express an orthodox view.⁵ Expressing any of the orthodox views results in the same gain of $v_{A},$ however, the misrepresentation costs $c (k_{i}^{e})$ varies for different orthodox views. Therefore heterodox individuals will only consider whether or not to express the closest orthodox view to their own view, namely $k_{R},$ since this view minimizes the misrepresentation costs $c (k_{i}^{e})$ . Denoting by $d_{i}$ the distance between the individual’s own view and the closest orthodox view (i.e. $d_{i} = k_{i} - k_{R}$ ), the value of self-denial is given by the following equation:

U_{i}^{S D} = v_{A} - d_{i} = v_{A} - (k_{i} - k_{R})

(2)

Equalizing (2) and (1), there is a unique cutoff distance

\tilde{d}

(for each given

μ

) from the closest acceptable view

k_{R}

below which individuals self-deny and above which they express dissent:

\tilde{d} = α μ - v_{T}

(3)

where

v_{T} = v_{S} - v_{A}

captures the utility of being truthful net of the disutility of expressing a heterodox view. We assume that

v_{T} > 0

. Taking into account that

\bar{k} - k_{R}

is the total mass of people who could potentially self-deny, the proportion of self-deniers is given by the following equation:

μ = {\begin{matrix} 0 \ if\ \tilde{d} < 0 \\ \frac{\tilde{d}}{\bar{k} - k_{R}} \ if\ 0 < \tilde{d} < \bar{k} - k_{R} \\ 1 \ if\ \tilde{d} > \bar{k} - k_{R} \end{matrix}

(4)

Introducing (3) into (4), the equilibrium level of self-denial is given by the following equation:

μ = {\begin{matrix} 0 \ if\ v_{T} > 0 \\ \frac{v_{T}}{α - (\bar{k} - k_{R})} \ if\ 0 < \frac{v_{T}}{α - (\bar{k} - k_{R})} < 1 \\ 1 \ if\ α - v_{T} > \bar{k} - k_{R} \end{matrix}

(5)

2.2. Equilibria in the simple model

Figure 1 presents the equilibrium proportion of self-deniers $μ$ as a function of the intensity of peer pressure $α$ . Given there is a positive value associated to being truthful ( $v_{T}$ ) and that the incentives for self-denial are always inexistent if nobody self-denies, a situation whereby all heterodox individuals truthfully express their views ( $μ = 0$ ) is always an equilibrium. In addition, if the peer pressure parameter $α$ is large enough to generate the self-denial of the most heterodox agent ( $α > \bar{k} - k_{R} + v_{T}$ ), an equilibrium with full self-denial ( $μ = 1$ ) becomes possible, and thus the area where $α > \bar{k} - k_{R} + v_{T}$ holds is characterized by multiple equilibria. As the cost of being truthful is increasing in the number of self-deniers, one equilibrium or the other arises depending on whether the agents coordinate on self-denial or instead on being truthful. In addition, there exists an interior equilibrium proportion of self-deniers so that the marginal mass member is indifferent between self-denying and being truthful.⁶ Note that as the span of heterodox views becomes larger (i.e. as $\bar{k} - k_{R}$ increases), political correctness can arise at equilibrium only for an increasingly stronger peer pressure.

Figure 1.

Equilibria in the simple model.

2.3. An increase in the orthodoxy range

One important question is to understand whether a society encompassing an increasingly large set of views (i.e. including more views as orthodox as $k_{R}$ becomes larger) automatically results in more individuals expressing their own views at equilibrium.

In the unique interior cutoff equilibrium, as represented in Figure 2, individuals express their views if they are orthodox—corresponding to the interval $(0, k_{R})$ —or alternatively if they are sufficiently far away from the orthodoxy—in $(k_{R} + \tilde{d}, \bar{k})$ —as self-denial would in that case entail too large a cost. Instead, the individuals with views closer to the orthodoxy—in $(k_{R}, k_{R} + \tilde{d})$ —self-deny by expressing the politically correct view $k_{R}$ .

Figure 2.

Individuals expressing their view (interior equilibrium).

As the total number of individuals who reveal their true views is given by the following equation:

h (k_{R,} \bar{k}) = \int_{0}^{k_{R}} \frac{d k_{i}}{\bar{k}} + \int_{k_{R} + \tilde{d}}^{\bar{k}} \frac{d k_{i}}{\bar{k}} = \frac{\bar{k} - \tilde{d}}{\bar{k}}

(6)

Using (3) and (5), it is easy to show that:

\frac{\partial h (k_{R,} \bar{k})}{\partial k_{R}} = \frac{α v_{T}}{{[α - (\bar{k} - k_{R})]}^{2} \bar{k}} > 0

(7)

so the number of individuals truthfully expressing their views unambiguously increases when more views become orthodox. Figure 3 represents the mechanisms at play. Initially, the extension of the Overton window to views in

(k_{R}, k_{R}^{'}]

removes any cost of expressing their own views for individuals holding views in this interval, and accordingly reduces self-denial. At the same time, the cost of self-denial goes down from

k_{i} - k_{R}

k_{i} - k_{R}^{'}

for individuals holding views to the right of

k_{R}^{'}

, and this results in individuals with views in

(k_{R} + \tilde{d}, k_{R}^{'} + {\tilde{d}}^{'})

changing their behavior from truthful expression to self-denial.⁷ However, from (7), this effect is dominated and overall more individuals truthfully express their view.⁸

Figure 3.

Effects of a larger Overton window (interior equilibrium).

So far, we have only considered how an increase in the orthodoxy range changes the characteristics of the interior equilibrium. However, from the analysis above, we know that the change in the orthodoxy range may also affect the set of equilibria. Assume that initially $k_{R}$ is quite small: in such a situation, the cost of self-denial is quite high for many individuals because their views are quite different from the closest orthodox view, which implies from Figure 1 that $μ = 0$ may be the unique equilibrium and thus nobody self-denies. As more views become orthodox, the individuals holding the newly orthodox views will still express their views, and this not as a result of an equilibrium, but simply because their views have become socially acceptable. At the same time, if the change in $k_{R}$ is such that it results in $α > \bar{k} - k_{R} + v_{T}$ , equilibria with positive self-denial become possible and the society potentially switches to a new equilibrium where either all individuals outside of the new orthodoxy range (in the $μ = 1$ equilibrium) or only individuals close enough to the orthodoxy range start self-denying. This implies that a society accepting more views as orthodox might paradoxically end up at an equilibrium with a smaller total number of individuals speaking up their views if the society coordinates on a different equilibrium when the Overton window increases.

3. A model with elites

We now introduce an elite in the model. The elite are the ‘guardians’ of the orthodoxy, either because orthodox ideas originated with them or because one of their functions in society is to represent/defend these ideas. However, in terms of actions, elite members still have a choice between behaving according to socially acceptable ideas or instead misbehaving—we also refer to this for simplicity as ‘being corrupt’. We assume that elites and masses play a simultaneous move Bayesian game. We first study the behavior of each elite member, who takes the behavior of the masses and the behavior of other elite members as given.

3.1. Behavior of the elites

An elite member who chooses to misbehave enjoys a fixed payoff from corruption, but this payoff is lowered by a penalty if the misbehavior is detected. We assume the probability of detection to be smaller the more prestigious the elite is, which is itself tied up to the prestige of orthodox views to the extent to which mass members self-deny. For simplicity, we assume the detection probability to be given by the proportion of dissenters $1 - μ$ , that is, the utility from misbehavior is

V_{M} (μ) = β - c (1 - μ)

(8)

where

β

is the payoff from corruption, and c the misbehavior penalty arising with detection probability

1 - μ

On the other hand, elite member j derives a fixed private utility $b_{j}$ from good behavior

V_{j}^{G} = b_{j}

(9)

and this fixed utility is uniformly distributed across the elites between

\underline{b}

and

\bar{b}

. This distribution is common knowledge. Putting together (8) and (9), the indifferent elite member is characterized by

\tilde{b} (μ) = β - c (1 - μ)

(10)

From (10), the proportion of honest elite members is determined by the following equation:

e (μ) = {\begin{matrix} 0 \ if\ \bar{b} < β - c (1 - μ) \\ \frac{\bar{b} - β + c - c μ}{\bar{b} - \underline{b}} \ if\ \bar{b} > β - c (1 - μ) > \underline{b} \\ 1 \ if\ β - c (1 - μ) < \underline{b} \end{matrix}

(11)

That is, depending on the characteristics of the distribution, on the behavior

μ

of the masses and on other parameters, we can have an interior value for the threshold (as in the second line of (11)) or a situation in which no elite members or all elite members are honest (respectively, the first and third lines). The bounds in (11) can be rewritten as bounds on the proportion of self-deniers as follows:

e (μ) = {\begin{matrix} 0 \ if\ μ > \frac{\bar{b} - β + c}{c} \\ \frac{\bar{b} - β + c - c μ}{\bar{b} - \underline{b}} \ if\ \frac{\underline{b} - β + c}{c} \leq μ \leq \frac{\bar{b} - β + c}{c} \\ 1 \ if\ μ < \frac{\underline{b} - β + c}{c} \end{matrix}

(12)

which defines the reaction function of the elite to the masses’ behavior. To allow the behavior of the masses to influence the elite we assume that

\underline{b} < β < \bar{b} + c

.⁹

3.2. Choice of the masses

We now determine the choice of a heterodox member of the masses who takes as given the honesty level of the elite as well as the choices of the other member of the masses.¹⁰ As in the simple model, we still assume (1) to hold, that is, expressing dissent entails a cost increasing in the number of self-deniers. However, the existence of the elite affects the value of self-denial, which we assume to be increasing in the prestige of the elite, given that a better behavior of the elites raises the prestige of orthodox views.¹¹ Specifically, we assume that

U_{i}^{S D} (μ) = v_{A} - d_{i} + v e

(13)

where v is a positive constant measuring the relevance of elite behavior for self-denial and thus ve can be interpreted as the intensity of elite-driven self-denial.¹²

Equalizing (1) and (13), there is still a cutoff distance $\tilde{d}$ from the closest acceptable view $k_{R}$ below which people self-deny and above which people express their point of view, namely:

\tilde{d} = v e + α μ - v_{T}

(14)

Introducing (14) into the proportion of self-deniers derived in (4), the reaction function of heterodox masses to the elite’s behavior becomes¹³:

μ (e) = {\begin{matrix} 0 if\ e < \frac{v_{T}}{v} \\ \frac{v_{T} - v e}{α - (\bar{k} - k_{R})} \ if\ 0 < \frac{v_{T} - v e}{α - (\bar{k} - k_{R})} < 1 \\ 1 \ if\ e > \frac{v_{T}}{v} + \frac{(\bar{k} - k_{R}) - α}{v} \end{matrix}

(15)

The reaction function (15) illustrates the new forces at play due to the existence of the elite. Without the elite, the only reason for self-denial is the fear of isolation captured by peer pressure

α

. Now, instead, mass members might choose to self-deny even in the absence of peer pressure if the elite’s good behavior sufficiently raises the value of self-denial, that is, if ve is sufficiently large.

Indeed, comparing the first branch of (15) with the first branch of (5), it appears that the equilibrium with no self-denial ( $μ = 0$ ) does not always exist anymore, and a condition for its existence is now that the value of elite-driven self-denial ve for the least heterodox individual (i.e. with a view $k_{i} \to k_{R}$ ) is smaller than the value of being truthful, $v_{T}$ .

At the other extreme, the third branch of (15) represents a situation of full self-denial ( $μ = 1$ ), which arises whenever the individual with the lowest utility from self-denial (i.e. the most unorthodox agent) has a payoff $v e - (\bar{k} - k_{R})$ from self-denial larger than the payoff from being truthful when everybody else self-denials ( $v_{T} - α$ ).

Finally, the intermediate branch of (15) indicates that there will be an interior equilibrium proportion $μ$ of self-deniers if the least unorthodox individual prefers to self-deny as $v e > v_{T} - μ α$ for her, while at the same time the most unorthodox agent does not self-deny given that $v_{T} - μ α > v e - (\bar{k} - k_{R})$ .

For $α < \bar{k} - k_{R}$ , that is, in a situation where peer pressure for political correctness is mild, the three branches of the reaction function (15) do not overlap and (if they exist) are connected. Thus, in contrast with the simple model where $μ = 0$ was the unique equilibrium for $α < \bar{k} - k_{R}$ , the presence of the elites makes self-denial possible at equilibrium as their prestige creates an extra cost for expressing dissent.

For $α >$ $\bar{k} - k_{R}$ , in turn, that is, when political correctness is ‘high-stakes’, the three branches of the reaction function overlap for $v_{T} + (\bar{k} - k_{R}) - α < v e < v_{T}$ , giving thus potentially rise to multiple equilibria. This is the same type of outcome that was observed in the simple model for $α >$ $\bar{k} - k_{R}$ when $α > \bar{k} - k_{R} + v_{T}$ .

4. Equilibria in the model with elites

A Bayesian equilibrium of our model is a proportion of self-deniers and a proportion of honest elite members $(e, μ)$ satisfying both the reaction function of the elite and that of the masses, that is, respectively, (12) and (15).¹⁴

As in the simple model, more self-denial raises the isolation of those that would express dissent and renders self-denial more attractive by lowering the value of being truthful, see (1). At the same time, more self-denial now generates incentives to misbehavior by the elites (lowers $\tilde{b} (μ)$ in (10)), which in turn weakens the value of self-denial as shown in (13).

Proposition 1 in Appendix B.3 fully characterizes the equilibrium when the reaction function of the masses is unique (i.e. $α < \bar{k} - k_{R}$ ) and shows the equilibrium to be always unique in that case.¹⁵ In turn, Proposition 2 in Appendix B.4 considers the case where several branches of the reaction function of the masses are simultaneously active ( $α > \bar{k} - k_{R}$ ) and shows that multiple equilibria arise in different parts of the parameter space.¹⁶

We now graphically present the equilibria as a function of peer pressure ( $α$ ) and the relevance of elite behavior for self-denial (v) with the equilibrium without elites corresponding to the particular case where $v = 0.$ When doing so, the different equilibrium configurations end up depending on the two parameters characterizing the payoff from corruption for the elite, namely the fixed payoff $β$ and the misbehavior penalty c linked to the number of self-deniers. The full characterization of the equilibrium in the $(α, v)$ space is presented in Proposition 3 (see Appendix B.2).

4.1. Case 1: Low corruption payoff

Consider first a situation where the payoff from corruption $β$ is low enough ( $\underline{b} < β < min [\underline{b} + c, \bar{b}]$ ) to discourage full corruption ( $e = 0$ ) from being a best response for the elite. The resulting equilibrium proportions of honest elites and self-deniers $(e, μ)$ are represented in Figure 4 as a function of the intensity of peer pressure ( $α$ ) and the relevance of elite behavior to the masses ( $υ$ ).

Figure 4.

Equilibria for a low corruption payoff ( $\underline{b} < β < min (\underline{b} + c, \bar{b})$ ).

In the model without elites (corresponding to $v = 0$ in Figure 4), multiple equilibria arise when a large peer-pressure ( $α > \bar{k} - k_{R} + v_{T}$ ) renders the marginal cost of being truthful very dependent on the proportion of self-deniers. When peer pressure is low, the equilibrium with no self-denial is unique.

As the behavior of the elite becomes more and more relevant to the masses, that is, v becomes positive and larger, multiple equilibria start arising also for smaller and smaller values of $α$ , as shown in the shaded area in the South-West of Figure 4. The intuition for this is simple: as the elite behaves (relatively) well in this area, this reinforces self-denial by the masses, and thus the equilibrium with self-denial remains now feasible despite a relatively low peer-pressure, implying that self-denial is here partly elite-driven.

At some point (for $v > v_{T}$ ), elite-driven self-denial is then strong enough to outweigh any incentives for a truthful expression of views resulting from a low peer pressure among the masses or from a coordination in the no self-denial equilibrium. Thus, relative to the simple model, multiple equilibria vanish in the stripped area in the North-East and full self-denial now becomes the unique equilibrium also in that area. Intuitively, in this area, the prestige of orthodox ideas stemming from full self-denial lowers the corruption detection probability and this systematically generates corruption at equilibrium.

More generally, corruption is more likely to arise whenever self-denial is higher, as higher self-denial makes it less likely for misbehavior to be detected.

4.2. Case 2: Intermediate corruption payoff

Consider next a situation characterized by a larger fixed corruption payoff $β$ , and specifically $min [\underline{b} + c, \bar{b}] < β < max [\underline{b} + c, \bar{b}]$ . We show that two equilibrium configurations arise depending on whether the misbehavior penalty c is small or large.

4.2.1. Case 2-low: Small misbehavior penalty

In the case of a larger corruption payoff $β$ , the elite has clearly stronger misbehavior incentives, and this tendency is reinforced if the misbehavior penalty is small ( $\underline{b} + c < β < \bar{b}$ ), see Figure 5. As a result, full honesty ( $e = 1$ ) is no longer an equilibrium outcome and this explains why the equilibrium in the South-West is now $(\tilde{e}, 0)$ instead of $(1, 0)$ in the preceding case, and why $(1, \bar{μ})$ is not anymore an equilibrium. Otherwise, the equilibrium configuration is qualitatively similar to the preceding case, with multiple equilibria still located in the South-East in a larger area than for $v = 0$ (see shaded triangle) but still a North-East quadrant (for $v > v_{T} \bar{b} - \underline{b} / \bar{b} - β + c$ and $α > \bar{k} - k_{R} + v_{T}$ ) in which multiplicity vanishes and full self-denial is the unique equilibrium (see stripped area).

Figure 5.

Equilibria for an intermediate corruption payoff and a low misbehavior penalty ( $\underline{b} + c < β < \bar{b}$ ).

4.2.2. Case 2-high: Large misbehavior penalty

A large misbehavior penalty ( $\bar{b} < β < \underline{b} + c$ ) instead restores no-corruption as a possible equilibrium outcome. Indeed, if the masses do not fully self-deny, they can detect elite corruption with a positive probability, and the large misbehavior penalty makes corruption a bad choice for the elite in some cases. At the same time, full corruption also becomes a possible equilibrium outcome for the first time, because in the presence of full self-denial by the masses, the probability of detection of elites’ corruption is zero, and the high misbehavior penalty becomes in that case irrelevant in a context where the corruption pay-off is non negligible.

In terms of the presence of multiple equilibria, the main difference with respect to the two preceding cases is that, for intermediate values of the intensity of peer pressure ( $α$ ), multiple equilibria now arise no matter the relevance of elite behavior to the masses (v), see the shaded area in Figure 6. Indeed, while all three cases are characterized by multiple equilibria for intermediate values of $υ$ , in Figure 6 multiple equilibria arise also for small and large values of $υ$ . For instance, a high v generates self-denial among the masses, reducing the corruption detection probability and pushing the elite towards full corruption, and we can then end up in an equilibrium in which the full adhesion of the masses to the ideas of the elite results in full corruption by the elite. In the two preceding cases, instead, while the high $υ$ still generated self-denial by the masses and lowered the detection probability, full corruption was never an equilibrium outcome for the elite.

Figure 6.

Equilibria for an intermediate corruption payoff and a high misbehavior penalty ( $\bar{b} < β < \underline{b} + c$ ).

Finally, in the areas characterized by a unique equilibrium, there is still an equilibrium with no-corruption by the elite and no self denial by the masses in the South-West, and in the North-East an equilibrium with no honest behavior and full self-denial—this time $(0, 1)$ instead of $(\hat{e}, 1)$ because the corruption payoff is larger ( $β > \bar{b}$ instead of $β < \bar{b}$ in the two preceding cases).

4.3. Case 3: Large corruption payoff

Finally, the case of a large corruption payoff ( $max [\underline{b} + c, \bar{b}] < β < \bar{b} + c$ ) is represented in Figure 7, with outcomes very similar to those of Figure 6 except for the fact that full honesty is not anymore an equilibrium outcome and at most a proportion $e_{i n t}$ or $\tilde{e}$ of the elites are honest. Appendix B.6 presents how the different cases arise as $β$ changes and shows there is a simple pattern explaining how the equilibrium configuration changes with $β$ .

Figure 7.

Equilibria for a high corruption payoff ( $max (\underline{b} + c, \bar{b}) < β < \bar{b} + c$ ).

4.4. An increase in the orthodoxy range with elites

We next examine how changes in the size of the Overton window $k_{R}$ affect the total number of individuals $h (k_{R})$ expressing their own views at equilibrium, see (6). To that purpose, we first need to understand which are the relevant equilibria for different values of $k_{R}$ . This is actually quite straightforward: since both $k_{R}$ and $α$ appear with a positive sign in the denominator of the reaction function of the masses (15), the figures characterizing the equilibrium in the space $(k_{R}, ν)$ are actually identical to those presented earlier on in this section in the space $(α, ν)$ .

Figure 8 represents how the total number of individuals expressing their own views at equilibrium (h) changes as the size of the Overton window ( $k_{R}$ ) increases, in a situation where the masses care little about the behavior of the elite (low $ν$ ) and the payoff from corruption is low (as represented in Figure 4).¹⁷ The horizontal line on top of the figure indicates that in the no self-denial equilibrium (which arises for every value of $k_{R}$ ) everybody expresses their view no matter the value of $k_{R}$ that is, independently on whether their view is orthodox or not. For larger values of $k_{R}$ , an equilibrium with full self-denial always exists: in that case, only the individuals with newly orthodox views join the set of those who express their views as $k_{R}$ grows, and $h (k_{R})$ corresponds for this reason to the $45^{\circ}$ line. Most interestingly, we see that in the case of the equilibrium with partial self-denial, $h (k_{R})$ is monotonically increasing in $k_{R}$ , that is, we get the same qualitative results as in the benchmark model, which is reassuring given that the elite matters little for the masses: a fall in h as $k_{R}$ goes up can only happen if there is a switch to a different equilibrium. Also, we can see that initial increases in $k_{R}$ are reinforced by lower self-denial among the heterodox, and instead further increases in $k_{R}$ have a smaller impact on h because the fraction of self-deniers starts to grow.

Figure 8.

Number of individuals expressing their views depending on Overton window size, low $ν$ .

Instead, when the corruption payoff is low and the masses care to an important extent about the elite, that is, $ν > ν_{T}$ .¹⁸ Figure 9 shows that the total number of individuals revealing their views is first decreasing and then increasing in the size of the Overton window. The graph represents the three unique equilibria that are relevant for different values of $k_{R}$ , namely two consecutive interior equilibria for low values of $k_{R}$ and then a full-self denial equilibrium when $k_{R}$ becomes larger (see Figure 4). In contrast with the simple model, we can see that for the two interior equilibria, the initial increase in the Overton range leads actually to an increase in self-denial that more than offsets it in the sense that the overall number of individual expressing their own views at equilibrium ends up falling. So although the group considers a larger set of views as being acceptable, a smaller variety of views is uttered at equilibrium. Intuitively, consider the situation of a ‘very heterodox’ individual. Initially, her views were so far away from the orthodoxy, that the cost of self-denying was prohibitive. Instead, when views closer to hers become socially acceptable because the Overton window becomes larger, self-denial becomes ‘cheaper’. While this type of effect exists also in the benchmark model, this second induced effect is never large enough to outweigh the initial effect. This becomes possible instead in the model with elite because the initial fall in the in the proportion of self-deniers boosts the corruption detection probability, which improves the behavior of the elite and further reinforces self-denial. Figures B1 and B2 in Appendix B.7 represent the cases for other areas of the parameter space, with a downward sloping $h (k_{R})$ arising in some cases.

Figure 9.

Number of individuals expressing their views depending on Overton window size, low $β$ and high $ν$ .

The next section considers some predictions of the model in light of the Survey of American Political Culture.

5. Empirical illustration based on the Survey of American Political Culture

The 2016 Survey of American Political Culture (Hunter and Bowman, 2016) includes a question on the perception of political correctness as a problem. This survey consists of 1904 telephone interviews by Gallup on a representative sample of American adults and was performed in late August 2016, that is, two months ahead of the election of Donald Trump. Respondents were questioned about the extent to which they agreed with the statement ‘political correctness is a serious problem in the US, making it hard for people to say what they really think’ (Hunter and Bowman, 2016: 67). In the sample, 40% of the individuals completely agree with the statement, 33% mostly agree, 17% mostly disagree and 10% completely disagree.

In our model, if all views were orthodox, everybody would simply express their own view and get a payoff $v_{S} + v_{A}$ . When not all views are orthodox, this generates a loss to some individuals through social pressure for political correctness. Figure 10 studies the size of this loss depending on the view held by the individual. Individuals holding an orthodox view—located in $(0, k_{R})$ —achieve $v_{S} + v_{A}$ as they can express their view at no cost. Moving towards the right in Figure 10, self-deniers—located in $(k_{R}, k_{R} + \tilde{d})$ —get a payoff $v_{A} - d_{i} + v e$ instead of the $v_{S} + v_{A}$ they would obtain if their view was orthodox, and thus lose $- d_{i} + v e - v_{S}$ . Political correctness is a cost to them because they end up expressing a view that provides them with a lower utility. Finally, heterodox individuals who choose to speak up their view—located in $(k_{R} + \tilde{d}, \bar{k})$ —have a pay-off of $v_{S} - α μ$ , and thus lose $- α μ - v_{A}$ from their view not being orthodox. As $d_{i} + v_{T} < v e + α μ$ from (14) for $d_{i} < \tilde{d}$ , the loss of heterodox individuals truthfully expressing their views is even larger than that of self-deniers. To them, political correctness is a problem because it stigmatizes the view they are expressing. Putting together these three different segments, our model thus predicts that the individual loss from the social pressure for political correctness is (weakly) increasing in the distance of the individual with respect to the orthodoxy.

Figure 10.

Loss to individual i associated to the existence of orthodoxy range [ $0, k_{R}$ ].

Our measure of distance of the individual i with respect to the orthodoxy range is constructed from a set of questions asking the respondent whether the beliefs and values of Americans like them are completely different, mostly different, mostly similar, or completely the same as those of different subgroups of the population. These subgroups are African Americans, Hispanic Americans, White Americans, Muslim or Islamic Americans, Conservative Christians, Non-religious people, and gays and lesbians. Specifically, let $s_{i g}$ be the level of similarity with respect to group g, we use $s_{i g} = 1$ whenever the individual says she shares the same values with group g, $s_{i g} = 2 / 3$ , when she shares mostly similar values, $s_{i g} = 1 / 3$ when the values are mostly different and $s_{i g} = 0$ when they are completely different. Then, aggregating across groups, we construct an average measure of distance for each individual i, namely

d_{i} = \frac{1}{G} \sum_{g = 1}^{G} 1 - s_{i g}

The second main variable of interest is the extent to which the respondent believes that ‘the most educated and successful people in America are more interested in serving themselves than in serving the common good’, for which again four possible answers are given (completely agree, mostly agree, mostly disagree, and completely disagree). In the sample, 21% of the individuals strongly agree with this statement, 40% mostly agree, 30% mostly disagree, and 8% completely disagree. Heterogeneity in the beliefs about the elite can be easily introduced in our model by assuming that the effort produced by the elite e is not perfectly observable and that individuals differ on their perception

η_{i}

of that effort. In that case, the utility from self-denial (13) becomes

U_{S D} (μ) = υ_{A} - d_{i} + v . η_{i} e

and self-denial is less costly for individuals with a more positive view on the elites (i.e. a higher

η_{i}

). Intuitively, individuals with a higher

η_{i}

value the orthodoxy because they have a higher opinion on the behavior of the guardians of the orthodoxy, that is, the elites. This immediately implies that the threshold distance for self-denial now depends on

η_{i}

and, as self-denial is less painful the larger

η_{i}

, individuals with a higher perception of the elite choose to self-deny even for higher distances to the orthodoxy, that is,

\partial {\tilde{d}}_{i} / {\partial η}_{i} > 0

. Figure 11 illustrates how the cost of holding/expressing a heterodox view changes depending on whether the individual has a high (

η_{H}

) or low (

η_{L}

) regard for the elite.

Figure 11.

Individual loss depending on view of the elite $η_{H} > η_{L}$ .

Note first that individuals holding a view in $(k_{R} + {\tilde{d}}_{H}, \bar{k})$ choose to speak up their mind no matter their perception of the elite, and thus their loss does not depend on this perception. Instead, for other heterodox individuals the loss from their views not being considered orthodox is increasing in their negative perception of the elite. Specifically, individuals holding views close to the orthodoxy—in $(k_{R}, k_{R} + {\tilde{d}}_{L})$ —self-deny no matter their $η_{i}$ , but the utility of self-denial is higher (and thus the loss lower) when $η$ is high because a positive view of the elite raises the utility of adhering to the orthodoxy. A similar argument holds for individuals in $(k_{R} + {\tilde{d}}_{L}, k_{R} + {\tilde{d}}_{H})$ , except that different perceptions of the elites for a given view result in different behaviors (self-denial for $η = η_{H}$ as opposed to truthfulness for $η = η_{L}$ ).

In order to be able to run a binomial probit, we pull together for the endogenous variable the individuals that completely agree and mostly agree, on the one hand, and those that completely or partly disagree with the statement. For exogenous variables, we keep separate the different existing categories. Column (1) in Table 1 shows that, as predicted by the model, individuals who view themselves as more distant from society are more likely to believe there is a political correctness problem in the US. In addition, in accordance too with our model, individuals considering that the elites are self-interested are also more likely to believe there is a political correctness problem.

Table 1.

Determinants of the perception of the existence of a political correctness problem, Survey of American Political Culture, 2016

	Dependent variable: Political correctness problem
	(1)	(2)	(3)	(4)	(5)	(6)
Distance	.939***	.878***	.923***	.888***	.659***	.537*
	(.250)	(.269)	(.261)	(.266)	(.280)	(.280)
Self-interested elites	.505***	.394***	.426***	.396***	.439***	.433***
	(.084)	(.090)	(.092)	(.093)	(.099)	(.101)
Age		.028	.0003	$- .008$	$- .041$	$- .030$
		(.024)	(.025)	(.025)	(.028)	(.028)
Male		.321***	.306***	.309***	.318***	.234**
		(.086)	(.088)	(.088)	(.092)	(.097)
Income		$- .002$	$- .002$	$- .002$	$- .004$	$- .005$
		(.005)	(.005)	(.005)	(.005)	(.005)
High school dropout		.355***	.601***	.590***	.500**	.500**
		(.181)	(.186)	(.192)	(.213)	(.216)
High school graduate		.624***	.726***	.739***	.653***	.650**
		(.117)	(.120)	(.120)	(.125)	(.128)
Some college		.372***	.408***	.391***	.309**	.280**
		(.112)	(.116)	(.117)	(.121)	(.123)
White			.383	.291	.328	.072
			(.248)	(.243)	(.256)	(.268)
Black			$- .154$	$- .266$	$- .219$	$- .215$
			(.271)	(.264)	(.279)	(.291)
Asian			$- .173$	$- .188$	$- .030$	$- .125$
			(.346)	(.348)	(.353)	(.368)
Hispanic			$- {.460}^{* * *}$	$- {.492}^{* * *}$	$- {.616}^{* * *}$	$- {.552}^{* * *}$
			(.137)	(.139)	(.149)	(.151)
Density neighborhood				$- {.002}^{* *}$	$- {.002}^{* * *}$	$- {.002}^{* *}$
				(.001)	(.001)	(.001)
Midwest				.008	$- .108$	$- .083$
				(.147)	(.153)	(.156)
South				${.276}^{* *}$	.080	.053
				(.134)	(.142)	(.147)
West				.200	.130	.104
				(.140)	(.151)	(.155)
Christian					.069	$- .019$
					(.163)	(.165)
No religion					$- .049$	$- .105$
					(.180)	(.180)
Catholic					${.277}^{* *}$	${.324}^{* *}$
					(.133)	(.136)
Evangelical					.203	.187
					(.140)	(.143)
Religion important					.041	.004
					(.205)	(.204)
Religious conservative					${1.09}^{* * *}$	${.735}^{* * *}$
					(.182)	(.192)
Republican						${.400}^{* * *}$
						(.149)
Democrat						$- {.456}^{* * *}$
						(.109)
Observations	1765	1692	1641	1628	1608	1598
Pseudo $R^{2}$	.04	.08	.11	.12	.18	.21

Note: The figures reported are the coefficients obtained from probit estimation.

Standard errors in parentheses. *, **, and *** denote significance at 10%, at 5%, and 1% levels, respectively. The data are from the Survey of American Political Culture, 2016.

Interestingly, columns (2) to (6) show the se two results to be robust to the introduction of additional variables. Specifically, in column (2), we introduce age, gender, income, and the education level. While income and age do not seem to play a role, males and individuals with less than college education¹⁹ are systematically more likely to consider there is a political correctness problem. In turn, race is introduced from column (3) and the only significant coefficient is a systematically lower perception of a political correctness problem by Hispanics. From column (4), individuals living in neighborhoods with a higher density are shown to systematically have a lower perception, while no significant regional effect is generally found. In columns (5) and (6) different variables on the religious beliefs/attitudes to religion are introduced, and both Catholics and individuals perceiving themselves as conservatives in terms of religion believe the problem to be more acute. Finally, and unsurprisingly, column (6) shows that individuals declaring to be Republicans (resp. Democrats) are more (resp. less) likely to perceive the existence of a political correctness problem.

6. Conclusion

If the views individuals hold are partly based on their own experiences or interests, and group belonging is correlated with specific experiences or interests, one would expect that an association is made between having certain views and belonging to a certain group. For this reason, this paper argues that the prevalence of political correctness cannot be fully understood if one abstracts from considering that the prestige of orthodox ideas is partly linked to the prestige of the elites perceived as representing them.

Specifically, while in a standard model self-denial is unlikely when peer pressure is low, the link with elite prestige makes self-denial a possible equilibrium outcome. Conversely, full self-denial becomes the unique equilibrium when the masses care sufficiently about the behavior of the elite in situations in which self-denial is just one of the possible outcomes in the standard model. In addition, we show that exogenous increases in the range of socially acceptable views do not necessarily imply anymore that a larger set of views are actually expressed at equilibrium.

Our model considers a linear distribution of views and assumes that heterodox views are only located at one extreme of the distribution. An extension of the model where heterodox views are situated at both ends of the distribution might be useful as in some cases observed changes in the Overton window may imply that formerly ‘very’ orthodox views become heterodox, as implied for instance by the cultural backlash theory (Norris and Inglehart, 2019).

Footnotes

Acknowledgements

We would like to thank Antonio Cabrales, Jonathan Hopkin, Julie Mallet, Simon Susen, two anonymous referees, the editor (Torun Dewan) and seminar participants at Kingston University, City University of London, the African Meeting of the Econometric Society (Rabat), NICEP Conference Nottingham and ETH Zürich Workshop on Democracy for helpful comments and discussions, as well as James Hunter and Carl Bowman from the UVA Institute for Advanced Studies for sharing the data on their 2016 Survey of American Political Culture and the Gallup Organization for fielding the study. J Ortega thanks the Department of Government and the European Institute at the LSE for their hospitality while revising this paper.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Spanish Agencia Estatal de Investigación, through the Severo Ochoa Programme for Centers of Excellence in R&D (Barcelona School of Economics CEX2019-000915-S); from the Spanish Ministry of Science, Innovation and Universities through project PGC2018-097898-B-100 MCIU/AEI/FEDER, EU; from the Government of Catalonia under project 2017 SGR 1571 AGAUR Generalitat de Catalunya and from the Spanish Ministry of Science and Innovation through research project PID2021-126209OB-I00 funded by MCIN-AEI/10.13039/501100011033 and by ERDF A way of making Europe.

ORCID iD

Javier Ortega

Appendix A. Increase in the orthodoxy range in the simple model

It is easy to show that $\partial \tilde{d} / \partial k_{R} = - α v_{T} / [α - (\bar{k} - k_{R})]^{2}$ and thus $\partial (k_{R} + \tilde{d}) / \partial k_{R} = ([α - (\bar{k} - k_{R})]^{2} - α v_{T}) / [α - (\bar{k} - k_{R})]^{2}$ , which is negative if and only if $v_{T} > [α - (\bar{k} - k_{R})]^{2} / α$ . As we are in an interior equilibrium, this needs to be compatible with $α > (\bar{k} - k_{R}) + v_{T} \Leftrightarrow v_{T} < α - (\bar{k} - k_{R})$ . A necessary condition for both conditions to hold simultaneously is $α - (\bar{k} - k_{R}) > [α - (\bar{k} - k_{R})]^{2} / α ⟺ 0 > - (\bar{k} - k_{R})$ , which is a contradiction, and thus $\partial (k_{R} + \tilde{d}) / \partial k_{R} > 0$ always.

Appendix B. Model with elites

Notes

References

Asch

(1955) Opinions and social pressure. Scientific American 193(5): 31–35.

Atkin

Colson-Sihra

Shayo

(2021) How do we choose our identity? A revealed preference approach using food consumption. Journal of Political Economy 129(4): 1193–1251.

Bernheim

(1994) A theory of conformity. Journal of Political Economy 102(5): 841–877.

Bourdieu

(1979) Symbolic power. Critique of Anthropology 4: 77–85.

Bourdieu

Boltanski

(1976) La production de l’idéologie dominante. Actes de la Recherche en Sciences Sociales 2-3: 3–73.

Braghieri

(2024) Political correctness, social image, and information transmission. American Economic Review 114(12): 3877–3904.

Bursztyn

Egorov

Fiorin

(2020) From extreme to mainstream: The erosion of social norms. American Economic Review 110(11): 3522–48.

Clarke

Newman

(2017) People in this country have had enough of experts’: Brexit and the paradoxes of populism. Critical Policy Studies 11(1): 101–16.

Conway

Repkea

Houck

(2017) Donald Trump as a cultural revolt against perceived communication restriction: Priming political correctness norms causes more Trump support. Journal of Social and Political Psychology 5(1): 244–259.

10.

Domínguez

Taing

Molenberghs

(2016) Why do some find it hard to disagree? An fMRI study. Frontiers in Human Neuroscience 9: 718.

11.

Duffy

Lafky

(2021) Social conformity under evolving private preferences. Games and Economic Behavior 128: 104–124.

12.

Fatas

Heap

SPH

Arjona

(2018) Preference conformism: An experiment. European Economic Review 105: 71–82.

13.

Funke

(2016) How accurate are surveyed preferences for public policies? Evidence from a unique institutional setup. Review of Economics and Statistics 98(3): 442–454.

14.

Goeree

Holt

Palfrey

(2016) Quantal Response Equilibrium: A Stochastic Theory of Games. Princeton, NJ: Princeton University Press.

15.

Goeree

Holt

Palfrey

(2020) Stochastic game theory for social science: A primer on quantal response equilibrium. In: Handbook of Experimental Game Theory. Edward Elgar Publishing, pp.8–47.

16.

Golman

Loewenstein

Moene

, et al. (2016) The preference for belief consonance. The Journal of Economic Perspectives 30(3): 165–187.

17.

Hallman

Spiro

(2023) A theory of hypocrisy. Journal of Economic Behavior & Organization 211: 401–410.

18.

Hopkin

(2020) Anti-System Politics: The Crisis of Market Liberalism in Rich Democracies. Oxford: Oxford University Press.

19.

Hunter

Bowman

(2016) The Vanishing Center of American Democracy: The 2016 Survey of American Political Culture. Charlottesville: University of Virginia, Institute for Advanced Studies in Culture.

20.

Kuran

(1987) Preference falsification, policy continuity and collective conservatism. The Economic Journal 97(387): 642–665.

21.

Loury

(1994) Self-censorship in public discourse: A theory of “political correctness” and related phenomena. Rationality and Society 6(4): 428–461.

22.

Michaeli

Spiro

(2015) Norm conformity across societies. Journal of Public Economics 132: 51–65.

23.

Morris

(2001) Political correctness. Journal of Political Economy 109(2): 231–265.

24.

Noelle-Neumann

(1974) The spiral of silence: A theory of public opinion. Journal of Communication 24(2): 43–51.

25.

Norris

Inglehart

(2019) Cultural Backlash. Trump, Brexit, and Authoritarian Populism. Cambridge: Cambridge University Press.

26.

Prentice

Miller

(1993) Pluralistic ignorance and alcohol use on campus: Some consequences of misperceiving the social norm. Journal of Personality and Social Psychology 64(2): 243.

27.